Performance prediction and parameter optimization method and system for proton exchange membrane fuel cell

Abstract

The invention discloses a proton exchange membrane fuel cell performance prediction and parameter optimization method and system. The method comprises the following steps: selecting characteristic parameters influencing the output performance of the PEMFC as input parameters according to experience, and selecting parameters capable of representing the output performance of the PEMFC as output parameters; obtaining multiple groups of input parameters and output parameters as training samples, and constructing a data set; establishing a PEMFC performance prediction model based on an RBF neural network, and optimizing a network structure by using an improved whale algorithm to obtain a PEMFC performance prediction model for predicting battery performance; and searching a characteristic parameter combination when the output performance of the PEMFC is optimal by using the improved whale algorithm in combination with the optimized PEMFC performance prediction model. According to the invention, under limited actual measurement data, accurate prediction of battery performance and optimization of key input parameters can be realized, and unnecessary test experiments are reduced.

Description

Technical Field

[0001] The present invention relates to the field of fuel cells, and more specifically, to a method, system, device and medium for performance prediction and parameter optimization of proton exchange membrane fuel cells based on an improved whale algorithm and RBF neural network. Background Art

[0002] In recent years, proton exchange membrane fuel cells (PEMFCs), as an important component of fuel cells, have shown broad application prospects in the automotive field and portable electronic devices due to their advantages such as high specific power, fast startup, and good environmental characteristics. PEMFC is a complex dynamic system involving the coupling of multiple disciplines such as electrochemistry, fluid mechanics, and thermodynamics. Its output performance depends on multiple input parameters such as operating pressure, operating temperature, intake humidity, stoichiometric ratio, and bipolar plate structure size. However, obtaining the various input parameters for optimal PEMFC output performance through experimental testing is costly and time-consuming. If a multi-input parameter PEMFC performance prediction model can be established before conducting the necessary experiments, it will not only provide a deep understanding of the PEMFC operating mechanism and obtain the optimal combination of input parameters corresponding to the optimal output performance, but also significantly reduce the cost of development and experimental testing, effectively improving development efficiency.

[0003] In the relevant patent schemes of existing PEMFC output performance prediction methods (such as China Invention Publication Patent: CN104133369A), artificial neural networks are generally used as the prediction model of its output performance, and characteristic parameters that affect the output performance of PEMFC are selected. The PEMFC power output under different characteristic parameter values is tested to obtain training samples, thereby achieving the purpose of training the artificial neural network. However, due to the uncertainty in the establishment of its network structure, the traditional artificial neural network model has large fluctuations in prediction accuracy. In addition, a large amount of experimental data is also required as support during the training process. Therefore, it is particularly important to develop a model that can accurately predict the output performance of PEMFC with the support of limited experimental data. Summary of the Invention

[0004] To address at least one of the issues with the prior art, the present invention provides a method for predicting and optimizing proton exchange membrane fuel cell (PEMFC) performance based on an improved whale algorithm and RBF neural networks. This method constructs an RBF neural network model as a predictive model for PEMFC output performance and optimizes its network structure using the improved whale algorithm, resulting in a model with higher prediction accuracy. Finally, the improved whale algorithm, combined with the optimized prediction model, identifies the characteristic parameter combination that optimizes PEMFC output performance, thereby overcoming the shortcomings of the prior art.

[0005] To achieve the purpose of the present invention, the present invention provides a proton exchange membrane fuel cell performance prediction and parameter optimization method based on an improved whale algorithm and RBF neural network, comprising the following steps:

[0006] Based on experience, characteristic parameters that affect the output performance of PEMFC are selected as input parameters, and parameters that can represent the output performance of PEMFC are selected as output parameters;

[0007] Obtain multiple sets of input parameters and output parameters as training samples and construct a data set;

[0008] A PEMFC performance prediction model based on RBF neural network was established, and the improved whale algorithm was used to optimize the network structure to obtain a PEMFC performance prediction model for predicting battery performance.

[0009] The improved whale algorithm is combined with the optimized PEMFC performance prediction model to find the characteristic parameter combination for the best PEMFC output performance.

[0010] Furthermore, the method includes the step of verifying the reliability of the optimal parameter combination and the accuracy of the optimized RBF neural network prediction model. The reliability of the optimization can be verified by comparing the optimal parameter combination with commonly used experimental parameters. The accuracy of the optimization can be verified by comparing the prediction accuracy of the RBF neural network before and after optimization.

[0011] Furthermore, the characteristic parameters affecting PEMFC output performance include at least the cell operating temperature, operating pressure, cathode and anode air humidity, and the cathode and anode stoichiometric ratio. Parameters representative of PEMFC output performance are selected as output parameters, including but not limited to peak power density, output voltage, and system efficiency.

[0012] Furthermore, through a combination of experimental testing and simulation, multiple sets of input and output data are obtained as a dataset. All data in the dataset is then preprocessed, converting each input variable to a uniform scale, thereby achieving data standardization. Preferably, the dataset is randomly divided into a training set and a test set in a 2:1 ratio.

[0013] The RBF neural network prediction model is trained with a data set obtained by combining experimental testing with simulation, and its network structure is optimized using the improved whale algorithm to improve its prediction accuracy.

[0014] Furthermore, the method for preprocessing the data is the Z-score method, and its mathematical expression is:

[0015]

[0016] in, is the average value of the feature vector in the data, X std is the standard deviation of the feature vector in the data, Y std is the target standard deviation after data processing, is the target average value after data processing.

[0017] Furthermore, the network structure parameters of the RBF neural network prediction model include the number of hidden layer nodes, the center and width of hidden layer neurons, the radial basis function expansion speed, the output layer weight, etc., which can all be optimized by improving the whale algorithm.

[0018] Furthermore, the optimization target of the improved whale algorithm is the characteristic parameters that affect the output performance of the PEMFC, and the objective function is a parameter that can represent the output performance of the PEMFC. The value range of each characteristic parameter can be determined based on experiments or experience.

[0019] Furthermore, compared with the traditional whale algorithm, the improved whale algorithm makes the following three improvements:

[0020] Improvement 1: The improved whale algorithm uses Circle Chaos Map as the method for generating the initial population. This method has strong randomness and will better travel the entire search range, making the initial population more evenly distributed, thereby effectively increasing the diversity of the population. The specific mathematical expression is as follows:

[0021]

[0022] Among them, mod is the remainder function.

[0023] Improvement 2: The improved whale algorithm solves the problem of incomplete global development in the early stage and slow convergence in the later stage through a nonlinear dynamic convergence factor. Its mathematical expression is as follows:

[0024]

[0025] Among them, t is the current iteration number, T max is the maximum number of iterations.

[0026] Improvement 3: The improved whale algorithm improves the algorithm's optimization ability through adaptive inertia weight. The weight value can be adaptively reduced as the number of iterations gradually increases. The specific mathematical expression is as follows:

[0027]

[0028] Where w is the adaptive inertia weight.

[0029] The present invention also provides a proton exchange membrane fuel cell performance prediction and parameter optimization system.

[0030] The present invention also provides a computer device.

[0031] The present invention also provides a computer-readable storage medium.

[0032] The advantages of the present invention compared with the prior art are:

[0033] 1. This paper addresses the problems of uneven population distribution and slow convergence speed in the later stage of the whale algorithm during its optimization and solution process. Based on three improvement strategies, an improved whale algorithm is proposed, which improves the optimization ability and convergence speed of the whale algorithm to a certain extent.

[0034] 2. The present invention adopts an improved whale algorithm to optimize the parameters of the RBF neural network model, such as the number of hidden layer nodes, the center and width of hidden layer neurons, the radial basis function expansion speed, and the output layer weight, which can greatly improve the output performance prediction accuracy.

[0035] 3. The improved whale algorithm of this invention uses Circle Chaos Mapping as the method for generating the initial population. This method has strong randomness and can better traverse the entire search range, making the initial population more evenly distributed, thereby effectively increasing the diversity of the population. Therefore, the improved prediction model can ensure the quality of the initial population using less experimental data as samples. The increased population diversity can reduce the demand for basic training sets. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a flowchart of a method for performance prediction and parameter optimization of a proton exchange membrane fuel cell based on an improved whale algorithm and RBF neural network in an embodiment of the present invention.

[0037] Figure 2 Schematic diagram of the correlation between the actual value of the RBF neural network prediction model and the model predicted value in the prior art.

[0038] Figure 3 Schematic diagram of the correlation between the actual value of the optimized RBF neural network prediction model and the model predicted value in the embodiment of the invention. DETAILED DESCRIPTION

[0039] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention, in conjunction with the accompanying drawings. It should be understood that the described embodiment is merely one specific embodiment of the present invention, and not all embodiments. The following description of at least one exemplary embodiment is merely illustrative and in no way limits the present invention, its application, or use. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are intended to be within the scope of protection of the present invention. Unless otherwise specified, the relative arrangement of components and steps, expressions, and numerical values described in these embodiments do not limit the scope of the present invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to scale. Technologies, methods, and devices known to persons of ordinary skill in the relevant art may not be discussed in detail, but, where appropriate, such technologies, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values should be interpreted as merely illustrative and not limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0040] The following describes examples of the present invention in detail.

[0041] An embodiment of the present invention provides a method for predicting the performance and optimizing the parameters of a proton exchange membrane fuel cell based on an improved whale algorithm and an RBF neural network, comprising the following steps:

[0042] Step 1: Select characteristic parameters that affect the output performance of PEMFC as input parameters, and select characteristic parameters that can represent the output performance of PEMFC as output parameters.

[0043] The characteristic parameters affecting PEMFC output performance include, but are not limited to, operating temperature, battery operating temperature, operating pressure, cathode and anode inlet air humidity, anode inlet air humidity, cathode stoichiometric ratio, and anode stoichiometric ratio. Parameters representative of PEMFC output performance include, but are not limited to, current density, battery peak power density, output voltage, and system efficiency. In one embodiment of the present invention, based on empirical experience, six factors—operating pressure, operating temperature, cathode inlet air humidity, anode inlet air humidity, cathode stoichiometric ratio, and anode stoichiometric ratio—are used as input parameters. The PEMFC current density at 0.4V is then used as the output parameter to construct a "multi-input, single-output" RBF neural network model.

[0044] Step 2: By combining experimental testing with simulation, multiple sets of input and output data are obtained as training samples.

[0045] In one embodiment of the present invention, training samples are accumulated, and 48 sets of data sets containing input data and output data are obtained through a combination of experimental testing and simulation. The 48 sets of data composed of different operating parameters are randomly divided into training sets and test sets in a ratio of 2:1, and the RBF neural network model is trained.

[0046] The simulation includes establishing a three-dimensional model of the direct current PEMFC, setting up the physical field, meshing, mesh independence verification and model accuracy verification.

[0047] Before training the RBF neural network model, all data in the dataset need to be preprocessed to convert each input parameter into a uniform scale range. The specific mathematical expression is as follows:

[0048]

[0049] in, is the average value of the feature vector in the data, X std is the standard deviation of the feature vector in the data, Y std is the target standard deviation after data processing, is the target average value after data processing, X i 、x i They refer to the input parameters before and after conversion to a uniform scale.

[0050] Step 3: Establish a PEMFC performance prediction model based on RBF neural network, and use the improved whale algorithm to optimize the network structure of the PEMFC performance prediction model based on RBF neural network, so as to obtain a PEMFC performance prediction model with higher prediction accuracy.

[0051] The network structure parameters of the PEMFC performance prediction model based on RBF neural network include the number of hidden layer nodes, the center and width of hidden layer neurons, the radial basis function expansion speed, the output layer weight, etc., which can all be optimized by improving the whale algorithm.

[0052] The objective function for optimizing the PEMFC performance prediction model based on the RBF neural network using the improved whale algorithm is set to the sum of the mean square error of the training set and the test set. The specific mathematical expression is as follows:

[0053] f=E1+E2

[0054]

[0055] Among them, E1 is the mean square error of the training set, E2 is the mean square error of the test set, n1 is the number of data groups in the training set, n2 is the number of data groups in the test set, and y′ iRepresents the average value of the data in the data group, y i Represents the i-th data in the data group, and n represents the number of data in the cumulative calculation.

[0056] The smaller the objective function value f is, the smaller the error of the RBF neural network prediction model under this network structure is, and the higher the prediction accuracy is.

[0057] In one embodiment of the present invention, the prediction accuracy of the PEMFC performance prediction model based on the RBF neural network can be calculated using the determination coefficient R 2 Evaluation, the calculation formula is as follows:

[0058]

[0059] Where n is the number of data sets, The average value of the actual output value. 2 The closer it is to 1, the higher the prediction accuracy of the prediction model.

[0060] The improved whale algorithm can effectively solve the problems of the traditional whale algorithm in dealing with more complex problems, such as uneven initial population distribution and slow convergence in the later stage. Compared with the traditional whale algorithm, it makes the following three improvements:

[0061] Improvement 1: The improved whale algorithm uses Circle Chaos Map as the method for generating the initial population. This method has strong randomness and will better travel the entire search range, making the initial population more evenly distributed, thereby effectively increasing the diversity of the population. The specific mathematical expression is as follows:

[0062]

[0063] Among them, mod is the remainder function, X k+1 、X k are the initial populations of the algorithm before and after using the Circle chaos mapping method, and k represents the number of the data in the training set.

[0064] Improvement 2: The improved whale algorithm solves the problem of incomplete global development in the early stage and slow convergence in the later stage through a nonlinear dynamic convergence factor. Its mathematical expression is as follows:

[0065]

[0066] Among them, a(t) is the convergence factor, t is the current number of iterations, T max is the maximum number of iterations, and T is the current number of iterations.

[0067] Improvement 3: The improved whale algorithm improves the algorithm's optimization ability through adaptive inertia weight. The weight value can be adaptively reduced as the number of iterations gradually increases. The specific mathematical expression is as follows:

[0068]

[0069] Where w is the adaptive inertia weight.

[0070] An improved whale algorithm is used to optimize the number of hidden layer nodes and the RBF expansion rate of a PEMFC performance prediction model based on an RBF neural network, thereby finding the optimal network structure. In one embodiment of the present invention, the number of hidden layer nodes ranges from 1 to 20, and the RBF expansion rate ranges from 0 to 10. After optimization, the optimal network structure of the RBF neural network prediction model is determined, with the number of hidden layer nodes being 10 and the RBF expansion rate being 2.

[0071] In one embodiment of the present invention, the PEMFC performance prediction model based on the RBF neural network before and after improvement is trained using training samples, and the training results are as follows: Figure 2 、 Figure 3 As shown. The traditional RBF neural network R 2 is 0.93542, while the optimized PEMFC performance prediction model based on RBF neural network in the embodiment of the present invention is R 2 As high as 0.99229, the prediction accuracy has been greatly improved compared with the traditional RBF neural network, by about 6.1%.

[0072] Step 4: Use the improved whale algorithm combined with the optimized PEMFC performance prediction model to find the characteristic parameter combination with the best PEMFC output performance.

[0073] In one embodiment of the present invention, based on experience, compared with the effects of operating pressure, operating temperature, cathode intake humidity, and cathode stoichiometric number on PEMFC performance, the effects of anode intake humidity and anode stoichiometric number are smaller. In order to improve the accuracy and efficiency of optimization, four parameters, namely operating pressure, operating temperature, cathode intake humidity, and cathode stoichiometric number, are selected as optimization variables, and the current density of PEMFC at 0.4V is selected as the optimization target.

[0074] The value ranges of the optimization variables are: operating pressure 0.1MPa-0.4MPa, operating temperature 280K-360K, cathode inlet air humidity 0-100%, and cathode stoichiometric coefficient 1-4.

[0075] Using the optimized RBF neural network model as the prediction model and the improved whale algorithm as the optimization model, the aforementioned optimization variables were optimized to determine the operating parameter combination that optimizes PEMFC output performance. The optimal parameter combination was: operating pressure of 0.386 MPa, operating temperature of 350.17 K, cathode inlet humidity of 87.49%, and cathode stoichiometric coefficient of 3.58.

[0076] Step 5: Verify the reliability of the optimal parameter combination and the accuracy of the optimized RBF neural network-based prediction model.

[0077] Comparing the best parameter combination with the commonly used experimental parameters can verify the reliability of the optimization. Figure 2 and Figure 3 It is verified that the PEMFC performance prediction model based on RBF neural network provided by the embodiment of the present invention has good prediction accuracy.

[0078] In one embodiment of the present invention, a proton exchange membrane fuel cell performance prediction and parameter optimization system is provided for implementing the method provided in the above embodiment. The system includes the following modules:

[0079] Based on experience, characteristic parameters that affect the output performance of PEMFC are selected as input parameters, and parameters that can represent the output performance of PEMFC are selected as output parameters;

[0080] A data set construction module is used to obtain multiple sets of input parameters and output parameters as training samples and construct a data set;

[0081] The model optimization module is used to establish a PEMFC performance prediction model based on the RBF neural network and optimize the network structure using the improved whale algorithm to obtain a PEMFC performance prediction model for predicting battery performance;

[0082] The parameter combination optimization module is used to use the improved whale algorithm combined with the optimized PEMFC performance prediction model to find the characteristic parameter combination with the best PEMFC output performance.

[0083] In one embodiment of the present invention, a computer device is further provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method provided in the aforementioned embodiment when executing the computer program.

[0084] In one embodiment of the present invention, a computer-readable storage medium is further provided. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the method provided in the above embodiment is implemented.

[0085] The scope of protection of the present invention is not limited to the above-described embodiments. Those skilled in the art and researchers may make various modifications to the present invention without departing from the scope of the present invention. If such modifications fall within the scope of the claims of the present invention and their equivalents, the present invention is intended to include such modifications.

Claims

1. A method for performance prediction and parameter optimization of proton exchange membrane fuel cells, characterized in that: The following steps are involved: Based on experience, characteristic parameters that affect the output performance of PEMFC are selected as input parameters, and parameters that can represent the output performance of PEMFC are selected as output parameters; Obtain multiple sets of input parameters and output parameters as training samples and construct a data set; A PEMFC performance prediction model based on RBF neural network was established, and the improved whale algorithm was used to optimize the network structure to obtain a PEMFC performance prediction model with higher prediction accuracy. The improved whale algorithm is combined with the optimized PEMFC performance prediction model to find the characteristic parameter combination for the best PEMFC output performance.

2. The method for performance prediction and parameter optimization of a proton exchange membrane fuel cell according to claim 1, characterized in that: The characteristic parameters that affect the output performance of PEMFC include but are not limited to battery operating temperature, operating pressure, cathode inlet humidity, anode inlet humidity, cathode stoichiometric ratio, and anode stoichiometric ratio; the parameters that can represent the output performance of PEMFC include but are not limited to current density, battery peak power density, output voltage, and system efficiency.

3. The method for performance prediction and parameter optimization of a proton exchange membrane fuel cell according to claim 1, characterized in that: Before training the PEMFC performance prediction model based on RBF neural network, all data in the dataset are preprocessed to convert each input parameter into a unified scale range.

4. The method for performance prediction and parameter optimization of a proton exchange membrane fuel cell according to claim 1, characterized in that: The parameters of the network structure of the PEMFC performance prediction model based on the RBF neural network include the number of nodes in the hidden layer, the center and width of the hidden layer neurons, the radial basis function expansion speed and the output layer weight, which are optimized by the improved whale algorithm.

5. The method for performance prediction and parameter optimization of a proton exchange membrane fuel cell according to claim 1, characterized in that: The objective function when optimizing the PEMFC performance prediction model based on the RBF neural network using the improved whale algorithm is set to the sum of the mean square errors of the training set and the test set. The mathematical expression is as follows: f=E1+E2 Among them, E1 is the mean square error of the training set, E2 is the mean square error of the test set, n1 is the number of data groups in the training set, n2 is the number of data groups in the test set, and y i Represents the average value of the data in the data group, y i Represents the i-th data in the data group, and n represents the number of data in the cumulative calculation.

6. A method for performance prediction and parameter optimization of a proton exchange membrane fuel cell according to any one of claims 1 to 5, characterized in that: The improvements of the improved whale algorithm include: using Circle chaos mapping to generate the initial population, using a nonlinear dynamic convergence factor, and improving the algorithm's optimization ability through adaptive inertia weights.

7. The method for performance prediction and parameter optimization of a proton exchange membrane fuel cell according to claim 6, characterized in that: The mathematical expression for generating the initial population using Circle chaos map is: Among them, mod is the remainder function, X k+1 、X k are the initial populations of the algorithm before and after the Circle chaos mapping method is adopted, and k represents the number of the data in the training set; The mathematical expression of the nonlinear dynamic convergence factor is: Among them, t is the current iteration number, T max is the maximum number of iterations; The mathematical expression for improving the optimization ability of the algorithm through adaptive inertia weight is: Where w is the adaptive inertia weight.

8. A proton exchange membrane fuel cell performance prediction and parameter optimization system, characterized in that: For implementing the method according to any one of claims 1 to 7, the system comprises the following modules: Based on experience, characteristic parameters that affect the output performance of PEMFC are selected as input parameters, and parameters that can represent the output performance of PEMFC are selected as output parameters; A data set construction module is used to obtain multiple sets of input parameters and output parameters as training samples and construct a data set; Model optimization module, which is used to establish a PEMFC performance prediction model based on RBF neural network and optimize the network structure using the improved whale algorithm to obtain a PEMFC performance prediction model with higher prediction accuracy; The parameter combination optimization module is used to use the improved whale algorithm combined with the optimized PEMFC performance prediction model to find the characteristic parameter combination with the best PEMFC output performance.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images