IOP Conference Series: Materials Science and Engineering

PAPER • OPEN ACCESS

Genetic Algorithm PID Control of Solid Oxide Fuel Cell

To cite this article: Jipeng Gu et al 2020 IOP Conf. Ser.: Mater. Sci. Eng. 768 042005

View the article online for updates and enhancements.

This content was downloaded from IP address 212.80.223.34 on 31/03/2020 at 18:13

SAMSE 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 768 (2020) 042005 doi:10.1088/1757-899X/768/4/042005

Genetic Algorithm PID Control of Solid Oxide Fuel Cell

Jipeng Gu1, Qianru Wang1, Wenting Lu1,*

1
School of Electrical Engineering, Northwest Minzu University, Lanzhou, China

*Corresponding author e-mail: [email protected]

Abstract. In this paper, a genetic algorithm PID controller is designed to control the
output voltage of SOFC single battery. The genetic algorithm is adopted and designed
to optimize the parameters of standard PID controller K P , K I and K D , so that the
difficulties in parameters setting of PID are solved. The simulation results show that
the designed genetic algorithm PID controller for the voltage of SOFC single battery
can achieve certain and required accuracy as well as fast response speed.

1. Introduction
Fuel cell is a new generation of power generation technology after hydropower generation, thermal
generation and atomic generation [1]. Because its power generation process is not restricted by the
Carnot cycle and no other mechanical transmission components, it has the advantages of quiet,
pollution-free, high energy efficiency [2].
In 2014, So-Ryeok Oh and Jing Sun analyzed the static and dynamic characteristics of the 5-
kilowatt class solid oxide fuel cell and gas turbine combined system under thermal constraints, and
proposed a model prediction controller based on state constraints and input constraints. The results
showed that the design of the controller was reasonable and the load could be transferred quickly
under the allowable range of electric reactor temperature [3]. When the load changes, the temperature
and fuel efficiency of solid oxide fuel cell will change accordingly, leading to the fission of internal
materials of battery and affecting its life. Therefore, R.Horalek and J.Hlava proposed a corresponding
model predictive control scheme for this problem in 2015. This scheme not only predicted and
controlled the battery temperature, fuel efficiency and oxygen efficiency, but also successfully limited
the thermal change rate of the system and effectively prevented the battery damage [4]. In 2016,
Tooran Emami and Alex Tsai applied robust PID controller to control the solid oxide fuel cell system
with a power of 300KW. The performance of the system output under three different stability
constraints was tested to prove the rationality of the controller design, and finally the effective
operating range of solid oxide fuel cell parameters was given [5].
In China, Wang Lijin proposed a control strategy based on the distributed lumped parameters
model of one-dimensional internal reforming solid oxide fuel cell in 2008, which mainly applied three
independent PID controllers to control the temperature, power and fuel efficiency of the battery in
parallel. The results showed that the established control system was correct and feasible, and could
maintain the constant output of the system in the case of load variation [6]. In 2014, Liu Xin
established the mathematical model of solid oxide fuel cell including two auxiliary components of
internal reforming device and combustion chamber, and controlled the composition, temperature,
pressure, concentration and flow rate of fuel gas by using the nonlinear model predictive control
strategy, making the whole power generation system in the expected normal operation state [7]. In
2016, Yang Ji established the mechanism model of the solid oxide fuel cell based on the dynamic
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
SAMSE 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 768 (2020) 042005 doi:10.1088/1757-899X/768/4/042005

change of reaction gas. Then the slide-mode interference observer based on the superhelix algorithm
was used to estimate and compensate the modeling errors and external uncertainties. Finally the slide-
mode controller was used to control the output of the battery. The results showed that the slide-mode
controller with slide-mode observer could effectively control the input of the reactive gas and ensure
the constant of the output voltage [8]. In 2018, Li Xiao obtained the battery mechanism model based
on the analysis of the power generation principle of solid oxide fuel cell, and then simulated the input
and output data in the mechanism model by using the autoregressive sliding average model to obtain
the battery identification model. Then the author designed a GPC algorithm based on the identification.
The results showed that the output performance index of the battery had a good tracking effect when
the load of the external circuit changed [9].
SOFC is a strongly nonlinear system with multiple inputs and outputs. In the design of the
controller, there are certain requirements on the real-time, adaptive and robust of the whole system, so
there are relatively few researches on SOFC control [10-13]. Since the research on control of SOFC is
still in the initial stage, it has important theoretical value and practical significance for the design of its
controller.

2. The principle of power generation for SOFC single battery

The power generation principle of SOFC single battery is that oxygen (O 2 ) in the air channel
penetrates through the porous electrode to the cathode side, and the reduction reaction occurs to
produce oxygen ions (O2-). Oxygen ions reach the anode side through the electrolyte due to diffusion
and oxidize with the hydrogen rich gas (H 2 ) in the fuel channel that enters the anode side through
osmosis to form water and release electrons. Electrons form a closed loop through the external load to
complete the external power supply behavior.
An adjustable parameter model of SOFC single battery is used in this paper. The model has a
certain flexibility due to it contains two adjustable parameters that can be changed moderately
according to the actual operating of the battery, so this paper continues to use this model for control
research. Fig.1 shows the adjustable parameter model of SOFC single battery by Simulink, q1 is the
hydrogen input molar flow, q3 is the water vapor input molar flow, U cell is the voltage of SOFC single
battery, P cell is the power of SOFC single battery .

Figure 1. Adjustable parameter model of SOFC single battery by Simulink.

3. Introduction to genetic algorithm

In 1975, professor J.Holland of the University of Michigan first proposed the concept of genetic
algorithm in his book Adaptation in Natural and Artificial Systems. Genetic algorithm is a heuristic
search algorithm which simulates the natural selection of Darwin's biological evolution theory and the
biological evolution process of genetic mechanism.
When genetic algorithm is applied to solve practical optimization problems, it is necessary to use
certain coding form to encode each solution in the problem solution space. Each solution after coding

2
SAMSE 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 768 (2020) 042005 doi:10.1088/1757-899X/768/4/042005

is called chromosome, and a certain number of chromosomes are randomly selected to form the initial
population. The fitness of each chromosome is calculated according to the pre-set evaluation function,
and the fitness of the better chromosome is larger. Then, according to the principle of "survival of the
fittest", the chromosomes with greater fitness are selected with a certain probability for replication,
crossover and mutation to produce a new generation of chromosomes and form a new population.
Through the multiple evolution of limited number times, the offspring population will be more
adapted to the environment, and the optimal individual in the last population after decoding will be the
optimal solution or approximate optimal solution of the problem.
4. Design of genetic algorithm PID controller
British scholars Collender and Stevenson first proposed the concept of PID controller in 1936. PID
controller is a series correction device composed of proportional, integral and differential units, which
is mainly used to control industrial systems with basic linear and dynamic characteristics that do not
change with time to achieve the desired operating state. Its working principle is that the error signal of
the system forms the control signal after the action of certain control law, and then the control signal is
used to control the output performance index of the controlled object. The time domain form of the
PID controller can be expressed by the following formula.
u(t) = KP ×e(t) + KI ×∫e(t)dt + KD ×(de(t)/dt) (1)
Where, e(t) is the error signal; u(t) is the control signal; K P , K I (K P /T I ), K D (K P T D ) are the weighted
values of the error signal and its integral and differential.

Genetic algorithm

KP KI KD
Ud e(t) u(t) Ur
PID controller SOFC single battery

Figure 2. Structure diagram of genetic algorithm PID control of SOFC single battery.

Begin End

A certain number of feasible solutions are randomly Output the population satisfying
selected within the reasonable value range of KP, KI the problem and decode the
and KD, and coded in real numbers to complete the optimal individuals to get the
population initialization optimal parameters KP, KI and KD

According to the predetermined

The chromosomes are assigned probability, the chromosomes
to the parameters KP, KI and KD with large output voltage are
in turn, and the SOFC single selected for replication, Y
battery adjustable parameter crossover and mutation to form
model is run a new population

The actual output voltage is compared Meet

with the expected output voltage N requirements

Figure 3. Flow chart of genetic algorithm PID control of SOFC single battery.

3
SAMSE 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 768 (2020) 042005 doi:10.1088/1757-899X/768/4/042005

The value of parameters K P , K I and K D will directly affect the effect of PID controller, so it is very
important to study the optimization tuning method of PID controller parameters. The optimization
tuning method in the thermal system are mainly simple value method and expert tuning method, but
the above methods have some defects in controlling SOFC single battery. The simple value method is
sensitive to the initial value set by the system, which easily leads to the failure of optimization. Expert
tuning method need to organize a large amount of expert empirical knowledge to form a special
knowledge base, which consumes a lot of time and cannot guarantee the accuracy of empirical
knowledge. The genetic algorithm can realize the PID controller parameter adaptive tuning and has the
rapidity. Therefore, genetic algorithm is used to optimize the PID controller parameters within a
reasonable range, and the optimal parameters K P , K I and K D are used to control the output voltage of
SOFC single battery. Fig.2 shows the structure diagram of controlling the output voltage of SOFC
single battery by using the genetic algorithm PID controller, and Fig.3 shows the corresponding flow
chart.

5. Experimental simulation and analysis

In the adjustable parameter model of SOFC single battery, when the water vapor input molar flow
remains constant, there is an optimal hydrogen input molar flow to maximize the output voltage of
SOFC single battery [14]. Therefore, in this paper, the water vapor input molar flow is set as 0.2mol/s,
and then the hydrogen input molar flow is controlled by genetic algorithm PID controller, so that the
output voltage of SOFC single battery can reach the user’s expected value of 0.85v.
The genetic algorithm used in this paper is realized by the genetic algorithm of MATLAB and the
function of direct search toolbox. The coding form is real number coding, the initial population size is
50, the number of elite is 10, the ratio of cross offspring is 0.6, the maximum evolution algebra is 30,
the selection function uses random consistent selection, the crossover function uses scatter crossover,
and the mutation function uses constrained adaptive mutation. The whole system is operated under the
above constraints. The variation of system error is shown in Fig.4, the variation of hydrogen input
molar flow is shown in Fig.5, the variation of output voltage of SOFC single battery is shown in Fig.6,
and the output power of SOFC single battery is shown in Fig.7.
e(t)

(mol/s)

Time (s) Time (s)

Figure 4. Diagram of system error and Figure 5. Diagram of hydrogen input
running time. molar flow and running time.
As can be seen from Fig.4, when the running time increases from 0s to 0.59s, the system error
decreases sharply. When the running time increases from 0.59s to 1.37s, the system error increases
slowly. When the running time exceeds 1.37s, the system error is zero and no longer changes. It
indicates that the control signal generated by the genetic algorithm PID controller can quickly and
effectively control the system error to decrease, and that is to make the output voltage of SOFC single
battery reach the expected value.
As can be seen from Fig.5, when the running time increases from 0s to 1.38s, the hydrogen input
molar flow decreases significantly. When the running time exceeds 1.38s, the hydrogen input molar
flow maintains at 1.33mol/s and does not change. Since the water vapor input molar flow has been set
as 0.2mol/s in advance, and the hydrogen input molar flow is 1.33mol/s when the system is running

4
SAMSE 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 768 (2020) 042005 doi:10.1088/1757-899X/768/4/042005

stably. This indicates that the ideal ratio of hydrogen input molar flow to water vapor input molar flow
during normal operation of SOFC single cell should be 6.65:1.

Pcell (w)
Ucell (V)

Time (s) Time (s)

Figure 6. Diagram of the output voltage of Figure 7. Diagram of the output power of
SOFC single battery and running time. SOFC single battery and running time.

As can be seen from Fig.6, the output voltage of SOFC single battery reaches the maximum value
of 0.854v at 0.63s, and the overshooting is 0.47%. After 1.19s, the output voltage of SOFC single
battery remains at 0.85v, and the system has no steady-state error. It shows that the genetic algorithm
PID controller can effectively control the output voltage of SOFC single battery to reach the user's
expected value of 0.85v.
As can be seen from Fig.7, when the running time increases from 0s to 1.46s, the output power of
SOFC single battery shows a downward trend. When the running time exceeds 1.46s, the output power
of SOFC single battery is maintained at 1.92w. The output voltage and output power of SOFC single
battery does not reach the constant value at the same time, and there is a small 0.09s delay, indicating
that the hydrogen input molar flow will affect the internal current of the battery and the delay of this
process is 0.09s.

6. Conclusions
In this paper, based on the input and output characteristics of SOFC single battery adjustable
parameter model, a corresponding genetic algorithm PID controller is designed to control the output
voltage of SOFC single battery. The following conclusions are obtained through simulation.
1) When the SOFC single battery is in normal operation, the ideal ratio of hydrogen input molar
flow to water vapor input molar flow is 6.65:1.
2) The hydrogen input molar flow affects the internal current of SOFC single battery, and this
process takes 0.09s.
3) The designed genetic algorithm PID controller can effectively control the hydrogen input molar
flow to the optimal value, so that the output voltage of SOFC single battery can be maintained
constant.

7. Acknowledgments
This work is financially supported by Fundamental Research Funds for the Central Universities for
postgraduate of Northwest Minzu University (Grant No.Yxm2019160), Fundamental Research Funds
for the Central Universities of Northwest Minzu University (Grant No.31920190067), and Gansu
Provincial First-class Discipline Program of Northwest Minzu University (Grant No.11080305).

5
SAMSE 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 768 (2020) 042005 doi:10.1088/1757-899X/768/4/042005

References
[1] Eva Frey, SOFC Powering-The Telecommunication in Future, SIEMENS. 2005: 305-309.
[2] Maryam Sadeghi Reineh, Faryar Jabbari, Enhanced Power Generation in SOFC Using Artificial
Limits on Actuator Control Signals, Journal of Electrochemical Energy Conversion and Storage.
2019, 16: 1-9.
[3] So-Ryeok Oh, Jing Sun, Model Predictive Control for Power and Thermal Management of
Integrated Solid Oxide Fuel Cell and Turbocharge System, IEEE transactions on control systems
technology. 2014, 22 (3): 911-920.
[4] R.Horalek, J.Hlava, Multiple Model Predictive Control of Grid Connected Solid Oxide Fuel Cell
for Extending Cell Life Time, 2015 23rd Mediterranean Conference on Control and Automation
(MED). 2015: 310-315.
[5] Tooran Emami, Alex Tsai, David Tucker, Robust PID controller design of a solid oxide fuel cell
gas turbine, Proceeding of the ASME 2016 14th International Conference on Fuel Cell Science.
2016: 1-8.
[6] Wang Lijin, Zhang Huisheng, Wen Shilie, Control Strategy Research of Direct Internal
Reforming Solid Oxide Fuel Cell, Proceedings of the CSEE. 2008, 128(20): 94-98.
[7] Liu Xin, Hao Xiaohong, Yang Xinhua, Nonlinear model predictive control of the direct internal
reforming solid oxide fuel cell system, Chemical industry and engineering progress. 2014, 33(4):
900 -906.
[8] Yang Ji, Dian Songyi, Pu Ming, Study on Performance of Sliding Mode Control of Solid Oxide
Fuel Cells, Computer Simulation. 2016, 33(3): 99-104.
[9] Li Xiao, Li Junhong, Shen Wenbing, Modeling and generalized predictive control of solid oxide
fuel cell, Battery Bimonthly. 2018, 48(6): 381-383.
[10] Matheus B. de O. e Silva, Jussara F.Fardin, Modeling and Grid Connection of Solid Oxide Fuel
Cell (SOFC) based on P-Q Theory for Stationary Loads, 2015 IEEE PES Innovative Smart Grid
Technologies Latin America (ISGT LATAM). 2015: 541-545.
[11] Wu Xiaolong, Xu Yuanwu, Hu Rong, Fault modeling and simulation comparative study of two
solid oxide fuel cell system, Control Theory & Applications. 2019, 36(3): 408-412.
[12] Huo Haibo, Zhu Xinjian, Cao Guangyi, Research status and development of modeling and
controlling for SOFC, Chinese Journal of Power Sources. 2017, 31(10): 833-835.
[13] Jipeng Gu, Caixia Wang, Modeling and Simulation of the Electrical Characteristic of Solid Oxide
Fuel Cells, 2019 4th International Symposium on Computer and Information Processing
Technology. Guizhou, China.
[14] Jipeng Gu, Caixia Wang, Particle Swarm PID Control of Solid Oxide Fuel Cell, 2019 The 2nd
International Conference on Robotics, Control and Automation Engineering (RACE 2019).
Lanzhou, China.