Sunteți pe pagina 1din 28

International Journal of Ambient Energy

ISSN: 0143-0750 (Print) 2162-8246 (Online) Journal homepage: https://www.tandfonline.com/loi/taen20

Solar assisted heat engine systems: multi-


objective optimisation and decision making

R. Venkata Rao & Hameer Singh Keesari

To cite this article: R. Venkata Rao & Hameer Singh Keesari (2019): Solar assisted heat engine
systems: multi-objective optimisation and decision making, International Journal of Ambient Energy,
DOI: 10.1080/01430750.2019.1636870

To link to this article: https://doi.org/10.1080/01430750.2019.1636870

Accepted author version posted online: 26


Jun 2019.
Published online: 03 Jul 2019.

Submit your article to this journal

Article views: 13

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=taen20
INTERNATIONAL JOURNAL OF AMBIENT ENERGY
https://doi.org/10.1080/01430750.2019.1636870

Solar assisted heat engine systems: multi-objective optimisation and decision making
R. Venkata Rao and Hameer Singh Keesari

Department of Mechanical Engineering, Sardar Vallabhbhai National Institute of Technology, Surat, India

ABSTRACT ARTICLE HISTORY


Identifying, analysing and optimising the effect of various parameters of a solar assisted heat engine sys- Received 11 March 2019
tem contribute significantly in assessing the system’s performance. Optimum parameters can be identified Accepted 24 June 2019
by performing multi-objective optimisation (MOO) considering multiple performance factors. Usually, non- KEYWORDS
dominated solutions set can be found for these MOO problems and selecting a most suitable solution from Solar-driven heat engine;
non-dominated solutions set is difficult. Hence, in this study, the applicability of multi attribute decision- solar assisted Brayton heat
making methods in MOO problems has been discussed and a decision making procedure is suggested on engine; multi-objective
the basis of average rank to identify the best solution in a Pareto front. Two MOO case-studies of solar- optimisation; multi-attribute
assisted engine systems are optimised by using the multi-objective-Jaya-algorithm and its variants known decision making; Jaya
as multi-team perturbation guiding Jaya algorithm and adaptive multi-team prturbation guiding Jaya algorithm
algorithm. In both the case studies, the proposed algorithms have improved the performance of the
considered systems. Furthermore, the performances of the proposed algorithms are compared using the
coverage, spacing, and hypervolume indicators.

1. Introduction Similarly, the applications of various heat engines in extract-


The limited availability and ecological impacts of energy ing solar energy have been widely studied. Ahmadi et al. “Multi-
resources such as fossil fuels and increased energy demand Objective Thermodynamic-Based Optimization,” (2013) investi-
during recent years have resulted in the development of new gated the application of the Stirling engine powered by solar
techniques and tools that could harvest the energy from the energy and proposed a mathematical model that can be utilised
renewable energy sources. Solar power is a low-grade renewable for finding the optimal design of the proposed system. The
energy resource which is vastly available. One of the best ways proposed system considers regenerative losses, thermal bridg-
of extracting solar energy is solar assisted heat engine. A typical ing losses, and finite rate heat transfer. Furthermore, process
solar aided heat engine includes a solar concentrator to focus parameters were obtained to attain optimum power, thermal
the sunlight at the focus point of the concentrator. The receiver efficiency, and entropy generation by employing simultaneous
located at the focus point of the concentrator absorbs the heat optimisation through the NSGA-II algorithm. The final reported
energy and transfers that to heat engine which in turn produces process parameters were selected by using the TOPSIS, LIN-
the useful work. MAP, and fuzzy decision-making methods. Also, Ahmadi et al.
Many researchers and scientists have focused upon devel- “Thermo-Economic Multi-Objective Optimization,” (2013) stud-
oping and optimising the solar aided heat engines in order to ied the effect of the various parameters such as regenerator
extract and utilise solar energy more efficiently. Furthermore, effectiveness, volumetric ratio, absorber temperature, and tem-
researchers have studied various characteristics of these sys- perature ratio on the performance of the solar-powered Stirling
tems considering system efficiency, output power, economic, system. Pareto optimal design points are obtained through the
and ecological aspects as performance measures and presented MOO using NSGA-II algorithm to achieve maximum power and
the optimal design and process parameters. Ahmadi et al. “Opti- thermal efficiency of the system. Three final optimum design
mal Design of a Solar Driven,” (2013) presented an optimisa- points were reported using decision making methods such
tion model of an irreversible solar-driven heat engine on the as the fuzzy bellman-Zadeh, TOPSIS, and LINMAP. Similarly,
basis of thermodynamic and thermo-economic criteria. Further- Ahmadi et al. “Designing a Solar Powered Stirling Heat Engine,”
more, optimum designs were obtained through multi-objective (2013) presented the optimum design of a solar-powered Stir-
optimisation (MOO) using the non-dominated sorting genetic ling system through the MOO of the thermo-economic function,
algorithm (NSGA-II) to achieve maximum power and thermal thermal efficiency, and dimensionless power output.
efficiency of the system. Finally, three best designs were selected Duan et al. (2014) presented the thermodynamic analysis
using the technique for order preference by similarity to ideal of a Stirling heat engine system with internal irreversibilities
solution (TOPSIS), linear programming technique for multidi- and regenerative losses. Then, the proposed model was opti-
mensional analysis of preference (LINMAP), and fuzzy decision- mised using the multi-objective particle swarm optimisation
making methods. (MOPSO) algorithm by taking power output, thermal efficiency,

CONTACT Hameer Singh Keesari hameerkeesari@yahoo.co.in

© 2019 Informa UK Limited, trading as Taylor & Francis Group


2 R. V. RAO AND H. S. KEESARI

and irreversibility parameter as the simultaneous optimisa- Ahmadi, Ahmadi, and Feidt (2016) presented a thermody-
tion goals. In addition, the best solution was identified from namic model of solar driven regenerative irreversible Bray-
the Pareto front obtained by the MOPSO algorithm using the ton cycle system and identified optimal process parameters
LINMAP method of decision making. Toghyani, Kasaeian, and of the system to achieve maximum power and thermal effi-
Ahmadi (2014) studied the power loss and efficiency of a Stir- ciency through MOO using the NSGA-II algorithm. Further-
ling engine through non-ideal adiabatic analysis and optimised more, the best Pareto optimal solutions using the TOPSIS, fuzzy,
the thermal efficiency and the power loss simultaneously using and LINMAP decision-making methods were reported. Ahmadi
the NSGA-II algorithm. Furthermore, the Pareto optimal solu- et al. “Thermodynamic Analysis and Multi Objective Optimiza-
tions obtained by the NSGA-II are compared using the TOPSIS, tion,” (2016) presented the thermodynamic analysis of a solar
LINMAP, and fuzzy methods. dish Stirling system with high-temperature differential, and
Sayyaadi, Ahmadi, and Dehghani (2015) presented the ther- simultaneous optimisation was carried out using the NSGA-II
mal and ecological models of a solar driven Carnot-like heat algorithm to obtain maximum thermal efficiency and power
engine with a finite-temperature difference. Also, optimisation output. Finally, three solutions were reported from the Pareto
was performed on single and MOO scenarios to identify the opti- solutions obtained by NSGA-II using the TOPSIS, fuzzy, and LIN-
mal design parameters which maximise the thermal efficiency, MAP methods on the basis of nearness to the preferred solution.
power output, and ecological function. Furthermore, three opti- Similarly, Ahmadi and Ahmadi (2016b) employed the NSGA-II
mal design points were selected in the MOO scenario by employ- algorithm for simultaneous optimisation of an irreversible radia-
ing the LINMAP, Bellman Zadeh, and TOPSIS decision making tive type heat engine, and decision making was performed using
methods. Li, Liao, and Liu (2015) presented the thermodynamic the TOPSIS and LINMAP methods.
model of an integrated solar-assisted irreversible open Bray- Arora, Kaushik, and Kumar (2016; Arora et al. “Multi-Objective
ton engine system with thermal bridging losses and an imper- Thermo-Economic Optimization,” 2016) performed multi-
fect solar collector to obtain power output, thermo-economic objective optimisation of a solar-driven Stirling system with
performance, and thermal efficiency. The proposed model was regenerative losses by employing the NSGA-II algorithm. The
optimised using NSGA-II algorithm, and optimum solutions best Pareto solutions were reported using the fuzzy, TOPSIS,
were reported using LINMAP, Shannon entropy, and TOPSIS Shannon’s entropy, and LINMAP decision making methods. In
decision making methods. Furthermore, the NSGA-II algorithm their another work, Arora, Kaushik, and Arora (2016) performed
was employed by Sadatsakkak, Ahmadi, and Ahmadi (2015) for the MOO of the two-stage thermo-electric generator using the
thermo-economic optimisation of an irreversible closed regen- NSGA-II algorithm and identified the best Pareto optimal solu-
erative Brayton cycle by taking the power output, ecological tions using the fuzzy, Shannon’s entropy, TOPSIS, and LINMAP
and thermo-economic functions as MOO goals. In addition, decision making methods. Luo et al. (2016) proposed a com-
the decision-making methods such as the TOPSIS and LIN- bined framework of differential evolution, genetic algorithms
MAP were employed to compare the Pareto optimal solutions and adaptive simulated annealing algorithms for the MOO of the
obtained. Ahmadi, Ahmadi, and Sadatsakkak (2015) presented performance of the GPU-3 Stirling engine system. A final opti-
thermodynamic analysis and multi-objective optimisation of an mal solution was identified from the Pareto front of the proposed
irreversible Carnot refrigerator system by considering internal approach using the TOPSIS and simple additive weighting (SAW)
and external irreversibilities of the system. The NSGA-II method decision-making methods. Bellos, Tzivanidis, and Antonopoulos
was employed for multi-objective optimisation considering the (2017) investigated the solar energy supported gas turbine sys-
ecological coefficient of performance, exergy input, cooling load tem and optimised the same to attain optimum collector area,
and exergetic performance as the objectives of optimisation. electricity production, and fuel consumption. Jokar et al. (2017)
Furthermore, by applying the LINMAP, TOPSIS, and fuzzy meth- presented a comprehensive modelling, parametric evaluation,
ods, optimal solutions were identified from the Paretofront. and multi-objective optimisation of a hybrid system consisting
Similarly, the optimum system parameters were determined for of a Brayton cycle and a molten carbonate fuel cell. Tiwari, Sher-
irreversible refrigeration systems by Ahmadi and Ahmadi (2015, wani, and Arora (2017) investigated the performance of solar-
2016a). driven organic Rankine cycle for different zeotropic mixtures and
Ahmadi et al. “Designing a Powered Combined Otto,” (2016) presented the optimum mass fraction by employing the multi-
presented a thermodynamic analysis of combined Stirling and objective genetic algorithm. Hooshang et al. (2017) employed
Otto cycle for power production, and multi-objective optimi- the NSGA-II algorithm for multi-objective optimisation of power
sation was performed for optimal power output and thermal output and regenerator pressure of a Stirling engine system and
efficiency of the Stirling engine using the NSGA-II algorithm. used TOPSIS, LINMAP, and fuzzy methods for decision making.
Also, three optimal solutions were reported by employing TOP- Carrillo Caballero et al. (2017) presented a mathematical
SIS, LINMAP, and fuzzy methods. Ahmadi et al. “Optimization model to determine the efficiency and energy production of a
of Powered Stirling Heat Engine,” (2016) presented a finite solar assisted Stirling engine system consisting of the directly
speed thermodynamic analysis of Stirling heat engine consid- illumined receiver. In addition, multi-objective optimisation
ering internal irreversibilities such as fluid friction and imper- was performed using NSGA-II algorithm and optimum system
fect regeneration. Furthermore, optimal power output and rota- parameters were identified to achieve maximum efficiency and
tional speed were determined for the proposed model. The power generation. Also, they had identified the best Pareto opti-
effects of the system parameters such as heat source temper- mal solution by employing the SAW decision-making method.
ature and volumetric efficiency on system performance were Shirazi et al. (2017) presented a MOO model of solar-driven
presented. heating and cooling chillers system on the basis of energetic,
INTERNATIONAL JOURNAL OF AMBIENT ENERGY 3

environmental and economic analysis aspects. By using the minimum investment cost and maximum thermal efficiency
multi-objective genetic algorithm, optimisation was performed and the best solution was identified from the Pareto front
to find optimal total annual cost and energy use of the plant. on the basis of non-dimensional Euclidian distances from the
From the Pareto fronts obtained, the best solution was reported ideal solution. Similarly, Khoshbazan et al. (2018) presented the
using the LINMAP method. Ahmadi et al. (2017) optimised the thermo-economic analysis of a micro-CHP system with solar-
thermodynamic model of a solar driven small scale transcrit- powered Stirling engine for three cities located in Iran and per-
ical CO2 power cycle system with the liquid natural gas heat formed multi-objective optimisation using evolutionary algo-
sink. The NSGA-II algorithm was employed for MOO considering rithms to optimise system efficiency, carbon dioxide emission
thermal efficiency, solar fraction and the total investment of the reduction and economic aspects. From the optimisation results,
plant as the objectives. In addition, the TOPSIS, fuzzy, and LIN- it was observed that the cities climates have a considerable
MAP decision-making methods were employed to identify the effect on the performance of the system. Ahmadi et al. “Multi-
best solution from the Pareto solutions obtained by the NSGA-II Objective Performance Optimization,” (2018) presented a ther-
algorithm. modynamic model of integrated molten carbonate fuel cell
Arora, Kaushik, and Kumar (2017) presented a finite time Braysson heat engine system and performed multi-objective
thermodynamic based mathematical model and performed optimisation using the NSGA-II algorithm. The four objectives
multi-objective optimisation for solar energy driven Stirling considered for the optimisation were energy efficiency, exergy
heat engine system with regenerative heat losses, finite rate destruction rate density, power density, and ecological function
heat transfer, thermal bridging losses, and regeneration pro- density. In addition, three final solutions were selected from the
cess time. Also, by employing the NSGA-II algorithm optimum Pareto front obtained by the NSGA-II using LINMAP, TOPSIS, and
design parameters of the system were identified for maximum fuzzy decision-making methods. Jing et al. (2018) presented a
power generation, ecological function and thermal efficiency. combined framework of MOO and multi-criteria decision mak-
Furthermore, the LINMAP, Shannon entropy, TOPSIS, and fuzzy ing for an optimal designing and planning of solar assisted solid
decision-making methods were employed to determine the best oxide fuel cell distributed power system. This framework con-
Pareto optimal solutions on the basis of closeness to the ideal sisted of two stages: the optimisation stage and the evaluation
solution. Ashouri et al. (2018) presented the exergo-economic stage. In the optimisation stage MOO was carried out by the
analysis of solar assisted organic Rankine cycle system and with mixed integer linear programming method and by employing
the aid of the proposed model showed the effect of parame- the TOPSIS, LINMAP and Shannon entropy methods three best
ters on the system performance. Also, bi-objective optimisation Pareto optimal solutions were identified. In the evaluation stage,
is performed through a multi-objective genetic algorithm con- six case studies of integrated systems were evaluated and ranked
sidering exergy efficiency and product cost rate as objectives. using an integrated evaluation model that combined the ana-
In addition, the best Pareto optimal point was identified using lytic hierarchy process (AHP), and the grey relational analysis
the TOPSIS decision-making method. Dai et al. (2018) presented (GRA) approaches.
the finite time thermodynamic model of Stirling heat engine by Ahmadi, Nazari, and Feidt (2018) presented a finite-time
considering finite rate heat transfer, conductive thermal bridg- thermodynamic model to obtain thermal efficiency, the output
ing losses, and regenerative heat losses. The proposed model power and the entropy generation rate of an endoreversible
was then optimised using MOPSO algorithm to find the opti- Lenoir heat engine cycle. Furthermore, multi-objective optimi-
mal parameters of the system. In addition, the TOPSIS method sation was carried out using NSGA-II approach to find optimal
was employed to identify the best solution from the Pareto front values of the ecological coefficient of performance and dimen-
which has a better tradeoff between the power, thermal effi- sionless thermo-economic function of the proposed model. Ge
ciency and economic coefficient of performance of the system. et al. (2018) presented the analytical expressions for-profit rate
Nazemzadegan et al. (2018) performed a many-objective and efficiency based on finite time thermodynamic analysis of an
optimisation of solar dish Stirling engine system. The dimen- irreversible intercooled regenerated gas turbine cycle. Ahmadi
sionless objectives of their study include the power gener- et al. “Multi-Objective Performance Optimization,” (2018) pre-
ated, efficiency, entropy generation and economic factors of sented the parametric evaluation and Multi-objective perfor-
the system. The optimisation is carried out by the branch and mance optimisation of an integrated system that contained a
bound method with priory articulation of preferences and by molten carbonate fuel cell, Stirling engine, and a reverse osmo-
the MOPSO algorithm with the posteriori approach. Also, by sis desalination unit. The objectives of optimisation were the
implementing the fuzzy, LINMAP, and TOPSIS decision-making energy efficiency, the exergy destruction rate density, the fresh
methods identified three best Pareto optimal solutions from the water production rate, and the ecological function density.
Pareto front. From the optimisation results, it was observed that Rashidi and Khorshidi (2018) presented a mathematical
the optimal dimensionless power and thermal efficiency are in model for a solar based multi-generation system consisting of
the range of 0.33–0.34 and 0.35–0.37 respectively. desalination unit, water heater, organic Rankine cycle evapora-
Furthermore, solar assisted systems applications related to tor, PV solar collectors, and a single effect absorption chiller unit
several integrated energy systems have been researched. Fer- by performing energy and exergy analysis of the system. In addi-
reira et al. (2017) presented a methodology to design and opti- tion, by taking total cost rate and energy efficiency as the objec-
mise a combined heat and power (micro-CHP) system with solar tives MOO was performed out using multi-objective differential
assisted Stirling heat engine on the basis of climatic data, refer- evolution (MODE) algorithm. From the Pareto front obtained
ence building characteristics and energy consumption. The pro- by the MODE algorithm, the best solution was identified using
posed model was optimised using NSGA-II algorithm to achieve fuzzy decision-making method. Harkouss, Fardoun, and Biwole
4 R. V. RAO AND H. S. KEESARI

(2018) presented the design parameters analysis of a net zero solutions from the Pareto optimal solutions. The Pareto optimal
energy building (NZEB) design problem consisting of a solar hot solutions were ranked by these methods. However, different
water system unit and PV array. They had presented a simulation decision-making methods follow different principles to rank the
model which was then optimised using NSGA-II algorithm. Multi- Pareto solutions. Hence, the ranking of the Pareto optimal solu-
ple objectives namely, life-cycle cost, auxiliary electric heater and tions by these methods will be different. Also, ranks given by
pump consumptions, thermal demands for cooling and heating, these methods vary depending upon the problem data. Thus,
and exports were considered for design optimisation of an NZEB implementing these decision-making methods may not be suffi-
system. In addition, a multi-criteria decision-making framework cient to address the problem of selecting the best solutions from
using ELECTRE III method was implemented to identify the best the Pareto front. Hence in this work, a selection method based on
solution from the Pareto optimal solutions. Habibollahzade et al. the average rank is presented by implementing it to two MOO
(2018) presented an integrated power generation system which case studies of solar assisted heat engine systems.
consisted of a biomass-based solid oxide fuel cell (BSOFC), a Stir- In this work, optimisation is carried out using the multi-
ling engine, and an electrolyser. In this system, the waste heat of objective Jaya algorithm and its variants known as multi-team
BSOFC unit was utilised to run the Stirling engine, and the excess perturbation guiding Jaya (MTPG-Jaya) algorithm, and adaptive
power generated by the Stirling engine was then used for hydro- multi-team perturbation guiding Jaya (AMTPG-Jaya) algorithm.
gen production by proton exchange membrane electrolyser. In The posteriori method is followed during the optimisation of
addition, by considering the system exergy efficiency, hydrogen these case studies. Due to the conflicting nature of the objec-
production and total product cost as the optimisation objectives tives considered in this work, selecting the best solution out of
MOO was performed to obtain system design parameters using these Pareto optimal solutions can be mimicked as a multiple
the genetic algorithm. attribute decision making (MADM) problem. To identify the best
Entezari, Manizadeh, and Ahmadi (2018) studied the inte- solution from a Pareto front, eight MADM methods namely sim-
grated gas turbine and Stirling engine power plant system ple additive weighting (SAW), weighted product method (WPM),
and obtained thermodynamic and economic optimal operat- preference ranking organisation method for enrichment evalua-
ing parameters of the system which optimised cost energy and tions (PROMETHEE), technique for order preference by similarity
exegetic efficiency. By employing the multi-objective genetic to ideal solution (TOPSIS), modified TOPSIS technique (MTOP-
algorithm, the optimisation was performed. Also, using the TOP- SIS), compromise ranking method (VIKOR), complex propor-
SIS method the best Pareto solution was identified that has tional assessment (COPRAS), and gray relational analysis (GRA)
16% more exegetic efficiency and 10% less Levelized cost of are employed. The average rank is calculated on the basis of
energy compared to stand alone gas turbine power plant. Mah- the rank given by the MADM methods. The contributions of this
moudimehr and Sebghati (2019) proposed a multi-objective work are:
dynamic programming based optimisation method to optimise
the performance of a solar thermal power plant. The daily power • Multi-objective optimisation of two case studies of solar
generation and daily revenue generated were considered as the assisted heat engines has been carried out to see if there
performance criteria. From the obtained Pareto front, a solu- can be any improvement possible in the performances of the
tion having better trade-off among the objectives was identified considered case studies.
on the basis of proximity to the ideal solution and compared • In both case studies, the proposed algorithms have attained
with that of the NSGA-II algorithm. Behzadi et al. (2019) pro- Pareto optimal solutions that have improved the perfor-
posed a solar based integrated energy system and presented mance of the solar assisted systems.
optimal design parameters to maximise exergy efficiency and • Decision making based on the average rank by the multi-
hydrogen production rate, and minimise total cost rate, through attribute decision-making methods has been suggested for
multi-objective optimisation of the system using the genetic selection of the best solutions from the Pareto front.
algorithm.
From the above literature, it can be observed that the effect The subsequent section presents the working principle of
of various parameters of a solar driven energy system had been the multi-objective optimisation algorithms considered in this
researched in assessing the performance of the system. The per- work.
formances of these systems had been studied and optimised
simultaneously on thermodynamic, thermo-economic, ecologi-
2. Multi-objective optimisation with Jaya algorithm
cal and economic aspects. The multi-objective optimisation con-
and its variants
sidering these aspects was performed by metaheuristics such as
NSGA-II, MOPSO, and MODE. For a MOO problem, these algo- Identifying a global best solution that gives optimum values for
rithms suggest a set of non-dominated Pareto optimal solutions. all the objectives of the multi-objective optimisation problem is
As these solutions are non-dominated in nature, identifying the difficult when the objectives are conflicting in nature. The values
best solution from these will be difficult due to the conflicting of decision variables which produce the most optimum value
nature of the objectives. for one objective may produce non-optimal values for the other
The solution with the most suitable compromise among the objectives. In such situations, the decision maker can convert
selected objectives may be considered as the best solution the MOO problem into a single objective optimisation problem
in a Pareto front. Various researchers had employed decision- using the priority information among the objectives and can
making methods such as the TOPSIS, ELECTRE-III, Shannon find the global best solution respective to the priorities. If the
entropy, AHP, LINMAP, SAW, fuzzy, and GRA to identify the best priorities of the objectives are not known, then Pareto optimal
INTERNATIONAL JOURNAL OF AMBIENT ENERGY 5

solutions can be found using the posteriori approach (Deb et al. algorithm. The Jaya algorithm is a population-based algorithm
2002) for MOO problems. The Pareto optimal solutions are the which moves the solutions towards the best solutions while
solutions in which at least one objective function value of each avoiding the movement towards the worst solution of the pop-
solution is better than the other solutions in the Pareto-front. ulation. For an optimisation problem consisting of D number of
In this work, the multi-objective Jaya algorithm and its independent variables, the movement of the population during
variants known as multi-objective MTPG-Jaya and AMTPG-Jaya the ith iteration is governed by the following equation:
algorithms are employed for optimisation. In multi-objective
optimisation using these algorithms, posteriori articulation of sv,p,i = sv,p,i + r1,v,i (sv,best,i − |sv,p,i |) − r2,v,i (sv,worst,i − |sv,p,i |)
preferences is followed to handle multiple objectives simulta- (1)
neously. By using dominance principles and crowding distance
measurements (Deb et al. 2002; Rao, Rai, and Balic 2016) solu- Where, sv,p,i and sv,p,i are the updated and old vth variable
tions are ranked during the iterative process of the algorithms. values for the pth candidate solution respectively. The terms
Furthermore, MADM methods are employed to identify the best sv,best,i and sv,worst,i are the vth variable values of best and worst
Pareto optimal solution from the Pareto fronts of these algo- solutions respectively. The terms r1,v,i and r2,v,i are two ran-
rithms. In addition, the performances of these algorithms are domly generated numbers between 0 and 1. For more details
evaluated using MOO performance indicators such as the cov- about working of the Jaya algorithm, the readers may refer to:
erage, hypervolume, and spacing. https://sites.google.com/site/Jayaalgorithm/.

2.1. Jaya algorithm 2.2. Multi-team perturbation guiding Jaya algorithm


Jaya algorithm developed by Rao (2016, 2019) is simple and has The Multi-team perturbation guiding Jaya algorithm is a multi-
no algorithm-specific parameters. The effectiveness of the Jaya population variant of the Jaya algorithm introduced by Rao and
algorithm had been confirmed using various single and multi- Keesari (2018). Unlike the basic Jaya algorithm, the MTPG-Jaya
objective optimizations applications (Du et al. 2018; Michailidis algorithm uses multiple perturbation equations simultaneously
2017; Rao and Saroj 2017, 2018; Rao et al. “Dimensional Opti-
mization of a Micro-Channel,” 2016; Rao et al. “A New Multi-
Objective Jaya Algorithm,” 2016; Rao et al. 2017, 2018; Wang and
Huang 2018; Wang et al. 2017, 2018; Warid et al. 2016, 2018;
Zhang et al. 2016). Figure 1 shows the flowchart of the Jaya

Figure 2. MTPG-Jaya algorithm flowchart for multi-objective optimisation prob-


Figure 1. Jaya algorithm flowchart for multi-objective optimisation problems. lems.
6 R. V. RAO AND H. S. KEESARI

during the iterative search process. The MTPG-Jaya algorithm become a tedious work when solving complex problems. The
makes use of six movement equations to direct the population AMTPG-Jaya algorithm adapts the number of teams in exploring
during the search process. Let the algorithm is initialised with the search domain. In this variant, the number of teams is mod-
NT (1 < NT < 7) number of teams, during the iterative search the ified online on the basis of the improvement of the best so far
population is guided by NT perturbation equations in each iter- solution. However, in this work, the quantity of teams is mod-
ation. This means that during any iteration the set of the popu- ified on the basis of the function evaluations consumed. Also,
lation will move in different directions of the search space based to further reduce the complexity of the algorithm the stagna-
on the perturbation equation guiding it. Furthermore, the MTPG- tion treatment is removed in the present work. Figure 3 shows
Jaya algorithm modifies the perturbation equations of a worst the flowchart of the AMTPG-Jaya algorithm which uses non-
performing team with a better performing team’s perturbation dominance principles and crowding distance measurement for
equation on the basis of the quality of the solutions produced multi-objective optimisation case studies. More details about
by it. In addition, when salutions are stagnated non-improved the working of the AMTPG-Jaya algorithm are available in Rao
solutions are introduced in the population during the iterative et al. (2019a, 2019b).
process. Figure 2 shows the flowchart of the multi-objective
MTPG-Jaya algorithm which uses non-dominance principles and
2.4. Performance indicators
crowding distance measurement to handle the multiple objec-
tives at the same time. For more details about the working of By employing the posteriori approach to solve MOO problems
the MTPG-Jaya algorithm readers may refer to Rao and Keesari diverse set of non-dominated solutions can be obtained. Hence,
(2018). to assess the performances of the algorithms in solving MOO
problems three performance indicators are adopted in this work,
and they are as follows:
2.3. Adaptive multi-team perturbation guiding Jaya
algorithm 2.4.1. Coverage
The AMTPG-Jaya algorithm proposed by Rao et al. (2019a) is a Zitzler, Deb, and Thiele (2000) proposed this performance indi-
modified version of the MTPG-Jaya algorithm. Depending on the cator to compare and calculate the percentage of solutions in a
optimisation case the number of teams and size of the popu- Pareto front (B) is dominated by the solutions in another Pareto
lation need to be tuned in the MTPG-Jaya algorithm. This may front (A). That is the coverage of the Pareto front A over the
Pareto front B is given by the following equation:

|{b ∈ B| ∃ a ∈ A : a ≺= b}|
Cov(A, B) = (2)
|B|

Where a ≺= b means a dominates b or is equal to b.


If all the solutions of Pareto front B are either dominated or
equal to all the solutions of the Pareto front A then the value
of Cov(A, B) is equal to 1. Similarly, if none of the solutions in
Pareto front B are covered by the Pareto front A then the value
of Cov(A, B) is equal to zero.

2.4.2. Spacing
This performance indicator proposed by Schott (1995) quantifies
the distribution of solutions along the Pareto front. The spac-
ing of a Pareto front consisting of n non-dominated solutions is
given by the following equation:


 1  n
S=
2
(d̄ − di ) (3)
|n − 1|
i=1


k
j
i
di = i,imin
=j |fm − fm |, i, j = 1, 2, . . . , n (4)
m=1


n
d̄ = (di /n) (5)
i=1

Where k is the number of objectives and fm is the objective


function value of the mth objective. The spacing indicator deter-
mines the uniformity of the distribution of the solutions along
Figure 3. AMTPG-Jaya algorithm flowchart for multi-objective optimisation prob- the Pareto front. For a Pareto front, zero spacing value indicates
lems. that the solutions in the Pareto front are equidistantly spread.
INTERNATIONAL JOURNAL OF AMBIENT ENERGY 7

The spacing indicator is an appropriate method when the Pareto of an alternative i, is given by the following equation:
fronts achieved by different algorithms are regular and contain

M
a set of remote solutions. Pi = (nmij )wj (8)
j=1
2.4.3. Hypervolume
The Hypervolume indicator (Rao 2019) for a Pareto front pro- Where, nmij is the normalised measure of performance, and the
vides the search space volume which is dominated by it with best alternative is the one with the highest Pi value.
respect to a specified reference point. Therefore in multi-
objective optimisation, when two are more algorithms perfor- 3.3. Preference ranking organisation method for
mances are compared higher values of Hypervolume indicate enrichment evaluations
better performance. For a Pareto front consisting of N solutions
Brans, Vincke, and Mareschal (1986) developed the PROMETHEE
the hypervolume indicator is given by the following equation:
method which belongs to the class of outranking methods. In
Hypervolume = volume(∪Ni=1 νi ) (6) this method, for each objective function, the Pareto optimal
solutions are compared pair-wise to determine the strength of
Where, vi is the hypervolume constructed with reference point
a solution ai over the solution ak . For the given weights of the
Ref and the solution i as the diagonal corners of the hyper-
objectives, in pair-wise comparison of the solutions usual pref-
cube. The following section presents the multi-attribute decision
erence function (Pj,aiak ) is considered in this work. The usual
making methods employed for decision making in this work.
preference function is the difference between the values of solu-
tions ai and ak for an objective bj . Now the objective preference
3. Multi-attribute decision-making methods (aiak ) index is given by:
Decision making in the presence of multiple conflicting criteria 
M
can be considered as multiple criteria decision making (MCDM). ajak = wj Pj,aiak (9)
Identifying the best Pareto optimal solution from a Pareto front j=1
is treated as a multiple attribute decision-making problem in this The value of ajak varies between zero and 1, and it denotes the
work. MADM methods are a class of MCDM methods which are intensity of the inclination towards the solution ai over ak , when
employed when a limited number of alternatives are available. concurrently all the objectives are compared. Now, for a solution
These methods suggest the best alternative using the attributes ai , the out ranking relations are calculated using the following
information of various alternatives. The set of Pareto optimal equations:
solutions are regarded as alternatives (Ai ∀ alternatives i = 1, 
2 . . . N), and the objective functions are regarded as attributes Leaving flow (Ø+ i )= ax ai (10)
(Bj ∀ attributes j = 1, 2 . . . M). The objective value of a non- x∈A

dominated solution i corresponding to an objective or attribute j Entering flow (Ø−
i ) = ai ax (11)
can be considered as the performance measure (mij ). The relative x∈A
importance or weights (wj ) of the objectives can be appraised
Net flow (Ø(ai )) = (Ø+ −
i ) − (Øi ) (12)
by the decision maker. The summation of weights of all the
attributes together is equal to 1. The objective functions con- The leaving flow denotes the supremacy of solution ai over all
sidered in this work have different units. Thus all the objective other solutions. The entering flow denotes the degree to which
values must be normalised (nmij ) to the same units before apply- all other solutions are dominating the solution ai . The net flow
ing the MADM methods. Based on the potential to find a best denotes outranking relationship of the solutions, and higher val-
alternative the following methods have been considered in this ues of the net flow mean the best Pareto optimal solution. For
work (Rao 2007, 2013): SAW, WPM, TOPSIS, MTOPSIS, VIKOR, more details about the PROMETHEE method, the readers may
PROMETHEE, COPRAS, and GRA. refer to Rao (2013).

3.1. Simple additive weighting 3.4. Technique for order preference by similarity to ideal
Fishburn (1967) developed the SAW method. For the given
solution
weights of the attributes, the cumulative performance (Pi ) of an The TOPSIS method developed by Hwang and Yoon (1981)
alternative i, is given by the following equation: provides the relative similarity index (SI) near to an ideal and

M non-ideal solution. The ideal and non-ideal solutions represent
Pi = Wj (nmij ) (7) the hypothetical best and worst scenarios of all the objectives
j=1 respectively. The normalised values (nmij ) of the objective func-
tion values are calculated by using the following equation:
Where, nmij is the normalised measure of performance, and the ⎛
best alternative is the one with the highest Pi value.  ⎞
  M

nmij = mij /⎝ m2ij ⎠ (13)
3.2. Weighted product method j=1

Miller and Starr (1969) developed the WPM method. For the Let, Rij be the matrix of normalised values and wj be the weights
given weights of the attributes, the cumulative performance (Pi ) of the objective functions, then the weighted normalised matrix
8 R. V. RAO AND H. S. KEESARI

can be calculated by using the following equation: 3.6. Compromise ranking method
Vij = wj Rij (14) The idea of compromise ranking was introduced by Yu (1973)
and Zeleny (1982) and later advanced by Opricovic and Tzeng
Now, identify the ideal (Oideal
j ) and non-ideal (Ononj
- ideal ) solu-
(2002, 2003). The procedure for VIKOR is described below:
tions with respect to all objective functions using the following Step 1: Identify the ideal (mideal ) and non-ideal (mnonideal )
ij ij
criteria respectively. For a maximisation-objective function (j), solutions for all the attributes based on the objective function
the solution with maximum Vij value will be considered as an values of the solutions for respective attributes using the follow-
ideal solution and the solution with minimum Vij value will be ing criteria. For a maximisation objective function (j), the solution
considered as the non-ideal solution. Similarly, for minimisation with maximum mij value will be considered as the ideal solution
objective function (j), the solution with minimum Vij value will be and the solution with minimum mij value will be considered as
considered as an ideal solution and the solution with maximum the non-ideal solution. Similarly, for minimisation objective func-
Vij value will be considered as the non-ideal solution. Then calcu- tion (j), the solution with minimum mij value will be considered
late the Pareto optimal solutions distances (di ) from the ideal and as the ideal solution and the solution with maximum mij value
non-ideal solution using the following equations respectively: will be considered as the non-ideal solution.

 M Step 2: Calculate Ei and Fi values for all Pareto optimal solu-

=
ideal 2 tions in the Paretofront.
di (Vij − Oideal
j ) (15)
j=1
 
M
 M
 Ei = wj [mideal − mij /mideal − mnonideal ] (20)
=
2 ij ij ij
dnonideal
i (Vij − Ononideal )
j (16) j=1
j=1

Where, i = 1, 2 . . . N (i.e. number of Pareto optimal candidate For a Pareto optimal solution i,
solutions or alternatives in a Pareto frontier), j = 1, 2 . . . M (i.e.
the number of objectives or attributes of the Pareto frontier). Fi = maximum of {wj [mideal − mij /mideal
ij ij
Now, the similarity index can be calculated using the following
equation: − mnonideal
ij ]| j = 1, 2 . . . M} (21)
dnonideal
i
SI = (17) Step 3: Calculate the performance index (Pi ) using the follow-
dideal
i + dnonideal
i
ing equation:
If the similarity index of a Pareto optimal solution is higher than
that of all other solutions in the Pareto-frontier, then it is the
nearest solution to the ideal solution. Similarly, if the similarity Pi = V[(Ei − Ei− )/(Ei+ − Ei− )] + (1 − V)[(Fi − Fi− )/(Fi+ − Fi− )]
index of a Pareto optimal solution is lower than that of other (22)
solutions in the Pareto-frontier, then it is the nearest solution to
a non-ideal solution. For more details about the TOPSIS method, Where Ei+ and Ei− are the maximum and minimum values of Ei
the readers may refer to Rao (2007). respectively. Similarly, Fi+ and Fi− are the maximum and mini-
mum values of Fi respectively. V is taken as 0.5.
3.5. Modified-TOPSIS Step 4: The solution with the least Pi will be considered
as the best alternative if it satisfies the following conditions
Deng, Yeh, and Willis (2000) proposed a modification to the TOP- (Tzeng, Lin, and Opricovic 2005). Condition-1 is Pimin − Pinext min ≥
SIS method by suggesting weighted Euclidean distances. In this (1/(N − 1)). Condition-2 is that the solution with least Pi must
modified version the ideal and non-ideal solutions are selected also be the best solution according to the Ei and/or Fi values.
from the normalised matrix (Rij ) and the weighted Euclidean dis- The solution with least Ei vales is considered best according to
tances are calculated from the ideal and non-ideal solutions. Ei values and the solution with least Fi value is considered as
According to the MTOPSIS method, the Pareto optimal solu- best according to Fi values. If these conditions failed, then a
tions distances (di ) from the ideal and non-ideal solution are compromise solutions set is proposed based on the following
calculated by the following equations: criteria. The solutions Pimin , Pinext min . . . Pik are compromise solu-
 tions if condition-1 failed, where Pik − Pinext min ≈ (1/(N − 1)).
 M

dideal =  wj (nmij − nmideal )
2
(18) The solutions Pimin and Pinextmin are compromise solutions if only
i j
j=1 condition-2 fails.

 M

=
2
dnonideal
i wj (nmij − nmnonideal ) j (19) 3.7. Complex proportional assessment
j=1
The complex proportional assessment method was introduced
Now the similarity index is calculated using the Equation (17). by Edmundas Kazimieras Zavadskas in 1994 (Razavi Hajiagha,
The solution with the most similarity index will be considered as Hashemi, and Zavadskas 2013). This method is simple to use and
the best solution, and the solution with the least similarity index stepwise procedure for ranking and evaluating the alternatives.
will be considered as the relatively worst solution. The stepwise procedure of COPRAS method is as follows:
INTERNATIONAL JOURNAL OF AMBIENT ENERGY 9

The normalised values (nmij ) of the objective function values Where,


are calculated by using the following equation:
ij = |nm0j − nmij |
⎛ ⎞
N
nm0j ∈ reference sequence {nm01 , nm02 , . . . , nm0M }
nmij = mij / ⎝ mij ⎠ (23)
j=1 min = min{ij , i = 1, 2, . . . , N; j = 1, 2, . . . , M}
max = max{ij , i = 1, 2, . . . , N; j = 1, 2, . . . , M}
Let, Rij be the matrix of normalised values and wj be the weights
of the objective functions, then the weighted normalised matrix Distinguishin coefficient (ξ ) = 0.5
can be calculated by using the following equation:
Now the gray relational grade can be calculated by the fol-
lowing equation:
Vij = wj Rij (24)

M
Now, calculate the summation Pi , only for the maximisation (nm0 , nmi ) = wj GR(nm0j , nmij ) for i = 1, 2, . . . , N
objectives using the following equation: j=1
(32)

k
Pi = Vij (25) If a solution has the highest gray relational grade with the ref-
j=1 erence sequence, it means that the comparability sequence is
most similar to the reference sequence and that solution would
Where k is the number of maximisation objectives. Now, calcu- be the best choice. The next section presents a brief description
late summation Ri , only for minimisation objectives using the of the MOO case studies of solar assisted heat engine systems.
following equation:


M
4. Description of solar systems considered
Ri = Vij (26)
j=k+1 Solar dish heat engine system extracts the energy from sunrays
and converts that into useful energy. A solar dish heat engine
Now, calculate the relative weight of each solution (Qi ) using the system typically consists of a heat engine and solar concentrator.
following equation: Solar concentrator reflects the sunlight at the focal point of the
 concentrator where the receiver is placed. The hot end region of

M 
M
the heat engine is placed at the receiver. The receiver then trans-
Q i = Pi + Ri / Ri (1/Ri ) (27)
mits the collected heat energy to the working fluid of the heat
i i
engine. The work output from the heat engine is then utilised to
Now, calculate the degree of utility (Ui ) for each solution using drive a generator that produces electric power. The heat engine
the following equation: of a solar dish system plays a vital role in the conversion of solar
energy into useful energy. Hence, in this work, two configura-
Ui = (Qi /Qimax ) × 100% (28) tions of solar systems which work on different thermodynamic
cycles are studied. First solar system configuration heat engine
works on a Carnot-like cycle and the latter works on the Bray-
The solution with a high degree of utility will be considered as
ton cycle. The following subsections present the thermodynamic
the best solution.
analysis of these solar system configurations.

3.8. Gray relational analysis


4.1. Solar-driven heat engine system
Gray relational analysis is one of the derived evaluation meth-
The solar-driven heat engine considered in this case study is
ods based on the concept of gray relational space (GRS). The
Carnot-like heat engine with a finite temperature difference
normalised values (nmij ) of the objective function values are
working in maximum ecological function conditions. This heat
calculated by using the following equations:
engine consists of one isothermal heat rejection process, one
mij − min{mij } isentropic heating process, one isothermal heat addition pro-
nmij = For maximization (29) cess, and one isentropic heat rejection process. The convec-
max{mij } − min{mij }
tion and radiation heat transfer between the heat source and
max{mij } − mij working fluid is considered during the isothermal heat addi-
nmij = For minimization (30)
max{mij } − min{mij } tion process. The convection heat transfer between the working
fluid and heat sink is considered during the isothermal heat
Now, calculate the gray relational coefficient (GR) using the fol- rejection process. The detailed description and thermodynamic
lowing equation: analysis of the considered solar-driven heat engine system was
presented by Sayyaadi, Ahmadi, and Dehghani (2015). Let, Th
min − ξ max
GR(nm0j , nmij ) = (31) be the working fluid temperature at the hot region of the heat
ij − ξ max engine, TH be the heat source temperature, TL be the heat sink
10 R. V. RAO AND H. S. KEESARI

Figure 4. Schematic diagram of a solar dish Brayton heat engine system.

temperature, α H,r be the radiation heat transfer coefficient at the


heat source, α H,c be convection heat transfer coefficient at the
heat source, α L,c be the convection heat transfer coefficient at
the heat sink, then the temperature ratio of working fluid and
heat source during the isothermal heat addition process (γ ),
the convective coefficient ratio (φ), the heat engine operating
temperature ratio (τ ), and heat source to heat sink allocation
parameter (β) are given by the Equations (33–36) respectively.

γ = Th /TH (33)
φ = αH,c /αL,c (34)
τ = TL /TH (35)
β = (αH,r /αL,c )TH3 (36)

Now, the thermal efficiency (η), non-dimensional power out-


put (W), and non-dimensional ecological function (E) are given
by Equations (37–39) respectively.

η = 1 − [τ/(γ − β(1 − γ 4 ) − φ(1 − γ ))] (37)


4 Figure 5. T-S diagram of a solar dish Brayton heat engine system.
W = η[β(1 − γ ) − φ(1 − γ )] (38)
E = [β(1 − γ 4 ) + φ(1 − γ )]
  engine is shown in Figure 5. The Brayton cycle (with ideal regen-
β(1 − γ 4 ) + φ(1 − γ ) − γ + 1
× η−τ (39) erator) consists of four processes. Process 1–2: an isobaric heat
γ − β(1 − γ 4 ) − φ(1 − γ )
addition. During this process, the working fluid is heated by the
regenerator (state 1–5) and then the heat source at TH temper-
4.2. Solar dish Brayton heat engine ature (state 5–2). Process 2–3: an isentropic expansion process,
Figure 4 shows the schematic diagram of solar assisted regener- in which the hot working fluid drives a turbine. Process 3–4: an
ative Brayton heat engine system considered in this work. This isobaric heat rejection process, in which heat released to the
system is driven by the combination of solar energy and fossil regenerator (state 3–6) and then to heat sink at TL temperature
fuel. The heat absorbed at the receiver is used to heat the work- (state 6–4). Process 4–1: an isentropic compression process, dur-
ing fluid before entering the combustion chamber. Hence, less ing this process the working fluid at state 4 is compressed to
fuel is burned to heat the working fluid. The solar energy from state 1 by the compressor. Let, T 1 be the temperature of the
the receiver is used as the supplement to the fuel of a solar dish working fluid at state 1, T 0 be the ambient temperature, H be
Brayton heat engine. Furthermore, the waste heat of working the hot side heat exchange effectiveness, L be the cold side
fluid exhausted from the gas turbine is used to preheat the work- heat exchange effectiveness, R be the regenerator effectiveness
ing fluid coming out of the compressor. The working fluid from and Cwf be the working fluid heat capacity rate, then the power
the combustion chamber expands rapidly and drives the turbine output is given by Equation (48).
which in turn drives the alternator. The detailed description and
thermodynamic analysis of the solar dish Brayton heat engine a1 = (1 − εH )(1 − εL )(1 − εR )εR (40)
system considered in this work was presented by Li, Liao, and
a2 = a4 T1 + a5 (41)
Liu (2015).
The T-S diagram of the solar assisted regenerative Brayton a3 = a1 T12 + a6 T1 + a7 (42)
INTERNATIONAL JOURNAL OF AMBIENT ENERGY 11

a4 = (1 − εH )(1 − εL )εR2 + (1 − εH )(1 − εL )(1 − εR )2 − 1 (43) Table 1. The algorithm parameters used in computational experiments of all case
studies.
a5 = εH (1 − εR )(1 − εL )TH + (1 − εH )εL εR TL (44)
Algorithm values
a6 = εH (1 − εL )εR TH + (1 − εH )εL (1 − εR )TL (45) Parameter Jaya MTPG-Jaya AMTPG-Jaya
a7 = εH εL TH TL (46) Population size 25 25 25
   Initial teams – 4 4
Maximum movement iterations – 20 20
a8 = −a2 − (a22 − 4a1 a3 ) /2a1 (47) Maximum function evaluations 40000 40000 40000

P = Cwf εH [TH − (1 − εR )T1 − εR a8 ]


− Cwf εL [(1 − εR )a8 + εR T1 − TL ] (48) multi-objective optimisation using these algorithms, the MADM
methods are employed to determine the best Pareto optimal
Now, the Brayton engine thermal efficiency (ηb ) and solar solution from the Pareto fronts obtained by the proposed algo-
concentrator efficiency (ηs ) are given by: rithms. Theses MADM methods include SAW, WPM, TOPSIS,
MTOPSIS, VIKOR, PROMETHEE, COPRAS, and GRA methods.
εH [TH − (1 − εR )T1 − εR a8 ] − εL [(1 − εR )a8 + εR T1 − TL ]
ηb = In the implementation of these MADM methods, the objec-
εH [TH − (1 − εR )T1 − εR a8 ] + ξ(TH − TL ) tive functions of the respective case studies are considered
(49) as attributes and Pareto optimal solutions as alternatives. In
1 both case studies, equal weights are taken for all the objec-
ηs = η0 − [hc (THavg − T0 ) + eC δ(TH4avg − T04 )] (50)
IRC tives. As described in the section-3, these MADM methods rank
the solutions differently. Therefore, to identify the final best
Where, THavg is average absorber temperature, RC is the collector solution, average ranks have been calculated. Furthermore, to
concentrating ratio, eC is the emissivity factor of the collector, hc assess the correlation and strength of the ranking given by the
is the convection heat transfer coefficient, I is the solar irradiance, MADM methods Spearman’s correlation coefficient is calculated
δ is the Stefan–Boltzmann constant, and η0 is the collector opti- between different pairs of the methods. The Spearman correla-
cal efficiency. Now, the thermal efficiency of the system is given tion coefficient is 1 when the ranking of two MADM methods
by: is identical and −1 when the ranking of two MADM meth-
η m = η s ηb (51) ods is entirely opposite. If the Spearman correlation coefficient
between two MADM methods is near to 1 means the ranking of
The non-dimensional thermo-economic performance function those two methods is similar to each other. Thus in this work,
is given by: the ranking given by different MADM methods are considered
εH [TH − (1 − εR )T1 − εR a8 ] − εL [(1 − εR )a8 + εR T1 − TL ] for calculating average ranks when they have positive Spearman
F=   correlation coefficient with all other methods. The best Pareto
H) L)
εH [TH − (1 − εR )T1 − εR a8 ] − k ln(1−ε
hH + ln(1−ε
hL optimal solution based on the average rank is then compared
(52) with the solutions achieved by other algorithms in a similar man-
ner. The results of the Jaya algorithm, MTPG-Jaya algorithm,
The following section presents the computational results of and AMTPG-Jaya algorithm are compared with those of the well
the MOO case studies obtained by the proposed algorithm. established non-dominated sorting genetic algorithm (NSGA-II).
Furthermore, the performances of the proposed algorithms are
5. Computational results of multi-objective compared using three performance indicators such as Coverage,
optimisation of selected case studies Spacing, and Hypervolume.
The proposed algorithms and MADM methods have been
5.1. Decision making in multi-objective optimisation case coded in MATLABR20016b. Computations are performed by
studies using a PC with 3.40-GHz Intel(R) Core(TM) i5-7500 processor
In this work, a MOO case study of a solar-driven heat engine and and 8GB of RAM. The algorithm parameters in the computational
a MOO case study of solar dish Brayton heat engine are con- experiments are kept constant for both the case studies and
sidered. These case studies are presented by Sayyaadi, Ahmadi, are presented in Table 1. The following subsections present the
and Dehghani (2015) and Li, Liao, and Liu (2015) respectively. computational results analysis related to the MOO case studies
Multi-objective optimisation of these case studies is performed considered.
using Jaya algorithm, MTPG-Jaya algorithm, and AMTPG-Jaya
algorithm. In the implementation of these algorithms for MOO
5.2. Multi-objective optimisation of the solar-driven heat
problem, the posteriori approach based on non dominated rank-
engine
ing method and crowding distance method (Deb et al. 2002;
Rao, Rai, and Balic 2016) is employed. However, selecting the This MOO case study is formulated by Sayyaadi, Ahmadi, and
best solution from a Pareto-front is a difficult task as each solu- Dehghani (2015). The three objective functions considered in
tion has the best value for at least one objective function. Thus this work are thermal efficiency (η), non-dimensional power out-
in this work, a selection methodology based on multi-attribute put (W), and non-dimensional ecological function (E) of the
decision-making methods is suggested to identify the best solar-driven heat engine system given by the Equations (37–39)
Pareto optimal solution in the Pareto front. After performing the respectively. All three objectives are maximisation functions.
12 R. V. RAO AND H. S. KEESARI

Table 2. The Pareto optimal solutions for the solar-driven heat engine case study Table 3. Ranks given by the MADM methods for AMTPG-Jaya algorithm solutions
obtained by the AMTPG-Jaya algorithm. for the solar-driven heat engine case study.
POS φ β τ γ η W E POS A B C D E F G H Average rank
1 0.228800 0.600000 0.2 0.8 0.500000 0.200000 0.080000 21 1 1 2 2 5 2 1 1 1.875
2 0.273105 0.387845 0.2 0.8 0.612700 0.173765 0.120645 2 2 2 4 4 1 1 4 2 2.5
3 0.101693 0.100000 0.2 0.8 0.722462 0.057348 0.051193 19 3 3 1 1 6 4 2 3 2.875
4 0.100000 0.242917 0.2 0.8 0.685822 0.112076 0.093417 20 4 4 3 3 7 5 3 4 4.125
5 0.284435 0.255345 0.2 0.8 0.662366 0.137536 0.108957 25 6 6 7 7 1 3 7 5 5.25
6 0.183648 0.555183 0.2 0.8 0.540748 0.197108 0.102608 9 5 5 5 5 10 7 5 6 6
7 0.230141 0.236822 0.2 0.8 0.674348 0.125326 0.101974 12 8 8 9 9 1 6 9 7 7.125
8 0.127370 0.169761 0.2 0.8 0.703396 0.088417 0.076274 14 7 7 6 6 12 8 6 8 7.5
9 0.107999 0.536161 0.2 0.8 0.566960 0.191717 0.112915 6 9 9 8 8 15 10 8 9 9.5
10 0.162404 0.184952 0.2 0.8 0.696198 0.098635 0.083928 5 11 11 12 12 4 9 12 11 10.25
11 0.100542 0.224090 0.2 0.8 0.691162 0.105341 0.088753 23 10 10 10 10 17 13 10 10 11.25
12 0.229402 0.327860 0.2 0.8 0.643208 0.154016 0.116472 18 13 13 14 14 8 11 14 13 12.5
13 0.212332 0.100000 0.2 0.8 0.713669 0.072442 0.063679 15 12 12 11 11 19 15 11 12 12.875
14 0.245761 0.507441 0.2 0.8 0.556791 0.194178 0.109360 7 14 14 15 15 9 12 15 14 13.5
15 0.157821 0.600000 0.2 0.8 0.517137 0.199513 0.090384 4 16 16 16 16 11 14 16 16 15.125
16 0.237864 0.100000 0.2 0.8 0.711561 0.075861 0.066433 1 15 15 13 13 24 19 13 15 15.875
17 0.100000 0.157808 0.2 0.8 0.708807 0.080215 0.069895 11 17 17 17 17 13 16 17 17 16.375
18 0.224482 0.239236 0.2 0.8 0.674192 0.125495 0.102077 24 18 18 18 18 14 17 18 18 17.375
19 0.207260 0.468425 0.2 0.8 0.585054 0.186053 0.117698 10 19 19 19 19 16 18 19 19 18.5
20 0.212070 0.472414 0.2 0.8 0.582178 0.187070 0.117077 8 20 20 20 20 18 20 20 20 19.75
21 0.101290 0.474100 0.2 0.8 0.599867 0.180060 0.119987 22 21 21 21 21 20 21 21 21 20.875
22 0.293125 0.100000 0.2 0.8 0.706889 0.083176 0.072220 17 22 22 22 22 21 22 22 22 21.875
23 0.403203 0.497998 0.2 0.8 0.529790 0.198490 0.097254 16 23 23 23 23 22 23 23 24 23
24 0.318330 0.137412 0.2 0.8 0.694752 0.100596 0.085357 13 24 24 24 24 23 24 24 25 24
25 0.119626 0.390510 0.2 0.8 0.633376 0.161183 0.118780 3 25 25 25 25 25 25 25 23 24.75
Note: ‘POS’-Pareto optimal solution number. Note: ‘POS’-Pareto optimal solution number; ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’-
MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’- COPRAS; ‘H’- GRA.

The decision variables of these objectives are the heat trans- The MADM methods ranking of the Pareto optimal solutions
fer ratio of the allocation parameter (φ), the heat source to sink achieved by the AMTPG-Jaya algorithm is presented in Table 3.
and the allocation parameter (β), the heat engine temperature Here an observation can be made that the ranking given by dif-
ratio (τ ), and the heat source temperature ratio (γ ) expressed by ferent MADM methods are different but consistent. The Spear-
the Equations (53–56). The lower and upper boundaries of the man correlation coefficients between the MADM methods are
decision variables are as follows: presented in Table 4. Here an observation can be made that
for different pairs of MADM methods the Spearman correlation
coefficients are higher. Therefore, for calculating the average
0.1 ≤ Ø ≤ 0.6 (53)
ranks of solutions, ranks given by all the methods can be con-
0.1 ≤ β ≤ 0.6 (54) sidered. Furthermore, it can be observed that the Spearman cor-
0.1 ≤ τ ≤ 0.6 (55) relation coefficient between SAW and WPM methods is 1, which
implies the rankings of these methods are identical. It can also
0.1 ≤ γ ≤ 0.6 (56) be observed from Table 3 that the rankings of SAW and WPM are
same. Similarly, the TOPSIS and MTOPSIS methods rankings are
Sayyaadi, Ahmadi, and Dehghani (2015) employed NSGA- identical with a Spearman correlation coefficient of 1. It can also
II to this MOO problem for finding the optimal decision vari- be observed that except the VIKOR method remaining all meth-
ables and documented three non-dominated solutions using ods have strong similarities in the ranking of the Pareto optimal
Bellman-Zadeh, TOPSIS, and LINMAP decision-making methods solutions. The VIKOR method has suggested the solutions 2, 12,
based on the similarity to the ideal solutions. The number of and 25 as compromise solutions which are ranked as 1. The
function evaluations taken by NSGA-II to achieve convergence Pareto optimal solution 21 is the best solution with an average
was not specified by Sayyaadi, Ahmadi, and Dehghani (2015); rank of 1.875. Four MADM methods (SAW, WPM, COPRAS, and
thus in this work, the proposed algorithms are executed for GRA) have suggested solution 21 as the best solution, and three
40,000 function evaluations. Furthermore, the best Pareto opti- MADM methods (TOPSIS, MTOPSIS, and PROMETHEE) suggested
mal solution from each algorithm’s Pareto front is identified solution 21 as the second-best solution. All the MADM methods
using the average rank as explained earlier. Then, the selected except GRA method have suggested solution 3 as the last choice.
best solutions are compared with those reported for the NSGA-II With an average rank of 24.75, solution 3 can be considered as
algorithm. the least good solution.
The Pareto optimal solutions achieved by the AMTPG-Jaya The Pareto optimal solutions achieved by the MTPG-Jaya
algorithm for tri-objective optimisation of a solar-driven heat algorithm for tri-objective optimisation of a solar-driven heat
engine are presented in Table 2. The thermal efficiency achieved engine are presented in Table 5. The thermal efficiency achieved
by AMTPG-Jaya algorithm is varied between 0.5 and 0.7225, by MTPG-Jaya algorithm is varied between 0.4999 and 0.7223,
the non-dimensional power output is varied between 0.0573 the non-dimensional power output is varied between 0.0576
and 0.2, and non-dimensional ecological function is varied from and 0.2, and non-dimensional ecological function is varied from
0.0511 to 0.1206. 0.0514 to 0.1206.
INTERNATIONAL JOURNAL OF AMBIENT ENERGY 13

Table 4. Spearman’s rank correlation coefficients between different pairs of Table 6. Ranks given by the MADM methods for MTPG-Jaya algorithm solutions
MADM methods for AMTPG-Jaya algorithm solutions ranking presented in Table 3. for the solar-driven heat engine case study.
Method A B C D E F G H POS A B C D E F G H Average rank
A 1 1 0.9915 0.9915 0.7983 0.9708 0.9931 0.9962 14 1 1 2 2 8 3 1 2 2.5
B 1 1 0.9915 0.9915 0.7983 0.9708 0.9931 0.9962 23 2 2 3 3 6 2 3 1 2.75
C 0.9915 0.9915 1 1 0.7375 0.9408 0.9992 0.9854 1 4 4 4 4 1 1 4 3 3.125
D 0.9915 0.9915 1 1 0.7375 0.9408 0.9992 0.9854 9 3 3 1 1 9 5 2 4 3.5
E 0.7983 0.7983 0.7375 0.7375 1 0.9138 0.7383 0.8083 10 6 6 7 7 1 4 7 5 5.375
F 0.9708 0.9708 0.9408 0.9408 0.9138 1 0.9423 0.9731 6 5 5 5 5 12 9 5 6 6.5
G 0.9931 0.9931 0.9992 0.9992 0.7383 0.9423 1 0.9869 15 7 7 8 8 1 6 8 7 6.5
H 0.9962 0.9962 0.9854 0.9854 0.8083 0.9731 0.9869 1 12 9 9 9 9 1 7 9 9 7.75
Note: ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’- MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’- 4 8 8 6 6 15 11 6 8 8.5
COPRAS; ‘H’- GRA. 16 10 10 10 10 1 8 10 10 8.625
7 11 11 13 13 7 10 11 11 10.875
8 12 12 14 14 10 12 14 12 12.5
17 14 14 15 15 11 13 15 14 13.875
Table 5. The Pareto optimal solutions for the solar-driven heat engine case study 5 13 13 11 11 22 17 12 13 14
obtained by the MTPG-Jaya algorithm. 21 16 16 16 16 13 14 16 16 15.375
2 15 15 12 12 24 20 13 15 15.75
POS φ β τ γ η W E 18 17 17 17 17 14 15 17 17 16.375
11 18 18 18 18 16 16 18 18 17.5
1 0.323446 0.370780 0.2 0.8 0.612705 0.173762 0.120645
13 19 19 19 19 17 18 19 19 18.625
2 0.244631 0.594646 0.2 0.8 0.499993 0.200000 0.079996
24 20 20 20 20 18 19 20 20 19.625
3 0.103845 0.100000 0.2 0.8 0.722296 0.057646 0.051444
25 21 21 21 21 19 21 21 21 20.75
4 0.229759 0.522619 0.2 0.8 0.551060 0.195354 0.107104
20 22 22 22 22 20 22 22 22 21.75
5 0.428259 0.516880 0.2 0.8 0.511220 0.199794 0.086934
22 23 23 23 23 21 23 23 23 22.75
6 0.348213 0.456602 0.2 0.8 0.565953 0.191983 0.112589
19 24 24 24 24 23 24 24 24 23.875
7 0.251107 0.267332 0.2 0.8 0.662131 0.137759 0.109075
3 25 25 25 25 25 25 25 25 25
8 0.238533 0.243317 0.2 0.8 0.671398 0.128479 0.103870
9 0.223359 0.458711 0.2 0.8 0.587208 0.185261 0.118126 Note: ‘POS’-Pareto optimal solution number; ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’-
10 0.178665 0.369729 0.2 0.8 0.633686 0.160969 0.118722 MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’- COPRAS; ‘H’- GRA.
11 0.275134 0.166603 0.2 0.8 0.690695 0.105945 0.089179
12 0.296682 0.288529 0.2 0.8 0.649317 0.149138 0.114528
13 0.422904 0.100000 0.2 0.8 0.695298 0.099859 0.084822 Table 7. Spearman’s rank correlation coefficients between different pairs of
14 0.358523 0.388633 0.2 0.8 0.599075 0.180414 0.119904 MADM methods for the MTPG-Jaya algorithm solutions ranking presented in
15 0.209627 0.344229 0.2 0.8 0.639537 0.156788 0.117449 Table 6.
16 0.288577 0.280720 0.2 0.8 0.653108 0.145939 0.113115
Method A B C D E F G H
17 0.100000 0.272262 0.2 0.8 0.677032 0.122369 0.100143
18 0.100000 0.242851 0.2 0.8 0.685841 0.112052 0.093401 A 1 1 0.9869 0.9869 0.7900 0.9569 0.9931 0.9977
19 0.226820 0.112481 0.2 0.8 0.709398 0.079292 0.069165 B 1 1 0.9869 0.9869 0.7900 0.9569 0.9931 0.9977
20 0.311024 0.100000 0.2 0.8 0.705343 0.085519 0.074042 C 0.9869 0.9869 1 1 0.7135 0.9154 0.9969 0.9823
21 0.364484 0.165171 0.2 0.8 0.682331 0.116279 0.096226 D 0.9869 0.9869 1 1 0.7135 0.9154 0.9969 0.9823
22 0.294771 0.100000 0.2 0.8 0.706747 0.083392 0.072389 E 0.7900 0.7900 0.7135 0.7135 1 0.9182 0.7390 0.8031
23 0.431395 0.344294 0.2 0.8 0.608189 0.176101 0.120562 F 0.9569 0.9569 0.9154 0.9154 0.9182 1 0.9300 0.9646
24 0.230898 0.151826 0.2 0.8 0.698878 0.094920 0.081186 G 0.9931 0.9931 0.9969 0.9969 0.7390 0.9300 1 0.9885
25 0.344280 0.100000 0.2 0.8 0.702427 0.089838 0.077358 H 0.9977 0.9977 0.9823 0.9823 0.8031 0.9646 0.9885 1
Note: ‘POS’-Pareto optimal solution number. Note: ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’- MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’-
COPRAS; ‘H’- GRA.

Table 6 presents the ranking of MTPG-Jaya algorithm Pareto as the best solution. All the MADM methods have suggested
optimal solutions using MADM methods. The Spearman corre- solution 25 as the least good solution.
lation coefficients between different pairs of MADM methods Table 8 presents the Pareto optimal solutions achieved by
ranking for MTPG-Jaya algorithm Pareto front is shown in Table the Jaya algorithm for tri-objective optimisation of solar-driven
7. Here an observation can be made that for different pairs heat engine case study. The thermal efficiency achieved by Jaya
of MADM methods the Spearman correlation coefficients are algorithm varies from 0.4999 to 0.7222, the non-dimensional
higher. Hence, the average ranks of solutions can be calculated power output varies between 0.0577 and 0.2, and the non-
by considering the ranks given by all the methods. Similar to dimensional ecological function varies between 0.0515 and
the ranking of AMTPG-Jaya algorithm solutions, the rankings 0.1206. The ranking of the Jaya algorithm Pareto optimal solu-
of MTPG-Jaya algorithm solutions given by the SAW and WPM tions by MADM methods is presented in Table 9. The Spearman
methods, and the TOPSIS and MTOPSIS methods are identical. correlation coefficient between different pairs of MADM meth-
Also, the VIKOR method has relatively lesser spearman coeffi- ods for the ranking of the Jaya algorithm solutions is presented
cients with other methods. The VIKOR method has suggested in Table 10. An observation can be made here that, the Spearman
the solutions 1, 10, 12, 15, and 16 as compromise solutions correlation coefficients between different pairs of MADM meth-
which are ranked as 1. The SAW, WPM, and COPRAS methods ods are higher. Hence, the ranking of all the MADM methods can
have suggested the solution 14 as the best solution, the TOP- be considered for calculating the average ranks of Jaya algorithm
SIS and MTOPSIS methods have suggested the solution 9 as the Pareto optimal solutions. Furthermore, the rankings given by the
best solution, and the PROMETHEE and GRA methods have sug- SAW and WPM methods are the same. These methods have sug-
gested the solutions 1 and 23 as the best solution respectively. gested solution 7 as the best solution. Similarly, the rankings
The solution 14 with an average rank of 2.5 can be considered given by the TOPSIS and MTOPSIS methods are identical. These
14 R. V. RAO AND H. S. KEESARI

Table 8. The Pareto optimal solutions for the solar-driven heat engine case study Table 10. Spearman’s rank correlation coefficients between different pairs of
obtained by the Jaya algorithm. MADM methods for the Jaya algorithm solutions ranking presented in Table 9.
POS φ β τ γ η W E Method A B C D E F G H
1 0.334445 0.367070 0.2 0.8 0.612698 0.173765 0.120645 A 1 1 0.9846 0.9846 0.8324 0.9746 0.9938 0.9862
2 0.379999 0.548782 0.2 0.8 0.499999 0.200000 0.079999 B 1 1 0.9846 0.9846 0.8324 0.9746 0.9938 0.9862
3 0.100000 0.101569 0.2 0.8 0.722235 0.057754 0.051536 C 0.9846 0.9846 1 1 0.7583 0.9308 0.9969 0.9654
4 0.316571 0.100000 0.2 0.8 0.704861 0.086243 0.074602 D 0.9846 0.9846 1 1 0.7583 0.9308 0.9969 0.9654
5 0.147506 0.105103 0.2 0.8 0.717692 0.065708 0.058172 E 0.8324 0.8324 0.7583 0.7583 1 0.9182 0.7853 0.8301
6 0.100000 0.314388 0.2 0.8 0.663518 0.136429 0.108366 F 0.9746 0.9746 0.9308 0.9308 0.9182 1 0.9500 0.9692
7 0.248270 0.425024 0.2 0.8 0.599529 0.180211 0.119952 G 0.9938 0.9938 0.9969 0.9969 0.7853 0.9500 1 0.9769
8 0.479569 0.100000 0.2 0.8 0.689945 0.106910 0.089856 H 0.9862 0.9862 0.9654 0.9654 0.8301 0.9692 0.9769 1
9 0.430832 0.141684 0.2 0.8 0.682632 0.115922 0.095991 Note: ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’- MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’-
10 0.118183 0.428923 0.2 0.8 0.617684 0.171020 0.120541 COPRAS; ‘H’- GRA.
11 0.100000 0.354326 0.2 0.8 0.649618 0.148889 0.114422
12 0.120594 0.100000 0.2 0.8 0.720998 0.059957 0.053388
13 0.320713 0.316124 0.2 0.8 0.635846 0.159459 0.118292
14 0.271803 0.495472 0.2 0.8 0.558609 0.193774 0.110039 Now the Pareto optimal solutions selected from the Pareto-
15 0.178149 0.373390 0.2 0.8 0.632299 0.161919 0.118974 fronts of the Jaya, MTPG-Jaya and AMTPG-Jaya algorithms based
16 0.210281 0.100000 0.2 0.8 0.713838 0.072166 0.063456
17 0.195508 0.479234 0.2 0.8 0.581554 0.187284 0.116936 on the average ranks of MADM methods are compared with
18 0.100000 0.154298 0.2 0.8 0.709683 0.078844 0.068810 those reported for NSGA-II algorithm (Sayyaadi, Ahmadi, and
19 0.271416 0.291440 0.2 0.8 0.651356 0.147434 0.113788 Dehghani 2015). The comparison of the Pareto optimal solutions
20 0.416989 0.100000 0.2 0.8 0.695846 0.099115 0.084279
21 0.356283 0.100000 0.2 0.8 0.701360 0.091385 0.078532 obtained by the different algorithms for the multi-objective opti-
22 0.376223 0.440349 0.2 0.8 0.569683 0.190973 0.113764 misation of solar-driven heat engine case study is presented in
23 0.270463 0.226217 0.2 0.8 0.673389 0.126362 0.102603 Table 11. Here an observation can be made that, the ecologi-
24 0.218147 0.100000 0.2 0.8 0.713192 0.073223 0.064310
25 0.220959 0.256002 0.2 0.8 0.669238 0.130726 0.105183 cal function is maximum for the AMTPG-Jaya algorithm solution,
the non-dimensional power output is maximum for the NSGA-
Note: ‘POS’-Pareto optimal solution number.
II (TOPSIS) solution, and thermal efficiency is maximum for the
NSGA-II (Bellman and Zadeh) solution. Figure 6 shows the Pareto
Table 9. Ranks given by the MADM methods for Jaya algorithm solutions for the fronts achieved by the Jaya algorithm, MTPG-Jaya algorithm,
solar-driven heat engine case study.
and AMTPG-Jaya algorithm along with the Pareto optimal solu-
POS A B C D E F G H Average rank tions of NSGA-II algorithm. The NSGA-II algorithm Pareto optimal
7 1 1 2 2 8 3 1 1 2.375 solutions are reported using the Bellman-Zadeh, TOPSIS, and
1 2 2 4 4 6 1 3 2 3 LINMAP decision making methods.
10 4 4 6 6 1 2 5 3 3.875
17 3 3 1 1 11 6 2 4 3.875
The ranking of the Pareto optimal solutions of these algo-
15 6 6 7 7 1 4 7 5 5.375 rithms by the MADM methods is presented in Table 12. Here an
22 5 5 3 3 12 9 4 6 5.875 observation can be made that the rankings given by the SAW,
13 7 7 8 8 1 5 8 7 6.375
WPM, and COPRAS methods are identical. Similarly, the rankings
11 9 9 9 9 1 7 9 9 7.75
14 8 8 5 5 14 10 6 8 8 given by the VIKOR, PROMETHEE, and GRA methods are identi-
19 10 10 10 10 1 8 10 10 8.625 cal, and the rankings given by the TOPSIS and MTOPSIS methods
6 11 11 12 12 7 11 11 11 10.75 are identical. The Spearman correlation coefficients of different
25 12 12 13 13 9 12 13 12 12
23 13 13 14 14 10 13 14 13 13 pairs of the MADM methods are presented in Table 13. All the
2 14 14 11 11 22 17 12 14 14.375 MADM methods have positive Spearman correlation coefficient
9 15 15 15 15 13 14 15 15 14.625 with the other methods. Therefore, the average ranks can be
8 16 16 16 16 15 15 16 16 15.75
20 17 17 17 17 16 16 17 17 16.75 calculated by considering the ranking of all the MADM meth-
21 18 18 18 18 17 18 18 18 17.875 ods. The Pareto optimal solution of the AMTPG-Jaya algorithm
4 19 19 19 19 18 19 19 19 18.875 can be considered as the best solution with an average rank
18 20 20 20 20 19 20 20 20 19.875
24 21 21 21 21 20 21 21 23 21.125 of 1.875. The AMTPG-Jaya algorithm solution has better com-
16 22 22 22 22 21 22 22 24 22.125 promise among the three objectives of this case study which
5 23 23 23 23 23 23 23 25 23.25 lead to an increase of the thermal efficiency by 2.63% and
12 24 24 24 24 24 24 24 22 23.75
3 25 25 25 25 25 25 25 21 24.5 2.73% when compared to that of the NSGA-II (LINMAP and TOP-
Note: ‘POS’-Pareto optimal solution number; ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’-
SIS) solutions respectively. Similarly, the ecological function is
MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’- COPRAS; ‘H’- GRA. increased by 3.79%, 0.66%, and 1% when compared to that of
the NSGA-II (Bellman and Zadeh, LINMAP, and TOPSIS) solu-
tions respectively. Also, the non-dimensional output power is
methods have suggested solution 17 as the best solution. The increased by 17.61% when compared to that of the NSGA-II (Bell-
COPRAS and GRA methods have suggested solution 7 as the best man and Zadeh) solution. The Jaya algorithm Pareto optimal
solution. The PROMETHEE method has suggested solution 1 as solution is the next best solution with an average rank of 2.375.
the best solution. The VIKOR method suggested the solutions 10, The MTPG-Jaya algorithm and NSGA-II (LINMAP) algorithm solu-
11, 13, 15, and 19 as the compromise solutions which are ranked tions have the same average rank value (2.875). However, it can
as 1. Solution 7 with an average rank of 2.375 can be considered be observed that five out of eight methods have suggested
as the best solution, and the solution 3 with an average rank of the MTPG-Jaya algorithm solution is better than the NSGA-II
24.5 can be considered as the least good solution. (LINMAP) solution. Also, the MTPG-Jaya algorithm solution has
INTERNATIONAL JOURNAL OF AMBIENT ENERGY 15

Table 11. Comparison of the Pareto optimal solutions of the NSGA-II, Jaya, MTPG-Jaya, AMTPG-Jaya algorithms obtained for the solar-driven heat engine case study.
Algorithm φ β τ γ η W E
NSGA-II (Bellman and Zadeh) 0.4401 0.2533 0.2 0.8 0.6444 0.1531 0.1156
NSGA-II (LINMAP) 0.3684 0.4149 0.2 0.8 0.5845 0.1862 0.1192
NSGA-II (TOPSIS) 0.4447 0.3903 0.2 0.8 0.5839 0.1865 0.1188
Jaya 0.24827 0.42502 0.2 0.8 0.599529 0.180211 0.119952
MTPG-Jaya 0.35852 0.38863 0.2 0.8 0.599075 0.180414 0.119904
AMTPG-Jaya 0.10129 0.4741 0.2 0.8 0.599867 0.180060 0.119987
Note: Results in the bold figure indicates better value; NSGA-II results are available in Sayyaadi, Ahmadi, and Dehghani (2015).

Figure 6. The Pareto optimal solutions of the NSGA-II (Bellman and Zadeh, TOPSIS, and LINMAP), Jaya, MTPG-Jaya, and AMTPG-Jaya algorithms obtained for multi-
objective optimisation of the solar-driven heat engine case study.

Table 12. Comparison of the ranking of the Pareto optimal solutions given by Table 13. Spearman’s rank correlation coefficients between different MADM
NSGA-II, Jaya, MTPG-Jaya, AMTPG-Jaya algorithms obtained by the MADM methods methods for the NSGA-II, Jaya, MTPG-Jaya, AMTPG-Jaya algorithm’s solutions rank-
for the solar-driven heat engine case study. ing presented in Table 12.
Algorithm A B C D E F G H Average rank Method A B C D E F G H
NSGA-II (Bellman and Zadeh) 6 6 6 6 6 6 6 6 6 A 1 1 0.4286 0.4286 0.6571 0.6571 1 0.6571
NSGA-II (LINMAP) 1 1 4 4 4 4 1 4 2.875 B 1 1 0.4286 0.4286 0.6571 0.6571 1 0.6571
NSGA-II (TOPSIS) 5 5 5 5 5 5 5 5 5 C 0.4286 0.4286 1 1 0.7714 0.7714 0.4286 0.7714
Jaya 3 3 2 2 2 2 3 2 2.375 D 0.4286 0.4286 1 1 0.7714 0.7714 0.4286 0.7714
MTPG-Jaya 4 4 1 1 3 3 4 3 2.875 E 0.6571 0.6571 0.7714 0.7714 1 1 0.6571 1
AMTPG-Jaya 2 2 3 3 1 1 2 1 1.875 F 0.6571 0.6571 0.7714 0.7714 1 1 0.6571 1
Note: ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’- MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’- G 1 1 0.4286 0.4286 0.6571 0.6571 1 0.6571
COPRAS; ‘H’- GRA. H 0.6571 0.6571 0.7714 0.7714 1 1 0.6571 1
Note: ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’- MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’-
COPRAS; ‘H’- GRA.

better thermal efficiency and non-dimensional ecological func-


tion than the NSGA-II (LINMAP) solution. Thus, the MTPG-Jaya be made that the AMTPG-Jaya algorithm has better spacing and
algorithm solution can be considered as a better solution than Hypervolume metrics. The AMTPG-Jaya algorithm has 85.9% and
the NSGA-II (LINMAP) solution. The NSGA-II (Bellman and Zadeh 12.03% lesser spacing value than that of the Jaya and MTPG-Jaya
and TOPSIS) solutions are ranked 6 and 5 respectively by all the algorithms respectively. Furthermore, Hypervolume achieved by
MADM methods. the AMTPG-Jaya algorithm is 3.5% and 1.5% higher than that of
Now the performances of the Jaya, MTPG-Jaya, and AMTPG- the Jaya and MTPG-Jaya algorithms respectively. Even though
Jaya algorithms are compared on the basis of coverage, spacing the Pareto fronts obtained by the AMTPG-Jaya, MTPG-Jaya and
and Hypervolume indicators. These performance indicators are Jaya algorithms appear overlapping each other, the Hypervol-
calculated for the Pareto fronts achieved by the AMTPG-Jaya ume of AMTPG-Jaya algorithm is more. This is due to the uniform
algorithm, MTPG-Jaya algorithm and Jaya algorithm presented distribution of the Pareto optimal solutions obtained by the
in Tables 2, 5, and 8 respectively. The coverage, spacing and AMTPG-Jaya algorithm along the Pareto front. The coverage of
Hypervolume values of the AMTPG-Jaya, MTPG-Jaya, and Jaya the Jaya algorithm over the MTPG-Jaya and AMTPG-Jaya algo-
algorithms are presented in Table 14. Here an observation can rithms are 0.04 and zero respectively. This means that 4% of the
16 R. V. RAO AND H. S. KEESARI

Table 14. The performance indicator values of the Jaya, MTPG-Jaya, AMTPG-Jaya the Hypervolume of the Pareto fronts of the Jaya (0.026656),
algorithm’s in the solar-driven heat engine case study.
MTPG-Jaya (0.026774), and AMTPG-Jaya (0.026726) algorithms
Coverage are competitive.
Algorithm Spacing Hypervolume Jaya MTPG-Jaya AMTPG-Jaya Similarly, Figure 8 presents the Pareto fronts obtained by the
AMTPG-Jaya, MTPG-Jaya, and Jaya algorithms for simultaneous
Jaya 0.139464013 0.004390526 – 0.04 0
MTPG-Jaya 0.084011836 0.004485021 0 – 0 optimisation of the thermal efficiency and non-dimensional eco-
AMTPG-Jaya 0.074990256 0.004550973 0.04 0.04 – logical function. The variation in thermal efficiency is ranged
Note: Results in the bold figure indicates better value. between 0.612689 and 0.722592 and variation in ecological
function is ranged between 0.050995 and 0.120645. Here an
MTPG-Jaya algorithm solutions are dominated by Jaya algorithm observation can be made that these two objectives are also con-
solutions. The MTPG-Jaya algorithm coverage values over the flicting in nature. The spacing values for the Pareto fronts of
Jaya and AMTPG-Jaya algorithms are zero. Similarly, the AMTPG- the Jaya, MTPG-Jaya, and AMTPG-Jaya algorithms are 0.037815,
Jaya algorithm coverage values over the Jaya and MTPG-Jaya 0.024866, and 0.023171 respectively. The spacing metric value of
algorithms are 0.04 and 0.04 respectively. This means that 4% the AMTPG-Jaya algorithm Pareto front is 63.2% and 7.3% bet-
of the solutions obtained by the Jaya algorithm and 4% of the ter than that of the Jaya and MTPG-Jaya algorithms respectively.
solutions obtained by the MTPG-Jaya algorithm are dominated The Hypervolume of the Pareto fronts obtained by the Jaya
by the AMTPG-Jaya algorithms solutions. (0.014621), MTPG-Jaya (0.01463), and AMTPG-Jaya (0.014637)
In addition to tri-objective optimisation of the solar-driven algorithms are competitive in this case also.
heat engine case study, dual-objective optimisation is also car- Similarly, Figure 9 presents the Pareto fronts obtained by
ried out using the Jaya, MTPG-Jaya and AMTPG-Jaya algorithms. the AMTPG-Jaya, MTPG-Jaya, and Jaya algorithms for simul-
Figure 7 presents the conflicting nature between thermal effi- taneous optimisation of the non-dimensional power output
ciency and non-dimensional power output. The thermal effi- and non-dimensional ecological function. The non-dimensional
ciency is varied between 0.49993 and 0.722592 for all the pro- power output is varied from 0.173759 to 0.199999, and the
posed algorithms. Similarly, the non-dimensional power out- non-dimensional ecological function varied from 0.079998 to
put varied between 0.057114 and 0.199999 for all the pro- 0.120645 for all the proposed algorithms. The Hypervolume
posed algorithms. Here an observation can be made that the of the Pareto fronts of the Jaya, MTPG-Jaya, and AMTPG-Jaya
Pareto fronts of the Jaya, MTPG-Jaya, and AMTPG-Jaya algo- algorithms are 0.0103976, 0.0103986, and 0.0104023 respec-
rithms are overlapping. The spacing values for the Pareto tively. The spacing values for the Pareto fronts of the Jaya, MTPG-
fronts of the Jaya, MTPG-Jaya, and AMTPG-Jaya algorithms are Jaya, and AMTPG-Jaya algorithms are 0.036454, 0.037813, and
0.034521, 0.030105, and 0.026662 respectively. The AMTPG-Jaya 0.03013 respectively. The AMTPG-Jaya algorithm has 20.9% and
algorithm has 29.4% and 12.9% better Spacing metric compared 25.4% lesser spacing values compared to the Jaya and MTPG-
to Jaya and MTPG-Jaya algorithms respectively. Furthermore, Jaya algorithms respectively. The following subsection presents

Figure 7. The Pareto frontiers obtained by the Jaya, MTPG-Jaya, and AMTPG-Jaya algorithms for dual objective (η−W) optimisation of the solar-driven heat engine case
study.
INTERNATIONAL JOURNAL OF AMBIENT ENERGY 17

Figure 8. The Pareto frontiers obtained by the Jaya, MTPG-Jaya, and AMTPG-Jaya algorithms for dual objective (η−E) optimisation of the solar-driven heat engine case
study.

Figure 9. The Pareto frontiers obtained by the Jaya, MTPG-Jaya, and AMTPG-Jaya algorithms for dual objective (W−E) optimisation of the solar-driven heat engine case
study.

the multi-objective optimisation results analysis related to the thermo-economic performance function (F) of the solar dish
solar dish Brayton heat engine case study. Brayton heat engine system, given by Equations (48–52) respec-
tively. These three objective functions are maximisation func-
tions. The decision variables of these objective functions are the
5.3. Multi-objective optimisation of the solar dish Brayton temperature of hot reservoir (TH ), the temperature of cold reser-
heat engine voir (TL ), the temperature of the working fluid at state 1 of Bray-
This MOO case study is presented by Li, Liao, and Liu (2015). ton cycle (T 1 ), hot side heat exchange effectiveness ( H ), cold
The three objective functions considered in this case study are side heat exchange effectiveness ( L ), and regenerator effec-
thermal efficiency (ηm ), power output (P), and non-dimensional tiveness ( R ). The lower and upper boundaries of the decision
18 R. V. RAO AND H. S. KEESARI

Table 15. The Pareto optimal solutions for the solar dish Brayton heat engine case study obtained by the AMTPG-Jaya algorithm.
POS H L R T H (K) T L (K) T 1 (K) P (kW) ηm F
1 0.7 0.69999999 0.8 1000 400 563.262087 67.322294 0.237706 0.314500
2 0.7 0.7 0.8 1000 400 566.311098 67.866710 0.237761 0.314428
3 0.7 0.69999999 0.8 1000 400 612.243687 71.603713 0.227684 0.299487
4 0.7 0.69999999 0.8 1000 400 606.776885 71.555795 0.229768 0.302385
5 0.7 0.7 0.8 1000 400 581.877339 70.030357 0.236444 0.312026
6 0.7 0.69999999 0.8 1000 400 575.446142 69.257679 0.237294 0.313407
7 0.7 0.7 0.8 1000 400 608.814537 71.585024 0.229014 0.301334
8 0.7 0.70000000 0.8 1000 400 610.682705 71.599910 0.228299 0.300340
9 0.7 0.69999999 0.8 1000 400 600.180493 71.366028 0.232008 0.305535
10 0.7 0.70000000 0.8 1000 400 604.344969 71.502920 0.230629 0.303591
11 0.7 0.7 0.8 1000 400 597.887942 71.265198 0.232712 0.306536
12 0.7 0.69999999 0.8 1000 400 583.955528 70.245045 0.236085 0.311473
13 0.7 0.69999999 0.8 1000 400 573.809546 69.034290 0.237444 0.313675
14 0.7 0.7 0.8 1000 400 578.761814 69.676734 0.236906 0.312760
15 0.7 0.69999999 0.8 1000 400 586.777835 70.509951 0.235535 0.310641
16 0.7 0.7 0.8 1000 400 594.830108 71.101949 0.233588 0.307792
17 0.7 0.69999999 0.8 1000 400 596.870004 71.214529 0.233012 0.306964
18 0.7 0.69999997 0.8 1000 400 565.279784 67.687161 0.237755 0.314468
19 0.7 0.69999999 0.8 1000 400 593.261898 71.005274 0.234009 0.308400
20 0.7 0.69999999 0.8 1000 400 571.915333 68.761858 0.237582 0.313939
21 0.7 0.7 0.8 1000 400 578.820212 69.683717 0.236899 0.312747
22 0.7 0.69999999 0.8 1000 400 595.617357 71.147151 0.233370 0.307477
23 0.7 0.70000000 0.8 1000 400 588.575682 70.662939 0.235148 0.310065
24 0.7 0.69999999 0.8 1000 400 589.840758 70.763336 0.234859 0.309638
25 0.7 0.69999999 0.8 1000 400 603.015736 71.465637 0.231083 0.304229
Note: ‘POS’-Pareto optimal solution number.

variables are as follows: Table 16. Ranks given by the MADM methods for AMTPG-Jaya algorithm solutions
for the solar dish Brayton heat engine case study.
700 K ≤ TH ≤ 1000 K (57) POS A B C D E F G H Average rank

400 K ≤ TL ≤ 500 K (58) 5 1 1 5 5 1 9 1 9 4


12 2 2 3 3 1 10 2 10 4.125
TL ≤ T1 ≤ TH (59) 15 5 5 1 1 1 11 5 11 5
21 3 3 7 7 6 8 3 8 5.625
0.5 ≤ εH ≤ 0.7 (60) 14 4 4 8 8 7 7 4 7 6.125
23 7 6 2 2 1 12 7 12 6.125
0.5 ≤ εL ≤ 0.7 (61) 24 8 8 4 4 5 13 8 13 7.875
6 6 7 12 12 9 6 6 6 8
0.5 ≤ εR ≤ 0.8 (62) 13 9 9 14 14 10 5 9 5 9.375
19 11 11 6 6 8 14 11 14 10.125
The characteristics of solar-driven engine system taken are 20 10 10 16 16 13 4 10 3 10.25
16 12 12 9 9 11 15 12 15 11.875
the same as given by Li, Liao, and Liu (2015). Those are as follows: 22 13 13 10 10 12 16 13 16 12.875
2 16 16 19 19 17 1 16 1 13.125
I = 1000Wm−2 , eC = 0.9, RC = 1300, η0 = 0.85, 17 14 14 11 11 14 17 14 17 14
18 17 17 21 21 18 2 17 2 14.375
δ = 5.67 × 10−8 WK −4 m−2 , 11 15 15 13 13 15 18 15 18 15.25
1 19 19 23 23 21 3 19 4 16.375
hC = 20Wm−2 K −1 , T0 = 300 K, k = 4, ξ = 0.02, 9 18 18 15 15 16 19 18 19 17.25
25 20 20 17 17 19 20 20 20 19.125
hH = hL = 2000 WK −1 m−2 , 10 21 21 18 18 20 21 21 21 20.125
4 22 22 20 20 22 22 22 22 21.5
and Cwf = 1050 WK −1 . 7 23 23 22 22 23 23 23 23 22.75
8 24 24 24 24 24 24 24 24 24
3 25 25 25 25 25 25 25 25 25
Li, Liao, and Liu (2015) employed NSGA-II to this MOO prob-
Note: ‘POS’-Pareto optimal solution number; ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’-
lem for finding the optimal decision variables and documented MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’- COPRAS; ‘H’- GRA.
three non-dominated solutions using Shannon Entropy, TOPSIS,
and LINMAP decision-making methods. The number of func- Brayton heat engine are presented in Table 15. The thermal effi-
tion evaluations taken by NSGA-II to achieve convergence was ciency achieved by AMTPG-Jaya algorithm is varied between
125000, but in this work, the proposed algorithms are tested 0.2276 and 0.2377, the power output is varied between 67.32
for less number (40000) of function evaluations. Furthermore, and 71.60 kW, and non-dimensional thermo-economic function
the best Pareto optimal solution from each algorithm’s Pareto is varied from 0.2994 to 0.3145.
front is identified using the average rank. Then, the selected The MADM methods ranking of the Pareto optimal solutions
best solutions are compared with those reported for the NSGA-II achieved by the AMTPG-Jaya algorithm for solar dish Brayton
algorithm. heat engine case study is presented in Table 16. The Spearman
The Pareto optimal solutions achieved by the AMTPG-Jaya correlation coefficients between the MADM methods are pre-
algorithm for tri-objective (ηm −P−F) optimisation of solar dish sented in Table 17. It can be observed here that two pairs of
INTERNATIONAL JOURNAL OF AMBIENT ENERGY 19

Table 17. Spearman’s rank correlation coefficients between different pairs of Table 19. Ranks given by the MADM methods for MTPG-Jaya algorithm solutions
MADM methods for the AMTPG-Jaya algorithm solutions ranking presented in for the solar dish Brayton heat engine case study.
Table 16.
POS A B C D E F G H Average rank
Method A B C D E F G H
8 1 1 3 3 1 8 1 8 3.25
A 1 0.9992 0.8723 0.8723 0.9622 0.5954 1 0.6023 15 3 3 2 2 1 9 3 9 4
B 0.9992 1 0.8800 0.8800 0.9672 0.5908 0.9992 0.5977 4 2 2 6 6 1 7 2 7 4.125
C 0.8723 0.8800 1 1 0.9607 0.2577 0.8723 0.2631 18 4 4 1 1 1 10 4 10 4.375
D 0.8723 0.8800 1 1 0.9607 0.2577 0.8723 0.2631 6 5 5 8 8 7 6 5 6 6.25
E 0.9622 0.9672 0.9607 0.9607 1 0.4694 0.9622 0.4755 25 7 6 4 4 5 11 7 11 6.875
F 0.5954 0.5908 0.2577 0.2577 0.4694 1 0.5954 0.9992 21 6 7 11 11 9 5 6 5 7.5
G 1 0.9992 0.8723 0.8723 0.9622 0.5954 1 0.6023 19 8 8 12 12 10 4 8 4 8.25
H 0.6023 0.5977 0.2631 0.2631 0.4755 0.9992 0.6023 1 22 9 9 5 5 6 12 9 12 8.375
Note: ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’- MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’- 9 10 10 7 7 8 13 10 13 9.75
COPRAS; ‘H’- GRA. 10 11 11 14 14 13 3 11 2 9.875
5 12 12 9 9 11 14 12 14 11.625
1 14 14 18 18 15 1 14 1 11.875
MADM methods (the SAW and COPRAS, and the TOPSIS and 17 13 13 10 10 12 15 13 15 12.625
2 16 16 21 21 17 2 16 3 14
MTOPSIS) have the Spearman correlation coefficients equal to 13 15 15 13 13 14 16 15 16 14.625
one. Furthermore, different pairs of MADM methods among the 7 17 17 15 15 16 17 17 17 16.375
SAW, WPM, TOPSIS, MTOPSIS, VIKOR, and COPRAS methods have 23 18 18 16 16 18 18 18 18 17.5
12 19 19 17 17 19 19 19 19 18.5
a strong correlation in the ranking of the solutions. In addition, 16 20 20 19 19 20 20 20 20 19.75
the PROMETHEE and GRA methods have a positive correlation 24 21 21 20 20 21 21 21 21 20.75
coefficient with the other methods. Hence, the average ranks 14 22 22 22 22 22 22 22 22 22
11 23 23 23 23 23 23 23 23 23
of solutions are calculated by taking the ranks given by all the 20 24 24 24 24 24 24 24 24 24
MADM methods and presented in Table 16. From Table 16 it can 3 25 25 25 25 25 25 25 25 25
be observed that the SAW, WPM, and COPRAS methods have Note: ‘POS’-Pareto optimal solution number; ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’-
suggested solution 5 as the best choice. The TOPSIS and MTOP- MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’- COPRAS; ‘H’- GRA.
SIS methods have suggested solution 15 as the best choice.
Similarly, the PROMETHEE and GRA methods have suggested
solution 2 as the best choice. The VIKOR method has suggested be made that the system thermal efficiency is varied between
the solutions 5, 12, 15, and 23 as compromise solutions which are 0.2276 and 0.2377, power output is varied between 67.32 and
ranked as 1. Finally, the Pareto optimal solution 5 can be consid- 71.60 kW, and non-dimensional thermo-economic function is
ered as the best solution with an average rank value of 4. The varied from 0.2995 to 0.3145. Table 19 presents the ranks
solutions 8 and 3 are ranked as 24 and 25 respectively by all the achieved by the MTPG-Jaya algorithm Pareto optimal solutions
MADM methods. using MADM methods. Table 20 presents the Spearman corre-
The Pareto optimal solutions achieved by the MTPG-Jaya lation coefficients between different pairs of MADM methods
algorithm for multi-objective optimisation of solar dish Brayton ranking for MTPG-Jaya algorithm solutions. Here an observation
heat engine are presented in Table 18. Here an observation can can be made that for different pairs of MADM methods have

Table 18. The Pareto optimal solutions for the solar dish Brayton heat engine case study obtained by the MTPG-Jaya algorithm.
POS H L R TH (K) T L (K) T 1 (K) P (kW) ηm F
1 0.7 0.7 0.8 1000 400 566.305268 67.865708 0.237761 0.314429
2 0.7 0.7 0.8 1000 400 563.264362 67.322717 0.237707 0.314500
3 0.7 0.7 0.8 1000 400 612.218083 71.603715 0.227694 0.299502
4 0.7 0.7 0.8 1000 400 580.194858 69.844165 0.236705 0.312437
5 0.7 0.7 0.8 1000 400 596.190295 71.178661 0.233207 0.307244
6 0.7 0.7 0.8 1000 400 576.298864 69.369714 0.237205 0.313254
7 0.7 0.7 0.8 1000 400 603.436399 71.478083 0.230941 0.304029
8 0.7 0.7 0.8 1000 400 582.349260 70.080585 0.236366 0.311905
9 0.7 0.7 0.8 1000 400 593.102844 70.994974 0.234051 0.308460
10 0.7 0.7 0.8 1000 400 570.061312 68.480602 0.237680 0.314151
11 0.7 0.7 0.8 1000 400 610.425523 71.598531 0.228399 0.300479
12 0.7 0.7 0.8 1000 400 605.748852 71.535849 0.230137 0.302901
13 0.7 0.7 0.8 1000 400 601.213221 71.405500 0.231678 0.305068
14 0.7 0.7 0.8 1000 400 609.303237 71.590012 0.228829 0.301077
15 0.7 0.7 0.8 1000 400 584.294658 70.278490 0.236023 0.311378
16 0.7 0.7 0.8 1000 400 607.632094 71.569725 0.229455 0.301948
17 0.7 0.7 0.8 1000 400 598.034553 71.272195 0.232668 0.306473
18 0.7 0.7 0.8 1000 400 586.086527 70.447876 0.235676 0.310853
19 0.7 0.7 0.8 1000 400 572.738576 68.882102 0.237527 0.313830
20 0.7 0.7 0.8 1000 400 612.141354 71.603704 0.227725 0.299544
21 0.7 0.7 0.8 1000 400 573.555178 68.998576 0.237464 0.313713
22 0.7 0.7 0.8 1000 400 592.057814 70.925008 0.234319 0.308850
23 0.7 0.7 0.8 1000 400 605.030026 71.519812 0.230391 0.303257
24 0.7 0.7 0.8 1000 400 608.024902 71.575315 0.229310 0.301745
25 0.7 0.7 0.8 1000 400 590.630281 70.822985 0.234671 0.309363
Note: ‘POS’-Pareto optimal solution number.
20 R. V. RAO AND H. S. KEESARI

Table 20. Spearman’s rank correlation coefficients between different pairs of Table 22. Ranks given by the MADM methods for Jaya algorithm solutions for the
MADM methods for the MTPG-Jaya algorithm solutions ranking presented in solar dish Brayton heat engine case study.
Table 19.
POS A B C D E F G H Average rank
Method A B C D E F G H
18 1 1 7 7 1 10 1 10 4.75
A 1 0.9992 0.9231 0.9231 0.9811 0.7554 1 0.7592 12 3 2 4 4 1 11 3 11 4.875
B 0.9992 1 0.9285 0.9285 0.9842 0.7508 0.9992 0.7546 17 4 4 3 3 1 12 4 12 5.375
C 0.9231 0.9285 1 1 0.9719 0.5238 0.9231 0.5292 16 5 5 1 1 1 13 5 13 5.5
D 0.9231 0.9285 1 1 0.9719 0.5238 0.9231 0.5292 6 2 3 9 9 7 9 2 9 6.25
E 0.9811 0.9842 0.9719 0.9719 1 0.6751 0.9811 0.6782 13 7 6 2 2 1 14 7 14 6.625
F 0.7554 0.7508 0.5238 0.5238 0.6751 1 0.7554 0.9992 24 6 7 10 10 10 8 6 8 8.125
G 1 0.9992 0.9231 0.9231 0.9811 0.7554 1 0.7592 7 9 8 5 5 6 15 9 15 9
H 0.7592 0.7546 0.5292 0.5292 0.6782 0.9992 0.7592 1 25 8 9 12 12 11 7 8 7 9.25
Note: ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’- MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’- 19 10 10 6 6 8 16 10 16 10.25
COPRAS; ‘H’- GRA. 21 12 12 15 15 12 6 12 5 11.125
5 11 11 8 8 9 17 11 17 11.5
15 13 13 16 16 14 5 13 3 11.625
22 14 14 18 18 16 3 14 2 12.375
strong correlation coefficients. The TOPSIS and MTOPSIS meth- 1 16 17 19 19 18 1 17 1 13.5
4 15 15 11 11 13 18 15 18 14.5
ods rankings are identical with a correlation coefficient equal to 10 19 19 21 21 20 2 19 4 15.625
one. Similarly, the rankings obtained by SAW and COPRAS meth- 11 17 16 13 13 15 19 16 19 16
odsare also identical. The PROMETHEE and GRA methods have 2 20 20 22 22 21 4 20 6 16.875
20 18 18 14 14 17 20 18 20 17.375
relatively less correlation coefficients but positive with other 14 21 21 17 17 19 21 21 21 19.75
methods. Hence, the average ranks of solutions can be calcu- 8 22 22 20 20 22 22 22 22 21.5
lated by considering the ranks given by all the methods. The 9 23 23 23 23 23 23 23 23 23
23 24 24 24 24 24 24 24 24 24
SAW, WPM, COPRAS methods have suggested solution 8 as the 3 25 25 25 25 25 25 25 25 25
best solution. The VIKOR method has suggested the solutions
Note: ‘POS’-Pareto optimal solution number; ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’-
4, 8, 15, and 18 as compromise solutions which are ranked as MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’- COPRAS; ‘H’- GRA.
1. The TOPSIS and MTOPSIS methods have suggested solution
18 as the best solution, and the PROMETHEE and GRA methods
have suggested solutions 1 as the best solution. Finally, the solu- 67.32 and 71.6 kW, and the non-dimensional thermo-economic
tion 8 with an average rank of 3.25 can be considered as the best function varies between 0.2994 and 0.3145. The ranking of the
solution. All the MADM methods have ranked the solutions 14, Jaya algorithm Pareto optimal solutions by MADM methods
11, 20, and 3 as 22, 23, 24, and 25 respectively. is presented in Table 22. The Spearman correlation coefficient
Table 21 presents the Pareto optimal solutions achieved by between different pairs of MADM methods for the ranking of
the Jaya algorithm for multi-objective optimisation of solar dish the Jaya algorithm solutions is presented in Table 23. Here an
Brayton heat engine case study. Here an observation can be observation can be made that the Spearman correlation coeffi-
made that, the thermal efficiency achieved by Jaya algorithm cients between different pairs of MADM methods are similar to
varies from 0.2276 to 0.2377, the power output varies between the correlation coefficients for the ranking of the AMTPG-Jaya

Table 21. The Pareto optimal solutions for the solar dish Brayton heat engine case study obtained by the Jaya algorithm.
POS H L R TH (K) TL (K) T 1 (K) P (kW) ηm F
1 0.7 0.7 0.8 1000 400 566.304785 67.865625 0.237761 0.314429
2 0.7 0.7 0.8 1000 400 563.266986 67.323203 0.237707 0.314500
3 0.7 0.7 0.8 1000 400 612.224590 71.603715 0.227692 0.299498
4 0.7 0.7 0.8 1000 400 596.313882 71.185304 0.233172 0.307194
5 0.7 0.7 0.8 1000 400 591.743513 70.903187 0.234398 0.308965
6 0.7 0.7 0.8 1000 400 580.194346 69.844107 0.236706 0.312437
7 0.7 0.7 0.8 1000 400 588.913509 70.690333 0.235072 0.309953
8 0.7 0.7 0.8 1000 400 606.486041 71.550508 0.229873 0.302532
9 0.7 0.7 0.8 1000 400 610.045866 71.596106 0.228546 0.300682
10 0.7 0.7 0.8 1000 400 564.352829 67.521785 0.237739 0.314491
11 0.7 0.7 0.8 1000 400 598.489145 71.293413 0.232531 0.306278
12 0.7 0.7 0.8 1000 400 583.930309 70.242541 0.236090 0.311480
13 0.7 0.7 0.8 1000 400 587.565832 70.578504 0.235369 0.310393
14 0.7 0.7 0.8 1000 400 602.978152 71.464497 0.231096 0.304247
15 0.7 0.7 0.8 1000 400 569.780469 68.436725 0.237692 0.314178
16 0.7 0.7 0.8 1000 400 586.321566 70.469185 0.235629 0.310782
17 0.7 0.7 0.8 1000 400 585.157537 70.361593 0.235860 0.311130
18 0.7 0.7 0.8 1000 400 581.821725 70.024380 0.236453 0.312041
19 0.7 0.7 0.8 1000 400 590.691499 70.827514 0.234656 0.309342
20 0.7 0.7 0.8 1000 400 599.284652 71.328806 0.232287 0.305932
21 0.7 0.7 0.8 1000 400 571.966010 68.769343 0.237579 0.313933
22 0.7 0.7 0.8 1000 400 568.977174 68.309364 0.237720 0.314252
23 0.7 0.7 0.8 1000 400 611.535289 71.602956 0.227965 0.299877
24 0.7 0.7 0.8 1000 400 576.677881 69.418559 0.237163 0.313183
25 0.7 0.7 0.8 1000 400 575.304708 69.238810 0.237308 0.313432
Note: ‘POS’-Pareto optimal solution number.
INTERNATIONAL JOURNAL OF AMBIENT ENERGY 21

Table 23. Spearman’s rank correlation coefficients between different pairs of the solution 18 with an average rank of 4.75 can be considered
MADM methods for the Jaya algorithm solutions ranking presented in Table 22.
as the best solution, and all the MADM methods have ranked the
Method A B C D E F G H solutions 9, 23, and 3 as 23, 24, and 25 respectively.
A 1 0.9969 0.8769 0.8769 0.9552 0.4262 0.9992 0.4462 Now the Pareto optimal solutions selected from the Pareto-
B 0.9969 1 0.8969 0.8969 0.9699 0.4000 0.9977 0.4200 fronts of the Jaya, MTPG-Jaya and AMTPG-Jaya algorithms based
C 0.8769 0.8969 1 1 0.9699 0.1100 0.8815 0.1262
on the average rank of the MADM methods are compared with
D 0.8769 0.8969 1 1 0.9699 0.1100 0.8815 0.1262
E 0.9552 0.9699 0.9699 0.9699 1 0.2703 0.9575 0.2903 those reported for NSGA-II algorithm. Figure 10 shows the Pareto
F 0.4262 0.4000 0.1100 0.1100 0.2703 1 0.4123 0.9946 fronts achieved by the Jaya algorithm, MTPG-Jaya algorithm
G 0.9992 0.9977 0.8815 0.8815 0.9575 0.4123 1 0.4323 and AMTPG-Jaya algorithm along with the Pareto optimal solu-
H 0.4462 0.4200 0.1262 0.1262 0.2903 0.9946 0.4323 1
tions of NSGA-II algorithm. The NSGA-II algorithm Pareto optimal
Note: ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’- MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’-
COPRAS; ‘H’- GRA. solutions are reported using the Shannon Entropy, TOPSIS, and
LINMAP decision making methods. Here an observation can be
made that the NSGA-II algorithm Pareto optimal solutions are in
and MTPG-Jaya algorithm solutions. Different pairs of MADM second level Paretofront which is dominated by the Paretofronts
methods excluding the pairs with the PROMETHEE or the GRA of the proposed algorithms.
methods have strong correlation coefficients. Also, the pairs with The comparison of the Pareto optimal solutions obtained by
the PROMETHEE or the GRA methods have positive correlation the different algorithms for the multi-objective optimisation of
coefficients. Hence, the ranking of all MADM methods can be solar dish Brayton heat engine system is presented in Table 24.
considered for calculating the average ranks of Jaya algorithm Here an observation can be made that, the AMTPG-Jaya, MTPG-
Pareto optimal solutions. From Table 22 it can be observed that Jaya, and Jaya algorithms solutions have better objective func-
the SAW, COPRAS and WPM methods have suggested solution tion values compared to the solutions reported for the NASGA-
18 as the best solution. The TOPSIS and MTOPSIS methods have II algorithm in terms of the power output, thermal efficiency,
suggested solution 16 as the best solution. The PROMETHEE and and thermo-economic function. This implies that the NSGA-II
GRA methods have suggested solution 1 as the best solution. solutions are dominated by the solutions of the AMTPG-Jaya,
The VIKOR method suggested the solutions 12, 13, 16, 17, and MTPG-Jaya, and Jaya algorithms. The Jaya algorithm solution has
18 as the compromise solutions which are ranked as 1. Finally, higher thermal efficiency and thermo-economic function value,

Figure 10. The Pareto optimal solutions of the NSGA-II (Shannon entropy, TOPSIS, and LINMAP), Jaya, MTPG-Jaya, and AMTPG-Jaya algorithms obtained for multi-
objective optimisation of the solar dish Brayton heat engine case study.

Table 24. Comparison of the Pareto optimal solutions of the NSGA-II, Jaya, MTPG-Jaya, AMTPG-Jaya algorithms obtained for the solar dish Brayton heat engine case study.
Algorithm H L R TH (K) TL (K) T 1 (K) P (kW) ηm F
NSGAII-Shannon 0.69 0.7 0.79 999.99 400.06 586.33 68.22 0.2339 0.3089
Entropy
NSGAII-LINMAP 0.69 0.69 0.79 999.99 400.01 586.87 68.65 0.2331 0.3077
NSGAII-TOPSIS 0.7 0.7 0.8 999.99 400.01 586.69 68.64 0.2332 0.3078
Jaya 0.7 0.7 0.8 1000 400 581.821725 70.0243801 0.2364535 0.3120406
MTPG-Jaya 0.7 0.7 0.8 1000 400 582.34926 70.0805847 0.2363664 0.3119052
AMTPG-Jaya 0.7 0.7 0.8 1000 400 581.877339 70.0303566 0.2364444 0.3120264
Note: Results in the bold figure indicates better value; NSGA-II results are available in Li, Liao, and Liu (2015).
22 R. V. RAO AND H. S. KEESARI

Table 25. Comparison of the ranking of the Pareto optimal solutions given Table 27. The performance indicator values of the Jaya, MTPG-Jaya, AMTPG-Jaya
by NSGA-II, Jaya, MTPG-Jaya, AMTPG-Jaya algorithms obtained by the MADM meth- algorithms in the solar dish Brayton heat engine case study.
ods for the solar dish Brayton heat engine case study.
Coverage
Algorithm A B C D E F G H Average rank
Algorithm Spacing Hypervolume Jaya MTPG-Jaya AMTPG-Jaya
NSGAII-Shannon Entropy 4 4 6 6 1 4 4 4 4.125
NSGAII-TOPSIS 6 6 4 4 6 6 6 6 5.5 Jaya 0.0562112 0.026508392 – 0 0.08
NSGAII-LINMAP 5 5 5 5 5 5 5 5 5 MTPG-Jaya 0.0454323 0.026508835 0 – 0.08
Jaya 2 2 3 3 1 1 3 1 2 AMTPG-Jaya 0.0401055 0.026510432 0 0 –
MTPG-Jaya 3 3 1 1 1 3 1 3 2 Note: Results in the bold figure indicates better value.
AMTPG-Jaya 1 1 2 2 1 2 2 2 1.625
Note: ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’- MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’-
COPRAS; ‘H’- GRA.
solution can be considered as a better solution than the Jaya
algorithm solution. The NSGA-II (Shannon Entropy, LINMAP, and
Table 26. Spearman’s rank correlation coefficients between different pairs of TOPSIS) solutions are ranked 4, 5, and 6 respectively based on
MADM methods for the NSGA-II, Jaya, MTPG-Jaya, AMTPG-Jaya algorithm’s solu- the average ranks of the MADM methods.
tions ranking presented in Table 25.
Now the performances of the Jaya, MTPG-Jaya, and AMTPG-
Method A B C D E F G H Jaya algorithms are compared on the basis of coverage, spac-
A 1 1 0.6 0.6 0.8452 0.9429 0.8286 0.9429 ing and Hypervolume indicators. These performance indicators
B 1 1 0.6 0.6 0.8452 0.9429 0.8286 0.9429 are calculated for the Pareto fronts achieved by AMTPG-Jaya,
C 0.6 0.6 1 1 0.3719 0.5429 0.7714 0.5429
D 0.6 0.6 1 1 0.3719 0.5429 0.7714 0.5429 MTPG-Jaya, and Jaya algorithms presented in Tables 2, 5, and 8
E 0.8452 0.8452 0.3719 0.3719 1 0.8452 0.8452 0.8452 respectively. The coverage, spacing and Hypervolume values of
F 0.9429 0.9429 0.5429 0.5429 0.8452 1 0.7714 1 the AMTPG-Jaya, MTPG-Jaya, and Jaya algorithms are presented
G 0.8286 0.8286 0.7714 0.7714 0.8452 0.7714 1 0.7714
H 0.9429 0.9429 0.5429 0.5429 0.8452 1 0.7714 1 in Table 27. Here an observation can be made that the spacing of
Note: ‘A’- SAW; ‘B’- WPM; ‘C’- TOPSIS; ‘D’- MTOPSIS; ‘E’- VIKOR; ‘F’- PROMETHEE; ‘G’- the AMTPG-Jaya algorithm is 13.28% and 40.1% better than that
COPRAS; ‘H’- GRA. of the Jaya and MTPG-Jaya algorithms respectively. In addition,
the Hypervolume indicator values are competitive for the pro-
posed algorithms. From the coverage values presented in Table
and MTPG-Jaya algorithm has higher power output value. The 27, it can be observed that 8% of the AMTPG-Jaya algorithm solu-
ranking of the Pareto optimal solutions of these algorithms by tions are dominated by both Jaya and MTPG-Jaya algorithms
the MADM methods is presented in Table 25. The Spearman cor- solutions respectively.
relation coefficients of different pairs of the MADM methods are In this case study also, along with the tri-objective opti-
presented in Table 26. Here an observation can be made that the misation of the solar dish Brayton heat engine system, dual-
pairs SAW-WPM, TOPSIS-MTOPSIS, and PROMETHEE-GRA have objective optimisation is carried out using the Jaya, MTPG-
correlation coefficient equal to one which indicates that the Jaya and AMTPG-Jaya algorithms. Figure 11 presents the con-
ranking by the methods in the respective pairs is identical. Fur- flicting nature between power output and thermal efficiency.
thermore, different pairs of MADM methods excluding the pairs The power output is varied between 67.86 and 71.6 kW for
with the TOPSIS or the MTOPSIS methods have strong correla- all the proposed algorithms and the thermal efficiency var-
tion coefficients. Also, the pairs with the TOPSIS or the MTOPSIS ied between 0.227693 and 0.237761 for all the proposed algo-
methods have positive correlation coefficients. Therefore, the rithms. The spacing values for the Pareto fronts of the Jaya,
average ranks can be calculated by considering the ranking of MTPG-Jaya, and AMTPG-Jaya algorithms are 0.041442, 0.032753,
all the MADM methods. and 0.023561 respectively. Here an observation can be made
The Pareto optimal solution of the AMTPG-Jaya algorithm that, even though the Paretofronts of the Jaya, MTPG-Jaya, and
can be considered as the best solution with an average rank AMTPG-Jaya algorithms are overlapping, the spacing metric of
of 1.625. The AMTPG-Jaya algorithm solution has better out- the AMTPG-Jaya algorithm has 75.8% and 39.01% lesser value
put power, thermal efficiency, and thermo-economic function compared to that of the Jaya and MTPG-Jaya algorithms respec-
values compared to those of the NSGA-II solutions. The output tively. Furthermore, the Hypervolume of the Pareto fronts of
power is increased by 2.65%, 2.01%, and 2.03% when compared the Jaya (56.792), MTPG-Jaya (56.793), and AMTPG-Jaya (56.794)
to that of NSGA-II (Shannon Entropy, LINMAP, and TOPSIS) solu- algorithms are competitive.
tions. Similarly, the thermal efficiency is increased by 1.09%, Similarly, Figure 12 presents the Pareto fronts obtained by
1.43%, and 1.39% when compared to that of The NSGA-II (Shan- the AMTPG-Jaya, MTPG-Jaya, and Jaya algorithms for simulta-
non Entropy, LINMAP, and TOPSIS) solutions respectively. The neous optimisation of the power output and non-dimensional
thermo-economic function is increased by 1.01%, 1.41%, and thermo-economic function. The power output is varied from
1.37% when compared to that of NSGA-II (Shannon Entropy, 67.3 to 71.6 kW, and the non-dimensional thermo-economic
LINMAP, and TOPSIS) solutions respectively. The Jaya and MTPG- function varied from 0.299492 to 0.3145 for all the proposed
Jaya algorithm Pareto optimal solutions are the next best solu- algorithms. The spacing of the AMTPG-Jaya algorithm Pareto
tions with an average rank of 2. However, it can be observed front (0.024322) is 37.1% and 51.7% lower than that of the Jaya
that the MTPG-Jaya algorithm solution has ranked 1 by four (0.036906) and MTPG-Jaya (0.033355) algorithms respectively.
MADM (TOPSIS, MTOPSIS, VIKOR, and COPRAS) methods and The Hypervolume of the Pareto fronts obtained by the Jaya
the Jaya algorithm solution ranked 1 by three MADM (VIKOR, (4.592881), MTPG-Jaya (4.593116), and AMTPG-Jaya (4.59315)
PROMETHEE, and GRA) methods. Thus, the MTPG-Jaya algorithm algorithms are competitive.
INTERNATIONAL JOURNAL OF AMBIENT ENERGY 23

Figure 11. The Pareto frontiers obtained by the Jaya, MTPG-Jaya, and AMTPG-Jaya algorithms for dual objective (P−ηm ) optimisation of the solar dish Brayton heat
engine case study.

Figure 12. The Pareto frontiers obtained by the Jaya, MTPG-Jaya, and AMTPG-Jaya algorithms for dual objective (P−F) optimisation of the solar dish Brayton heat engine
case study.

Figure 13 presents the Pareto fronts obtained by the Jaya, economic function is also increased. Here it can be inferred
MTPG-Jaya and AMTPG-Jaya algorithms for simultaneous opti- that the effect of design parameters on the thermal efficiency
misation of the thermal efficiency and non-dimensional thermo- and thermo-economic function is similar in nature. Furthermore,
economic function. The variation in thermal efficiency is ranged the spacing values for the Pareto fronts of the Jaya, MTPG-
between 0.237706 and 0.237761 and variation in thermo- Jaya, and AMTPG-Jaya algorithms are 0.058665, 0.043054, and
economic function is ranged between 0.314429 and 0.3145. 0.039572 respectively. The spacing metric value of the AMTPG-
Here an observation can be made that the variations of both the Jaya algorithm Pareto front is 48.2% and 8.7% better than that of
objectives are relatively very less. Also, it can be observed from the Jaya and MTPG-Jaya algorithms respectively. The Hypervol-
Figure 10, that as the thermal efficiency increases the thermo ume of the Pareto fronts obtained by the Jaya, MTPG-Jaya, and
24 R. V. RAO AND H. S. KEESARI

Figure 13. The Pareto frontiers obtained by the Jaya, MTPG-Jaya, and AMTPG-Jaya algorithms for dual objective (ηm −F) optimisation of the solar dish Brayton heat
engine case study.

AMTPG-Jaya algorithms are the same, which is 0.002435598. The fronts obtained by the Jaya, MTPG-Jaya, and AMTPG-Jaya algo-
following section presents the conclusions of this work. rithms are dominating the solutions of NSGA-II algorithm. The
AMTPG-Jaya algorithm’s solution has attained the best average
rank of 1.625 and has resulted in an increase of the output power,
6. Conclusions thermal efficiency, and thermo-economic function by 2.65%,
1.43%, and 1.37% respectively when compared to those of the
In this work, the optimisation of a solar-driven heat engine sys-
NSGA-II solutions.
tem and a solar assisted Brayton engine system is carried out by
In addition, the performances of the proposed algorithms
using the multi-objective Jaya algorithm and its variants known
are compared using the coverage, spacing, and Hypervolume
as MTPG-Jaya and AMTPG-Jaya algorithms. In both the case
performance indicators in bi-objective and tri-objective opti-
studies, optimal Pareto fronts are obtained by the proposed
misation scenarios. In both the case studies, hyper volumes
algorithms through the posteriori method of multi-objective
generated by the Pareto fronts of the proposed algorithms are
optimisation. As the Pareto optimal solutions in Pareto front are
same, but the spacing metric values are varying. The AMTPG-
non-dominant in nature, selecting the most suitable solution
Jaya algorithm attained superior spacing values when compared
is difficult. Hence, the applicability of decision-making meth-
to those of MTPG-Jaya and Jaya algorithms.
ods such as SAW, WPM, PROMETHEE, TOPSIS, MTOPSIS, VIKOR,
COPRAS, and GRA in multi-objective optimisation problems has
been discussed. As these methods rank the alternatives employ- Disclosure statement
ing different principles, the ranks given by each method would No potential conflict of interest was reported by the authors.
vary. Therefore, a decision making procedure based on the aver-
age rank of the aforementioned methods is proposed to identify ORCID
the best Pareto optimal solution in a Pareto front. Hameer Singh Keesari http://orcid.org/0000-0001-7633-1170
Furthermore, by employing the proposed procedure, he pro-
cedure, the best solutions are identified from the Pareto fronts
obtained by Jaya, MTPG-Jaya, and AMTPG-Jaya algorithms and
References
compared with the renowned NSGA-II algorithm solutions. The Ahmadi, Mohammad H., and Mohammad Ali Ahmadi. 2015. “Thermody-
AMTPG-Jaya algorithm, in both the case studies, has attained namic Analysis and Optimization of an Irreversible Ericsson Cryogenic
Refrigerator Cycle.” Energy Conversion and Management 89: 147–155.
the best average rank and healthier compromise among the Ahmadi, Mohammad H., and Mohammad Ali Ahmadi. 2016a. “Multi Objec-
objective functions. In the solar-driven heat engine case study, tive Optimization of Performance of Three-Heat-Source Irreversible
the AMTPG-Jaya algorithm’s solution attained the best aver- Refrigerators Based Algorithm NSGAII.” Renewable and Sustainable Energy
age rank of 1.875. The AMTPG-Jaya algorithm’s solution leads Reviews 60: 784–794.
to an increase in the thermal efficiency, the ecological function, Ahmadi, Mohammad H., and Mohammad Ali Ahmadi. 2016b. “Thermody-
namic Analysis and Optimisation of an Irreversible Radiative-Type Heat
and non-dimensional power output by 2.63%, 1%, and 17.61% Engine by Using Non-Dominated Sorting Genetic Algorithm.” Interna-
respectively when compared to those of NSGA-II solutions . Sim- tional Journal of Ambient Energy 37 (4): 403–408. doi:10.1080/01430750.
ilarly, in solar dish Brayton heat engine case study, the Pareto 2014.977498.
INTERNATIONAL JOURNAL OF AMBIENT ENERGY 25

Ahmadi, Mohammad Hosein, Mohammad Ali Ahmadi, and Michel Feidt. Generator in Electrically Series and Parallel Configuration.” Applied Ther-
2016. “Performance Optimization of a Solar-Driven Multi-Step Irreversible mal Engineering 103: 1312–1323. doi:10.1016/j.applthermaleng.2016.05.009.
Brayton Cycle Based on a Multi-Objective Genetic Algorithm.” Oil & Arora, Rajesh, S. C. Kaushik, and Raj Kumar. 2016. “Multi-Objective Thermo-
Gas Science and Technology – Revue d’IFP Energies Nouvelles 71 (1): 16. dynamic Optimization of Solar Parabolic Dish Stirling Heat Engine With
doi:10.2516/ogst/2014028. Regenerative Losses Using NSGA-II and Decision Making.” Applied Solar
Ahmadi, Mohammad Hossein, Mohammad Ali Ahmadi, Adel Mellit, Fathol- Energy 52 (4): 295–304. doi:10.3103/S0003701X16040046.
lah Pourfayaz, and Michel Feidt. 2016. “Thermodynamic Analysis and Arora, Rajesh, S. C. Kaushik, and Raj Kumar. 2017. “Multi-Objective Thermo-
Multi Objective Optimization of Performance of Solar Dish Stirling dynamic Optimisation of Solar Parabolic Dish Stirling Heat Engine Using
Engine by the Centrality of Entransy and Entropy Generation.” Inter- NSGA-II and Decision Making.” International Journal of Renewable Energy
national Journal of Electrical Power and Energy Systems 78: 88–95. Technology 8 (1): 64–92.
doi:10.1016/j.ijepes.2015.11.042. Arora, Rajesh, S. C. Kaushik, Raj Kumar, and Ranjana Arora. 2016. “Multi-
Ahmadi, Mohammad H., Mohammad Ali Ahmadi, Fathollah Pourfayaz, Objective Thermo-Economic Optimization of Solar Parabolic Dish Stirling
Mokhtar Bidi, Hadi Hosseinzade, and Michel Feidt. 2016. “Optimization Heat Engine With Regenerative Losses Using NSGA-II and Decision Mak-
of Powered Stirling Heat Engine With Finite Speed Thermodynamics.” ing.” International Journal of Electrical Power and Energy Systems 74: 25–35.
Energy Conversion and Management 108: 96–105. doi:10.1016/j.ijepes.2015.07.010.
Ahmadi, Mohammad H., Mohammad Ali Ahmadi, Fathollah Pourfayaz, Ashouri, Milad, Mohammad H. Ahmadi, S. Mohsen Pourkiaei, Fatemeh Razi
Hadi Hosseinzade, Emin Acıkkalp, Iskander Tlili, and Michel Feidt. 2016. Astaraei, Roghaye Ghasempour, Tingzhen Ming, and Javid Haj Hemati.
“Designing a Powered Combined Otto and Stirling Cycle Power Plant 2018. “Exergy and Exergo-Economic Analysis and Optimization of a Solar
Through Multi-Objective Optimization Approach.” Renewable and Sus- Double Pressure Organic Rankine Cycle.” Thermal Science and Engineering
tainable Energy Reviews 62: 585–595. Progress 6: 72–86. doi:10.1016/j.tsep.2017.10.002.
Ahmadi, Mohammad H., Mohammad Ali Ahmadi, and Seyed Abbas Sadat- Behzadi, Amirmohammad, Ali Habibollahzade, Pouria Ahmadi, Ehsan Gho-
sakkak. 2015. “Thermodynamic Analysis and Performance Optimization lamian, and Ehsan Houshfar. 2019. “Multi-Objective Design Optimization
of Irreversible Carnot Refrigerator by Using Multi-Objective Evolution- of a Solar Based System for Electricity, Cooling, and Hydrogen Produc-
ary Algorithms (MOEAs).” Renewable and Sustainable Energy Reviews 51: tion.” Energy 169: 696–709. doi:10.1016/j.energy.2018.12.047.
1055–1070. Bellos, Evangelos, Christos Tzivanidis, and Kimon A. Antonopoulos. 2017.
Ahmadi, Mohammad H., Saeed Dehghani, Amir H. Mohammadi, Michel Feidt, “Parametric Analysis and Optimization of a Solar Assisted Gas Turbine.”
and Marco A. Barranco-Jimenez. 2013. “Optimal Design of a Solar Driven Energy Conversion and Management 139: 151–165. doi:10.1016/j.encon
Heat Engine Based on Thermal and Thermo-Economic Criteria.” Energy man.2017.02.042.
Conversion and Management 75: 635–642. doi:10.1016/j.enconman.2013. Brans, J. P., Ph Vincke, and B. Mareschal. 1986. “How to Select and How to
07.078. Rank Projects: The Promethee Method.” European Journal of Operational
Ahmadi, Mohammad H., Mohammad Ali Jokar, Tingzhen Ming, Michel Research 24: 228–238. doi:10.1016/0377-2217(86)90044-5.
Feidt, Fathollah Pourfayaz, and Fatemeh Razi Astaraei. 2018. “Multi- Carrillo Caballero, Gaylord Enrique, Luis Sebastian Mendoza, Arnaldo Mar-
Objective Performance Optimization of Irreversible Molten Carbonate tin Martinez, Electo Eduardo Silva, Vladimir Rafael Melian, Osvaldo José
Fuel Cell–Braysson Heat Engine and Thermodynamic Analysis with Eco- Venturini, and Oscar Almazán del Olmo. 2017. “Optimization of a Dish
logical Objective Approach.” Energy 144: 707–722. doi:10.1016/j.energy. Stirling System Working with DIR-Type Receiver Using Multi-Objective
2017.12.028. Techniques.” Applied Energy 204: 271–286. doi:10.1016/j.apenergy.2017.
Ahmadi, Mohammad H., Mehdi Mehrpooya, Sara Abbasi, Fathollah Pour- 07.053.
fayaz, and Joan Carles Bruno. 2017. “Thermo-Economic Analysis and Dai, Dongdong, Fang Yuan, Rui Long, Zhichun Liu, and Wei Liu. 2018.
Multi-Objective Optimization of a Transcritical CO2 Power Cycle Driven “Performance Analysis and Multi-Objective Optimization of a Stirling
by Solar Energy and LNG Cold Recovery.” Thermal Science and Engineering Engine Based on MOPSOCD.” International Journal of Thermal Sciences
Progress 4: 185–196. doi:10.1016/j.tsep.2017.10.004. 124: 399–406. doi:10.1016/j.ijthermalsci.2017.10.030.
Ahmadi, Mohammad H., Amir H. Mohammadi, Saeed Dehghani, and Marco Deb, Kalyanmoy, Amrit Pratap, Sameer Agarwal, and T. Meyarivan. 2002. “A
A. Barranco-Jiménez. 2013. “Multi-Objective Thermodynamic-Based Opti- Fast and Elitist Multiobjective Genetic Algorithm: NSGA-II.” IEEE Trans-
mization of Output Power of Solar Dish-Stirling Engine by Implement- actions on Evolutionary Computation 6 (2): 182–197. doi:10.1109/4235.
ing an Evolutionary Algorithm.” Energy Conversion and Management 75: 996017.
438–445. doi:10.1016/j.enconman.2013.06.030. Deng, Hepu, Chung Hsing Yeh, and Robert J. Willis. 2000. “Inter-Company
Ahmadi, Mohammad H., Mohammad Alhuyi Nazari, and Michel Feidt. 2018. Comparison Using Modified TOPSIS with Objective Weights.” Comput-
“Thermodynamic Analysis and Multi-Objective Optimisation of Endore- ers and Operations Research 27: 963–973. doi:10.1016/S0305-0548(99)
versible Lenoir Heat Engine Cycle Based on the Thermo-Economic Per- 00069-6.
formance Criterion.” International Journal of Ambient Energy 0 (0): 1–10. Du, Dinh-Cong, Ho-Huu Vinh, Vo-Duy Trung, Ngo-Thi Hong Quyen, and
doi:10.1080/01430750.2017.1423386. Nguyen-Thoi Trung. 2018. “Efficiency of Jaya Algorithm for Solving the
Ahmadi, Mohammad H., Mohammad Sameti, Seyed M. Pourkiaei, Tingzhen Optimization-Based Structural Damage Identification Problem Based on
Ming, Fathollah Pourfayaz, Ali J. Chamkha, Hakan F. Oztop, and Moham- a Hybrid Objective Function.” Engineering Optimization 50 (8): 1233–1251.
mad Ali Jokar. 2018. “Multi-Objective Performance Optimization of doi:10.1080/0305215X.2017.1367392.
Irreversible Molten Carbonate Fuel Cell–Stirling Heat Engine–Reverse Duan, Chen, Xinggang Wang, Shuiming Shu, Changwei Jing, and Huawei
Osmosis and Thermodynamic Assessment with Ecological Objective Chang. 2014. “Thermodynamic Design of Stirling Engine Using Multi-
Approach.” Energy Science and Engineering 6 (6): 783–796. doi:10.1002/ Objective Particle Swarm Optimization Algorithm.” Energy Conversion and
ese3.252. Management 84: 88–96. doi:10.1016/j.enconman.2014.04.003.
Ahmadi, Mohammad Hossein, Hoseyn Sayyaadi, Saeed Dehghani, and Hadi Entezari, Ashkan, Ali Manizadeh, and Rouhollah Ahmadi. 2018. “Energetical,
Hosseinzade. 2013. “Designing a Solar Powered Stirling Heat Engine Exergetical and Economical Optimization Analysis of Combined Power
Based on Multiple Criteria: Maximized Thermal Efficiency and Power.” Generation System of Gas Turbine and Stirling Engine.” Energy Conversion
Energy Conversion and Management 75: 282–291. doi:10.1016/j.encon and Management 159: 189–203. doi:10.1016/J.ENCONMAN.2018.01.012.
man.2013.06.025. Ferreira, Ana C., Manuel L. Nunes, José C. F. Teixeira, Luís A. S. B. Mar-
Ahmadi, Mohammad H., Hoseyn Sayyaadi, Amir H. Mohammadi, and tins, Senhorinha F. C. F. Teixeira, and Silvia A. Nebra. 2017. “Design
Marco A. Barranco-Jimenez. 2013. “Thermo-Economic Multi-Objective of a Solar Dish Stirling Cogeneration System: Application of a Multi-
Optimization of Solar Dish-Stirling Engine by Implementing Evolu- Objective Optimization Approach.” Applied Thermal Engineering 123:
tionary Algorithm.” Energy Conversion and Management 73: 370–380. 646–657. doi:10.1016/j.applthermaleng.2017.05.127.
doi:10.1016/j.enconman.2013.05.031. Fishburn, P. C. 1967. “Letter to the Editor—Additive Utilities with Incom-
Arora, Ranjana, S. C. Kaushik, and Rajesh Arora. 2016. “Thermodynamic plete Product Sets: Application to Priorities and Assignments.” Operations
Modeling and Multi-Objective Optimization of Two Stage Thermoelectric Research 15 (3): 537–542.
26 R. V. RAO AND H. S. KEESARI

Ge, Yanlin, Wenhua Wang, Lingen Chen, and Fankai Meng. 2018. “Exer- Springer Series in Advanced Manufacturing. London: Springer London.
goeconomic Performance Analysis and Optimisation of an Irreversible- doi:10.1007/978-1-4471-4375-8.
Closed Intercooled Regenerated Gas Turbine Cycle.” International Journal Rao, R. Venkata. 2016. “Jaya: A Simple and New Optimization Algorithm
of Ambient Energy 0 (0): 1–9. doi:10.1080/01430750.2018.1531259. for Solving Constrained and Unconstrained Optimization Problems.”
Habibollahzade, Ali, Ehsan Gholamian, Ehsan Houshfar, and Amirmoham- International Journal of Industrial Engineering Computations 7 (1): 19–34.
mad Behzadi. 2018. “Multi-Objective Optimization of Biomass-Based Solid doi:10.5267/j.ijiec.2015.8.004.
Oxide Fuel Cell Integrated with Stirling Engine and Electrolyzer.” Energy Rao, Ravipudi Venkata. 2019. Jaya: An Advanced Optimization Algorithm
Conversion and Management 171: 1116–1133. doi:10.1016/j.enconman. and Its Engineering Applications. Cham: Springer International Publishing.
2018.06.061. doi:10.1007/978-3-319-78922-4.
Harkouss, Fatima, Farouk Fardoun, and Pascal Henry Biwole. 2018. “Multi- Rao, R. Venkata, and Hameer Singh Keesari. 2018. “Multi-Team Perturbation
Objective Optimization Methodology for Net Zero Energy Buildings.” Guiding Jaya Algorithm for Optimization of Wind Farm Layout.” Applied
Journal of Building Engineering 16: 57–71. doi:10.1016/j.jobe.2017.12.003. Soft Computing 71: 800–815. doi:10.1016/j.asoc.2018.07.036.
Hooshang, Mazdak, Somayeh Toghyani, Alibakhsh Kasaeian, Reza Askari Rao, R. Venkata, Hameer Singh Keesari, P. Oclon, and Jan Taler. 2019a.
Moghadam, and Mohammad Hossein Ahmadi. 2017. “Enhancing and “An Adaptive Multi-Team Perturbation-Guiding Jaya Algorithm for
Multi-Objective Optimising of the Performance of Stirling Engine Using Optimization and Its Applications.” Engineering with Computers, 1–29.
Third-Order Thermodynamic Analysis.” International Journal of Ambient doi:10.1007/s00366-019-00706-3.
Energy 39 (4): 382–391. doi:10.1080/01430750.2017.1303638. Rao, R. Venkata, Hameer Singh Keesari, P. Oclon, and Jan Taler. 2019b.
Hwang, Ching-Lai, and Kwangsun Yoon. 1981. Multiple Attribute Decision “Improved Multi-Objective Jaya Optimization Algorithm for a Solar Dish
Making: Methods and Applications A State-of-the-Art Survey. New York: Stirling Engine.” Journal of Renewable and Sustainable Energy 11 (2):
Springer. doi:10.1007/978-3-642-48318-9. 025903. doi:10.1063/1.5083142.
Jing, Rui, Xingyi Zhu, Zhiyi Zhu, Wei Wang, Chao Meng, Nilay Shah, Ning Rao, R. V., K. C. More, L. S. Coelho, and V. C. Mariani. 2017. “Multi-
Li, and Yingru Zhao. 2018. “A Multi-Objective Optimization and Multi- Objective Optimization of the Stirling Heat Engine Through Self-Adaptive
Criteria Evaluation Integrated Framework for Distributed Energy System Jaya Algorithm.” Journal of Renewable and Sustainable Energy 9 (3),
Optimal Planning.” Energy Conversion and Management 166: 445–462. doi:10.1063/1.4987149.
doi:10.1016/j.enconman.2018.04.054. Rao, R. V., K. C. More, J. Taler, and P. Ocłoń. 2016. “Dimensional Optimiza-
Jokar, Mohammad Ali, Mohammad H. Ahmadi, Mohsen Sharifpur, Josua P. tion of a Micro-Channel Heat Sink Using Jaya Algorithm.” Applied Thermal
Meyer, Fathollah Pourfayaz, and Tingzhen Ming. 2017. “Thermodynamic Engineering 103: 572–582. doi:10.1016/j.applthermaleng.2016.04.135.
Evaluation and Multi-Objective Optimization of Molten Carbonate Fuel Rao, R. Venkata, Dhiraj P. Rai, and J. Balic. 2016. “Multi-Objective Optimiza-
Cell-Supercritical CO2 Brayton Cycle Hybrid System.” Energy Conversion tion of Machining and Micro-Machining Processes Using Non-Dominated
and Management 153: 538–556. Sorting Teaching–Learning-Based Optimization Algorithm.” Journal of
Khoshbazan, Maryam, Mohammad Hossein Ahmadi, Tingzhen Ming, Jamal Intelligent Manufacturing, 1–23. doi:10.1007/s10845-016-1210-5.
Tabe Arjmand, and Mohammad Rahimzadeh. 2018. “Thermo-Economic Rao, R. V., D. P. Rai, J. Ramkumar, and J. Balic. 2016. “A New Multi-Objective
Analysis and Multiobjective Optimization of Micro-CHP Stirling System for Jaya Algorithm for Optimization of Modern Machining Processes.”
Different Climates of Iran.” International Journal of Low-Carbon Technolo- Advances in Production Engineering & Management 11 (4): 271–286.
gies 13 (4): 388–403. doi:10.1093/ijlct/cty050. doi:10.14743/apem2016.4.226.
Li, Yuqiang, Shengming Liao, and Gang Liu. 2015. “Thermo-Economic Multi- Rao, R. Venkata, and Ankit Saroj. 2017. “Economic Optimization of Shell-and-
Objective Optimization for a Solar-Dish Brayton System Using NSGA-II Tube Heat Exchanger Using Jaya Algorithm with Maintenance Consider-
and Decision Making.” International Journal of Electrical Power and Energy ation.” Applied Thermal Engineering 116: 473–487. doi:10.1016/j.applther
Systems 64: 167–175. doi:10.1016/j.ijepes.2014.07.027. maleng.2017.01.071.
Luo, Zhongyang, Umair Sultan, Mingjiang Ni, Hao Peng, Bingwei Shi, Rao, R. Venkata, and Ankit Saroj. 2018. “Multi-Objective Design Optimiza-
and Gang Xiao. 2016. “Multi-Objective Optimization for GPU3 Stirling tion of Heat Exchangers Using Elitist-Jaya Algorithm.” Energy Systems 9 (2):
Engine by Combining Multi-Objective Algorithms.” Renewable Energy 94: 305–341. doi:10.1007/s12667-016-0221-9.
114–125. doi:10.1016/j.renene.2016.03.008. Rao, Ravipudi Venkata, Ankit Saroj, Pawel Ocloń, Jan Taler, and Dawid
Mahmoudimehr, Javad, and Parvin Sebghati. 2019. “A Novel Multi-Objective Taler. 2018. “Single- and Multi-Objective Design Optimization of Plate-
Dynamic Programming Optimization Method: Performance Manage- Fin Heat Exchangers Using Jaya Algorithm.” Heat Transfer Engineering 39:
ment of a Solar Thermal Power Plant as a Case Study.” Energy 168: 1201–1216. doi:10.1080/01457632.2017.1363629.
796–814. doi:10.1016/j.energy.2018.11.079. Rashidi, Halimeh, and Jamshid Khorshidi. 2018. “Exergoeconomic Analysis
Michailidis, Panagiotis D. 2017. “An Efficient Multi-Core Implementation and Optimization of a Solar Based Multigeneration System Using Multi-
of the Jaya Optimisation Algorithm.” International Journal of Parallel, objective Differential Evolution Algorithm.” Journal of Cleaner Production
Emergent and Distributed Systems 0: 1–33. doi:10.1080/17445760.2017. 170: 978–990. doi:10.1016/j.jclepro.2017.09.201.
1416387. Razavi Hajiagha, Seyed Hossein, Shide Sadat Hashemi, and Edmundas Kaz-
Miller, D. W., and M. K. Starr. 1969. Executive Decisions with Operations imieras Zavadskas. 2013. “A Complex Proportional Assessment Method
Research. Englewood Cliffs, NJ: Prentice Hall. for Group Decision Making in an Interval-Valued Intuitionistic Fuzzy Envi-
Nazemzadegan, Mohammad Reza, Alibakhsh Kasaeian, Somayeh Toghyani, ronment.” Technological and Economic Development of Economy 19 (1):
Mohammad Hossein Ahmadi, R. Saidur, and Tingzhen Ming. 2018. “Multi- 22–37. doi:10.3846/20294913.2012.762953.
Objective Optimization in a Finite Time Thermodynamic Method for Dish- Sadatsakkak, Seyed Abbas, Mohammad Hossein Ahmadi, and Moham-
Stirling by Branch and Bound Method and MOPSO Algorithm.” Frontiers in mad Ali Ahmadi. 2015. “Thermodynamic and Thermo-Economic Analysis
Energy, 1–17. doi:10.1007/s11708-018-0548-0. and Optimization of an Irreversible Regenerative Closed Brayton Cycle.”
Opricovic, Serafim, and Gwo Hshiung Tzeng. 2002. “Multicriteria Planning of Energy Conversion and Management 94: 124–129. doi:10.1016/j.encon
Post-Earthquake Sustainable Reconstruction.” Computer-Aided Civil and man.2015.01.040.
Infrastructure Engineering 17: 211–220. doi:10.1111/1467-8667.00269. Sayyaadi, Hoseyn, Mohammad Hossein Ahmadi, and Saeed Dehghani. 2015.
Opricovic, Serafim, and Gwo-Hshiung Tzeng. 2003. “Fuzzy Multicriteria “Optimal Design of a Solar-Driven Heat Engine Based on Thermal and
Model for Postearthquake Land-Use Planning.” Natural Hazards Review 4: Ecological Criteria.” Journal of Energy Engineering 141 (3): 04014012.
59–64. doi:10.1061/(asce)1527-6988(2003)4:2(59). doi:10.1061/(ASCE)EY.1943-7897.0000191.
Rao, R. Venkata. 2007. Decision Making in the Manufacturing Environment. Schott, Jason R. 1995. Fault Tolerant Design Using Single and Multicriteria
Springer Series in Advanced Manufacturing. Springer Series in Advanced Genetic Algorithm Optimization. Boston, MA: Massachusetts Institute of
Manufacturing. London: Springer London. doi:10.1007/978-1-84628- Technology.
819-7. Shirazi, Ali, Robert A. Taylor, Graham L. Morrison, and Stephen D. White.
Rao, R. Venkata. 2013. Decision Making in Manufacturing Environment Using 2017. “A Comprehensive, Multi-Objective Optimization of Solar-Powered
Graph Theory and Fuzzy Multiple Attribute Decision Making Methods. Absorption Chiller Systems for Air-Conditioning Applications.” Energy
INTERNATIONAL JOURNAL OF AMBIENT ENERGY 27

Conversion and Management 132: 281–306. doi:10.1016/j.enconman. Wang, Long, Zijun Zhang, Chao Huang, and Kwok Leung Tsui. 2018.
2016.11.039. “A GPU-Accelerated Parallel Jaya Algorithm for Efficiently Estimating
Tiwari, Deepak, Ahmad Faizan Sherwani, and Akhilesh Arora. 2017. “Ther- Li-Ion Battery Model Parameters.” Applied Soft Computing 65: 12–20.
modynamic and Multi-Objective Optimisation of Solar-Driven Organic doi:10.1016/j.asoc.2017.12.041.
Rankine Cycle Using Zeotropic Mixtures.” International Journal of Ambient Warid, Warid, Hashim Hizam, Norman Mariun, and Noor Izzri Abdul-Wahab.
Energy 40 (2): 135–151. doi:10.1080/01430750.2017.1378718. 2016. “Optimal Power Flow Using the Jaya Algorithm.” Energies 9 (9): 678.
Toghyani, Somayeh, Alibakhsh Kasaeian, and Mohammad H. Ahmadi. Warid, Warid, Hashim Hizam, Norman Mariun, and Noor Izzri Abdul-Wahab.
2014. “Multi-Objective Optimization of Stirling Engine Using Non-Ideal 2018. “A Novel Quasi-Oppositional Modified Jaya Algorithm for Multi-
Adiabatic Method.” Energy Conversion and Management 80: 54–62. Objective Optimal Power Flow Solution.” Applied Soft Computing 65:
doi:10.1016/j.enconman.2014.01.022. 360–373. doi:10.1016/j.asoc.2018.01.039.
Tzeng, Gwo Hshiung, Cheng Wei Lin, and Serafim Opricovic. 2005. “Multi- Yu, P. L. 1973. “A Class of Solutions for Group Decision Problems.” Manage-
Criteria Analysis of Alternative-Fuel Buses for Public Transportation.” ment Science 19: 936–946. doi:10.1287/mnsc.19.8.936.
Energy Policy 33: 1373–1383. doi:10.1016/j.enpol.2003.12.014. Zeleny, M. 1982. Multiple Criteria Decision Making. New York: McGraw Hill.
Wang, Long, and Chao Huang. 2018. “A Novel Elite Opposition-Based Jaya Zhang, Yudong, Xiaojun Yang, Carlo Cattani, Ravipudi Venkata Rao, Shui-
Algorithm for Parameter Estimation of Photovoltaic Cell Models.” Optik hua Wang, and Preetha Phillips. 2016. “Tea Category Identification Using
155: 351–356. doi:10.1016/j.ijleo.2017.10.081. a Novel Fractional Fourier Entropy and Jaya Algorithm.” Entropy 18 (3): 77.
Wang, Shuihua, Ravipudi Venkata Rao, Peng Chen, Yu-Dong Zhang, Liuaijun doi:10.3390/e18030077.
Aijun, and Ling Wei. 2017. “Abnormal Breast Detection in Mammogram Zitzler, E., K. Deb, and L. Thiele. 2000. “Comparison of Multiobjective Evo-
Images by Feed-Forward Neural Network Trained by Jaya Algorithm.” lutionary Algorithms: Empirical Results.” Evolutionary Computation 8 (2):
Fundamenta Informaticae 151 (1–4): 191–211. 173–195. doi:10.1162/106365600568202.

S-ar putea să vă placă și