Documente Academic
Documente Profesional
Documente Cultură
Turbine
Houman Hanachia,b, Jie Liua,c,1, Ping Dinga,c , Il Yong Kimb, Chris K Mechefskeb
d
ite
a
National Research Base of Intelligent Manufacturing Service, Chongqing Technology and
ed
py
b
Department of Mechanical and Materials Engineering, Queen’s University, Kingston, ON, K7L
Co
3N6, Canada
c
Department of Mechanical and Aerospace Engineering, Carleton University, Ottawa, ON, K1S
5B6, Canada
ot
tN
hanachih@lifepredictiontech.com
rip
jie.liu@carleton.ca
sc
ping.ding@carleton.ca
nu
kimiy@queensu.ca
Ma
chris.mechefske@queensu.ca
ed
pt
ce
Ac
1
Corresponding author. Tel.: +1 6135202600, E-mail address: jie.liu@carleton.ca
1
GTP-18-1117 Liu
Abstract
Gas turbine engines are widely used for power generation, ranging from stationary power plants
to airplane propulsion systems. Compressor fouling is the dominant degradation mode in gas
turbines that leads to economic losses due to power deficit and extra fuel consumption. Washing
d
ite
of the compressor removes the fouling matter and retrieves the performance, while causing a
ed
variety of costs including loss of production during service time. In this paper the effect of fouling
py
and washing on the revenue of the power plant is studied, and a general solution for the optimum
Co
time between washes of the compressor under variable fouling rates and demand power is
presented and analyzed. The framework calculates the savings achievable with optimization of
ot
time between washes during a service period. The methodology is utilized to optimize total costs
tN
of fouling and washing and analyze the effects and sensitivities to different technical and economic
rip
factors. As a case study, it is applied to a sample set of cumulative gas turbine operating data for a
sc
time-between-overhauls and the potential saving has been estimated. The results show
2
GTP-18-1117 Liu
Nomenclature
Symbols
𝐶𝐶 cost 𝜃𝜃 temperature
d
ite
EGT exhaust gas temperature 𝜙𝜙 relative humidity
ed
𝑁𝑁 shaft speed 𝑎𝑎 actual
py
𝑃𝑃 pressure 𝑎𝑎𝑎𝑎 ambient
Co
𝑝𝑝 price 𝐶𝐶 compressor
𝑡𝑡 time 𝐹𝐹 fuel
𝑣𝑣 𝑓𝑓
nu
𝜂𝜂 𝑇𝑇
ed
3
GTP-18-1117 Liu
1. Introduction
Performance of gas turbine engines (GTEs) gradually deteriorates during the operation from two
compressor parts and hot corrosion in the turbine section cause permanent deterioration in GTE
d
ite
performance. At the same time, fouling and congestion of particulates on the aerofoils in the
ed
compressor section, lead to disturbance of gas-path aerodynamics and affect the compressor
py
performance. Isentropic efficiency and mass flow capacity of the compressor decrease due to
Co
fouling and the matching point with the turbine section deviates from the design point. Power
capacity declines because of elevation of the exhaust gas temperature (EGT) and fuel flow
ot
increases due to loss of cycle efficiency. It is estimated that fouling accounts for around 80% of
tN
the GTE performance deterioration during the service life [1]. As a result of fouling, power deficit
rip
leads to loss of revenue, and extra fuel consumption adds to the costs of power generation. That is
sc
why compressors should be cleaned from fouling on a regular basis. There are two methods to
clean the compressor of fouling matter. Spraying water into the compressor while the GTE is
nu
operating is called on-line washing, which partly recovers the performance of the compressor. A
Ma
more effective washing method is offline washing, in which the full compressor section is hand
ed
washed with water and special additives [2]. Offline washing (or on-crank washing) can effectively
pt
reverse the fouling and restore the compressor performance, however, it requires the GTE to be
ce
shut down, while no power can be generated. The service requires a skilled labor force as well as
Ac
consumable materials [3]. Another cost associated with off-line washing is accounting for the
equivalent life consumption for the cold starts after each wash [4].
Operating the GTE under fouling is associated with significant cost, while the compressor wash
service also adds costs. The question to answer is therefore, how to schedule the wash services, so
4
GTP-18-1117 Liu
that the total revenue loss of the operation is minimized? Some works in the literature have tried
to address this question based on the cost factors and the information on the compressor fouling
rate [5][6][7][8][9][10]. In some studies, the fouling increase with the operating time of the GTE,
is assumed to have a linear relation [8][11]. In reality, growth of the fouling is not linear with time
d
[12], moreover, the rate of fouling depends on different factors such as the filter quality, mass flow
ite
rate of the GTE and the humidity of the intake air [2][3][6]. As a result, optimization of the
ed
compressor washing that requires realistic prediction models for the fouling, becomes a complex
py
task to address in practice.
Co
The main objective of this paper is to incorporate the results of fouling prediction modeling into
ot
tN
the fouling and washing total cost function, to optimize the offline washing schedule under variable
nonlinear fouling rates and maximise the power production revenue. In Section 2, a cost model for
rip
the compressor fouling and washing is developed and the optimum duration of washing period
sc
(i.e., time between washes) is parametrically calculated. In Section 3, variation of the GTE
nu
performance with respect to deterioration of the fouling over time is studied, and quantitative
Ma
results for a single spool SGT100 GTE are acquired. The effects of performance deterioration and
other economic variables on the corresponding costs of the fouling and washing are analysed in
ed
Section 4. A case study is presented and analysed in Section 5 and a conclusive discussion on the
pt
Operation of the GTE with compressor fouling condition decreases the achievable power capacity
and increases the fuel consumption. Both of these effects, increase the operating cost and reduces
5
GTP-18-1117 Liu
the revenue of power plants. Likewise, compressor offline washing has its own costs due to missed
opportunity of power production, required resources for the washing service and accelerated
degradation of the GTE during re-starting. In this section different factors contributing to the total
loss of revenue are discussed and a cost function is presented for the fouling/washing process.
d
ite
2.1. Costs of fouling and washing
ed
py
The demand power in most power plants has no cut-off limit and the goal is to achieve as much
production as possible from the assets. However, this is not the case for some local power
Co
generators where the demand power fluctuates from season to season. The GTE power capacity
ot
also gradually decreases due to non-recoverable degradation mechanisms. With fouling of the
tN
compressor, the actual power under fouling becomes even less than expected long-term capacity.
rip
With compressor offline washing the actual power can return to the long-term capacity if the
washing completely removes the fouling. Figure 1(a) shows the demand power 𝑃𝑃𝑃𝑃𝑑𝑑𝑑𝑑 (𝑡𝑡), GTE
sc
long-term clean capacity 𝑃𝑃𝑃𝑃𝑐𝑐𝑐𝑐 (𝑡𝑡) and the actual capacity under fouling 𝑃𝑃𝑃𝑃𝑓𝑓 (𝑡𝑡) for a typical cycle.
nu
GTE actual power 𝑃𝑃𝑃𝑃𝑎𝑎 (𝑡𝑡) is limited to the capacity of the GTE under fouling, and at the same
Ma
time, it is limited to the demand power, i.e., 𝑃𝑃𝑃𝑃𝑎𝑎 = Min(𝑃𝑃𝑃𝑃𝑓𝑓 , 𝑃𝑃𝑃𝑃𝑑𝑑𝑑𝑑 ), as shown with the thick
ed
purple curve in Fig. 1(a). The power deficit due to fouling of the compressor is therefore,
pt
where Min(𝑃𝑃𝑃𝑃𝑐𝑐𝑐𝑐 , 𝑃𝑃𝑃𝑃𝑑𝑑𝑑𝑑 ) equals the ideal power generation with no compressor fouling. For the
Ac
example cycle presented in Fig. 1, part (b) shows the power deficit. Power deficit is the missed
opportunity for power production that leads to loss of profit. For a plant operating with partial load,
it can be shown that the corresponding loss of profit (or cost of power deficit) 𝑐𝑐𝛥𝛥𝑃𝑃𝑃𝑃 is proportional
where 𝑝𝑝𝑠𝑠𝑠𝑠 is the sales price and 𝑝𝑝𝑐𝑐 is the cost price of electricity (excluding the investment
depreciation) at a given economic condition. Compressor fouling also leads to extra fuel
consumption. For a given power level, 𝑃𝑃𝑃𝑃𝑎𝑎 , the actual fuel flow under fouling 𝑊𝑊𝐹𝐹𝑎𝑎 is more than
d
ite
the hypothetical fuel consumption 𝑊𝑊𝐹𝐹𝑐𝑐𝑐𝑐 under the compressor clean condition.
ed
𝛥𝛥𝑊𝑊𝐹𝐹 (𝑡𝑡) = 𝑊𝑊𝐹𝐹𝑎𝑎 − 𝑊𝑊𝐹𝐹𝑐𝑐𝑐𝑐 (𝑃𝑃𝑃𝑃𝑎𝑎 ). (3)
py
A typical variation of extra fuel consumption due to fouling is presented in Fig. 1(c). For extra fuel
Co
consumption, the rate of extra fuel cost 𝑐𝑐𝛥𝛥𝑊𝑊𝐹𝐹 (𝑡𝑡) is proportional to extra fuel consumption rate in
most cases. The rate of the cost per unit of time ($/h) attributed to compressor fouling comprises
ot
both the costs of the power deficit 𝑐𝑐𝛥𝛥𝑃𝑃𝑃𝑃 (𝑡𝑡) and the rate of extra fuel consumption 𝑐𝑐𝛥𝛥𝑊𝑊𝐹𝐹 (𝑡𝑡).
tN
rip
Washing the compressor comprises a fixed cost 𝐶𝐶𝑠𝑠0 that is independent from the duration of the
sc
wash. The cost of consumable material and the equivalent life consumption for cold starts are
nu
among the fixed costs. At the same time, there are time-dependent costs for loss of production
Ma
𝑐𝑐𝑃𝑃𝑃𝑃 (𝑡𝑡) and variable service costs 𝑐𝑐𝑠𝑠 (𝑡𝑡) such as labor that may be proportional to the duration of
service time. From Eqs. 1 and 2, loss of production while the GTE is out of service (𝑃𝑃𝑃𝑃𝑎𝑎 = 0)
ed
Considering all cost components, the total cost of fouling and washing is calculated as,
ce
𝑇𝑇 𝑇𝑇 +𝑇𝑇𝑠𝑠
𝐶𝐶𝑓𝑓,𝑠𝑠 = ∫0 𝑏𝑏𝑏𝑏 (𝑐𝑐𝛥𝛥𝑃𝑃𝑃𝑃 + 𝑐𝑐𝛥𝛥𝑊𝑊𝐹𝐹 ) 𝑑𝑑𝑑𝑑 + 𝐶𝐶𝑠𝑠0 + ∫𝑇𝑇 𝑏𝑏𝑏𝑏 (𝑐𝑐𝑃𝑃𝑃𝑃 + 𝑐𝑐𝑠𝑠 ) 𝑑𝑑𝑑𝑑 ,
Ac
(4)
𝑏𝑏𝑏𝑏
where 𝑇𝑇𝑏𝑏𝑏𝑏 and 𝑇𝑇𝑠𝑠 are the duration of the washing period (i.e., time between two consecutive
washes) and the duration of washing service respectively. The average rate of total costs during a
d
Fouling and washing costs include different components, mostly out of control of the power plants.
ite
For instance, it is clear that the shorter the washing time is, the smaller the loss of production will
ed
be. However, a reasonable time is required for compressor washing that cannot be further reduced.
py
The only variable the power plants effectively have under control is the duration of the washing
Co
period. To find the optimum washing time of the compressor, the minimum of the cost rate function
in Eq. 5 should be found with respect to 𝑇𝑇𝑏𝑏𝑏𝑏 , where 𝑑𝑑 𝑐𝑐𝑓𝑓,𝑠𝑠 /𝑑𝑑 𝑇𝑇𝑏𝑏𝑏𝑏 = 0. Using Eqs. 4 and 5, the
ot
tN
optimum 𝑇𝑇𝑏𝑏𝑏𝑏 is found as the root of following equation.
(6)
nu
Eq. 6 provides a general solution for the optimum time of compressor washing. To find the
numerical results, time-based functions of 𝑐𝑐𝛥𝛥𝑃𝑃𝑃𝑃 , 𝑐𝑐𝛥𝛥𝑊𝑊𝐹𝐹 , 𝑐𝑐𝑃𝑃𝑃𝑃 , 𝑐𝑐𝑠𝑠 and the fixed service costs 𝐶𝐶𝑠𝑠0 need
Ma
to be available. The rate of total costs 𝑐𝑐𝑓𝑓,𝑠𝑠 can then be found from Eqs. 4 and 5, and utilized in Eq.
ed
6.
pt
ce
The fouling rate depends on various factors such as humidity and filter condition, however, it is
Ac
acceptable to assume that fouling accumulates with a linear rate in the early stages [13].
Considering linearly increasing rates for 𝑐𝑐𝛥𝛥𝑃𝑃𝑃𝑃 (𝑡𝑡) = 𝛼𝛼1 𝑡𝑡 and 𝑐𝑐𝛥𝛥𝑊𝑊𝐹𝐹 (𝑡𝑡) = 𝛼𝛼2 𝑡𝑡 during the fouling
and fixed rates for 𝑐𝑐𝑃𝑃𝑃𝑃 and 𝑐𝑐𝑠𝑠 during washing, the optimum washing time from Eq. 6 can be found
as:
8
GTP-18-1117 Liu
where, 𝐶𝐶𝑠𝑠 = 𝐶𝐶𝑠𝑠0 + (𝑐𝑐𝑃𝑃𝑃𝑃 + 𝑐𝑐𝑠𝑠 )𝑇𝑇𝑠𝑠 is the total cost of washing. Analysis of the resulting washing
time and its dependency to the rate of fouling and economic variables is presented in section 4.
d
ite
3. GTE performance deterioration
ed
py
Gas-path parameters of the GTE are defined by the operating condition (control setting and
ambient condition) and the health state of the parts including the compressor fouling and turbine
Co
degradation level. In the economic study of the GTE performance, fuel consumption and power
ot
capacity are the two parameters of interest. In this section, the quantitative relation between
tN
compressor fouling level and the gas-path parameters are analyzed.
rip
Gas-path parameters can be predicted with a heat balance model of the GTE. The model employs
Ma
the performance maps of the compressor and the turbine and finds the matching point of the
components such that the energy conservation and continuity equations both hold [14]. In the gas-
ed
path model, the compressor map provides the functional relation between the corrected shaft speed
pt
𝑁𝑁𝐶𝐶𝑐𝑐𝑐𝑐𝑐𝑐 , and the pressure ratio 𝑃𝑃𝑃𝑃𝐶𝐶 as inputs and the corrected air mass flow 𝑊𝑊𝐶𝐶𝑐𝑐𝑐𝑐𝑐𝑐 and the isentropic
ce
Under fouling condition, by reduction in swallowing capacity of the compressor due to reduction
of the flow area, the compressor supplies less mass flow, and at the same time, due to increased
9
GTP-18-1117 Liu
surface roughness on the blades and vanes, the isentropic efficiency of the component decreases.
Fouling severity can be quantified with coefficients of mass flow decrease 𝜌𝜌𝑊𝑊𝐶𝐶 and isentropic
efficiency decline 𝜌𝜌𝜂𝜂𝐶𝐶 with respect to the expected values predicted by the map.
d
ite
𝜌𝜌𝜂𝜂𝐶𝐶 = 1 − (𝜂𝜂𝐶𝐶𝑎𝑎 /𝜂𝜂𝐶𝐶 ) . (9)
ed
Several studies have reported a highly correlated relation between 𝜌𝜌𝑊𝑊𝐶𝐶 and 𝜌𝜌𝜂𝜂𝐶𝐶 [15][16], as a
py
result, the fouling factor can reduce to a single parameter 𝜌𝜌𝐶𝐶 = [𝜌𝜌𝑊𝑊𝐶𝐶 , 𝜌𝜌𝜂𝜂𝐶𝐶 ]. With a similar
Co
approach, 𝜌𝜌𝑇𝑇 is defined as the degradation in the hot section of the GTE. Experiments show
ot
isentropic efficiency of the turbine declines with degradation, while the mass flow capacity
tN
increases due to widening of the nozzle area. Efficiency decline and mass flow increase of the
turbine have close variation rates. During the compressor fouling, as the thermal efficiency drops,
rip
the GTE consumes more fuel and the exhaust gas temperature (EGT) rises. Power capacity of the
sc
GTE 𝑃𝑃𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚 is limited with the EGT predefined based on endurance of the hot section material
nu
during GTE design. Power deficit 𝛥𝛥𝛥𝛥𝛥𝛥 is therefore predicted by the gas-path model as the
Ma
𝛥𝛥𝛥𝛥𝛥𝛥 = 𝑃𝑃𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚 ([𝑣𝑣, 𝑁𝑁, 𝜌𝜌𝑇𝑇 ], 0) − 𝑃𝑃𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚 ([𝑣𝑣, 𝑁𝑁, 𝜌𝜌𝑇𝑇 ], 𝜌𝜌𝐶𝐶 ),
ed
(10)
pt
where 𝑣𝑣 = [𝜃𝜃𝑎𝑎𝑎𝑎 , 𝑃𝑃𝑎𝑎𝑎𝑎 , 𝜙𝜙𝑎𝑎𝑎𝑎 ] is the ambient condition vector including the temperature, pressure
ce
and humidity and 𝑁𝑁 is the shaft speed. Similarly, rise of fuel consumption at a given power 𝑃𝑃𝑃𝑃
Ac
(full load or partial load) is calculated by the gas-path model by comparing the fuel consumption
𝛥𝛥𝑊𝑊𝐹𝐹 = 𝑊𝑊𝐹𝐹 ([𝑣𝑣, 𝑁𝑁, 𝜌𝜌𝑇𝑇 ], 𝑃𝑃𝑃𝑃, 𝜌𝜌𝐶𝐶 ) − 𝑊𝑊𝐹𝐹 ([𝑣𝑣, 𝑁𝑁, 𝜌𝜌𝑇𝑇 ], 𝑃𝑃𝑃𝑃, 0). (11)
10
GTP-18-1117 Liu
For the case study, a single spool SGT100 GTE is selected. The GTE consumes natural gas and is
used for power generation for a local community. Table 1 provides the corresponding technical
specifications. There were major overhauls once every three years of operation and the operating
data between two overhauls have been utilized in this study. The records show the compressor is
d
periodically washed in uneven periods. The shaft speed is accurately controlled during the
ite
operation to maintain the frequency of the generated current. The records also show a limited
ed
variation for the ambient pressure that is expected for the stationary GTE.
py
Co
A nonlinear heat balance model was developed and verified for the gas path parameter prediction
by this research group in previous research work [14]. The developed gas path model is employed
ot
tN
in this study to quantify the functional relation of the fouling factor with power deficit and fuel
consumption increase stated in Eqs. 10 and 11. Variation of the gas-path parameters are studied
rip
under different ambient temperature and humidity and with varying level of turbine degradation.
sc
Simulation results are generated for the fouling factor from 0 to 5%. This is a wide range for
nu
fouling. In practice, the fouling seldom exceeds 2%, and the compressor is usually washed long
Ma
Figure 2 shows the results for power deficit and fuel consumption increase versus fouling severity
pt
at different condition. The results indicate that compressor fouling in cold temperatures has a
ce
relatively smaller effect on the power deficit and fuel consumption increase rates. This can be
Ac
explained by the fact that despite the decrease of flow capacity of the compressor, the high density
of the air in cold temperature partly compensates the mass flow decrease, therefore, the power
does not change as much as the case with warm ambient air. In our analysis, the effect of relative
humidity of air was found to be negligible. The results with different degradation levels of the
11
GTP-18-1117 Liu
turbine show that the more the turbine section degrades, the more the power deficit changes with
the fouling. This can be explained with the divide of the matching point between compressor and
turbine when both degrade. The mass flow capacity decreases in the compressor whereas it
increases with degradation of the turbine. For fuel consumption however, the dependency is
d
negligible, especially with smaller levels of fouling.
ite
ed
When the GTE operates at partial load, the demand power might be achieved while the
py
corresponding EGT remains within the acceptable limit, however, the fuel consumption should
Co
increase to compensate the extra losses. In Fig. 3, the effect of fouling on fuel consumption
increase for partial load operating condition at standard ambient conditions is presented. It shows
ot
tN
that the load effect on the fuel consumption rate is negligible.
rip
The results show the ambient temperature is the only factor with tangible effect on the variation
sc
rate of the power deficit and fuel consumption increase versus fouling factor. The correlations at
nu
standard ambient temperature show for each percent increase of the fouling factor, the GTE power
Ma
Susceptibility of the compressor to fouling depends on the design parameters, location and layout
Ac
of the plant, ambient air pollution, filtering quality, compressor mass flow and air humidity
[2][3][6]. In a study of six different power plants, Thames et al. reported power deficit rates ranging
from 6% to 10% per month in five power plants and about 20% per month in one case [17].
Experience shows that fouling has a fast rate after the compressor wash and it stabilizes after 1000
12
GTP-18-1117 Liu
to 2000 hours of operation [18][19]. It is also observed that condensation of humidity due to
temperature drop at the compressor inlet contributes to the fouling rate, however, with rise of
condensation rate, the fouling matter becomes unstable, which leads to self-cleaning of the
compressor [2]. In a previous work, the authors developed an adaptive neuro-fuzzy framework to
d
identify the effective parameters and the corresponding effects on the compressor fouling [13]. For
ite
the GTE under study, the compressor mass flow and the rate of humidity condensation were
ed
identified as the parameters with tangible effects on the fouling. In summary the results showed
py
the fouling factor grows steadily with the operating time for dry air conditions with a rate of
Co
𝑑𝑑𝜌𝜌𝐶𝐶 /𝑑𝑑𝑑𝑑 = 0.73% ∶ 0.95% per month, depending on the air mass flow during the GTE operation.
The fouling rate increases with the humidity condensation, reaching to 1.44% per month at the
ot
tN
highest, before the self-cleaning initiates as shown in Fig. 4. As a result, the fouling can be
estimated as linearly increasing with time, with a slope dependant on the air mass flow and
rip
With integration of the gas-path model results from the previous section and the fouling growth
Ma
model, dependency of the power deficit as well as the fuel consumption increase to the operating
time are acquired. For full-load operation of the GTE with no turbine degradation at standard
ed
ambient condition, the results are 𝑑𝑑(𝛥𝛥𝑃𝑃𝑃𝑃)/𝑑𝑑𝑑𝑑 = 1.91% and 𝑑𝑑(𝛥𝛥𝑊𝑊𝐹𝐹 )/𝑑𝑑𝑑𝑑 = 0.58% per month
pt
for the power deficit and the fuel consumption increase respectively. The values indicate that the
ce
GTE in this study has been operating in a clean environment compared to the reported power
Ac
13
GTP-18-1117 Liu
As reported by different researchers, linear trend is an acceptable estimate, especially in the early
stages of fouling, low fouling rate and low humidity [13][17][18][19]. With linear fouling, the
optimum washing time can be found with Eq. 7, however, the optimum washing period would
d
ite
vary depending on cost factors. This section studies dependency of the optimum washing period
ed
on different cost factors. Table 1 provides the technical information of the GTE, and the reference
py
assumptions for cost factors are provided in table 2.
Co
With the reference assumptions, the annual cost of fouling and washing as well as the effect on
ot
specific cost price are calculated using Eqs. 4, 5, 10 and 11 as shown in Fig. 5. In this case, the
tN
optimum washing period is 47 days, which still creates a total cost over $38,600 per year. If the
rip
compressor is washed earlier, the profit gained by the performance improvement would be covered
by the expenses of frequent washings. For instance, for 𝑇𝑇𝑏𝑏𝑏𝑏 =20 days, total costs of fouling and
sc
washing reach over $53,200, which is about $14,500 more than the minimum costs possible. With
nu
longer times between washing, on the other hand, total costs would increase due to dominance of
Ma
Average power generated in a cycle is the measure of GTE availability. With the reference data,
ce
the expected demand power is 4550 kW, whereas the maximum availability is 4493 kW if the
Ac
The specific cost of fouling and washing (costs per unit of energy) in Fig. 5(b) closely follows the
variations of total cost. There is a slight difference between the trends, such that the lowest specific
14
GTP-18-1117 Liu
cost happens at 𝑇𝑇𝑏𝑏𝑏𝑏 = 46 days; one day shorter than the optimum time for total costs. The reason is,
the average generated power declines as the 𝑇𝑇𝑏𝑏𝑏𝑏 increases beyond 20 days, therefore for shorter
𝑇𝑇𝑏𝑏𝑏𝑏 , the total costs would be depreciated over more units of generated energy, as a result, the
d
ite
Variation of total costs near the optimum washing time is subtle. When other constraints such as
ed
limitation of maintenance resources necessitate, the washing time can be shifted by a few days.
py
For instance, within ±5 days around the optimum time, the extra cost will be about $200 per year.
Co
ot
In Section 3, it was shown that air temperature, degradation of the turbine section and the rate of
tN
humidity condensation in the compressor affect the fouling rate and the resulting power deficit and
fuel consumption increase. In the following sections, the effects of the mentioned factors on the
rip
The rate of fouling has a significant impact on the fouling and washing costs as it is shown in Fig.
ed
6. In a hypothetical operation with no fouling, obviously there will be no need for compressor
pt
washing. As the fouling rate increases, the corresponding operating costs grow significantly,
ce
therefore, an earlier wash is needed to remove the fouling and avoid extra costs. With the fouling
Ac
rate assumed in the reference conditions (0.95% per month), the optimum washing is at 47 days
with an annual cost of $38,600. If the fouling rate rises to 2% per month, this washing schedule
leads to over $60,000 a year in total costs. In this case, the costs can be reduced to $55,900 if the
compressor is washed every 32 days instead. In a highly polluted environment with a fouling rate
15
GTP-18-1117 Liu
of 5% per month (similar to cases reported in [17]), if the compressor is washed at the reference
washing schedule, the yearly cost rises to $121,400. In this case, the costs can be significantly
reduced by washing the compressor every 20 days, where the yearly costs reduce to $88,300.
d
The specific cost of fouling and washing per unit of energy closely follows the same trend as the
ite
total costs. At the reference condition, the minimum specific cost is $0.99/MWh, which is 1.2%
ed
of the sales price. With the same washing schedule, under a high rate of 5% fouling per month, the
py
specific cost rises to $3.26/MWh. This can reduce to $2.31/MWh (2.9% of the sales price) by
Co
shortening the washing periods to 20 days that is the optimum washing period under this fouling
rate.
ot
tN
rip
Regardless of the fouling condition, non-recoverable degradation in the turbine section leads to
performance deterioration of the GTE. At the same time, turbine degradation amplifies the effects
sc
of fouling in overall performance of the GTE as shown in Fig. 2(c). Figure 7 shows variation of
nu
the costs at different degradation levels of the turbine, attributed only to the fouling effects. At the
Ma
reference condition, the turbine section is assumed with no degradation (𝜌𝜌𝑇𝑇 = 0). As the turbine
ed
section undergoes nonrecoverable degradation during long-term operation of the GTE, the running
costs of the GTE increase due to overall performance deterioration. Consequently, the profit
pt
margin is reduced given a fixed sales price of the electricity. In this condition, a power deficit will
ce
not lead to loss of profit, as much as it was the case for the reference condition with a new turbine
Ac
section, and the compressor can be washed in longer periods than the reference case. For a highly
degraded turbine (𝜌𝜌𝑇𝑇 = 6%) for instance, the optimum washing period increases to 𝑇𝑇𝑏𝑏𝑏𝑏 =61 days.
16
GTP-18-1117 Liu
In practice, a GTE with degraded turbine section generates much less power and consumes more
fuel than the one with a new turbine section. It should be noted that the illustrated losses in Fig.
d
ite
4.2. Effect of economic variables
ed
As it was discussed in section 2.1, the cost of fouling and washing is a function of different
py
economic variables. This section studies the quantitative effect of economic factors on the yearly
Co
costs of fouling and washing, to find the optimum washing time in each case.
ot
One may intuitively infer that if the washing service costs too much, early washing of the
tN
compressor cannot be justified. The cost of washing per service depends on several factors such
rip
as consumable material, service contracts and washing duration. As a general goal, power plants
sc
attempt to reduce their service costs, however, the cost cannot be practically reduced less than
nu
certain values and the only remaining choice is to optimize the washing time to minimize the total
Ma
costs. As shown in Fig. 8, with a $1,000 service cost, the optimum washing period is every 33 days
and the annual cost is $27,200. Incremental increases of the service cost elongates the washing
ed
period such that with $3,000 and $5,000 service cost, the optimum washing period increases to 53
pt
and 68 days. Meanwhile, the corresponding annual costs rise to $44,100 and $56,100 respectively.
ce
Ac
As Eq. 4 suggests, the total costs of fouling and washing increase when the duration of washing
gets longer. The longer service duration translates to higher service cost due to the time-dependent
costs, however, the dominant cost incurred with long service periods is due to loss of production.
With costlier washing service, we expect longer washing periods. Fig. 9 shows how the total annual
17
GTP-18-1117 Liu
cost varies with the length of washing period for different durations of washing services. An
unrealistically short washing duration of two hours leads to a total annual cost of $36,000 at its
optimum washing period of 43 days. The optimum washing period gets longer almost
proportionally with the increase of the duration of washing up to 12 hours. At this condition the
d
optimum washing period is at 69 days and the total cost rises to $56,800 per year.
ite
ed
Fuel price is one the most important factors in the operating cost of power plants. This factor is
py
normally out of control of the power plants as consumers, however, the companies can adjust their
Co
scheduling of the maintenance services to minimize the corresponding costs. It is obvious that
higher fuel price causes higher cost price, which leads to a narrower profit margin for the operation.
ot
tN
Consequently, generation of more power does not return such a large margin of profit to justify
costs of frequent washing services. In other words, we should expect longer washing periods as
rip
the fuel price increases. This may seem counter-intuitive, as the first impression with longer
sc
washing periods is burning more fuel that would cost more with the more expensive fuel. In reality
nu
however, the downside of shorter washing periods is higher total washing cost that cannot be
Ma
compensated with the gained profit achieved by increased production rate. Figure 10 shows how
the total annual cost attributed to fouling and washing changes with washing period at different
ed
fuel prices. With the lowest fuel price of $.09/m3 the compressor should be washed every 41 days.
pt
This will lead to an annual cost of $49,000. When the fuel price rises to $0.21/m3, the optimum
ce
washing period reaches to 59 days, where the total cost of fouling and washing is $27,100 per year.
Ac
It should be noted that Fig. 10 and the above-mentioned costs refer to power deficit and extra fuel
consumption attributed to the fouling only. Obviously, with the increase of the fuel price, the
overall operating cost of the power plant will grow, which is not included in this discussion.
18
GTP-18-1117 Liu
When the electricity sales price increases while the other economic factors remain unchanged, the
profit margin of the production would grow. This is an incentive to generate more power. As a
result, the maintenance plan should consider maximising the production. This objective translates
to shortening the washing periods. Compared to the reference case with $80/MWh sales price and
d
the optimum 𝑇𝑇𝑏𝑏𝑏𝑏 = 47 days, if the sales price increase to $100/MWh, the optimum washing period
ite
reduces to 39 days. On the contrary, low sales price of electricity (similar to high price of the fuel)
ed
diminishes the profit margin that in turn, reduces the necessity of a high rate of production. With
py
the sales price of $60/MWh, the production becomes so unprofitable that the washing period
Co
extends up to 71 days. It should be noted that the total annual costs correspond to the fouling and
washing only. Therefore, the losses become costlier as the sales price of electricity increases as
ot
tN
illustrated in Fig. 11.
rip
In most power-plants, variation of the demand power is managed with individual units brought
nu
into or out of service while attempting to operate the individual units under full-load. In this way,
Ma
the GTEs operate at on-design condition with the highest efficiency and the lowest specific fuel
ed
consumption. Running at partial load is not a desirable choice of operation, because the GTE
components then operate at off-design condition which leads to a drop in the overall cycle
pt
ce
efficiency [20]. Despite the disadvantages of the partial load operation, it is the only choice of
maintaining the demand power in some local power-plants with a single GTE. In partial load
Ac
operation, the corresponding fouling and washing costs would be different than those of the full
load, because the rate of fuel consumption and the amount of power deficit both vary when the
19
GTP-18-1117 Liu
GTE power alters. The corresponding variations in total annual costs of fouling and washing are
When the demand power is slightly smaller than the full load capacity, (for example 3% smaller),
d
except for the early days after the wash, the demand power cannot be maintained by the GTE
ite
degraded due to fouling. In this condition total cost of fouling comprises the cost of extra fuel
ed
consumption as well as the cost associated with power deficit. For the condition that the demand
py
power is considerably below the full load capacity, (90% of the full load for instance), the demand
Co
power is achievable even if the GTE is severely degraded by fouling. In such occasions, the losses
are attributed to extra fuel consumption only, as there is no deviation from the demand power with
ot
tN
the degraded GTE. Because of the mentioned explanations, there are two distinct regimes for the
total annual cost of fouling and washing under partial load condition as shown in Fig. 12.
rip
sc
With the assumptions provided as the reference case, the small deviation regime takes place from
nu
full load condition down to 95% of the full load. In this condition, the cost profile is highly
Ma
sensitive to variations of demand power as shown in Fig. 12. That is mainly due to the deficit in
power generation with respect to the demand power. With decrease of the demand power from full
ed
load to 95%, the optimum washing period increases from 47 days up to 69 days. Elongation of the
pt
optimum washing period can be explained as follows: operation under partial load leaves a smaller
ce
margin of profit compared to the full load. In this condition, the cost of washing service
Ac
outnumbers the gains of power recovery with that small margin. As a result, it would be more
economical to wash the compressor after longer periods than the original full load condition.
20
GTP-18-1117 Liu
From partial load below 95%, there is a separate regime for the cost profiles with less sensitivity
to variation of the load percentage. As shown in Fig. 12, the optimum washing period of 69 days
at 95% of the load increases to 88 days for 45% partial load condition. The reason for extension
of the washing period is similar to that of the small deviation regime. However, the cost profiles
d
and the corresponding optimum washing periods are less sensitive to the load level, because the
ite
only difference is caused by growth of the fuel consumption, whereas the demand power is
ed
accommodated by the GTE.
py
Co
4.4. Sensitivity analysis on cost factors
ot
In the previous section, effects of different factors in the total cost of fouling and washing was
tN
investigated, and it was shown that with variation of the factors, the compressor washing period
rip
should be modified to minimize the costs. This section studies the sensitivity of the optimum
sc
washing period to different factors. With small changes of a factor (such as service fixed cost), the
nu
optimum washing period would change as much as 𝜕𝜕𝑇𝑇𝑏𝑏𝑏𝑏 /𝜕𝜕𝐶𝐶𝑠𝑠0 . Sensitivity analysis has been
Ma
carried out at full load and partial load, i.e., 75% load, for the GTE with a healthy hot section at
reference condition (see table 2). With a healthy turbine, the optimum washing period at full load
ed
of the GTE is 46.6 days, whereas at 75% load, the compressor should be washed as late as every
pt
75 days. The study is repeated for the case with 2% degradation of the turbine. Table 3 presents
ce
the sensitivity results for all factors enumerated in 4.1 and 4.2. For instance, 1% increment in the
Ac
fouling rate (from 0.950% to 0.959% per month) calls for a 0.5% shorter washing period. The
highest sensitivity at the full load condition is due to the electricity sales price, such that a 1%
increase in the sales price requires 0.98% shorter washing period for the compressor. Unlike the
full load scenario, when operating the GTE at 75% load, the same change in the electricity sales
21
GTP-18-1117 Liu
price leads to even longer washing periods (0.17% longer). As it was explained in section 4.2 for
large variations of the demand power, the target power is achievable even with fouled condition
of the compressor and therefore, there is no power deficit costs for such partial load condition.
Consequently, the loss of profit due to operation under fouled condition will be much smaller than
d
at full power demand and it would not be necessary to wash the compressor as early.
ite
Contrary to the case of electricity sales price, sensitivity of the optimum washing period to the fuel
ed
cost and the operating labor cost is positive at full load condition. It means, with higher costs of
py
fuel and labor, it is more profitable to slightly postpone the washing time of the compressor. Under
Co
this condition, specific profit of the power generation becomes such small that it will not be
economical to rush for the costly compressor wash to retrieve the power level. Under partial load
ot
tN
condition of 75%, as explained earlier, the demand power can be achieved by the GTE, and the
increasing cost of fouling is only attributed to the increasing fuel consumption. In this condition,
rip
washing the compressor will not change the power level and it merely reduces the fuel rate. This
sc
justifies why earlier wash of the compressor reduces the costs and increases the profit margin, as
nu
represented with negative values for the corresponding sensitivities in Table 3. It should be noted
Ma
that the sensitivity factors are dependent on the technical and economic parameters of the system,
and the values should be found for the individual case in the study.
ed
When the turbine section is degraded by 2%, the optimum washing periods are 50.1 and 72.7 days
pt
for full load and 75% load condition respectively. Sensitivity of the washing periods to different
ce
factors are mostly similar to the healthy turbine. For variation of the sales price however, washing
Ac
periods should decrease for 1.2% for the degraded turbine versus 1.0% for the new turbine. This
can be explained with the fact that the profit margin for the GTE with degraded turbine is smaller
than that with healthy turbine. Increase of the sales price improves the sales margin, which at the
same time depends on the production capacity. Consequently, the compressor wash should be
22
GTP-18-1117 Liu
scheduled earlier in order to minimize the losses due to power deficit. In partial load condition,
power deficit is not a concern, and sensitivities of the washing period to the sales price hold close
d
5. Case study on GTE data
ite
ed
In Section 4, cost profiles at representative operating and economic conditions were calculated and
py
optimized. In practice, the GTE operating conditions, including the ambient condition and control
Co
settings, vary with time, and the total cost of fouling and washing should take into account such
variations. For the GTE operating data with compressor fouling and turbine degradation, hourly
ot
ambient data including temperature 𝜃𝜃𝑎𝑎𝑎𝑎 , pressure 𝑃𝑃𝑎𝑎𝑎𝑎 and humidity 𝜙𝜙𝑎𝑎𝑎𝑎 for a four-year time-
tN
between-overhaul period (35,064 hours from 2011 to 2014) are obtained from Environment
rip
Canada [21]. Demand power and control settings are obtained from a GTE (SGT100) in a local
sc
power plant along with the control cut-off condition of EGT=512°C based on the technical
specifications of the GTE. Compressor washing periods vary from one to three months base on the
nu
service logs of the GTE under study [22]. Figure 13 shows some operating data of the GTE and
Ma
The process flow diagram in Fig. 14 shows the algorithm used to implement the cost model
Ac
developed in Section 2 to the GTE data in the study. The operating data, the economic factors and
the washing data are received and processed, and the cumulative costs of fouling and washing
23
GTP-18-1117 Liu
Based on the process algorithm, Fig. 15 shows the results of power generation for the first two
washing periods. Part (a) shows the achieved power and the calculated power with clean
compressor under the same operating condition. The power deficit is the difference between the
actual and the ideal powers that are shown in part (b). The more time passes from the compressor
d
wash, the more power deficit takes place due to the fouling effects. The increasing trend of power
ite
deficit is clearly visible especially for the second period from day 81 to 167, where the power
ed
deficit increases from 0 to 175 kW within less than three months. At some operating times the
py
power deficit becomes zero, as we see at day 20 for instance. In these occasions the demand power
Co
has been lower than the GTE capacity and it is achievable even with the fouled GTE. This
ot
tN
Throughout the operating time, the power deficit leads to an accumulation of missed opportunities
rip
for power generation and sales. Fig. 15(c) shows the cumulative energy production deficit for the
sc
first two washing periods. Technically, it is the integral of the power deficit plotted in part (b). The
nu
more the power deficit increases, the faster the energy production deficit grows. Compressor
Ma
washing rectifies the power deficit and sets back the loss rate to zero, as illustrated for day 81.
Within 167 days of two washing periods, the cumulative energy production deficit reaches over
ed
261 MWh, that amounts to $20,880 merely for the missed opportunity of production.
pt
ce
Figure 16 shows the results for the fuel consumption for two periods between the 12th and 14th
Ac
washes. The actual fuel consumption and the expected fuel consumption with a clean compressor
are shown in part (a). Their difference is the extra fuel consumption due to the fouling that is
plotted in part (b). It shows that longer times between compressor washes result in more extra fuel
would be consumed. For instance, within 79 days after the compressor wash on day 770, the extra
24
GTP-18-1117 Liu
fuel consumption rate reaches to 20 m3/h. Extra fuel consumption accumulates throughout the
operating time, presented in Fig. 16(c). By washing the compressor, the rate of extra fuel
consumption slows down as we see after the wash on day 770. For the two periods shown in the
figure from day 734 to 849, the wasted fuel reaches to 17,600 m3 in total. This amounts to $2,640
d
lost during less than four months.
ite
ed
In the four-year (35,064 hours) operation of the GTE, the total energy production deficit and total
py
extra fuel consumption reach to 2,254.3 MWh and 239,840 m3 respectively. For these losses, the
Co
total cost of fouling amounts to $216,300. At the same time, the compressor has been washed 24
times that cumulatively costs $94,300. As a result, a total of $310,600 loss of profit has incurred
ot
tN
during the four-year period. Figure 17 shows how each of the costs accumulate through the
operating time.
rip
sc
Hypothetically, the GTE can be stopped for compressor washing anytime after the previous wash.
For such a hypothetical period, the total cost includes the cumulative fouling cost and the cost of
ed
washing according to Eq. 4. The average rate of total costs during that time can be found by Eq.
pt
5. Figure 18 shows the variation of the average daily costs based on the GTE data. In part (b), the
ce
variation of the average cost rate for the first washing period is presented as an example. If the
Ac
compressor is washed only ten days after the previous wash, the corresponding average cost rate
amounts to $400 per day, or $4000 in ten days. Under the corresponding operating condition, the
later the compressor is washed the smaller the average cost rate that results. This trend continues
until 𝑇𝑇𝑏𝑏𝑏𝑏 =54.4 days, where the average daily cost minimizes at $120. After this point, it gradually
25
GTP-18-1117 Liu
increases if the washing period gets longer. According to the GTE data, the first wash is after 80.9
days from the start, where the average daily cost reaches $148.2. The average rate of the fouling
and washing cost is $28.2/day more than the optimum value. In other words, there was a potential
d
ite
Fig. 18(a) shows the variation of the average rate of costs for the four-year operating period. For a
ed
better comparison, Fig. 19 shows the optimum duration of washing periods beside the actual times
py
in the GTE records. In three cases (wash numbers 5, 8 and 11) the compressor is washed nearly at
Co
the optimum time and in four cases (wash numbers 10, 12, 13 and 18) it is washed slightly earlier.
In all other cases, the compressor is washed later than the optimum time. Similar to the first period
ot
tN
explained above, for all cases of non-optimal washing periods, there is a missed potential for
saving. In a single washing period, if the compressor is washed sometime in the middle (for
rip
instance at day 54.4 for the first period), a new hypothetical period would be started from then.
sc
Measurement data are not available for such hypothetical periods, because the data corresponds to
nu
the actual washing times. To estimate the potential saving over the entire operating period, it is
Ma
assumed that the optimum average rate of cost accumulation is constant for the sub-periods
resulting from breaking the actual periods between washes. Table 4 provides potential savings in
ed
each period, based on the mentioned explanation. With shortening the washing periods in most
pt
cases, the number of optimized compressor washes during the four-year operation is found 35.9
ce
(36 washes in practice), and the total saving during this period would be about $48,700. The
Ac
Comparison of variation of the costs in Fig. 18(a) with the ambient temperature in Fig. 13(a)
reveals that the optimum average rate of cost is highly correlated with the average temperature. In
26
GTP-18-1117 Liu
other words, in warm weather, fouling and washing leads to higher average cost rates, while in
cold ambient temperature, the corresponding cost rate is smaller. In fact, in cold temperatures,
power generation capacity of the GTE that is limited with EGT, increases. As a result, the demand
power can potentially be achieved even when the compressor is slightly fouled. This reduces the
d
losses associated with power deficit and the total cost would decrease. Whereas in warm weather,
ite
the demand power is out of reach with the fouled compressor and the resulting costs attributed to
ed
power deficit will grow higher.
py
Co
6. Conclusion and discussion
ot
This article investigates the effects of ambient and operating condition on the economic losses of
tN
the GTE operation due to compressor fouling and the consequent washing, and proposes an
rip
optimization method to modify the washing schedule and minimize the resulting loss of profit. A
sc
cost model for the economic effects of fouling and washing is developed and its optimal solution
for cost minimization is provided. The effect of fouling on performance deterioration of the GTE
nu
in terms of power deficit and extra fuel consumption is studied. The resulting costs given different
Ma
durations of washing periods are calculated in the next step. Sensitivity of the cost to the operating
ed
conditions and the economic factors is analysed and the optimum washing durations under each
pt
condition are calculated. A case study on a SGT100 GTE is presented and the potential savings
ce
Based on the GTE data in the study, a monthly fouling rate of 0.95% causes a power deficit and
extra fuel consumption of 1.91% and 0.58% per month respectively. With the assumptions on the
27
GTP-18-1117 Liu
economic factors such as fuel cost and washing expenses, the study shows the corresponding
annual cost minimizes at $38,600 in the case where the compressor is washed every 47 days.
Fouling rate is highly depended on the plant location, ambient pollution and filtering condition and
d
it varies case by case. Increase of the fouling rate calls for shorter washing periods. Under similar
ite
conditions, if the fouling rate increases from 0.95% to 5%, the optimum washing period reduces
ed
from 47 days to 20 days. The case is the opposite with a degradation level of the turbine section,
py
such that the optimum washing period reaches to 50, 54 and 61 days for 2%, 4% and 6%
Co
degradations in the turbine section respectively.
ot
tN
The study shows with higher cost of washing, the compressor should be washed at longer periods.
For instance, an optimum washing period is 33 days in the case where the total cost of washing is
rip
only $1000, whereas a $5000 washing cost leads to 68 days washing periods. Similarly, the quicker
sc
the compressor is washed, the shorter the washing periods should be scheduled. A fast washing
nu
within two hours leads to a washing period of 43 days, while a slow washing of 12 hours extends
Ma
Cost of fuel and electricity sales price have opposite effects on washing period; cheaper fuel and
pt
higher electricity price both entail shortening the washing period, because they create a larger
ce
profit margin, and the plant can profit from generating more power. On the contrary, a rise of the
Ac
fuel cost ($0.06/m3 for instance) and a decrease of the electricity sales price (for example
$20/MWh) extends the optimum washing period for 12 and 24 more days respectively.
28
GTP-18-1117 Liu
The washing period increases if the GTE operates with partial load for the demand power. Small
changes in demand power lead to large variations in optimum washing period, such that a deviation
of 5% from the full load operation extends the washing period from 47 days to 69 days.
Application of the developed framework on a four-year operating data set of a GTE verifies its
d
effectiveness in cost reduction with optimization of washing schedules. It identifies the optimum
ite
washing periods given the varying operating conditions. With application of the framework, the
ed
result shows the total cost of fouling and washing can be reduced by $48,700 that is 15.7% of the
py
actual costs incurred during the four-year operating period. For this optimization, the total number
Co
of compressor washes should increase to 36 instead of the original 24 washes during four years.
ot
tN
This work shows a great potential of cost reduction through optimization of the compressor wash
scheduling using the GTE operating data and performance prediction model. Case study results
rip
for a small local power plant shows the potential of saving in larger power plants. It signifies the
sc
With some adjustments, the developed framework can be extended for maintenance optimization
ed
Acknowledgements
Ac
This project was financially supported by Fond de Recherche Nature et Technologies (FRQNT)
from the Quebec government in Canada, the Natural Sciences and Engineering Research Council
29
GTP-18-1117 Liu
(NSERC) of Canada, Life Prediction Technologies Inc. (LPTi), Ottawa, Canada, and Chongqing
d
ite
ed
py
Co
ot
tN
rip
sc
nu
Ma
ed
pt
ce
Ac
30
GTP-18-1117 Liu
References
[1] Diakunchak, I. S., 1992, “Performance Deterioration in Industrial Gas Turbines,” ASME
[2] Stalder, J. P., 2001, “Gas Turbine Compressor Washing State of the Art: Field
d
ite
Experiences,” ASME J. Eng. Gas Turbines Power, 123(2), pp. 363-370.
ed
[3] Meher-Homji, C., and Bromley, A., 2004, “Gas Turbine Axial Compressor Fouling and
py
[4] Bohrenkämper, G., Bals, H., Wrede, U., and Umlauft, R., 2000, “Hot-Gas-Path Life
Co
Extension Options for the V94.2 Gas Turbine,” ASME Paper No. 2000-GT-0178.
ot
[5] Boyce, M. P., and Gonzalez, F., 2005, “A Study of On-Line and Off-Line Turbine
tN
Washing to Optimize the Operation of a Gas Turbine,” ASME J. Eng. Gas Turbines
rip
[6] Sánchez, D., Chacartegui, R., Becerra, J. A., and Sánchez, T., 2009, “Determining
sc
compressor wash programmes for fouled gas turbines,” Proc. Inst. Mech. Eng. Part A, J.
nu
[7] Rao, P. N. S., and Naikan, V. N. A., 2008, “An Optimal Maintenance Policy for
Compressor of a Gas Turbine Power Plant,” ASME J. Eng. Gas Turbines Power, 130(2),
ed
p. 21801.
pt
[8] Aretakis, N., Roumeliotis, I., Doumouras, G., and Mathioudakis, K., 2012, “Compressor
ce
washing economic analysis and optimization for power generation,” Appl. Energy, 95, pp.
Ac
77–86.
[9] Basendwah, A. A., Pilidis, P., and Li, Y. G., 2006, “Turbine Off-Line Water Wash
31
GTP-18-1117 Liu
[10] Hovland, G., and Antoine, M., 2004, “Economic Optimisation of Gas Turbine Compressor
[11] Martín-Aragón, J., and Valdés, M., 2014, “A method to determine the economic cost of
d
fouling of gas turbine compressors,” Appl. Therm. Eng., 69(1-2), pp. 261–266.
ite
[12] Veer, T., Haglero̸d, K. K., and Bolland, O., 2004, “Measured Data Correction for
ed
Improved Fouling and Degradation Analysis of Offshore Gas Turbines,” ASME Paper
py
No. GT2004-53760.
Co
[13] Hanachi, H., Liu, J., Banerjee, A., and Chen, Y., 2016, “Prediction of Compressor Fouling
Rate Under Time Varying Operating Conditions,” ASME Paper No. GT2016-56242.
ot
tN
[14] Hanachi, H., Liu, J., Banerjee, A., Chen, Y., and Koul, A., 2015, “A physics-based
modeling approach for performance monitoring in gas turbine engines,” IEEE Trans.
rip
[15] Melino, F., Morini, M., Peretto, A., Pinelli, M., and Ruggero Spina, P., 2012,
nu
Gas Turbine Operation Time,” ASME J. Eng. Gas Turbines Power, 134(5), p. 52401.
[16] Tarabrin, A. P., Schurovsky, V. A., Bodrov, A. I., and Stalder, J. P., 1998, “Influence of
ed
Axial Compressor Fouling on Gas Turbine Unit Perfomance Based on Different Schemes
pt
[17] Thames, J. M., Stegmaier, J. W., and Ford, J. J., 1989, “On-line compressor washing
Ac
practices and benefits,” in Gas Turbine and Aeroengine Congress and Exposition.
[18] Tarabrin, A. P., Schurovsky, V. A., Bodrov, A. I., and Stalder, J. P., 1996, “An Analysis
of Axial Compressors Fouling and a Cleaning Method of Their Blading,” ASME Paper
No. 96-GT-363.
32
GTP-18-1117 Liu
[19] Tarabrin, A. P., Schurovsky, V. A., Bodrov, A. I., and Stalder, J. P., 1998, “An analysis of
[20] Saravanamuttoo, H.I.H., Rogers, G.F.C., Cohen, H., and Straznicky, P., 2008, Gas
d
Turbine Theory, 6th ed. Pearson, Canada.
ite
[21] “Environment Canada,” 2017. [Online]. Available: https://weather.gc.ca/canada_e.html.
ed
[22] Hanachi, H., Liu, J., Banerjee, A., Chen, Y., and Koul, A., 2014, “A Physics-Based
py
Performance Indicator for Gas Turbine Engines Under Variable Operating Conditions,”
Co
ASME Paper No. GT2014-26367.
ot
tN
rip
sc
nu
Ma
ed
pt
ce
Ac
33
GTP-18-1117 Liu
Tables
Table 1
Specifications of the GTE
Power 4.55 MW
Fuel Natural gas
d
Shaft speed 16500 RPM
ite
Pressure ratio 14.7
EGT 512 °C
ed
Exhaust flow 18.8 kg/s
Gross efficiency 31.1%
py
Table 2
Co
Degradation and washing information
Degradation
ot
Rate of fuel consumption increase (𝛥𝛥𝑊𝑊𝐹𝐹 ) 0.58% per month
Turbine degradation (𝜌𝜌𝑇𝑇 ) 0% (new condition)
tN
Fuel cost $0.15/m3
Operating
3h
nu
Table 3
Sensitivity of optimum washing period to different factors
ed
34
GTP-18-1117 Liu
Table 4
Washing periods and potential savings in four years
Cycle Actual Optimum Potential
number duration duration saving
(days) (days) ($)
1 80.9 54.4 2,273
d
2 86.0 35.7 6,673
ite
3 70.7 36.8 3,204
4 75.4 42.0 2,758
5 74.5 74.0 56
ed
6 53.4 43.8 515
7 69.2 44.3 3,026
py
8 40.2 40.2 12
9 72.3 34.9 3,496
Co
10 31.8 33.4 97
11 46.5 46.5 9
12 32.7 35.2 145
13 35.8 37.5 103
ot
14 79.3 44.6 5,043
15 71.6 34.2 2,964
tN
16 49.0 41.5 385
17 86.9 35.5 6,598
rip
18 32.0 33.6 95
19 56.2 50.4 224
20 52.8 37.0 1,177
sc
Total: $48,697
ed
pt
ce
Ac
35
GTP-18-1117 Liu
List of Figures
Fig. 1. Variation of power and fuel consumption with time; (a) demand power, clean capacity and
actual power, (b) power deficit percentage, (c) extra fuel consumption rate.
d
Fig. 2. Effects of fouling on power deficit (𝛥𝛥𝛥𝛥𝛥𝛥) and fuel consumption increase (𝛥𝛥𝑊𝑊𝐹𝐹 ) at different
ite
(a) ambient temperatures, and (b) turbine degradation levels.
ed
Fig. 3. Effects of fouling on fuel consumption increase (𝛥𝛥𝑊𝑊𝐹𝐹 ) at different loads.
py
Fig. 4. Compressor fouling at 16.5 kg/s compressor flow under different humidity condensation
Co
rates
Fig. 5. Variation of the cost and availability with different washing times; (a) total annual costs,
ot
and (b) specific cost per unit of energy and the average available power.
tN
Fig. 6. Effects of the fouling rate on the cost of fouling and washing; (a) total annual costs, and (b)
rip
Fig. 7. Effects of the turbine degradation level on the cost of fouling and washing; (a) total annual
nu
Fig. 8. Effect of the total cost of washing per service on the total annual costs
Fig. 10. Effect of the fuel price on the annual cost of fouling and washing
pt
Fig. 11. Rise of electricity sales price calls for shorter washing periods
ce
Fig. 12. Effects of small and large deviations of the demand power on the annual cost of fouling
Ac
and washing
Fig. 13. Operating condition, (a) ambient temperature, (b) ambient pressure, and (c) GTE power
under degradation
36
GTP-18-1117 Liu
Fig. 14. The process flow to calculate the total cost and the average cost rate for fouling and
Fig. 15. Power generation results, (a) actual power and expected power with clean compressor, (b)
d
Fig. 16. Fuel consumption results, (a) actual fuel flow and expected fuel flow with clean
ite
compressor, (b) extra fuel consumption, and (c) cumulative extra fuel consumed
ed
Fig. 17. Cumulative fouling and washing costs
py
Fig. 18. Variation of average rate of fouling and washing cost for different durations of washing
Co
periods in: (a) four-year operation, and (b) the first washing period
ot
tN
rip
sc
nu
Ma
ed
pt
ce
Ac
37
GTP-18-1117 Liu