Fuel cell output voltage control method and system based on fuzzy neural network PID

Abstract

The invention discloses a fuel cell output voltage control method and system based on fuzzy neural network PID, and the method comprises the steps: building an air-cooled fuel cell voltage model based on an MATLAB / Simulink simulation platform, and obtaining the temperature changes of an air-cooled fuel cell under different working conditions; establishing an air-cooled fuel cell test bench, and obtaining a polarization curve of output voltage changing along with current; according to experimental data, calibrating the semi-empirical equation by adopting a nonlinear least square method, and comparing experimental results to verify the accuracy of the model; the input hydrogen flow is changed through the PID controller, so that stable control of the output voltage of the galvanic pile is realized, and the power generation performance of the battery is improved; a fuzzy neural network is adopted to optimize a PID controller, control parameters of PID are adjusted in a self-adaptive mode according to the error between expected voltage and actual voltage and the error change rate, and therefore a better temperature control effect is achieved. The method has the advantages of high robustness, high response speed and the like, and ensures that the voltage output of the fuel cell stack is maintained at a fixed preset value.

Description

Technical Field

[0001] The present invention belongs to the technical field of fuel cell voltage control, and in particular relates to a fuel cell output voltage control method and system based on fuzzy neural network PID. Background Art

[0002] With the increasing severity of environmental pollution and energy crisis, the development of clean energy is in full swing. Among them, the proton exchange membrane fuel cell (PEMFC) is considered to be the most promising energy conversion device due to its advantages such as low operating temperature, high energy conversion efficiency, no pollution, and high power density. It has received extensive attention and research.

[0003] Voltage is the most important indicator of fuel cell output characteristics and has a significant impact on battery performance. The output voltage of PEMFC is directly related to its power output. By controlling the voltage, the output power of the fuel cell can be adjusted to meet the needs of different application scenarios; the working voltage of PEMFC is directly related to its energy conversion efficiency. Within a certain voltage range, increasing the output voltage can reduce battery heat loss and improve energy utilization efficiency; excessive voltage will cause the catalyst in the membrane electrode to decay faster, thereby shortening the service life of the fuel cell; fuel cells will inevitably experience frequent changes in operating parameters such as start-stop, load change, operating environment, and operating parameters during operation. Voltage control helps maintain the stable operation of the fuel cell and avoid performance degradation due to voltage fluctuations.

[0004] The current PEMFC voltage control methods have the following shortcomings. Traditional PID control and model predictive control (MPC) need to be designed based on accurate mathematical models. However, the PEMFC system is highly nonlinear and time-varying, and it is difficult to establish an accurate model, which limits the effectiveness of these control methods. The parameters of the traditional PID controller usually need to be manually adjusted by trial and error, which is a time-consuming process and may not achieve the optimal control effect. In addition, when system conditions change, these parameters may need to be readjusted, and they are not sufficiently adaptive and cannot automatically adjust the control strategy during system operation to adapt to changing working conditions. Although model predictive control can provide good control effects, it has high computational complexity and requires powerful computing resources, which may be limited in practical applications. Summary of the invention

[0005] In order to solve the problems in the prior art, the present invention provides an air-cooled PEMFC voltage control method and system based on fuzzy neural network PID, which realizes effective control of the PEMFC temperature.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: The fuel cell output voltage control method based on fuzzy neural network PID includes the following steps: S1, based on the electrochemical characteristics of the voltage loss of the proton exchange membrane fuel cell PEMFC, a dynamic semi-empirical model of the output voltage of the air-cooled PEMFC stack was built based on the MATLAB / Simulink simulation platform; S2, in order to optimize the unknown parameters in the dynamic semi-empirical model described in S1, an air-cooled PEMFC voltage test platform was built to test and obtain the polarization curve of the PEMFC stack output voltage under different operating conditions; S3, according to the polarization curve measured by S2, using the nonlinear least square method, calibrating the unknown parameters in the dynamic semi-empirical model described in S1, comparing the simulation data generated by the dynamic semi-empirical model with the experimental data, and verifying the accuracy of the established dynamic semi-empirical model; S4, based on the dynamic semi-empirical model verified by S3, a PID controller is used to adjust the hydrogen flow rate input to the PEMFC stack and control the output voltage of the PEMFC stack to a given value; S5, using a fuzzy neural network to optimize the three control parameters in the PID controller described in S4, and obtain a fuzzy neural network PID controller for improving the output voltage control of the air-cooled PEMFC stack, the three control parameters are Kp, Ki and Kd.

[0007] A further improvement of the present invention is that in step S1, according to the electrochemical characteristics of the voltage loss of the proton exchange membrane fuel cell PEMFC, a dynamic semi-empirical model of the output voltage of the air-cooled PEMFC stack is built based on the MATLAB / Simulink simulation platform, including: Nernst voltage: The maximum electrical energy that a fuel cell can generate is related to the Gibbs free energy. The thermodynamic electromotive force of a fuel cell is expressed as:

[0008] in is the thermodynamic electromotive force of the fuel cell under standard conditions, i.e., 25°C operating temperature and 1 atm pressure, ΔS is the entropy free energy change of the battery reaction, nF is the Faraday constant, T is the actual temperature of the battery during operation, is the standard reference temperature of the battery, R is the ideal gas constant; Activation loss: Activation loss represents the voltage sacrificed to overcome the activation energy barrier associated with the electrochemical reaction, which is caused by the slow electrochemical reaction kinetics and is expressed by the equation:

[0009] Where: I is the battery current, i represents the actual working current, in represents the no-load current; ξ is the constant coefficient calculated from experimental data; C O2 It represents the oxygen concentration on the contact surface between the cathode membrane and oxygen, calculated using Henry's law; Ohmic losses: Ohmic losses are caused by the resistance of protons passing through the electrolyte and electrons flowing through fuel cell components such as bipolar plates and gas diffusion layers. They are calculated according to Ohm's law:

[0010] Where: Rm represents the equivalent impedance of protons passing through the exchange membrane, and Rc represents the equivalent resistance of the fuel cell when electrons pass through the circuit. Since Rc is relatively fixed, it is treated as a constant. Concentration loss: When reactants undergo electrochemical reactions at the electrodes and are rapidly consumed, concentration polarization occurs, which produces a concentration gradient and affects the Nernst voltage and kinetic reaction rate of the single cell, resulting in a decrease in voltage. The calculation formula for concentration loss is:

[0011] Where: b represents the coefficient of concentration loss, J represents the current density, and Jmax represents the limiting current density; Double-layer model: There is a "double-layer charge layer phenomenon" inside the PEMFC, which forms a potential difference between the electrolyte and the electrode. An equivalent capacitor is used to replace it, and the following equation is used to describe it:

[0012] Where Vd is the voltage of the double layer, I is the current through the battery, C is the equivalent capacitance, and Rd is the equivalent resistance; The above establishes a complete mathematical model of PEMFC single-chip battery; In order to increase the total output voltage of the PEMFC stack system and meet the power requirements, multiple single cells are connected in series. The resulting PEMFC stack output voltage is expressed as:

[0013] Where Vs is the total output voltage of the battery stack, Ns is the number of battery cells connected in series in the battery stack, and Vcell is the voltage of a single battery cell; Therefore, a complete semi-empirical model of PEMFC stack output voltage was established.

[0014] A further improvement of the present invention is that in step S2, an air-cooled PEMFC voltage test platform is constructed, including: Stack: Open cathode air-cooled PEMFC stack Fan + PWM speed signal regulation generator: Use PWM speed signal to control fan speed, and use fan speed to maintain stack temperature and oxygen supply; Gas supply subsystem: Provides hydrogen as a reactant for the experimental fuel cell, and consists of the following parts: Hydrogen tanks - provide 99.99% pure hydrogen to the stack; Pressure reducing valve; Flow meter: connected to the pipeline before the anode hydrogen inlet to measure the flow rate and volume of hydrogen; Pressure sensor: connected to the pipeline in front of the anode hydrogen inlet to measure the hydrogen pressure at the anode inlet; Anode purification subsystem: manages the emission of anode gas and has the following functions: When opened, the gas and water in the anode flow channel are discharged; When closed, the anode reaches dead-end mode and maintains a certain pressure; The experiment used constant opening and closing times of 12.5s:0.5s throughout; Data acquisition subsystem: responsible for collecting data such as battery stack voltage and temperature, consisting of a data collector and a test system; Output subsystem: used to record the output current and dissipate the output power of the battery stack. It consists of an ammeter and an electronic load: Ammeter: record output current; Electronic load: used to dissipate the output power of the battery stack.

[0015] A further improvement of the present invention is that in step S3, calibrating the unknown parameters of the PEMFC stack output voltage model includes: Firstly, the experimentally measured polarization curve is used as reference data, and the nonlinear least squares method is used to fit the unknown parameter values ​​of the semi-empirical model to form an accurate dynamic semi-empirical model of the PEMFC stack output voltage. Then, the PEMFC stack output voltage model is run to obtain the voltage output of the PEMFC stack output voltage model under different input conditions, and the voltage output of the real stack under the corresponding conditions is compared. The mean square error (MSE) is used as the evaluation standard to verify the accuracy of the model. The formula of MSE is:

[0016] Where: and Represent the corresponding voltage values ​​of simulation and experiment respectively, and n represents the amount of data.

[0017] A further improvement of the present invention is that in step S4, based on the dynamic semi-empirical model verified in S3, a PID controller is used to adjust the hydrogen flow rate input to the PEMFC stack to control the output voltage of the PEMFC stack to a given value, including: The output voltage of PEMFC is stabilized by controlling the anode hydrogen flow rate; the hydrogen pressure is determined by the inflow, consumption and outflow of hydrogen inside the fuel cell; the following equation is obtained based on the gas state equation and the law of conservation of mass:

[0018] Where: , , Respectively represent the inflow, outflow and consumption of hydrogen, the unit is mol / l, Represents the total volume of the anode flow channel, taking 0.005m 3 ; The PID controller is used to control the output voltage of the PEMFC system to be stable by changing the hydrogen flow rate. The PID controller consists of three basic parts: proportional P, integral I and differential D. The PID controller can generate the corresponding control quantity according to the error. The PID control algorithm is expressed as:

[0019] Where: Kp is the proportional gain, Ki is the integral gain, Kd is the differential gain; u(t) represents the control action in the time domain; The error e(t)=Vs(t)-V(t) between the expected voltage value Vs(t) and the actual voltage value V(t) is selected as the input of the PID voltage controller. Through proportional, integral and differential operations, the output control action u(k) is obtained as the hydrogen volume flow rate; The error e(t)=Vs(t)-V(t) between the expected voltage value Vs(t) and the actual voltage value V(t) is selected as the input of the PID voltage controller. Through proportional, integral and differential operations, the output control action u(t) is obtained as the hydrogen volume flow rate.

[0020] A further improvement of the present invention is that, in step S5, a fuzzy neural network is used to optimize three control parameters in the PID controller to obtain a fuzzy neural network PID controller for controlling the output voltage of the air-cooled PEMFC stack, including: Firstly, the PID is optimized by using a fuzzy controller, and then the fuzzy rules of the fuzzy controller are optimized by using a fuzzy neural network. The voltage control effects of the traditional PID controller, the fuzzy controller, and the fuzzy neural network PID controller are compared. Optimize PID through fuzzy control, realize the real-time adjustment function of the important influencing parameters Kp, Ki, Kd of the PID controller, and obtain better control effect; make control decisions by observing the size of the error E and the error change rate EC, combining experience and technical knowledge; then realize the real-time adjustment of the important influencing parameters Kp, Ki, Kd of the PID controller through the computer; Neural network is used as a mathematical tool, and the input and output of the neural network are used to represent the input and output of the fuzzy system. The membership function and fuzzy rules of the fuzzy system are added to the implicit nodes of the neural network to optimize the membership function and fuzzy rules of the fuzzy controller.

[0021] Fuel cell output voltage control system based on fuzzy neural network PID, including: Dynamic semi-empirical model building module, based on the MATLAB / Simulink simulation platform, builds a dynamic semi-empirical model of the output voltage of the air-cooled PEMFC stack according to the electrochemical characteristics of the voltage loss of the proton exchange membrane fuel cell PEMFC; The polarization curve measurement module is used to optimize the unknown parameters in the dynamic semi-empirical model described in the dynamic semi-empirical model building module, build an air-cooled PEMFC voltage test platform, and test and obtain the polarization curve of the PEMFC stack output voltage under different operating conditions; The model verification module uses the nonlinear least square method to calibrate the unknown parameters in the dynamic semi-empirical model in the dynamic semi-empirical model building module according to the polarization curve measured by the polarization curve measurement module, compares the simulation data generated by the dynamic semi-empirical model with the experimental data, and verifies the accuracy of the established dynamic semi-empirical model; The parameter adjustment module uses a PID controller to adjust the hydrogen flow rate input to the PEMFC stack based on the dynamic semi-empirical model verified by the model verification module, and controls the output voltage of the PEMFC stack to a given value; The output voltage control module uses a fuzzy neural network to optimize the three control parameters in the PID controller in the parameter adjustment module to obtain a fuzzy neural network PID controller, which is used to improve the output voltage control of the air-cooled PEMFC stack. The three control parameters are Kp, Ki and Kd.

[0022] Compared with the prior art, the present invention has at least the following beneficial technical effects: The air-cooled PEMFC voltage control method and system based on fuzzy neural network PID provided by the present invention first establishes an air-cooled PEMFC system model, builds an air-cooled PEMFC voltage test experimental platform, and uses the measured experimental data to calibrate the unknown parameters of the voltage model in the air-cooled PEMFC system; on this basis, the present invention uses a PID controller to adjust the hydrogen flow rate, controls the PEMFC stack output voltage to a given value, and then uses a fuzzy neural network to optimize the three control parameters in the PID controller to obtain a fuzzy neural network PID controller, thereby further improving the control effect of the air-cooled PEMFC voltage. The present invention realizes the adaptive change of PID control parameters Kp, Ki, and Kd by combining fuzzy control, solves the problem that the control performance of the PID controller cannot meet the control requirements of the complex and strongly nonlinear air-cooled PEMFC stack output voltage model under the change of working conditions, and improves the control effect; the present invention complements the advantages of neural network and fuzzy control, and the proposed fuzzy neural network PID controller can determine the relatively optimal fuzzy rules and membership functions, thereby further improving the fuzzy PID control effect.

[0023] In summary, the present invention adopts neural network as a mathematical tool. This scheme combines the advantages of neural network and fuzzy control: it solves the problem that fuzzy control cannot determine the optimal fuzzy rules and membership functions, and at the same time, through fuzzy theory, it makes up for the shortcomings of neural network that cannot process language variables and cannot combine expert knowledge, and gives full play to the parallel processing capability of neural network and the reasoning ability of fuzzy system. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is the structural schematic diagram of the test bench; Figure 2 It is a schematic diagram of the structure of an open cathode air-cooled PEMFC and the appearance of a fan; Figure 3 (a) is the fan static pressure and air volume characteristic curve, (b) is the fan power under different PWM values; Figure 4 This is a schematic diagram of the coefficients of the least squares method for fitting the semi-empirical equation; Figure 5 This is the principle block diagram of the PID controller; Figure 6 This is the principle block diagram of the fuzzy PID controller; Figure 7 It is a schematic diagram of the input membership function of fuzzy PID; Figure 8 It is a schematic diagram of the output membership function of fuzzy PID; Fig. 9 It is the schematic diagram of fuzzy rules of fuzzy PID; Fig.10This is the principle block diagram of the fuzzy neural network PID controller; Fig.11 It is the principle diagram of the standard fuzzy neural network; Fig.12 This is the fuzzy neural network training effect diagram; Fig.13 It is the curve of the proportional coefficient Kp of the network output changing with the actual value; Fig.14 It is the curve of the proportional coefficient Ki of the network output changing with the actual value; Fig.15 It is the curve of the proportional coefficient Kd of the network output changing with the actual value; Fig.16 Schematic diagram of the input membership function and fuzzy rules of ΔKp; Fig.17 Schematic diagram of the input membership function and fuzzy rules of ΔKp; Fig.18 Schematic diagram of the input membership function and fuzzy rules of ΔKp; Fig.19 Schematic diagram of the input membership function and fuzzy rules of ΔKp; Fig. 20 Schematic diagram of the input membership function and fuzzy rules of ΔKp; Fig.21 Schematic diagram of the input membership function and fuzzy rules of ΔKp; Fig. 22 A schematic diagram showing the comparison of the control effects of different control methods on the air-cooled PEMFC voltage; Fig.23 It is a structural block diagram of the fuel cell output voltage control system based on fuzzy neural network PID of the present invention. DETAILED DESCRIPTION

[0025] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and descriptions are considered to be exemplary and non-restrictive in nature.

[0026] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0027] It should also be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0028] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0029] Various structural schematic diagrams of the embodiments disclosed in the present invention are shown in the accompanying drawings. These figures are not drawn to scale, and some details are magnified and some details may be omitted for the purpose of clear expression. The shapes of various regions and layers shown in the figures and the relative sizes and positional relationships therebetween are only exemplary, and may deviate in practice due to manufacturing tolerances or technical limitations, and those skilled in the art may additionally design regions / layers with different shapes, sizes, and relative positions according to actual needs.

[0030] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0031] Example 1 The air-cooled PEMFC voltage control method based on fuzzy neural network PID provided by the present invention comprises the following steps: S1, based on the electrochemical characteristics of the voltage loss of the proton exchange membrane fuel cell PEMFC, a dynamic semi-empirical model of the output voltage of the air-cooled PEMFC stack was built based on the MATLAB / Simulink simulation platform; S2, in order to optimize the unknown parameters in the dynamic semi-empirical model described in S1, an air-cooled PEMFC voltage test platform was built to test and obtain the polarization curve of the PEMFC stack output voltage under different operating conditions; S3, according to the polarization curve measured by S2, using the nonlinear least square method, calibrating the unknown parameters in the dynamic semi-empirical model described in S1, comparing the simulation data generated by the dynamic semi-empirical model with the experimental data, and verifying the accuracy of the established dynamic semi-empirical model; S4, based on the dynamic semi-empirical model verified by S3, a PID controller is used to adjust the hydrogen flow rate input to the PEMFC stack and control the output voltage of the PEMFC stack to a given value; S5, using a fuzzy neural network to optimize the three control parameters in the PID controller described in S4, and obtain a fuzzy neural network PID controller for improving the output voltage control of the air-cooled PEMFC stack, the three control parameters are Kp, Ki and Kd.

[0032] In step S1, in order to design a control strategy and verify the effectiveness of the control strategy, it is necessary to model the controlled object. The dynamic semi-empirical model of the PEMFC stack output voltage includes: a) Nernst voltage: The maximum electrical energy that a fuel cell can generate is related to the Gibbs free energy. In this sense, the thermodynamic electromotive force of the fuel cell can be expressed as:

[0033] in is the thermodynamic electromotive force of the fuel cell under standard conditions, i.e., 25°C operating temperature and 1 atm (101315 Pa) pressure, ΔS is the entropy free energy change of the battery reaction, nF is the Faraday constant, T is the actual temperature of the battery during operation, is the standard reference temperature of the battery, and R is the ideal gas constant.

[0034] b) Activation loss: Activation loss represents the voltage sacrificed to overcome the activation energy barrier associated with the electrochemical reaction, which is caused by the slow electrochemical reaction kinetics and can be expressed as:

[0035] Where: I is the battery current, i represents the actual working current, and in represents the no-load current. The term ξ is a constant coefficient that can be calculated from experimental data. CO2 represents the oxygen concentration on the contact surface between the cathode membrane and oxygen, which can be calculated using Henry's law.

[0036] c) Ohmic losses: Ohmic losses are caused by the resistance of protons passing through the electrolyte and electrons flowing through fuel cell components such as bipolar plates and gas diffusion layers. They can be calculated according to Ohm's law:

[0037] Where: Rm represents the equivalent impedance of protons passing through the exchange membrane, and Rc represents the equivalent resistance of the fuel cell when electrons pass through the circuit. Since Rc is relatively fixed, it is treated as a constant.

[0038] d) Concentration loss: When reactants undergo electrochemical reactions at the electrodes and are rapidly consumed, concentration polarization occurs, which produces a concentration gradient and affects the Nernst voltage and kinetic reaction rate of the single cell, resulting in a decrease in voltage. The calculation formula for concentration loss is:

[0039] Where: b represents the coefficient of concentration loss, J represents the current density, and Jmax represents the limiting current density.

[0040] e) Double-layer model: There is a "double-layer charge layer phenomenon" inside the PEMFC, which forms a certain potential difference between the electrolyte and the electrode. It can be replaced by an equivalent capacitor to "smooth" the output voltage. The resulting impact is combined with the activation loss and concentration loss terms to produce a first-order model, which can be described by the following equation:

[0041] Where Vd is the voltage of the double layer, I is the current through the battery, C is the equivalent capacitance, and Rd is the equivalent resistance.

[0042] f) Connecting multiple single cells in series can increase the total output voltage of the system and meet the power requirements. Therefore, the output voltage of the battery stack can be expressed as:

[0043] Where Vs is the total output voltage of the battery stack, Ns is the number of battery cells connected in series in the battery stack, and Vcell is the voltage of a single battery cell.

[0044] In this way, a complete semi-empirical model of air-cooled fuel cell is established, but some parameters are unknown: ξ1, ξ2, ξ3, ξ4, Rc, b, λ, Jmax, Jn, which need to be fitted through experimental data.

[0045] In step S2, the air-cooled PEMFC voltage test platform is as follows: Figure 1 As shown, it includes the following subsystems: a) Stack: This experiment uses an open cathode air-cooled PEMFC stack, which looks like Figure 2 The electrical parameters of the battery stack are shown in Table 1: Table 1: Electrical parameters of the open cathode air-cooled PEMFC used in the experiment

[0046] b) Fan + PWM speed signal generator: This experiment uses PWM speed signal to control the fan speed, and uses the fan speed to maintain the temperature of the battery stack and the oxygen supply. An axial flow fan (model cht9248by-w38) produced by Zhaoqing Shenghui Electronic Technology Co., Ltd. is installed on the lightweight housing. The fan characteristic curve, also known as the PQ curve, is as follows Figure 3 (a) can be used to evaluate the cooling capacity of the fan. The fan can provide a maximum airflow of 173.46 CFM (cubic feet per minute) or a maximum static pressure of 65.3 mmAq (millimeters of water column) at full speed. The fan speed is regulated by a pulse width modulation (PWM) signal, where an increase in the PWM value causes a corresponding increase in the fan speed. Figure 3 (b) shows the fan power at different PWM values.

[0047] c) Gas supply subsystem: Provides hydrogen as a reactant for the experimental stack and consists of the following parts: d) Anode purification subsystem: manages the emission of anode gas. At the anode outlet of the stack, a solenoid valve produced by Foshan Weilizi Electronic Technology Co., Ltd. is used for regular purge. This valve can be switched on and off through a control signal and has the following functions: This experiment used a constant opening and closing time of 12.5s:0.5s throughout.

[0048] e) Data acquisition subsystem: responsible for collecting data such as the voltage and temperature of the fuel cell stack, and consists of a data collector and a test system. 40 sets of voltage acquisition cables are attached to each single cell and connected to the test system to measure the voltage of each single cell. In order to measure the stack temperature, 75 sets of K-type thermocouples are placed in the air flow channel and connected to the test system. The test system is produced by Guangdong Hydrogen Energy Technology Co., Ltd. and can collect various parameters such as voltage, current, temperature, ambient temperature, humidity, etc. during the operation of the fuel cell. It also outputs control signals to the fan and solenoid valve.

[0049] In step S3, the purpose of calibrating the parameters is to make the simulated polarization curve as consistent as possible with the experimentally measured polarization curve, so obtaining experimental data is the basis for parameter optimization. First, the experimentally measured polarization curve is used as reference data, and the nonlinear least squares method is used to fit the unknown parameter values ​​of the semi-empirical model to form an accurate dynamic semi-empirical model of the PEMFC stack output voltage; then the model is run to obtain the voltage output of the stack model under different input conditions, and the voltage output of the real stack under the corresponding conditions is compared, and the mean square error (MSE) is used as the evaluation standard to verify the accuracy of the model. The formula of MSE is:

[0050] Where: and Represent the corresponding voltage values ​​of simulation and experiment respectively, and n represents the amount of data.

[0051] The semi-empirical model was verified based on the experimental data. The results are as follows: Figure 4 The maximum relative error of the corresponding voltage is 1.34%, and the minimum relative error is 0.059%. Within the allowable error range, it can be considered that the two are highly consistent, which verifies the feasibility of using this model for simulation research.

[0052] In step S4, PEMFC is a nonlinear time-varying system with complex internal parameters and multiphase flow. Under different operating conditions such as increasing or decreasing load, the battery voltage will change sharply and fluctuate, which reduces the stability of the output voltage. The output voltage of the PEMFC system is a key indicator for evaluating the performance of battery power generation, and it is particularly important to achieve stability control. The anode hydrogen flow rate is used to control the output voltage of PEMFC to stabilize it. Hydrogen pressure has a direct impact on the output voltage, but in actual engineering, the pressure is not easy to control directly, so the relationship between hydrogen flow and pressure must be established to achieve stable control of the output voltage. The hydrogen pressure is determined by the inflow, consumption and outflow of hydrogen inside the fuel cell. According to the gas state equation and the law of conservation of mass, the following equation can be obtained:

[0053] The PID controller is used to control the output voltage of the PEMFC system to be stable by changing the hydrogen flow rate. The PID controller is widely used in practical engineering due to its simple structure, strong robustness and easy implementation. The PID controller consists of three basic parts: proportional (P), integral (I) and differential (D). Each part has its specific function: a) Proportional (P): Proportional to the current error, used to reduce the error.

[0054] b) Integral (I): Proportional to the integral of the error and used to eliminate steady-state errors.

[0055] c) Differential (D): Proportional to the rate of change of the error and used to predict the future trend of the error.

[0056] The PID controller can generate the corresponding control quantity according to the error. The PID control algorithm can be expressed as:

[0057] The present invention selects the error e(t)=Vs(t)-V(t) between the expected voltage value Vs(t) and the actual voltage value V(t) as the input of the PID voltage controller, and obtains the output control action u(k) as the hydrogen volume flow rate through proportional, integral and differential operations; The present invention selects the error e(t)=Vs(t)-V(t) between the expected voltage value Vs(t) and the actual voltage value V(t) as the input of the PID voltage controller, and obtains the output control action u(t) as the hydrogen volume flow rate through proportional, integral and differential operations; PID embedded interface: such as Figure 5As shown, the input of the PID controller is the error e(t)=Vs(t)-V(t) between the expected voltage value Vs(t) and the actual voltage value V(t). In the PID controller, through proportional, integral and differential operations, the output control action u(t) is obtained as the hydrogen volume flow rate, and the output signal is input into the open cathode air-cooled PEMFC gas supply subsystem to affect the actual working performance of the control object (fuel stack), and obtain a new output voltage V(t) to achieve real-time control. The control amount generated by the operation of the PID controller can gradually reduce the error e(t) to 0, that is, automatic control without steady-state error can be achieved. In step S5, the present invention first uses a fuzzy controller to optimize the PID, and then further uses a fuzzy neural network to optimize it, and compares the voltage control effects of the traditional PID controller, the fuzzy PID controller, and the fuzzy neural network PID controller.

[0058] Since the control parameters Kp, Ki, and Kd of the traditional PID control algorithm are fixed and cannot be adaptively changed according to the system state, they are not suitable for strongly nonlinear systems such as the PEMFC thermal management system.

[0059] Fuzzy control is a mathematical method that uses human knowledge and experience to process uncertainty and fuzzy information. It is often used in complex nonlinear systems. Fuzzy control is based on fuzzy set theory, fuzzy language and fuzzy logic control. It realizes the application of fuzzy mathematics in control systems and belongs to nonlinear intelligent control.

[0060] By optimizing PID through fuzzy control, the function of real-time adjustment of control parameters can be realized to obtain better control effect, that is, fuzzy PID control, also known as fuzzy adaptive PID; Fuzzy control systems generally consist of the following five parts: (1) Define variables: that is, determine the input and output variables of the fuzzy control system. In general, in the fuzzy control optimization PID problem, the system error E and the error change rate EC are selected as input variables, and the control parameter changes ∆Kp, ∆Ki, and ∆Kd are selected as outputs.

[0061] (2) Fuzzification: Convert the input value to the numerical value of the fuzzy domain in an appropriate proportion, describe it using colloquial vocabulary, and calculate the corresponding membership. For example, the input error E can be described by the following words: {negative large, negative medium, negative small, zero, positive small, positive medium, positive large}; the corresponding English abbreviations are: {NB, NM, NS, ZO, PS, PM, PB}; the input and output membership functions required for fuzzification are as follows: Figure 7 , Figure 8 shown.

[0062] (3) Knowledge base: It consists of a database and a rule base. The database defines how to process fuzzy data, and the rule base defines control strategies through language control rules. Control rules are often expressed in the form of "if condition, then result".

[0063] (4) Logical judgment: Imitate the fuzzy concepts of human judgment and use fuzzy inference to obtain fuzzy control signals.

[0064] (5) Defuzzification: Convert the fuzzy value obtained by fuzzy inference into a clear control signal.

[0065] In order to optimize the output voltage of the classic PID-controlled PEMFC, a fuzzy PID controller is designed. The controller takes the deviation E between the output voltage and the given voltage and the deviation change rate EC as input, and dynamically adjusts the PID control parameters Kp, Ki, and Kd in real time to stabilize the output voltage. Position PID control is adopted, and the discrete expression of fuzzy PID is as follows:

[0066]

[0067]

[0068]

[0069] Where: k represents the sampling time, ∆Kp, ∆Ki, ∆Kd For fuzzy control systems Kp, Ki, Kd The structure of the PEMFC fuzzy PID control system is as follows: Figure 6 As shown, firstly, the error E and error rate of change EC As the input of the fuzzy control system, it is mapped to the fuzzy domain through the quantization factor, and then the corresponding membership value is calculated according to the corresponding membership function, and then the fuzzy rule reasoning is used to obtain ∆Kp, ∆Ki, ∆Kd Corresponding to the membership degree, the defuzzification is finally performed to obtain ∆ Kp、∆Ki、∆ Kd The final value is applied to the PID controller.

[0070] Fuzzy control rules are gradually formed based on human learning, experiments and long-term experience accumulation. E and error rate of change EC The size of the control is combined with experience and technical knowledge to make control decisions and formulate control rules as shown in Tables 2 to 4.

[0071] Table 2: Proportional gain change ∆Kp Fuzzy control rule table

[0072] Table 3: Integral gain change ∆Ki Fuzzy control rule table

[0073] Table 4: Differential gain change ∆Kd Fuzzy control rule table

[0074] The corresponding fuzzy rules of fuzzy PID are detailed in Fig. 9 .

[0075] Then, using the fuzzy set theory and the concept of linguistic variables, the fuzzy PID control related operations are realized through computers, so that the system can achieve the expected goals.

[0076] However, fuzzy control has the following problems: fuzzy rules and membership functions are determined by the experience of human experts and can be used as feasible solutions with reference value, but they are not optimal solutions, that is, humans cannot directly obtain the optimal solution; in the face of this optimization problem, neural networks are introduced for improvement, namely fuzzy neural networks (FNN for short).

[0077] Using neural network as a mathematical tool, the input and output of the neural network are used to represent the input and output of the fuzzy system, and the membership function and fuzzy rules of the fuzzy system are added to the implicit nodes of the neural network. The membership function and fuzzy rules of the fuzzy controller are optimized, which solves the problem that fuzzy control cannot determine the optimal fuzzy rules and membership function. At the same time, the fuzzy theory makes up for the shortcomings of the neural network that cannot process language variables and cannot combine expert knowledge, and gives full play to the parallel processing ability of the neural network and the reasoning ability of the fuzzy system. Fuzzy neural networks combine the advantages of both: using the learning ability of neural networks to optimize fuzzy control rules and membership functions; pre-distributing expert knowledge into neural networks to give full play to the parallel processing capabilities of neural networks and the reasoning capabilities of fuzzy systems.

[0078] The PID parameters are automatically tuned online by using a fuzzy neural network to obtain a fuzzy neural network PID voltage controller; the error e(t) = Vs(t) - V(t) between the expected output voltage value Vs(t) and the actual output voltage value V(t) of the PEMF stack and its error change rate ec(t) are selected as input variables of the fuzzy PID temperature controller, and the parameter adjustment values ​​ΔKP, ΔKI, ΔKD of the fuzzy PID temperature controller are used as output values, where ΔKP, ΔKI, ΔKD are the changes of the three parameters KP, KI, KD corresponding to the proportional integral P, the integral I and the differential D respectively; The system structure of the fuzzy neural PID controller used in the present invention is as follows: Fig.10As shown in the figure, the parameter tuning of the PID controller based on fuzzy neural network is to use the learning function of fuzzy neural network to adjust the output layer weight of the network, the center value and width of the Gaussian membership function online, modify the three parameters of PID, and obtain a set of appropriate control parameters, namely Kp, Ki, Kd, ​​so as to realize parameter self-tuning.

[0079] The standard fuzzy neural network structure is as follows Fig.11 As shown in the figure, it is essentially a 5-layer feedforward network, which includes input layer, membership function generation layer, inference layer, normalization layer and output layer. It realizes nonlinear mapping from input to output through error back propagation learning method.

[0080] a) The first layer: input layer, the input nodes are linear, composed of n neurons, which convert the input signal of the network Transfer to the next layer, the number of which is the number of input variables; b) The second layer: the membership function generation layer, each node of this layer represents a language variable value, so as to calculate the corresponding membership function ,in: , n is the dimension of the input variable; mi is the number of fuzzy rules of xi; there are n groups in total, each with membership functions. In the present invention, n=3, =49. The membership function generally uses the Gaussian function, and the formula is as follows:

[0081] in, and They represent the center value and width value of the membership function respectively.

[0082] The total number of nodes in this layer .

[0083] c) The third layer: fuzzy reasoning layer, using TS-type fuzzy reasoning. Each node in this layer represents a fuzzy rule. Each corresponding fuzzy rule is used to match the antecedent of the fuzzy rule, so as to calculate the fitness of each rule. The following formula is generally used for calculation:

[0084] in:

[0085] The total number of nodes in this layer .

[0086] d) The fourth layer: normalization layer, which realizes the normalization calculation of fitness. Since the fitness values ​​of each rule are different and cannot be directly compared, normalization processing is required to facilitate mutual comparison; the following formula is used for normalization:

[0087] in:

[0088] The number of nodes is the same as the third layer. .

[0089] e) The fifth layer: output layer, which realizes the defuzzification calculation, that is, the change of the output parameters Kp, Ki, and Kd ;

[0090] Wherein, k = 1, 2, 3; ; is the connection weight of the output layer.

[0091] This fuzzy neural network is trained using the historical adjustment data of PID, and the parameters of the membership function layer and the output layer are updated through regression training. Fig.12 It can be seen that after fuzzy neural network training, as the number of iterations increases, the error gradually decreases until it is less than the set value.

[0092] Depend on Figure 13-15 It can be seen that the values ​​of the parameters Kp, Ki, and Kd output by the trained neural network can follow the corresponding actual value fluctuations well.

[0093] ΔKp, ΔKi, and ΔKd are trained separately to obtain three fuzzy inference systems. Since TS-type fuzzy inference has no output membership function, each fuzzy inference system only has input membership function and fuzzy rules: the input membership function and fuzzy rules of ΔKp are as follows: Fig.16 and Fig.17 As shown, the input membership function and fuzzy rule of ΔKp are as follows Fig.18 and Fig.19 As shown, the input membership function and fuzzy rule of ΔKp are as follows Fig. 20 and Fig.21 shown.

[0094] like Fig. 22 As shown in the figure, the voltage control effects of the three proposed control methods are compared: traditional PID controller, fuzzy controller, and fuzzy neural network PID controller. It can be seen that the voltage overshoot is significantly reduced each time the control method is optimized; the integral absolute error IAE (Integral Absolute Error) is taken as the evaluation index, and its calculation expression is:

[0095] The control optimization effects of the fuzzy PID controller, fuzzy neural network PID controller and traditional PID controller are shown in Table 5.

[0096] Table 5:

[0097] The fuzzy neural network PID controller used in the present invention has a significant improvement in control effect compared with the traditional PID controller and the fuzzy PID in solving the proposed air-cooled PEMFC voltage control problem, and has certain advancement.

[0098] Example 2 like Fig.23 As shown, the fuel cell output voltage control system based on fuzzy neural network PID provided by the present invention includes: Dynamic semi-empirical model building module, based on the MATLAB / Simulink simulation platform, builds a dynamic semi-empirical model of the output voltage of the air-cooled PEMFC stack according to the electrochemical characteristics of the voltage loss of the proton exchange membrane fuel cell PEMFC; The polarization curve measurement module is used to optimize the unknown parameters in the dynamic semi-empirical model described in the dynamic semi-empirical model building module, build an air-cooled PEMFC voltage test platform, and test and obtain the polarization curve of the PEMFC stack output voltage under different operating conditions; The model verification module uses the nonlinear least square method to calibrate the unknown parameters in the dynamic semi-empirical model in the dynamic semi-empirical model building module according to the polarization curve measured by the polarization curve measurement module, compares the simulation data generated by the dynamic semi-empirical model with the experimental data, and verifies the accuracy of the established dynamic semi-empirical model; The parameter adjustment module uses a PID controller to adjust the hydrogen flow rate input to the PEMFC stack based on the dynamic semi-empirical model verified by the model verification module, and controls the output voltage of the PEMFC stack to a given value; The output voltage control module uses a fuzzy neural network to optimize the three control parameters in the PID controller in the parameter adjustment module to obtain a fuzzy neural network PID controller, which is used to improve the output voltage control of the air-cooled PEMFC stack. The three control parameters are Kp, Ki and Kd.

[0099] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic features of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the attached claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any figure mark in the claims should not be regarded as limiting the claims involved.

[0100] In addition, it should be understood that although this specification is described in accordance with the implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation modes that can be understood by those skilled in the art. The above content is only to illustrate the technical idea of ​​the present invention, and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution according to the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A fuel cell output voltage control method based on fuzzy neural network PID, characterized in that: The following steps are involved: S1, based on the electrochemical characteristics of the voltage loss of the proton exchange membrane fuel cell PEMFC, a dynamic semi-empirical model of the output voltage of the air-cooled PEMFC stack was built based on the MATLAB / Simulink simulation platform; S2, in order to optimize the unknown parameters in the dynamic semi-empirical model described in S1, an air-cooled PEMFC voltage test platform was built to test and obtain the polarization curve of the PEMFC stack output voltage under different operating conditions; S3, according to the polarization curve measured by S2, using the nonlinear least square method, calibrating the unknown parameters in the dynamic semi-empirical model described in S1, comparing the simulation data generated by the dynamic semi-empirical model with the experimental data, and verifying the accuracy of the established dynamic semi-empirical model; S4, based on the dynamic semi-empirical model verified by S3, a PID controller is used to adjust the hydrogen flow rate input to the PEMFC stack and control the output voltage of the PEMFC stack to a given value; S5, using a fuzzy neural network to optimize the three control parameters in the PID controller described in S4, and obtain a fuzzy neural network PID controller for improving the output voltage control of the air-cooled PEMFC stack, the three control parameters are Kp, Ki and Kd.

2. The method for controlling the output voltage of a fuel cell based on a fuzzy neural network PID according to claim 1, characterized in that: In step S1, according to the electrochemical characteristics of the voltage loss of the proton exchange membrane fuel cell PEMFC, a dynamic semi-empirical model of the output voltage of the air-cooled PEMFC stack is built based on the MATLAB / Simulink simulation platform, including: Nernst voltage: The maximum electrical energy that a fuel cell can generate is related to the Gibbs free energy. The thermodynamic electromotive force of a fuel cell is expressed as: in is the thermodynamic electromotive force of the fuel cell under standard conditions, i.e., 25°C operating temperature and 1 atm pressure, ΔS is the entropy free energy change of the battery reaction, nF is the Faraday constant, T is the actual temperature of the battery during operation, is the standard reference temperature of the battery, R is the ideal gas constant; Activation loss: Activation loss represents the voltage sacrificed to overcome the activation energy barrier associated with the electrochemical reaction, which is caused by the slow electrochemical reaction kinetics and is expressed by the equation: Where: I is the battery current, i represents the actual working current, in represents the no-load current; ξ is the constant coefficient calculated from experimental data; C O2 It represents the oxygen concentration on the contact surface between the cathode membrane and oxygen, calculated using Henry's law; Ohmic losses: Ohmic losses are caused by the resistance of protons passing through the electrolyte and electrons flowing through fuel cell components such as bipolar plates and gas diffusion layers. They are calculated according to Ohm's law: Where: Rm represents the equivalent impedance of protons passing through the exchange membrane, and Rc represents the equivalent resistance of the fuel cell when electrons pass through the circuit. Since Rc is relatively fixed, it is treated as a constant. Concentration loss: When reactants undergo electrochemical reactions at the electrodes and are rapidly consumed, concentration polarization occurs, which produces a concentration gradient and affects the Nernst voltage and kinetic reaction rate of the single cell, resulting in a decrease in voltage. The calculation formula for concentration loss is: Where: b represents the coefficient of concentration loss, J represents the current density, and Jmax represents the limiting current density; Double-layer model: There is a "double-layer charge layer phenomenon" inside the PEMFC, forming a potential difference between the electrolyte and the electrode, which is replaced by an equivalent capacitor and described by the following equation: Where Vd is the voltage of the double layer, I is the current through the battery, C is the equivalent capacitance, and Rd is the equivalent resistance; The above establishes a complete mathematical model of PEMFC single-chip battery; In order to increase the total output voltage of the PEMFC stack system and meet the power requirements, multiple single cells are connected in series. The resulting PEMFC stack output voltage is expressed as: Where Vs is the total output voltage of the battery stack, Ns is the number of battery cells connected in series in the battery stack, and Vcell is the voltage of a single battery cell; Therefore, a complete semi-empirical model of PEMFC stack output voltage was established.

3. The fuel cell output voltage control method based on fuzzy neural network PID according to claim 2 is characterized in that: In step S2, an air-cooled PEMFC voltage test platform is constructed, including: Stack: Open cathode air-cooled PEMFC stack Fan + PWM speed signal regulation generator: Use PWM speed signal to control fan speed, and use fan speed to maintain stack temperature and oxygen supply; Gas supply subsystem: Provides hydrogen as a reactant for the experimental fuel cell, and consists of the following parts: Hydrogen tanks - provide 99.99% pure hydrogen to the stack; Pressure reducing valve; Flow meter: connected to the pipeline before the anode hydrogen inlet to measure the flow rate and volume of hydrogen; Pressure sensor: connected to the pipeline in front of the anode hydrogen inlet to measure the hydrogen pressure at the anode inlet; Anode purification subsystem: manages the emission of anode gas and has the following functions: When opened, the gas and water in the anode flow channel are discharged; When closed, the anode reaches dead-end mode and maintains a certain pressure; The experiment used constant opening and closing times of 12.5s:0.5s throughout; Data acquisition subsystem: responsible for collecting data such as battery stack voltage and temperature, consisting of a data collector and a test system; Output subsystem: used to record the output current and dissipate the output power of the battery stack. It consists of an ammeter and an electronic load: Ammeter: record output current; Electronic load: used to dissipate the output power of the battery stack.

4. The method for controlling the output voltage of a fuel cell based on a fuzzy neural network PID according to claim 3, characterized in that: In step S3, the unknown parameters of the PEMFC stack output voltage model are calibrated, including: Firstly, the experimentally measured polarization curve is used as reference data, and the nonlinear least squares method is used to fit the unknown parameter values ​​of the semi-empirical model to form an accurate dynamic semi-empirical model of the PEMFC stack output voltage. Then, the PEMFC stack output voltage model is run to obtain the voltage output of the PEMFC stack output voltage model under different input conditions, and the voltage output of the real stack under the corresponding conditions is compared. The mean square error (MSE) is used as the evaluation standard to verify the accuracy of the model. The formula of MSE is: Where: and Represent the corresponding voltage values ​​of simulation and experiment respectively, and n represents the amount of data.

5. The method for controlling the output voltage of a fuel cell based on a fuzzy neural network PID according to claim 4, characterized in that: In step S4, based on the dynamic semi-empirical model verified in S3, a PID controller is used to adjust the hydrogen flow rate input to the PEMFC stack to control the output voltage of the PEMFC stack to a given value, including: The output voltage of PEMFC is stabilized by controlling the anode hydrogen flow rate; the hydrogen pressure is determined by the inflow, consumption and outflow of hydrogen inside the fuel cell; the following equation is obtained based on the gas state equation and the law of conservation of mass: Where: , , Respectively represent the inflow, outflow and consumption of hydrogen, the unit is mol / l, Represents the total volume of the anode flow channel, taking 0.005m 3 ; The PID controller is used to control the output voltage of the PEMFC system to be stable by changing the hydrogen flow rate. The PID controller consists of three basic parts: proportional P, integral I and differential D. The PID controller can generate the corresponding control quantity according to the error. The PID control algorithm is expressed as: Where: Kp is the proportional gain, Ki is the integral gain, Kd is the differential gain; u(t) represents the control action in the time domain; The error e(t)=Vs(t)-V(t) between the expected voltage value Vs(t) and the actual voltage value V(t) is selected as the input of the PID voltage controller. Through proportional, integral and differential operations, the output control action u(k) is obtained as the hydrogen volume flow rate; The error e(t)=Vs(t)-V(t) between the expected voltage value Vs(t) and the actual voltage value V(t) is selected as the input of the PID voltage controller. Through proportional, integral and differential operations, the output control action u(t) is obtained as the hydrogen volume flow rate.

6. The method for controlling the output voltage of a fuel cell based on a fuzzy neural network PID according to claim 5, characterized in that: In step S5, a fuzzy neural network is used to optimize three control parameters in the PID controller to obtain a fuzzy neural network PID controller for controlling the output voltage of the air-cooled PEMFC stack, including: Firstly, the PID is optimized by using a fuzzy controller, and then the fuzzy rules of the fuzzy controller are optimized by using a fuzzy neural network. The voltage control effects of the traditional PID controller, the fuzzy controller, and the fuzzy neural network PID controller are compared. Optimize PID through fuzzy control, realize the real-time adjustment function of the important influencing parameters Kp, Ki, Kd of the PID controller, and obtain better control effect; make control decisions by observing the size of the error E and the error change rate EC, combining experience and technical knowledge; then realize the real-time adjustment of the important influencing parameters Kp, Ki, Kd of the PID controller through the computer; Neural network is used as a mathematical tool, and the input and output of the neural network are used to represent the input and output of the fuzzy system. The membership function and fuzzy rules of the fuzzy system are added to the implicit nodes of the neural network to optimize the membership function and fuzzy rules of the fuzzy controller.

7. A fuel cell output voltage control system based on fuzzy neural network PID, characterized in that: include: Dynamic semi-empirical model building module, based on the MATLAB / Simulink simulation platform, builds a dynamic semi-empirical model of the output voltage of the air-cooled PEMFC stack according to the electrochemical characteristics of the voltage loss of the proton exchange membrane fuel cell PEMFC; The polarization curve measurement module is used to optimize the unknown parameters in the dynamic semi-empirical model described in the dynamic semi-empirical model building module, build an air-cooled PEMFC voltage test platform, and test and obtain the polarization curve of the PEMFC stack output voltage under different operating conditions; The model verification module uses the nonlinear least square method to calibrate the unknown parameters in the dynamic semi-empirical model in the dynamic semi-empirical model building module according to the polarization curve measured by the polarization curve measurement module, compares the simulation data generated by the dynamic semi-empirical model with the experimental data, and verifies the accuracy of the established dynamic semi-empirical model; The parameter adjustment module uses a PID controller to adjust the hydrogen flow rate input to the PEMFC stack based on the dynamic semi-empirical model verified by the model verification module, and controls the output voltage of the PEMFC stack to a given value; The output voltage control module uses a fuzzy neural network to optimize the three control parameters in the PID controller in the parameter adjustment module to obtain a fuzzy neural network PID controller, which is used to improve the output voltage control of the air-cooled PEMFC stack. The three control parameters are Kp, Ki and Kd.

8. The fuel cell output voltage control system based on fuzzy neural network PID according to claim 7, characterized in that: In the dynamic semi-empirical model building module, according to the electrochemical characteristics of the voltage loss of the proton exchange membrane fuel cell PEMFC, based on the MATLAB / Simulink simulation platform, a dynamic semi-empirical model of the output voltage of the air-cooled PEMFC stack is built, including: Nernst voltage: The maximum electrical energy that a fuel cell can generate is related to the Gibbs free energy. The thermodynamic electromotive force of a fuel cell is expressed as: in is the thermodynamic electromotive force of the fuel cell under standard conditions, i.e., 25°C operating temperature and 1 atm pressure, ΔS is the entropy free energy change of the battery reaction, nF is the Faraday constant, T is the actual temperature of the battery during operation, is the standard reference temperature of the battery, R is the ideal gas constant; Activation loss: Activation loss represents the voltage sacrificed to overcome the activation energy barrier associated with the electrochemical reaction, which is caused by the slow electrochemical reaction kinetics and is expressed by the equation: Where: I is the battery current, i represents the actual working current, in represents the no-load current; ξ is the constant coefficient calculated from experimental data; C O2 It represents the oxygen concentration on the contact surface between the cathode membrane and oxygen, calculated using Henry's law; Ohmic losses: Ohmic losses are caused by the resistance of protons passing through the electrolyte and electrons flowing through fuel cell components such as bipolar plates and gas diffusion layers. They are calculated according to Ohm's law: Where: Rm represents the equivalent impedance of protons passing through the exchange membrane, and Rc represents the equivalent resistance of the fuel cell when electrons pass through the circuit. Since Rc is relatively fixed, it is treated as a constant. Concentration loss: When reactants undergo electrochemical reactions at the electrodes and are rapidly consumed, concentration polarization occurs, which produces a concentration gradient and affects the Nernst voltage and kinetic reaction rate of the single cell, resulting in a decrease in voltage. The calculation formula for concentration loss is: Where: b represents the coefficient of concentration loss, J represents the current density, and Jmax represents the limiting current density; Double-layer model: There is a "double-layer charge layer phenomenon" inside the PEMFC, forming a potential difference between the electrolyte and the electrode, which is replaced by an equivalent capacitor and described by the following equation: Where Vd is the voltage of the double layer, I is the current through the battery, C is the equivalent capacitance, and Rd is the equivalent resistance; The above establishes a complete mathematical model of PEMFC single-chip battery; In order to increase the total output voltage of the PEMFC stack system and meet the power requirements, multiple single cells are connected in series. The resulting PEMFC stack output voltage is expressed as: Where Vs is the total output voltage of the battery stack, Ns is the number of battery cells connected in series in the battery stack, and Vcell is the voltage of a single battery cell; Therefore, a complete semi-empirical model of PEMFC stack output voltage was established.

9. The fuel cell output voltage control system based on fuzzy neural network PID according to claim 8, characterized in that: In the polarization curve measurement module, an air-cooled PEMFC voltage test platform is built, including: Stack: Open cathode air-cooled PEMFC stack Fan + PWM speed signal regulation generator: Use PWM speed signal to control fan speed, and use fan speed to maintain stack temperature and oxygen supply; Gas supply subsystem: Provides hydrogen as a reactant for the experimental fuel cell, and consists of the following parts: Hydrogen tanks - provide 99.99% pure hydrogen to the stack; Pressure reducing valve; Flow meter: connected to the pipeline before the anode hydrogen inlet to measure the flow rate and volume of hydrogen; Pressure sensor: connected to the pipeline in front of the anode hydrogen inlet to measure the hydrogen pressure at the anode inlet; Anode purification subsystem: manages the emission of anode gas and has the following functions: When opened, the gas and water in the anode flow channel are discharged; When closed, the anode reaches dead-end mode and maintains a certain pressure; The experiment used constant opening and closing times of 12.5s:0.5s throughout; Data acquisition subsystem: responsible for collecting data such as battery stack voltage and temperature, consisting of a data collector and a test system; Output subsystem: used to record the output current and dissipate the output power of the battery stack. It consists of an ammeter and an electronic load: Ammeter: record output current; Electronic load: used to dissipate the output power of the battery stack.

10. The fuel cell output voltage control system based on fuzzy neural network PID according to claim 9, characterized in that: In the model verification module, the unknown parameters of the PEMFC stack output voltage model are calibrated, including: Firstly, the experimentally measured polarization curve is used as reference data, and the nonlinear least squares method is used to fit the unknown parameter values ​​of the semi-empirical model to form an accurate dynamic semi-empirical model of the PEMFC stack output voltage. Then, the PEMFC stack output voltage model is run to obtain the voltage output of the PEMFC stack output voltage model under different input conditions, and the voltage output of the real stack under the corresponding conditions is compared. The mean square error (MSE) is used as the evaluation standard to verify the accuracy of the model. The formula of MSE is: Where: and Represent the corresponding voltage values ​​of simulation and experiment respectively, and n represents the amount of data.

Images