Professional Documents
Culture Documents
ScienceDirect
highlights
An integrated system consists of geothermal and solar energy, an electrolyzer, and a fuel cell unit.
Thermoeconomic analysis and the ANN-based GA optimization are evaluated.
The overall system energy and exergy efficiencies are calculated at 4.97% and 16.0%.
The optimum cost of produced hydrogen and electricity are calculated as 1.576 $/kg H2 and 0.091 $/kWh, respectively.
Article history: The study aims to optimize the geothermal and solar-assisted sustainable energy and
Received 13 September 2021 hydrogen production system by considering the genetic algorithm. The study will be useful
Received in revised form by integrating hydrogen as an energy storage unit to bring sustainability to smart grid
4 January 2022 systems. Using the Artificial Neural Network (ANN) based Genetic Algorithm (GA) optimi-
Accepted 14 March 2022 zation technique in the study will ensure that the system is constantly studied in the most
Available online 6 April 2022 suitable under different climatic and operating conditions, including unit product cost and
the plant's power output. The water temperature of the Afyon Geothermal Power Plant
Keywords: varies between 70 and 130 C, and its mass flow rate varies between 70 and 150 kg/s. In
Geothermal energy addition, the solar radiation varies between 300 and 1000 W/m2 for different periods. The
Solar energy net power generated from the region's geothermal and solar energy-supported system is
Hydrogen production calculated as 2900 kW. If all of this produced power is used for hydrogen production in the
Genetic optimization electrolysis unit, 0.0185 kg/s hydrogen can be produced. The results indicated that the
overall energy and exergy efficiencies of the integrated system are 4.97% and 16.0%,
respectively. The cost of electricity generated in the combined geothermal and solar power
plant is 0.027 $/kWh if the electricity is directly supplied to the grid and used. The opti-
mized cost of hydrogen produced using the electricity produced in geothermal and solar
power plants in the electrolysis unit is calculated as 1.576 $/kg H2. The optimized unit cost
of electricity produced due to hydrogen in the fuel cell is calculated as 0.091 $/kWh.
© 2022 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.
* Corresponding author.
E-mail address: ceyhunyilmaz@aku.edu.tr (C. Yilmaz).
https://doi.org/10.1016/j.ijhydene.2022.03.140
0360-3199/© 2022 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 1 6 4 2 4 e1 6 4 3 9 16425
Fig. 1 e Operating diagram of combined power and hydrogen production plant [8].
Table 1 e Operating parameters of AFJES [15]. Table 2 e Solar radiation values of Afyonkarahisar [16].
rate, h is specific enthalpy, S_gen is the rate of entropy genera- kJ/elec V ¼ 96,485 kJ/kmol V). Electrical potential of the fuel
tion, and Ex_ dest is the rate of exergy destruction. cell:
Hydrogen production and fuel cell unit are preferred for DG
energy storage and conversion. Fuel cells are thermoelectric V0 ¼ (11)
96485ne
devices that convert the chemical energy of fuels to electricity
Generally, when fuel cells are examined, the fuel cell
without any combustion process. One of the essential ad-
changes efficiency depending on the temperature. Also, there
vantages of fuel cells is that they are environmentally friendly,
are losses in this theoretically obtained potential value. When
contributing to very low to zero emissions. Environmental
the types of fuel cells are examined, polymer electrode fuel
concerns have focused on developing and utilizing clean and
cells and phosphoric acid fuel cells stand out at low temper-
sustainable energy resources. Therefore, it is essential to
atures. PEM fuel cell operates at temperatures close to 80 C,
develop environmentally benign energy utilization and con-
and phosphoric acid fuel cell operates around 200 C. Melted
version processes to replace conventional fossil fuel-based
carbonate and solid oxide fuel cells can reach temperatures up
power generation systems. Fuel cells are considered one of
to 600e1000 C. In low-temperature Polymer Electrode and
the most promising clean and efficient energy conversion
phosphoric acid fuel cells, sensitivity to gases such as carbon
technologies.
monoxide increases, and losses may occur. According to the
Along with their pure nature, fuel cells hold many other
operating temperature of the systems, it is seen in this study
advantages such as efficient, immobile, quiet, and reliable
that the PEM fuel cell is the most efficient system [19].
operation. PEM fuel cell has been adapted for this study due to
Although the PEM fuel cell is an optimum system, the
suitable temperature ranges and operating conditions. The
system's theoretical potential can never be achieved. There
energy released from the system is the difference between the
are various irreversibilities in the system. These can be
enthalpy of the products reacting and the enthalpy of the
grouped under three main headings: activation losses, ohmic
products released.
losses, and concentration losses [20].
X X o X o
DH ¼ Hr Hp ¼ Nr hf þ h ho Np hf þ h ho
r p Vloss ¼ Vact:loss þ Vohm:loss þ Vcon:loss (12)
(5)
Activation losses are seen on both the anode and the
o
hf is the standard enthalpy of formation, and the expres- cathode electrode. This loss is seen as both voltage and cur-
sion h ho is the difference between incoming and outgoing rent loss. Ohmic losses result from the membrane's electrical
products (25 C 100 kPa). Since the products entering and resistance and other conduction materials. On the other hand,
concentration losses can reach high levels with an excessive
leaving an ideal fuel cell are in standard conditions, h ho
increase in the concentration of chemical products at high
expression is zero.
current values. The actual voltage value can be found by
X o X o
DH ¼ Nr hf Np hf (6) subtracting all these losses from the theoretical voltage value
r p
obtained. V voltage is the net output voltage of the fuel cell.
If it is written instead of inputs and outputs in PEM Fuel
V ¼ V0 Vloss (13)
cell:
X o X o X o
DH ¼ NH2 hf NO2 hf NH2 O hf (7) i þ ileak iL
V ¼ V0 b ln i:Rohm c ln (14)
H2 O2 H2 O i0 iL ði þ ileak Þ
The maximum work output that can be obtained here: where b and c cell constants, i0 reference current value and iL is
X o
X o limit current value. The basic assumptions of the PEM model
Wmax ¼ Nr hf þ h ho T0 s Np hf þ h ho T0 s
r p considered are given in Table 3.
(8) The thermal efficiency of the overall system is defined as
the ratio of the total electrical power output and the total heat
The maximum potential of the electrochemical reaction is
input of geothermal and solar power plants [8].
1.229 V at standard conditions (25 C, 100 kPa), and it corre-
sponds to the Gibbs free energy. Consider a reversible reaction
occurring at a constant temperature equal to that of its envi-
ronment. The work output of the fuel cell is [18]:
Table 3 e PEM fuel cell model assumptions [20].
X X
welec ¼ ne ge ni gi ¼ DG (9) Fuel Cell Value Unit
Operating temperature 350 K
where DG is the change in Gibbs function for the overall
i0 0.013 A/cm2
chemical reaction. Besides, the term work can also be 0.01 A/cm2
Ileak
expressed in terms of electrical potential. iL 2 A/cm2
b 0,08 V
w ¼ Vne N0e (10) c 0.1 V
R 0,01 U
V is an indicator of electrical potential. ne is the number of
V0 1.22 V
electrons in kmol. (N0e ¼ 6.022136 1026 elec/kmol x 1.602177
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 1 6 4 2 4 e1 6 4 3 9 16429
_ elec _ elec during the production of the product. These costs consist of
W W
h¼ ¼ (15)
fuel cost C_ F , initial investment cost Z_ and operation and
CI
_ _
Q sol þ Q geo Colr Cols Aap G þ m_ geo hgeo h0
maintenance Z_ . The expression of the cost ratio related to
OM
are calculated using the Aspen Plus program. In this study, for
the thermodynamic and thermoeconomic analysis of the
system, calculations are performed by coding the required and evaluated, and the actual analysis and prediction results
data into EES and Aspen Plus software programs. To be used in are shown in the figure. In the design of thermal systems,
economic analysis, the annual interest rate (i) we use within parameters such as which process or equipment will be
the program's scope is assumed as 10%. The system's ther- selected, how they will be placed, the system's dimensions,
moeconomic analysis is calculated in the EES program by the optimum temperature, pressure, flow rate, and chemical
writing the necessary thermoeconomic connections, and composition of the system's fluids are essential. The answer
thermoeconomic relationship tables are created [25,26]. these questions, it is necessary to choose the most suitable
The thermoeconomic analysis of the system is performed optimization method for your system at the system design
under three main headings. The thermoeconomic analyses of stage. The first step in defining an optimization problem is
solar and geothermal power plants are evaluated separately, defining the system's boundaries to be optimized. All sub-
and the thermoeconomic analyses of the electrolysis unit and systems that will affect the system's operating performance
fuel cells are evaluated separately. The purpose of performing should be expressed in the optimization problem. The criteria
this is that the two separate power plants' annual working by which the system will be fundamentally examined and
hours and economic lives differ. In the thermoeconomic optimized are the critical elements of the optimization prob-
analysis of the electrolysis unit and fuel cell, the annual lem [28].
operating hours and economic life of equipment are consid- Initially, individuals are usually randomly generated in the
ered the same as the geothermal power plant. Because the genetic algorithm, but this is not a requirement. Especially in
geothermal power plant provides most of the electrical energy very constrained optimization problems, better candidates
obtained in this model and the geothermal power plant's can be created by paying attention to some of the defined
annual working hours are longer. constraints to create the starting individuals. As a result of
subjecting the individuals to the fitness function process, the
fitness value is determined, evaluating how close the solution
System application of ANN based GA is to the optimal solution. Genetic algorithm with initial
optimization population generated works with three evolution operators.
These are selection, crossover, and mutation operators.
Artificial Neural Networks (ANN) can be defined as the Generally, each of these operators is applied to every popu-
mathematical modeling of the human brain's learning and lation formed in the next generation. The selection process
decision-making skills. ANN is a technique inspired by the involves selecting individuals based on their fitness values to
way the human brain processes information. The way the create new individuals. The crossover operator is applied after
simple biological nervous system works with ANN is modeled the selection process and refers to the reciprocal displace-
mathematically. ANN can artificially solve problems requiring ment of certain parts of the chromosomes belonging to the
people's natural ability to think and observe in a mathematical parent individuals, thus creating new characteristics. The
environment. Learning is carried out by training the created mutation process is changing a gene in any of the newly
mathematical artificial neural network using examples. An formed individual chromosomes, depending on the possibility
ANN, whose training process has been completed, can of mutation. There are various methods to terminate the ge-
mathematically perform operations that are difficult to do netic algorithm process. These methods; when the desired
with pure arithmetic methods such as data classification, solution is found during the operation of the algorithm when
recognition, optimization, data association, and predictions the total number of iterations defined at the beginning of the
for the advanced operating processes system [27]. genetic algorithm is reached, or the fitness value remains
In the equations in Table 4, the data set's size is the output constant, the solution represented by the best individual
values of the EES program and the output values of the ANN- found is presented as the most suitable solution found for the
based Matlab program. Power and cost estimates are per- problem [29].
formed due to the network trained on the Matlab program In this study, net power generation, electricity generation,
using the ANN method. To measure the prediction success of cost of electricity, and hydrogen are considered the objective
the network trained in ANN, mean absolute error, mean functions. The environmental temperature, geothermal
square error, root mean square error and mean absolute source temperature, the mass flow of the geothermal fluid,
percentage error values are calculated. The actual analysis and solar radiation are selected as decision parameters to
results and the estimation results of the ANN are compared maximize the system's net power and electricity generation
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 1 6 4 2 4 e1 6 4 3 9 16431
According to the thermodynamic analysis, 25,542 kW of en- Table 8 e Cost equations of equipments of the
geothermal power plant depending on exergy.
ergy and 6800 kW of exergy are provided to the system from
the geothermal plant with a temperature of 130 C and a mass Equipments Cost balance Auxiliary
flow rate of 85 kg/s The net power obtained from the equations equations
geothermal power plant is calculated as 2693 kW. The energy Pump C_ 1 þ Z_ P þ C_ WP ¼ C_ 2 c1 (is known)
efficiency of the geothermal power plant was 10.5%. The exit c2 (variable)
In this section, the system presented in the study is modeled Newton's algorithm and the stability of the steepest descent
using ANN. The system has four inputs and four outputs. In- method, is used effectively in network training today [32].
puts are environment temperature, geothermal fluid tem- In Artificial Neural Networks, information and algorithms
perature, geothermal fluid flow, and solar radiation. Outputs are not precise; experiences are used to achieve the desired
are power generation of the system, the fuel cell's power success. Various learning algorithms are used to provide the
generation, the cost of the electricity, and hydrogen produced. necessary experience. In this study, Artificial Neural Network
In the modeling progress, 100 8 data sets from the system is trained with the Levenberg-Marquardt method, aiming to
are divided into 80 8 training data and 20 8 test data. After predict system outputs. The Levenberg-Marquardt training
this process, the feed-forward multi-layer ANN structure is algorithm is used during the training phase. Performance re-
created on Matlab with four inputs and four outputs. The feed- sults for the ANN-based system are presented in Fig. 3. As a
forward multi-layer ANN block diagram of the 4-8-4 structure result of 2000 epochs in the training phase, the mean square
is conducted and given in Fig. 2 for the system. error value of 7.20 10E-1 MSE is reached.
There are eight neurons in one hidden layer in the ANN After the training phase, the test process is carried out with
structure, and the Tangent Sigmoid transfer function is used the 20 8 dataset. MAE, MSE, RMSE, and MAPE values are
as a transfer function in each neuron. The neurons in the calculated and given in Table 17 due to the performance
output layer have a Purelin transfer function. The parameters evaluation performed by using the test data set containing the
related to the ANN-based system are given in Table 16. EES program's accurate analysis results and the ANN results.
In Artificial Neural Networks, a back propagation algorithm The error analysis calculated for the power generation of
is preferred to model the input-output matching with the the system, the fuel cell's power generation, the generated
appropriate data set. This algorithm, which requires first- electricity, and hydrogen costs is given in Table 15. As un-
order derivative information, has disadvantages such as low derstood from the error analysis results, ANN's estimation
training efficiency and poor convergence speed. Therefore, results are close to the actual results. These results show that
this algorithm is not suitable for practical applications. The the level of confidence in the study is relatively high. The
Levenberg-Marquardt (LM) algorithm significantly increases comparison of actual power generation and ANN-based esti-
the learning speed for operations that require quadratic de- mation results is shown in Fig. 4. The accuracy level of the
rivatives. The LM algorithm, which combines the speed of ANN-based estimation result for power generation has been
16434 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 1 6 4 2 4 e1 6 4 3 9
measured as 99.9%. ANN-based estimation results almost of electricity cost estimates ANN-based is 97.9%. This result
corresponded to actual values. This result shows that the ANN shows that the accuracy of ANN estimates for electricity cost
estimates the power generation with high sensitivity. is quite realistic. Actual system ANN integration has a good
The accuracy level of the ANN estimation result for the fuel potential for renewable system cases.
cell power generation is 99.9%. This means that the reliability In this system, energy storage material is via hydrogen
of the estimation results is very high. So that can be used for production. So the unit cost of hydrogen production from
the smart grid's actual power system energy transformation. renewable surplus electricity is the most critical parameter.
Fig. 5 shows the actual power generation of the fuel cell and The comparison of the actual hydrogen cost with ANN-based
the ANN estimation data. estimation results is shown in Fig. 7. The accuracy level of
Fuel cell electricity cost is another important parameter of hydrogen cost estimates ANN-based is measured as 98.6%.
the system. Fig. 6 shows the actual electricity cost and ANN- This result means that the reliability of ANN estimation re-
based electricity cost estimation results. The accuracy level sults is relatively high. The comparison of the actual hydrogen
Fig. 4 e Variation of actual and ANN-based estimated power generation with test data set.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 1 6 4 2 4 e1 6 4 3 9 16435
Fig. 5 e Variation of actual and ANN-based estimated power generation of fuel cell with test data set.
Fig. 6 e Variation of actual and ANN-based estimated electricity cost with test data set.
Fig. 7 e Variation of actual and ANN-based estimated hydrogen cost with test data set.
16436 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 1 6 4 2 4 e1 6 4 3 9
Fig. 12 e Exergy and exergy based cost flow diagram of combined power and hydrogen system.
Fig. 9 shows the best solution of the system to net power differences between the streams exchanging heat usually re-
output and hydrogen cost. As a result of the multi-objective duces the specific irreversibility of the process. As a result of
optimization, the best power output and the best hydrogen these optimization improvements, the net electrical work
cost are 4092 kW and 1.343 $/kg, respectively. produced from the fuel cell was increased from 1360 kW to
The best solution according to the electricity generated in 1706 kW, an increase of 25%. In addition, the unit cost of the
the fuel cell and the electricity cost is shown in Fig. 10. The electricity ultimately produced from the system was reduced
best electricity generation and the electricity cost are 1842 kW from 0.152 $/kWh to 0.091 $/kWh, thus reducing the cost by
and 0.043 $/kWh, respectively. approximately 40%. Finally, the unit cost of hydrogen pro-
Fig. 11 shows the system's best solution to the electricity duced to store excess electricity was decreased from 2.332
generated in the fuel cell and the hydrogen cost. As a result of $/kg to 1.576 $/kg, thus reducing its cost by about 32%. More
the multi-objective optimization, the best electricity genera- details of the exergy and exergy cost analysis optimum values
tion and the best hydrogen cost are 1842 kW and 1.343 $/kg, of the combined hydrogen and power generation plant can be
respectively. shown in Fig. 12.
The design and optimum values of the decision parameters
of the system are given in Table 18. The net power output, the
electricity generation, the electricity cost and hydrogen cost of Conclusion
the design stage, the system's optimum solution, and the de-
viation between the developed and optimum solutions are The renewable energy-driven hydrogen and electric power
given in Table 19. generation system is considered, and thermoeconomic and
Listing component exergy loss may help in assessing the optimization are performed. The system considered is a novel
performance of the hydrogen and power production plant. configuration consisting of a binary geothermal power plant
The proper representation of exergy and exergy cost flows and and a parabolic trough concentrating solar plant for electricity
destructions is the flow diagram that may adapt the plant's production and water electrolysis and fuel cell unit for
energy and economic analyses used for exergy and exergy hydrogen storage and utilization. The optimization studies
cost transfer. The flow diagram in the combined power sys- are performed to investigate the effects of geothermal source
tem can be particularly valuable since it shows exergy de- temperature and solar radiation. As a result of the optimiza-
structions, the splitting of exergy cost streams, and tion using the genetic algorithm method, a significant
recirculation of exergy. It also shows graphically how a part of improvement has been observed in the system's thermody-
the original exergy input is dissipated in the successive stages namic performance. Some conclusions that can be drawn
of energy transformation. Fig. 2 can be presented either in a from this study can be summarized as follows:
dimensional form of exergy and exergy cost flows of the
combined hydrogen and power production plant. As will be The system's net power output has been increased from
seen from Fig. 12, the exergy efficiency of the plant takes into 2900 kW to 3858 kW, and the electricity generation from
account three types of losses: The intrinsic irreversibility of 1361 kW to 1706 kW.
the geothermal and solar energies injection and reinjection As a result of the optimization using the genetic algorithm
process. It causes irreversibility due to heat transfer from method, it has also improved the system's economic per-
finite temperature differences. Irreversibility is due to the formance. The cost of electricity generated and hydrogen
dissipation of the exergy of the products of the overall plant. generated have been optimized 40.13% and 32.41%. The
The success of the optimization in improving the efficiency cost of electricity generated in the combined geothermal
and product costs is due to the strong interaction between the and solar power plant is 0.027 $/kWh if the electricity is
plant components. A reduction in the mean temperature directly supplied to the grid and used.
16438 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 1 6 4 2 4 e1 6 4 3 9
[16] Hocaog lu FO. Stochastic approach for daily solar radiation [25] Yilmaz C, Kanoglu M, Abusoglu A. Thermoeconomic cost
modeling. Sol Energy 2011;85(2):278e87. evaluation of hydrogen production driven by binary
[17] Cengel YA, Boles MA, Kanoglu M. Thermodynamics: an geothermal power plant. Geothermics 2015;57:18e25.
engineering approach. 9th ed. New York, A.B.D.: McGraw- [26] Balta MT, Dincer I, Hepbasli A. Exergoeconomic analysis of a
Hill; 2019. hybrid copperechlorine cycle driven by geothermal energy
[18] Arsalis A. Modeling and simulation of a 100 kWe HT-PEMFC for hydrogen production. Int J Hydrogen Energy
subsystem integrated with an absorption chiller subsystem. 2011;36(17):11300e8.
Int J Hydrogen Energy 2012;37(18):13484e90. [27] Beasley D, Bull DR, Martin RR. An overview of genetic
[19] Yilmaz C, Kanoglu M. Thermodynamic evaluation of algorithms: Part 1, fundamentals. Univ Comput
geothermal energy powered hydrogen production by PEM 1993;15(2):56e69.
water electrolysis. Energy 2014;69:592e602. [28] Starke AR, Cardemil JM, Escobar R, Colle S. Multi-objective
€
[20] Ozgür T, Yakaryılmaz AC. A review: exergy analysis of PEM optimization of hybrid CSPþ PV system using genetic
and PEM fuel cell based CHP systems. Int J Hydrogen Energy algorithm. Energy 2018;147:490e503.
2018;43(38):17993e8000. [29] Ehyaei MA, Ahmadi A, Rosen MA, Davarpanah A.
[21] Dhillon BS. Life cycle costing for engineers. CRC Press; 2009. Thermodynamic optimization of a geothermal power plant
[22] Bejan A, Tsatsaronis G, Moran M. Thermal design with a genetic algorithm in two stages. Processes
and optimization. New York, A.B.D.: John Wiley 2020;8(10):1277.
and Sons; 1996. [30] F-Chart Software. EES, engineering equation solver. In: F-
[23] Frangopoulos CA, Nakos LG. Development of a model for chart software. Inter-net Website; 2021. www.fchart.com/
thermoeconomic design and operation optimization of a ees/ees.shtml.
PEM fuel cell system. Energy 2006;31(10e11):1501e19. [31] Aspen PlusV8.4. Engineering Economic Analysis Library; 2015.
[24] Yilmaz C. Thermoeconomic modeling and optimization of a [32] Wilamowski BM, Yu H. Improved computation for
hydrogen production system using geothermal energy. levenbergemarquardt training. IEEE Trans Neural Network
Geothermics 2017;65:32e43. 2010;21(6):930e7.