Documente Academic
Documente Profesional
Documente Cultură
copying and redistribution of the article or any adaptations for non-commercial purposes.
Article
pubs.acs.org/IECR
Department of Chemical Engineering & Materials Science, University of Minnesota, Minneapolis, Minnesota 55455, United States
VerdeNero LLC, 955 Highway 169 North, Plymouth, Minnesota 55441, United States
ABSTRACT: This paper presents a model-based dynamic optimization study of the operation of an outdoor open pond for
microalgae cultivation. A nonlinear mathematical model based on rst-principles for predicting the growth of microalgae in open
ponds is developed and validated against literature data. To account for the impact of weather vagaries on the cultivation of
microalgae, data for local climatic conditions is incorporated into the model. The supply of dissolved CO2 to the algal culture
from a CO2 rich gas is modeled as well. Optimal monthly operating proles for the dilution rate, CO2 gas ow rate, and makeup
water ow rate are determined based on minimization of the cultivation cost. The case study included in the analysis is for
cultivating Nannochloropsis Salina over an annual production cycle in California, United States of America. The dynamic
optimization identied a set of operation proles that reduced the cultivation cost by at least 15% as opposed to relying on
heuristic approaches for improving operation.
and optimization based on computational uid dynamics
(CFD) modeling have been carried out as well.10
The economics of algae cultivation are highly sensitive to the
productivity of microalgae and maintaining steady operation in
outdoor cultivation at commercial scale is challenging, because
varying weather conditions greatly aect the growth rate.7 The
identication of optimal operating conditions mitigating the
impact of such environmental factors can maximize the productivity of microalgae over a production cycle. Most of the
studies on optimization of microalgae cultivation are based on
trial-and-error and/or design heuristics.1114 Some model-based
optimization studies have appeared focusing on laboratory algae
growth systems without accounting for weather variations.15,16
The objective of this study is to develop a dynamic
optimization formulation for determining optimal operating
conditions for improving outdoor algae production over a
production period. Specically, a rst-principles model is
developed for algae cultivation in outdoor open ponds
considering the eect of daily varying local climatic conditions
on algae growth. The model accounts for the eect of medium
temperature, irradiance level, and nutrient availability on the
growth of microalgae as well as the transfer of CO2 from a CO2
rich gas to the growth culture. Model validation against
experimental results from the literature is then conducted.
Finally, a dynamic optimization problem is formulated to
determine the optimal dilution rate, makeup water ow rate,
and CO2 gas ow rate monthly proles that minimize the cost
of producing microalgae in a representative location (Imperial
County in California, USA) over the course of a year.
1. INTRODUCTION
1
September 1, 2015
October 26, 2015
October 27, 2015
October 27, 2015
DOI: 10.1021/acs.iecr.5b03209
Ind. Eng. Chem. Res. 2016, 55, 33273337
Article
Figure 2. Modeling the open pond as a cascade of continuous stirredtank reactors (CSTRs) with an internal and an external recycle stream.
(4)
(5)
(6)
where U is the wind speed above the pond water and Tw is the
water temperature.25 The saturation vapor pressure at the water
temperature (Psat) and the vapor pressure in the overlaying air
(Pair) are computed using Antoines equation.
The temperature of the water greatly aects the growth rate
of algae. Hence, an energy balance for water is derived to track
the water temperature in each compartment
n
i=1
Le, i
where v is the water velocity, and i [1, n]. The pond feed and
harvest are taken into account by adding Ffeed and subtracting
Fharvest from eq 1 for the rst and nth CSTRs, respectively. The
dilution rate (D) dictates the rate of algal biomass harvest
Fharvest = DLW Hi
LWEv, i
LWc p
(3)
d
(HiTw, i) = c p(Fin, iTw, i 1 Fout, iTw, i) + LWEi
dt
(7)
3328
DOI: 10.1021/acs.iecr.5b03209
Ind. Eng. Chem. Res. 2016, 55, 33273337
Article
(8)
(9)
(10)
Es = 3Ia
(11)
description
value
cp
1
2
3
5
2526
1.0
1.0
11.7 108
0.9725
0.625
0.47
0.927
0.0325
fT, i
fI, i average =
d
(HiCalgae, i) = Fin, iCalgae, i 1 Fout, iCalgae, i
dt
+ (i Bi )LWHiCalgae, i
feed
= (1 Heff )Calgae, n
Fharvest
Ffeed
(12)
(13)
i=1
Hi
tp
t p
fI, i dt dz
(18)
(19)
(14)
1
Hi
HiCalgae,i
(17)
k T,1(Tw,i Topt)2
, Tw, i > Topt
e
=
ek T,2(Topt Tw,i)2 , T T
w, i
opt
where Topt is the optimal temperature for algae growth and kT,1
and kT,2 are tting constants.23 The optimal temperature for
Spirulina is within the range of 2442 C depending on the
particular strain of Spirulina, and for N. Salina it is around
28 C.28,31 Autotrophic microalgae drive photosynthesis by the
energy from light photons, and due to the shading eect, the
microalgae closer to the pond surface are exposed to higher
light intensities; hence, the mean value of f I is considered
(16)
(15)
fI, i =
Ii
Imax
e1 Ii / Imax
(20)
DOI: 10.1021/acs.iecr.5b03209
Ind. Eng. Chem. Res. 2016, 55, 33273337
Article
derived in ref 24
yout, i =
I
Ii = 34 a eke,iz
(21)
(22)
ai = db2
N0, i =
(23)
g, i =
PC =
R gTg
(yin yout, i )
6 (1 g, i)
(28)
6Q g, i
6Q g, i + vWW
s
(29)
t=1
(24)
(30)
Q g, i
db3
Tc
(i Bi )HiR CO2Calgae, i
PgQ g, i
(27)
WW
s (v b v )
2
CCO
2, i
Ks
No, i
CCO2, i
K C + CCO2, i +
(26)
y Pg
HeCCO2, i + in
CCO2, i
Pg
HeR gTg
R gTg
(31)
Q g,i ,t
(32)
i=1
(33)
costenergy, t = xeEmixing, t
(34)
where xCO2, xw, and xe are the prices of CO2, freshwater, and
electricity, respectively. The CO2 gas density (CO2) is
calculated using the ideal gas law. The daily energy requirement
for mixing the open pond water using a paddle wheel (Emixing) is
given by
(25)
Emixing =
3330
89(hfriction + hbend)vW i = 1 Hi
nMeff
(35)
DOI: 10.1021/acs.iecr.5b03209
Ind. Eng. Chem. Res. 2016, 55, 33273337
Article
hbend
nL
= v 2n02 4/3
r
37v 2
=
g
description
value
(36)
(37)
where n0 is a roughness factor also known as the GaucklerManning coecient, r is the channel hydraulic radius, 7 is the
kinetic loss coecient, and g is the acceleration of gravity.41
The mathematical model described above was coded in
gPROMS ModelBuilder v4.0 installed in a 64-bit Windows 7
CPU equipped with an i7 processor at 3.4 GHz and 16 GB of
RAM.42
3. MODEL VALIDATION
3.1. Experimental Setup and Model Assumptions.
Spirulina was cultivated in a 450 m2 outdoor open pond for 10
months in Malaga, Spain.43,44 The experiment resembles a
scenario for commercial scale algal biomass production,
especially given the prolonged cultivation period and varying
environmental conditions. Therefore, the water temperature,
biomass areal concentration, and productivity proles reported
in refs 43 and 44 were used to validate the ones predicted by
the developed model.
In the experiment, the depth of water was maintained at
30 cm, and the growth culture was prepared with modied
Zarrouks Medium providing adequate nutrient levels.43,45
Moreover, the pond was inoculated with an areal concentration
of 15 g DW m2 and harvest was started after 13 days and only
interrupted during February due to heavy rains.44 The
corresponding model assumptions and parameters are (1) the
open pond was discretized into n = 18 compartments each
having a length and width of L = W = 5 m; (2) the makeup
water ow rate was set to match the evaporation losses to x
the depth at H = 0.3 m; (3) the nutrient concentrations were
at their optimal values; therefore, eq 23 was set to f N,i = 1 and
eqs 2429 were excluded; (4) after the inoculation period
(13 days), the dilution rate was set to an estimate D = 0.10 d1
based on ref 46 except for 17 days during February where there
was no harvest.
The initial conditions and values assigned to the kinetic
parameters for Spirulina growth are shown in Table 2. Figure 3
shows the solar radiation and air temperature data for the
period of the experiment at Malaga, which were obtained from
the European Database of daylight and Solar Radiation47 and
the Tutiempo Network,48 respectively. These data, including
data for humidity, wind speed, and photoperiod, were used in
the simulation of the model consisting of eqs 118 and 2023.
The time integration of the ordinary dierential equations was
performed using the DASOLV solver which is based on variable
time step Backward Dierentiation Formulas (BDFs).49
This solver embeds the MA48 subsolver which uses a direct
LU-factorization algorithm to solve a set of linear algebraic
equations. To increase numerical stability the default value of
the PivotStabilityFactor was changed to 0.9, which is one of
the setting parameters of the MA48 subsolver.
0.0423
2023
0.06923
0.5
24
0.325
2.0550
Spirulina
1.451
27.523
N. Salina
1.334
2752
0.00523
0.0152
0.00423
0.0352
20053
5834
6.643
1754
Spirulina
0.3
N. Salina
0.3
22.1a
12.4a
15
15
0.02a
DOI: 10.1021/acs.iecr.5b03209
Ind. Eng. Chem. Res. 2016, 55, 33273337
Article
description
value
KC
Ks
CCO2,atm
9 104 24
18024
0.02
RCO2
vb
Tg
Pg
yin
Rg
He
Ws
db
6
Katm
KL
0.04256
30 assumed
31626
1.226
1
8.2 105
0.8317
0.326
257,58
0.9640
2.45
9.5959
description
value
xw
xe
xCO2
0.01660
0.0461
4062
Meff
n0
r
7
40% assumed
2.08 107 26
0.2926
226
4. DYNAMIC OPTIMIZATION
4.1. Problem Denition. Imperial County in California is
one of the suitable places for cultivating N. Salina in USA owing
to the warmer weather and availability of resources including
land, water and CO2.26 This site was selected for the
optimization case study and the daily weather conditions in a
typical meteorological year for this site were obtained from the
National Solar Radiation Data Base.55 The 4 ha open pond
proposed in ref 26 for the production of algal biofuels at
commercial scale was adopted herein. Therefore, the pond was
discretized into 44 compartments with L = W = 30 m and the
CO2 gas was set to be introduced at the bottom of the 12th and
34th compartments. The parameter values selected for the
growth kinetics of N. Salina are shown in Table 2. The values
assigned to the parameters of modeling the CO2 transfer and
process economics are shown in Tables 3 and 4, respectively.
For a production cycle of 1 y, the program, comprising
eqs 119 and 2137, contains 1043 variables and 1876
parameters.
4.2. Base Case. The dilution rate (D) is one of the main
operating parameters aecting the productivity of an algal
system. A low D creates a dense culture where light availability
becomes limited due to the shading eect. This reduces the
growth rate of microalgae and consequently leads to a lower
productivity. On the other hand, a high D would reduce the
3332
DOI: 10.1021/acs.iecr.5b03209
Ind. Eng. Chem. Res. 2016, 55, 33273337
Article
that the cost of the CO2 gas constitutes 59% of the improved
base case PC; hence, Qg needs to be optimized as well. Also,
although the cost of makeup water is insignicant as Chart 1
shows, optimizing Fmakeup would aect the CO2 absorption.
4.3. Optimization Problem. The optimization problem
considered determines the proles of the dilution rate, CO2 gas
ow rate, and makeup water ow rate that
minimize PC eq 30
D , Q g , Fmakeup
Figure 7. Eect of the attenuation factors on the growth rate during the search for the optimal dilution rate and biomass algal concentration in the
pond using the heuristic approach.
3333
DOI: 10.1021/acs.iecr.5b03209
Ind. Eng. Chem. Res. 2016, 55, 33273337
Article
Figure 9. Comparing results from the improved base case and the optimization study: (a) algal biomass areal concentration; (b) dilution rate;
(c) CO2 gas ow rate; (d) makeup water ow rate.
Figure 10. Eect of attenuation factors on the growth rate for the optimal case scenario.
Q g 0,
Fmakeup 0,
0 D 0.5,
0.2 H 0.4
assuming that the CO2 gas and makeup water availability are
unlimited. The bounds on the dilution rate and depth were set
based on typical operation in commercial algae facilities.4 The
improved base case values for the decision variables were used
as an initial guess and the dynamic optimization was performed
using the CVP_SS solver which implements a control vector
parametrization algorithm based on the single-shooting
method.65 It took the optimizer 1443 s to nd the optimal
operating proles shown in Figures 9bd corresponding to a
PC of $167 tonne1 which is 15% less than in the improved
DOI: 10.1021/acs.iecr.5b03209
Ind. Eng. Chem. Res. 2016, 55, 33273337
Article
5. CONCLUSIONS
A mathematical model for estimating the growth of microalgae
in an outdoor open pond based on local climatic conditions was
developed in this work. The model was validated against
literature data for the production of Spirulina in an outdoor
open pond in Malaga, Spain. The simulated algal biomass
productivity agreed with experimental data, with a mean
percent error of approximately 16%. A dynamic optimization
problem was formulated for determining the location-specic
optimal monthly operating proles for the dilution rate, CO2
gas ow rate, and makeup water ow rate. A case study was
conducted for the cultivation of N. Salina in California, USA.
The operating proles generated by the optimization lowered
the cultivation cost by at least 15% when compared with other
case scenarios where best practice operation and sensitivity
analysis were employed.
Based on the analysis presented, algae facilities adopting
outdoor open ponds and having access to local meteorological
data can use the formulation presented in this work to lower
the cultivation cost by identifying optimal operating conditions.
Furthermore, the proposed formulation can be useful for
determining facility locations for algal biomass production by
comparing candidate sites based on optimized operations.
AUTHOR INFORMATION
Corresponding Author
*E-mail: daout001@umn.edu.
Notes
ACKNOWLEDGMENTS
The authors acknowledge the nancial support from Abu Dhabi
National Oil Company (ADNOC).
NOMENCLATURE
a = total interfacial area of CO2 gas bubbles (m1)
1 = atmospheric attenuation coecient
2 = Bowens coecient (mmHg C1)
3 = radiation absorption factor
4 = fraction accounting for the visible portion of solar
irradiance
5 = reection coecient
6 = correction factor for the compression of gas under water
7 = kinetic loss coecient
8 = conversion factor (9.8 W s kg1 m1)
3335
DOI: 10.1021/acs.iecr.5b03209
Ind. Eng. Chem. Res. 2016, 55, 33273337
Article
REFERENCES
DOI: 10.1021/acs.iecr.5b03209
Ind. Eng. Chem. Res. 2016, 55, 33273337
Article
DOI: 10.1021/acs.iecr.5b03209
Ind. Eng. Chem. Res. 2016, 55, 33273337