Method and system for evaluating and predicting working parameter performance of electrolytic cell

Abstract

The invention provides an electrolytic cell working parameter performance evaluation and prediction method and system, and the method comprises the steps: taking an external operation parameter and an internal structure parameter of an electrolytic cell as input variables; based on the input variables and the variation range of each input variable, determining performance evaluation indexes of the electrolytic cell, and taking the performance evaluation indexes as output variables; obtaining multiple groups of data sets based on the input variables and the output variables, and constructing a BP neural network prediction model based on the multiple groups of data sets; optimizing structural parameters of the BP neural network prediction model by using a genetic algorithm to obtain an optimized BP neural network prediction model; the current external operation parameters and the internal structure parameters of the electrolytic cell are obtained, prediction is carried out according to the current external operation parameters and the internal structure parameters of the electrolytic cell and the optimized BP neural network prediction model, the performance prediction result of the electrolytic cell is obtained, and the reliability of PEMEC working parameter performance evaluation prediction is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrogen production by electrolysis of water, and in particular to a method and system for evaluating and predicting the performance of working parameters of an electrolyzer. Background Art

[0002] Hydrogen energy is considered to be a renewable energy with development prospects. With the growth of the world's population and the advancement of science and technology, the demand for hydrogen energy is also rising rapidly. To develop hydrogen energy, we must start with a complete electric hydrogen system and hydrogen production technology. Proton exchange membrane electrolyzer (PEMEC) is a new type of hydrogen production device that converts surplus renewable electricity into hydrogen energy. Compared with traditional hydrogen production methods (hydrogen production from fossil chemical raw materials, coal, oil, and natural gas reforming), PEMEC is more environmentally friendly and efficient. However, due to its high cost and low durability, PEMEC has not yet been developed and utilized on a large scale.

[0003] At present, domestic proton exchange membrane electrolyzers mostly use polarization curves (current density-voltage) to evaluate the performance impact of changes in PEMEC working parameters, and lack research on other performance aspects (such as temperature, gas production, etc.); moreover, the type of working parameters that affect the performance of the PEMEC to be evaluated is single, and there is a lack of application of changes in multiple working parameters, resulting in low reliability in the prediction of the performance evaluation of PEMEC working parameters.

[0004] In view of this problem, the present invention provides a method and system for evaluating and predicting the performance of electrolytic cell operating parameters to solve the above problem. Summary of the invention

[0005] In order to solve the problems existing in the prior art, the present invention innovatively proposes a method and system for evaluating and predicting the performance of the working parameters of an electrolytic cell, which effectively solves the problem that the reliability of the performance evaluation and prediction of the working parameters of PEMEC is low due to the single type of working parameters that affect the performance of the PEMEC to be evaluated in the prior art, and effectively improves the reliability of the performance evaluation and prediction of the working parameters of PEMEC.

[0006] A first aspect of the present invention provides a method for evaluating and predicting the performance of electrolytic cell operating parameters, comprising:

[0007] Taking the external operating parameters and internal structural parameters of the electrolyzer as input variables; determining the performance evaluation index of the electrolyzer based on the input variables and the variation range of each input variable, and taking the performance evaluation index as the output variable; wherein the electrolyzer is a proton exchange membrane electrolyzer;

[0008] Acquire multiple sets of data sets based on input variables and output variables, and build a BP neural network prediction model based on the multiple sets of data sets;

[0009] The genetic algorithm is used to optimize the structural parameters of the BP neural network prediction model to obtain the optimized BP neural network prediction model;

[0010] The current external operating parameters and internal structural parameters of the electrolytic cell are obtained, and prediction is performed based on the current external operating parameters, internal structural parameters and optimized BP neural network prediction model to obtain the performance prediction result of the electrolytic cell.

[0011] Optionally, the external operating parameters of the electrolytic cell include the temperature of water during the operation of passing deionized water into the anode flow channel of the electrolytic cell, the flow rate of water during the operation of passing deionized water into the anode flow channel of the electrolytic cell, and the voltage between two poles of the electrolytic cell respectively connected to a power supply; the internal structural parameters include the membrane thickness of the electrolytic cell membrane electrode assembly, the effective porosity of the anode diffusion layer of the electrolytic cell, the thickness of the anode diffusion layer of the electrolytic cell, and the conductivity of the anode catalyst layer of the electrolytic cell.

[0012] Furthermore, the performance evaluation indexes of the electrolytic cell include the average temperature of the anode catalyst layer of the electrolytic cell, the current density of the anode catalyst layer of the electrolytic cell, and the average hydrogen mole fraction of the cathode diffusion layer and the catalyst layer of the electrolytic cell.

[0013] Optionally, the temperature of the water during the operation of the electrolytic cell anode channel is in the range of 20 to 80°C, the flow rate of the water during the operation of the electrolytic cell anode channel is in the range of 0.00015 to 0.0007 kg / s, the voltage between the two electrodes of the electrolytic cell respectively connected to the power supply is in the range of 1.8 to 2.5 V; the membrane thickness of the electrolytic cell membrane electrode assembly is in the range of Wherein, h is the initial value of the membrane thickness of the electrolytic cell membrane electrode assembly; the effective porosity of the electrolytic cell anode diffusion layer varies in the range of 0.2 to 0.8; the thickness of the electrolytic cell anode diffusion layer varies in the range of Wherein, l is the initial value of the thickness of the anode diffusion layer of the electrolytic cell; the conductivity of the anode catalyst layer of the electrolytic cell varies in the range of 0.1 to 1000 S / m.

[0014] Optionally, obtaining multiple sets of data sets based on the input variables and the output variables, and constructing a BP neural network prediction model based on the multiple sets of data sets specifically includes:

[0015] For the electrolytic cell to be predicted, the parameter values ​​of the input variables are changed equidistantly within the variation range of each input variable, and only one parameter value of an input variable is changed each time, and a set of output variable data is correspondingly generated through experimental or simulation operations; multiple parameter values ​​of each input variable are cross-changed, thereby correspondingly generating multiple sets of output data to form a data set;

[0016] The data set is divided into a training set and a test set according to a preset division ratio, wherein the training set is used as a sample for predictive modeling training of the BP neural network prediction model, and the test set is used as a sample for predictive model testing of the BP neural network prediction model.

[0017] Optionally, the structural parameters of the BP neural network prediction model include the number of input layer nodes, the number of output layer nodes, the number of hidden layer nodes, the link weight from the input layer to the hidden layer, the hidden layer weight, the link weight from the hidden layer to the output layer, the output layer weight, the link bias between the input layer and the hidden layer, the hidden layer bias, the link bias between the hidden layer and the output layer, and the output layer bias.

[0018] Furthermore, the number of input layer nodes is the number of input variables, the number of output layer nodes is the number of output variables, and the number of hidden layer nodes is determined as follows:

[0019]

[0020] Among them, P is the number of hidden layer nodes, m is the number of input layer nodes, n is the number of output layer nodes, and o is the grid accuracy prediction coefficient.

[0021] Optionally, the optimizing the structural parameters of the BP neural network prediction model by using a genetic algorithm to obtain an optimized BP neural network prediction model specifically includes:

[0022] The link weights from the input layer to the hidden layer, the hidden layer weights, the link weights from the hidden layer to the output layer, the output layer weights, the link biases between the input layer and the hidden layer, the hidden layer biases, the link biases between the hidden layer and the output layer, and the output layer biases in the structural parameters of the BP neural network prediction model are encoded into a chromosome in the genetic algorithm, and the error function of the BP neural network prediction model is used as the fitness function of the genetic algorithm. Each chromosome gene represents the value of a weight parameter or a threshold parameter; wherein the error functions corresponding to different output variables are the same;

[0023] In the genetic algorithm, selection, crossover, mutation and generation of new populations are carried out in sequence, and the fitness corresponding to the new population is iteratively calculated until the maximum evolutionary generation is reached or the fitness function is less than the preset error threshold, and the structural parameters of the optimized BP neural network prediction model are obtained. Based on the structural parameters of the optimized BP neural network prediction model, the optimized BP neural network prediction model is obtained.

[0024] Furthermore, the error verification index corresponding to each output variable of the optimized BP neural network prediction model is calculated through the test set, and the error verification of the optimized BP neural network prediction model is performed. If any error verification index fails, the structural parameters of the BP neural network prediction model are updated again until all error verification indexes pass, and the optimized BP neural network prediction model that passes the error verification is obtained. Based on the optimized BP neural network prediction model that passes the error verification, the external operating parameters and internal structural parameters of the current electrolytic cell are predicted, wherein the error verification indicators include MSE, MAE, RMSE, MAPE and R 2 ;

[0025] The calculation method of the error verification index corresponding to each output variable of the optimized BP neural network prediction model is as follows:

[0026]

[0027] Where N is the number of samples in the test set; X i represents the predicted value of each output variable of the optimized BP neural network prediction model corresponding to the i-th sample; Y i represents the actual simulated value of the output variable corresponding to the i-th sample; Represents the average value of the actual simulated value of the output variable corresponding to all samples.

[0028] The technical solution of the present invention also provides a method for evaluating and predicting the performance of electrolytic cell working parameters, comprising:

[0029] A determination module is provided, wherein the external operating parameters and the internal structural parameters of the electrolyzer are used as input variables; based on the input variables and the variation range of each input variable, a performance evaluation index of the electrolyzer is determined, and the performance evaluation index is used as an output variable; wherein the electrolyzer is a proton exchange membrane electrolyzer;

[0030] A construction module is used to obtain multiple sets of data sets based on input variables and output variables, and a BP neural network prediction model is constructed based on the multiple sets of data sets;

[0031] The optimization module uses a genetic algorithm to optimize the structural parameters of the BP neural network prediction model to obtain an optimized BP neural network prediction model;

[0032] The prediction module obtains the current external operating parameters and internal structural parameters of the electrolytic cell, and performs prediction based on the current external operating parameters, internal structural parameters and the optimized BP neural network prediction model to obtain the performance prediction result of the electrolytic cell.

[0033] The technical solution adopted by the present invention includes the following technical effects:

[0034] 1. The technical solution of the present invention takes the external operating parameters and internal structural parameters of the electrolytic cell as input variables; determines the performance evaluation index of the electrolytic cell based on the input variables and the variation range of each input variable, and takes the performance evaluation index as the output variable; obtains multiple sets of data sets based on the input variables and the output variables, and constructs a BP neural network prediction model based on the multiple sets of data sets; optimizes the structural parameters of the BP neural network prediction model using a genetic algorithm to obtain an optimized BP neural network prediction model; obtains the current external operating parameters and internal structural parameters of the electrolytic cell, and predicts based on the current external operating parameters, internal structural parameters and the optimized BP neural network prediction model of the electrolytic cell to obtain a performance prediction result of the electrolytic cell, which effectively solves the problem of low reliability of performance evaluation prediction of the working parameters of PEMEC due to the single type of working parameters that affect the performance of the PEMEC to be evaluated in the prior art, and effectively improves the reliability of performance evaluation prediction of the working parameters of PEMEC.

[0035] 2. The external operating parameters of the electrolyzer described in the technical solution of the present invention include the temperature of water during the operation of passing deionized water into the anode flow channel of the electrolyzer, the flow rate of water during the operation of passing deionized water into the anode flow channel of the electrolyzer, and the voltage between the two poles of the electrolyzer connected to the power supply respectively; the internal structural parameters include the membrane thickness of the electrolyzer membrane electrode assembly, the effective porosity of the anode diffusion layer of the electrolyzer, the thickness of the anode diffusion layer of the electrolyzer, and the conductivity of the anode catalyst layer of the electrolyzer; the performance evaluation index of the electrolyzer includes the average temperature of the anode catalyst layer of the electrolyzer, the current density of the anode catalyst layer of the electrolyzer, and the average hydrogen mole fraction of the cathode diffusion layer and the catalyst layer of the electrolyzer; it includes not only 3 external operating parameters of the electrolyzer and 4 internal structural parameters of the electrolyzer, but also 3 performance evaluation indicators, thereby avoiding a single performance evaluation indicator or a single type of working parameter, which results in low accuracy in the performance evaluation prediction of the working parameters of PEMEC.

[0036] 3. In the technical solution of the present invention, for the electrolytic cell to be predicted, the parameter values ​​of the input variables are equidistantly changed within the variation range of each input variable, and only one parameter value of one input variable is changed each time, and a set of output variable data is correspondingly generated through experimental or simulation operations; multiple parameter values ​​of each input variable are cross-changed, thereby correspondingly generating multiple sets of output data to constitute a data set, which not only ensures the diversity of the data set, but also enables the data set to better train and test the BP neural network prediction model.

[0037] 4. In the technical solution of the present invention, the link weights from the input layer to the hidden layer, the hidden layer weights, the link weights from the hidden layer to the output layer, and the output layer weights in the structural parameters of the BP neural network prediction model are encoded into a chromosome in a genetic algorithm, and the error function of the BP neural network prediction model is used as the fitness function of the genetic algorithm, and each chromosome gene represents the value of a weight parameter or a threshold parameter; in the genetic algorithm, selection, crossover, mutation and generation of new populations are performed in sequence, and the fitness corresponding to the new population is iteratively calculated until the maximum evolutionary generation is reached or the fitness function is less than a preset error threshold, and the structural parameters of the optimized BP neural network prediction model are obtained. Based on the structural parameters of the optimized BP neural network prediction model, the optimized BP neural network prediction model is obtained, which ensures the effectiveness and reliability of the performance evaluation prediction of the working parameters of PEMEC.

[0038] 5. In the technical solution of the present invention, the error verification index corresponding to each output variable of the optimized BP neural network prediction model is calculated through the test set, and the error verification is performed on the optimized BP neural network prediction model. If any error verification index fails, the structural parameters of the BP neural network prediction model are updated again until all error verification indexes pass, and the optimized BP neural network prediction model that passes the error verification is obtained. Based on the optimized BP neural network prediction model that passes the error verification, the external operating parameters and internal structural parameters of the current electrolytic cell are predicted, which further ensures the effectiveness and reliability of the performance evaluation prediction of the working parameters of PEMEC.

[0039] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0041] Figure 1 This is a schematic diagram of a process of a method in Example 1 of the present invention;

[0042] Figure 2 This is a schematic diagram of the structure of a PEMEC single-channel model in the method of Example 1 of the present invention;

[0043] Figure 3 This is a schematic diagram of the process of optimizing the BP neural network through a genetic algorithm in the method of Example 1 of the present invention;

[0044] Figure 4A comparison chart of the predicted value and the actual value of the current density in the method of Example 1 of the present invention;

[0045] Figure 5 A comparison chart of the predicted value and the actual value of the temperature in the method of Example 1 of the present invention;

[0046] Figure 6 A comparison chart of the predicted value and the actual value of the hydrogen mole fraction in the method of Example 1 in the scheme of the present invention;

[0047] Figure 7 It is a schematic diagram of the structure of the system of Example 2 in the solution of the present invention.

[0048] Figure 2 In: 1. Anode plate; 2. Anode flow channel; 3. Anode diffusion layer; 4. Anode catalyst layer; 5. Proton exchange membrane; 6. Cathode catalyst layer; 7. Cathode diffusion layer; 8. Cathode flow channel; 9. Cathode plate; 10. Deionized water. DETAILED DESCRIPTION

[0049] In order to clearly illustrate the technical features of the present solution, the present invention is described in detail below through specific implementation methods and in conjunction with the accompanying drawings. The disclosure below provides many different embodiments or examples for realizing different structures of the present invention. In order to simplify the disclosure of the present invention, the components and settings of specific examples are described below. In addition, the present invention may repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed. It should be noted that the components illustrated in the accompanying drawings are not necessarily drawn to scale. The present invention omits the description of known components and processing techniques and processes to avoid unnecessary limitations on the present invention.

[0050] Embodiment 1

[0051] like Figure 1 As shown, the present invention provides a method for evaluating and predicting the performance of electrolytic cell operating parameters, comprising:

[0052] S1, taking the external operating parameters and internal structural parameters of the electrolyzer as input variables; determining the performance evaluation index of the electrolyzer based on the input variables and the variation range of each input variable, and taking the performance evaluation index as the output variable; wherein the electrolyzer is a proton exchange membrane electrolyzer;

[0053] S2, obtaining multiple sets of data sets based on input variables and output variables, and building a BP neural network prediction model based on the multiple sets of data sets;

[0054] S3, optimizing the structural parameters of the BP neural network prediction model by using a genetic algorithm to obtain an optimized BP neural network prediction model;

[0055] S4, obtaining the current external operating parameters and internal structural parameters of the electrolytic cell, and performing prediction based on the current external operating parameters and internal structural parameters of the electrolytic cell and the optimized BP neural network prediction model to obtain a performance prediction result of the electrolytic cell.

[0056] Among them, in step S1, the external operating parameters of the electrolytic cell may include the temperature of water during the operation of passing deionized water into the anode flow channel of the electrolytic cell, the flow rate of water during the operation of passing deionized water into the anode flow channel of the electrolytic cell, and the voltage between the two poles of the electrolytic cell respectively connected to the two poles of the power supply; the internal structural parameters include the membrane thickness of the electrolytic cell membrane electrode assembly, the effective porosity of the anode diffusion layer of the electrolytic cell, the thickness of the anode diffusion layer of the electrolytic cell, and the conductivity of the anode catalyst layer of the electrolytic cell.

[0057] like Figure 2 As shown, taking a single-channel PEMEC model as an example, during the operation of the PEMEC anode flow channel 2 passing deionized water 10, the water temperature and flow rate mainly affect the electrolysis performance of the PEMEC; during the process of the anode and cathode of the overall electrolyzer being connected to the power supply, the voltage is also a major factor affecting the electrolysis performance of the PEMEC; in the electrolysis process of the PEMEC, in addition to external operating parameters, since the deionized water 10 passing into the anode undergoes electrolysis reaction in the anode catalyst layer 4 through the anode diffusion layer 3 to produce large bubble oxygen, the change in the anode structure mainly affects the electrolysis performance of the PEMEC; in the electrochemical reaction process of the PEMEC, many complex physical and chemical processes occur on the membrane electrode assembly, which directly affect the electrolysis efficiency. It is also crucial to adjust the structure and composition of the membrane electrode assembly. Therefore, under the above conditions, other parameters are guaranteed to remain unchanged, and the changing parameters of PEMEC and the input variables of the input prediction model are determined, namely, the water temperature during the operation of deionized water passing through the anode flow channel of the electrolyzer, the water flow rate during the operation of deionized water passing through the anode flow channel of the electrolyzer, the voltage between the two poles of the electrolyzer connected to the power supply, the membrane thickness of the electrolyzer membrane electrode assembly, the effective porosity of the anode diffusion layer of the electrolyzer, the thickness of the anode diffusion layer of the electrolyzer, and the conductivity of the anode catalyst layer of the electrolyzer.

[0058] After determining the working parameters of the PEMEC to be changed, based on the design of the electrolytic cell, membrane performance, electrolysis conditions, electrode catalytic performance and other factors of the specific experiment or simulation operation, the respective ranges of the above-mentioned parameters are determined, that is, the range of the water temperature during the operation of the electrolytic cell anode flow channel passing deionized water is 20 to 80°C, the range of the water temperature flow rate during the operation of the electrolytic cell anode flow channel passing deionized water is 0.00015 to 0.0007 kg / s, the range of the voltage between the two poles of the electrolytic cell connected to the power supply is 1.8 to 2.5 V; the range of the membrane thickness of the electrolytic cell membrane electrode assembly is Wherein, h is the initial value of the membrane thickness of the electrolytic cell membrane electrode assembly; the effective porosity of the electrolytic cell anode diffusion layer varies in the range of 0.2 to 0.8; the thickness of the electrolytic cell anode diffusion layer varies in the range of Wherein, l is the initial value of the thickness of the anode diffusion layer of the electrolytic cell; the conductivity of the anode catalyst layer of the electrolytic cell varies in the range of 0.1 to 1000 S / m.

[0059] Among the multiple factors in evaluating the performance of the electrolyzer, in order to comprehensively consider the working efficiency, stability and product quality of the electrolyzer, and thus comprehensively evaluate its performance, three evaluation indicators for evaluating the electrolysis performance of PEMEC were determined, namely temperature (stability and working efficiency), current density (stability and working efficiency) and hydrogen mole fraction (product quality). More specifically, the average temperature of the anode catalyst layer of the electrolyzer, the current density of the anode catalyst layer of the electrolyzer, and the average hydrogen mole fraction of the cathode diffusion layer and catalyst layer of the electrolyzer. Through these three evaluation indicators (output variables), the working performance of the electrolyzer, as well as the working efficiency, stability and product quality of the electrolyzer can be more comprehensively evaluated.

[0060] In step S2, multiple sets of data sets are obtained based on the input variables and the output variables, and the BP neural network prediction model is constructed based on the multiple sets of data sets, specifically including:

[0061] For the electrolytic cell to be predicted, the parameter values ​​of the input variables are changed equidistantly within the variation range of each input variable, and only one parameter value of an input variable is changed each time, and a set of output variable data is correspondingly generated through experimental or simulation operations; multiple parameter values ​​of each input variable are cross-changed, thereby correspondingly generating multiple sets of output data to form a data set;

[0062] Among the seven input variables, each input variable takes a value to form a data sample. After the electrolyzer is in operation, since the chemical reaction mainly occurs in the anode catalyst layer and the hydrogen production is mainly in the cathode part, the anode catalyst layer temperature, current density and the average value of the hydrogen mole fraction of the cathode catalyst layer and the gas diffusion layer are calculated to evaluate the temperature safety, electrochemical reaction rate and hydrogen evolution efficiency of the electrolyzer. Seven variable parameters (input variables) were changed equidistantly within their respective ranges, and three values ​​were analyzed for each variable parameter. The three values ​​of the seven variable parameters were changed crosswise, thus generating 3 7 = 2187 sets of data and their corresponding different temperatures, current densities and hydrogen mole fraction values, eliminating the combinations that affect the electrolytic cell material and cannot produce hydrogen under extreme conditions (such as extremely low speed and very small membrane thickness), a total of one set of multi-sample (2061 sets of data) data sets were generated in one electrolytic cell.

[0063] The data set is divided into a training set and a test set according to a preset division ratio, wherein the training set is used as a sample for predictive modeling training of the BP neural network prediction model, and the test set is used as a sample for predictive model testing of the BP neural network prediction model.

[0064] In this data set, seven variable parameters are used as input variables, and three evaluation indicators are used as output variables. Multiple different data sets are divided into training sets and test sets. According to actual conditions, all data sets can be divided into 80% training sets and 20% test sets. Among them, the training set is used as a sample for predictive modeling training, and the test set is used as a sample for predictive model testing.

[0065] Before using the genetic algorithm to optimize the structural parameters of the BP neural network prediction model, the structural parameters, training parameters, genetic algorithm parameters, etc. of the BP neural network prediction model are set; wherein, the structural parameters of the BP neural network prediction model are the number of input layer nodes, the number of output layer nodes and the number of hidden layer nodes, the input layer and hidden layer link weights, the hidden layer weights, the hidden layer and output layer link weights, the output layer weights, the input layer and hidden layer link bias, the hidden layer bias, the hidden layer and output layer link bias, and the output layer bias; the training parameters of the BP neural network prediction model are the learning rate, the number of training times and the training target error; the genetic algorithm parameters are the initial population size, the maximum evolutionary generations, the crossover and mutation probabilities.

[0066] The structural parameters of the BP neural network prediction model include the number of input layer nodes, the number of output layer nodes, the number of hidden layer nodes, the link weight from the input layer to the hidden layer, the hidden layer weight, the link weight from the hidden layer to the output layer, the output layer weight, the link bias between the input layer and the hidden layer, the hidden layer bias, the link bias between the hidden layer and the output layer, and the output layer bias.

[0067] The number of input layer nodes is the number of input variables, the number of output layer nodes is the number of output variables, and the number of hidden layer nodes is determined as follows:

[0068]

[0069] Among them, P is the number of hidden layer nodes, m is the number of input layer nodes, n is the number of output layer nodes, and o is the grid accuracy prediction coefficient, o = 1, 2, 3..., 10. By taking different values ​​of o within the range, the prediction accuracy of the BP neural network is compared, and an optimal o value with high accuracy is selected.

[0070] In step S3, Figure 3 As shown, the genetic algorithm is used to optimize the structural parameters of the BP neural network prediction model, and the optimized BP neural network prediction model specifically includes:

[0071] The link weights from the input layer to the hidden layer, the hidden layer weights, the link weights from the hidden layer to the output layer, the output layer weights, the link biases between the input layer and the hidden layer, the hidden layer biases, the link biases between the hidden layer and the output layer, and the output layer biases in the structural parameters of the BP neural network prediction model are encoded into a chromosome in the genetic algorithm, and the error function of the BP neural network prediction model is used as the fitness function of the genetic algorithm. Each chromosome gene represents the value of a weight parameter or a threshold parameter; wherein the error functions corresponding to different output variables are the same;

[0072] BP neural networks are often limited by local minima during the training phase, and their convergence results often fail to reach the global optimum. In contrast, genetic algorithms, with their excellent global search strategies, can effectively explore in a wide range of solution spaces, helping BP neural networks break through the limitations of local optimal solutions, and then discover and approach the global optimal solution, significantly improving the training effect and performance of neural networks. In addition, genetic algorithms simulate the biological evolution mechanism of nature, including selection, crossover, and mutation, and can optimize the weights and thresholds of BP neural networks. It enhances the adaptability and learning ability of neural networks, significantly improves their overall performance, and enables them to show more superior results when dealing with complex problems.

[0073] The input layer and hidden layer link weight, hidden layer weight, hidden layer and output layer link weight, output layer weight, input layer and hidden layer link bias, hidden layer bias, hidden layer and output layer link bias, output layer bias, etc. are encoded into a chromosome in the genetic algorithm, and the error function of the BP neural network prediction model is used as the fitness function of the genetic algorithm. Each chromosome gene represents the value of a weight parameter or a threshold parameter. Among them, the error functions corresponding to the three output variables are the same, all of which are MSE functions.

[0074] In the genetic algorithm, selection, crossover, mutation and generation of new populations are carried out in sequence, and the fitness corresponding to the new population is iteratively calculated until the maximum evolutionary generation is reached or the fitness function is less than the preset error threshold (the first judgment on whether the end condition is met), and the structural parameters of the optimized BP neural network prediction model are obtained. Based on the structural parameters of the optimized BP neural network prediction model, the optimized BP neural network prediction model is obtained.

[0075] The specific forms of selection, crossover, and mutation operations can be as follows:

[0076] Selection: The roulette wheel method adopts a selection strategy based on the ratio of fitness values ​​to determine the selection probability of each individual (chromosome) x.

[0077]

[0078] Among them, f s is an individual xs fitness; M is the number of individuals in the group.

[0079] Crossover: involves randomly selecting an individual chromosome from a set of chromosomes and crossing it with another chromosome to produce a new individual using a specific method.

[0080] Mutation: Select the jth gene to mutate in the sth individual. At a specific gene position, the gene is replaced by an allele, resulting in the generation of a new individual.

[0081] Calculate the model error in the optimized BP neural network and check whether the maximum evolutionary generation (maximum number of iterations) is reached or the fitness function is less than the preset error threshold. If the conditions are met, simulate the prediction and output the prediction results. If the end conditions cannot be met, repeat the genetic algorithm selection, crossover, mutation and generation of new populations and calculate the fitness operations to obtain the optimal weights and deviations (update weights and deviations), simulate the prediction and output the prediction results. The flowchart of the BP neural network optimized by the genetic algorithm is as follows: Figure 3 shown.

[0082] Preferably, if Figure 3 As shown in the figure, the error verification index corresponding to each output variable of the optimized BP neural network prediction model is calculated through the test set, and the error verification of the optimized BP neural network prediction model is performed (the model error is calculated to determine whether the error verification indexes have passed the verification). If any error verification index fails, the structural parameters of the BP neural network prediction model are updated again until all error verification indexes pass (i.e., the model error is calculated and it is checked again whether the end condition is met). The optimized BP neural network prediction model that passes the error verification is obtained. Based on the optimized BP neural network prediction model that passes the error verification, the external operating parameters and internal structural parameters of the current electrolytic cell are predicted. Among them, the error verification indicators include MSE, MAE, RMSE, MAPE and R 2 ;

[0083] The calculation method of the error verification index corresponding to each output variable of the optimized BP neural network prediction model is as follows:

[0084]

[0085] Where N is the number of samples in the test set; X i represents the predicted value of each output variable of the optimized BP neural network prediction model corresponding to the i-th sample; Y i represents the actual simulated value of the output variable corresponding to the i-th sample; Represents the average value of the actual simulated value of the output variable corresponding to all samples.

[0086] MSE, MAE, RMSE, and MAPE verification pass means that the calculated values ​​are all less than the preset corresponding indicator thresholds. 2 Verification passed specifically means that the calculated values ​​are all greater than the preset corresponding indicator thresholds.

[0087] For the same data set, different evaluation indicators describe the results differently. MSE, RMSE, MAE and MAPE are used to evaluate the accuracy of the prediction model (BP neural network prediction model). The smaller the value, the closer the simulated prediction result is to the accurate value, and the higher the model accuracy. 2 It is used to evaluate the degree of simulation fit. The closer its value is to 1, the better the degree of simulation fit is and the higher the accuracy of the model is. 2 The closer it is to 0, the lower the model accuracy. In addition, the five accuracy evaluation indicators have the same rules for the evaluation of the three output variables. The higher the model accuracy, the higher the accuracy of the model's prediction of its corresponding output variable.

[0088] The results of applying this method to predict the working parameters of PEMEC can be obtained from Figure 4 , Figure 5 and Figure 6 Get. Figure 4 As shown in the results of current density prediction, the training set and test set of actual current density and predicted current density basically overlap, and the MSE, MAE, RMSE, MAPE and R 2 The values ​​are 0.005, 0.057, 0.071, 4.028% and 0.996 respectively. The MSE, RMSE, MAE and MAPE are small enough and R 2 is large enough, the model accuracy has high accuracy and effectiveness. Figure 5 and Figure 6 The same results were obtained in the above studies. The MSE, MAE, RMSE, MAPE and R 2 The values ​​are 0.00003, 0.00137, 0.00173, 0.251%, 0.997 and 0.009, 0.067, 0.094, 0, 0.993 respectively. The BP model optimized by genetic algorithm has significant effect. The model is suitable for the prediction of performance evaluation indexes of PEMEC working parameters and can provide theoretical support for the prediction and subsequent optimization of PEMEC working parameters.

[0089] The technical solution of the present invention takes the external operating parameters and internal structural parameters of the electrolytic cell as input variables; determines the performance evaluation index of the electrolytic cell based on the input variables and the variation range of each input variable, and takes the performance evaluation index as the output variable; obtains multiple sets of data sets based on the input variables and the output variables, and constructs a BP neural network prediction model based on the multiple sets of data sets; optimizes the structural parameters of the BP neural network prediction model using a genetic algorithm to obtain an optimized BP neural network prediction model; obtains the current external operating parameters and internal structural parameters of the electrolytic cell, and predicts based on the current external operating parameters, internal structural parameters and the optimized BP neural network prediction model of the electrolytic cell to obtain a performance prediction result of the electrolytic cell, which effectively solves the problem of low reliability of performance evaluation prediction of the working parameters of PEMEC due to the single type of working parameters that affect the performance of the PEMEC to be evaluated in the prior art, and effectively improves the reliability of performance evaluation prediction of the working parameters of PEMEC.

[0090] The external operating parameters of the electrolytic cell described in the technical solution of the present invention include the temperature of water during the operation of passing deionized water into the anode flow channel of the electrolytic cell, the flow rate of water during the operation of passing deionized water into the anode flow channel of the electrolytic cell, and the voltage between the two poles of the electrolytic cell respectively connected to the power supply; the internal structural parameters include the membrane thickness of the electrolytic cell membrane electrode assembly, the effective porosity of the anode diffusion layer of the electrolytic cell, the thickness of the anode diffusion layer of the electrolytic cell, and the conductivity of the anode catalyst layer of the electrolytic cell; the performance evaluation index of the electrolytic cell includes the average temperature of the anode catalyst layer of the electrolytic cell, the current density of the anode catalyst layer of the electrolytic cell, and the average hydrogen mole fraction of the cathode diffusion layer and the catalyst layer of the electrolytic cell; it not only includes 3 kinds of external operating parameters of the electrolytic cell, 4 kinds of internal structural parameters of the electrolytic cell, but also 3 kinds of performance evaluation indexes, thereby avoiding the situation that a single performance evaluation index or a single type of working parameter causes the low accuracy of the performance evaluation prediction of the working parameters of PEMEC.

[0091] In the technical solution of the present invention, for the electrolytic cell to be predicted, the parameter values ​​of the input variables are equidistantly changed within the variation range of each input variable, and only one parameter value of one input variable is changed each time, and a set of output variable data is correspondingly generated through experimental or simulation operations; multiple parameter values ​​of each input variable are cross-changed, thereby correspondingly generating multiple sets of output data to constitute a data set, which not only ensures the diversity of the data set, but also enables the data set to better train and test the BP neural network prediction model.

[0092] In the technical solution of the present invention, the link weights from the input layer to the hidden layer, the hidden layer weights, the link weights from the hidden layer to the output layer, and the output layer weights in the structural parameters of the BP neural network prediction model are encoded into a chromosome in a genetic algorithm, and the error function of the BP neural network prediction model is used as the fitness function of the genetic algorithm, and each chromosome gene represents the value of a weight parameter or a threshold parameter; in the genetic algorithm, selection, crossover, mutation and generation of new populations are continuously performed in sequence, and the fitness corresponding to the new population is iteratively calculated until the maximum evolutionary generation is reached or the fitness function is less than a preset error threshold, and the structural parameters of the optimized BP neural network prediction model are obtained; based on the structural parameters of the optimized BP neural network prediction model, the optimized BP neural network prediction model is obtained, thereby ensuring the effectiveness and reliability of the performance evaluation prediction of the working parameters of PEMEC.

[0093] In the technical solution of the present invention, the error verification index corresponding to each output variable of the optimized BP neural network prediction model is calculated through the test set, and the error verification is performed on the optimized BP neural network prediction model. If any error verification index fails, the structural parameters of the BP neural network prediction model are updated again until all error verification indexes pass, and the optimized BP neural network prediction model that passes the error verification is obtained. Based on the optimized BP neural network prediction model that passes the error verification, the external operating parameters and internal structural parameters of the current electrolytic cell are predicted, which further ensures the effectiveness and reliability of the performance evaluation prediction of the working parameters of the PEMEC.

[0094] Embodiment 2

[0095] like Figure 7 As shown, the technical solution of the present invention also provides a method for evaluating and predicting the performance of electrolytic cell working parameters, comprising:

[0096] Determination module 101, taking external operating parameters and internal structural parameters of the electrolyzer as input variables; determining the performance evaluation index of the electrolyzer based on the input variables and the variation range of each input variable, and taking the performance evaluation index as the output variable; wherein the electrolyzer is a proton exchange membrane electrolyzer;

[0097] A construction module 102 is used to obtain multiple sets of data sets based on input variables and output variables, and to construct a BP neural network prediction model based on the multiple sets of data sets;

[0098] The optimization module 103 optimizes the structural parameters of the BP neural network prediction model using a genetic algorithm to obtain an optimized BP neural network prediction model;

[0099] The prediction module 104 obtains the current external operating parameters and internal structural parameters of the electrolytic cell, and performs prediction based on the current external operating parameters and internal structural parameters of the electrolytic cell and the optimized BP neural network prediction model to obtain the performance prediction result of the electrolytic cell.

[0100] The determination module 101, the construction module 102, the optimization module 103, and the prediction module 104 in the technical solution of the present invention correspond to the steps in the embodiment, and will not be described in detail in this embodiment.

[0101] The technical solution of the present invention takes the external operating parameters and internal structural parameters of the electrolytic cell as input variables; determines the performance evaluation index of the electrolytic cell based on the input variables and the variation range of each input variable, and takes the performance evaluation index as the output variable; obtains multiple sets of data sets based on the input variables and the output variables, and constructs a BP neural network prediction model based on the multiple sets of data sets; optimizes the structural parameters of the BP neural network prediction model using a genetic algorithm to obtain an optimized BP neural network prediction model; obtains the current external operating parameters and internal structural parameters of the electrolytic cell, and predicts based on the current external operating parameters, internal structural parameters and the optimized BP neural network prediction model of the electrolytic cell to obtain a performance prediction result of the electrolytic cell, which effectively solves the problem of low reliability of performance evaluation prediction of the working parameters of PEMEC due to the single type of working parameters that affect the performance of the PEMEC to be evaluated in the prior art, and effectively improves the reliability of performance evaluation prediction of the working parameters of PEMEC.

[0102] The external operating parameters of the electrolytic cell described in the technical solution of the present invention include the temperature of water during the operation of passing deionized water into the anode flow channel of the electrolytic cell, the flow rate of water during the operation of passing deionized water into the anode flow channel of the electrolytic cell, and the voltage between the two poles of the electrolytic cell respectively connected to the power supply; the internal structural parameters include the membrane thickness of the electrolytic cell membrane electrode assembly, the effective porosity of the anode diffusion layer of the electrolytic cell, the thickness of the anode diffusion layer of the electrolytic cell, and the conductivity of the anode catalyst layer of the electrolytic cell; the performance evaluation index of the electrolytic cell includes the average temperature of the anode catalyst layer of the electrolytic cell, the current density of the anode catalyst layer of the electrolytic cell, and the average hydrogen mole fraction of the cathode diffusion layer and the catalyst layer of the electrolytic cell; it not only includes 3 kinds of external operating parameters of the electrolytic cell, 4 kinds of internal structural parameters of the electrolytic cell, but also 3 kinds of performance evaluation indexes, thereby avoiding the situation that a single performance evaluation index or a single type of working parameter causes the low accuracy of the performance evaluation prediction of the working parameters of PEMEC.

[0103] In the technical solution of the present invention, for the electrolytic cell to be predicted, the parameter values ​​of the input variables are equidistantly changed within the variation range of each input variable, and only one parameter value of one input variable is changed each time, and a set of output variable data is correspondingly generated through experimental or simulation operations; multiple parameter values ​​of each input variable are cross-changed, thereby correspondingly generating multiple sets of output data to constitute a data set, which not only ensures the diversity of the data set, but also enables the data set to better train and test the BP neural network prediction model.

[0104] In the technical solution of the present invention, the link weights from the input layer to the hidden layer, the hidden layer weights, the link weights from the hidden layer to the output layer, and the output layer weights in the structural parameters of the BP neural network prediction model are encoded into a chromosome in a genetic algorithm, and the error function of the BP neural network prediction model is used as the fitness function of the genetic algorithm, and each chromosome gene represents the value of a weight parameter or a threshold parameter; in the genetic algorithm, selection, crossover, mutation and generation of new populations are continuously performed in sequence, and the fitness corresponding to the new population is iteratively calculated until the maximum evolutionary generation is reached or the fitness function is less than a preset error threshold, and the structural parameters of the optimized BP neural network prediction model are obtained; based on the structural parameters of the optimized BP neural network prediction model, the optimized BP neural network prediction model is obtained, thereby ensuring the effectiveness and reliability of the performance evaluation prediction of the working parameters of PEMEC.

[0105] In the technical solution of the present invention, the error verification index corresponding to each output variable of the optimized BP neural network prediction model is calculated through the test set, and the error verification is performed on the optimized BP neural network prediction model. If any error verification index fails, the structural parameters of the BP neural network prediction model are updated again until all error verification indexes pass, and the optimized BP neural network prediction model that passes the error verification is obtained. Based on the optimized BP neural network prediction model that passes the error verification, the external operating parameters and internal structural parameters of the current electrolytic cell are predicted, which further ensures the effectiveness and reliability of the performance evaluation prediction of the working parameters of the PEMEC.

[0106] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A method for evaluating and predicting the performance of electrolytic cell operating parameters, characterized in that: include: The external operating parameters and internal structural parameters of the electrolyzer are taken as input variables; Based on the input variables and the variation range of each input variable, determining the performance evaluation index of the electrolyzer, and taking the performance evaluation index as the output variable; wherein the electrolyzer is a proton exchange membrane electrolyzer; Acquire multiple sets of data sets based on input variables and output variables, and build a BP neural network prediction model based on the multiple sets of data sets; The genetic algorithm is used to optimize the structural parameters of the BP neural network prediction model to obtain the optimized BP neural network prediction model; The current external operating parameters and internal structural parameters of the electrolytic cell are obtained, and prediction is performed based on the current external operating parameters, internal structural parameters and optimized BP neural network prediction model to obtain the performance prediction result of the electrolytic cell.

2. The method for evaluating and predicting the performance of electrolytic cell operating parameters according to claim 1, wherein: The external operating parameters of the electrolytic cell include the temperature of water during the operation of passing deionized water into the anode flow channel of the electrolytic cell, the flow rate of water during the operation of passing deionized water into the anode flow channel of the electrolytic cell, and the voltage between the two poles of the electrolytic cell respectively connected to the two poles of the power supply; the internal structural parameters include the membrane thickness of the electrolytic cell membrane electrode assembly, the effective porosity of the anode diffusion layer of the electrolytic cell, the thickness of the anode diffusion layer of the electrolytic cell, and the conductivity of the anode catalyst layer of the electrolytic cell.

3. The method for evaluating and predicting the performance of electrolytic cell operating parameters according to claim 2, wherein: The performance evaluation indexes of the electrolytic cell include the average temperature of the anode catalyst layer of the electrolytic cell, the current density of the anode catalyst layer of the electrolytic cell, and the average hydrogen mole fraction of the cathode diffusion layer and the catalyst layer of the electrolytic cell.

4. The method for evaluating and predicting the performance of an electrolytic cell operating parameter according to claim 2, wherein The temperature of the water in the electrolytic cell anode flow channel during operation is in the range of 20 to 80°C, the flow rate of the water in the electrolytic cell anode flow channel during operation is in the range of 0.00015 to 0.0007 kg / s, the voltage between the two electrodes of the electrolytic cell connected to the power supply is in the range of 1.8 to 2.5 V; the membrane thickness of the electrolytic cell membrane electrode assembly is in the range of Wherein, h is the initial value of the membrane thickness of the electrolytic cell membrane electrode assembly; the effective porosity of the electrolytic cell anode diffusion layer varies in the range of 0.2 to 0.8; the thickness of the electrolytic cell anode diffusion layer varies in the range of Wherein, l is the initial value of the thickness of the anode diffusion layer of the electrolytic cell; the conductivity of the anode catalyst layer of the electrolytic cell varies in the range of 0.1 to 1000 S / m.

5. The method for evaluating and predicting the performance of electrolytic cell operating parameters according to claim 1, characterized in that: Based on the input variables and output variables, multiple sets of data sets are obtained, and a BP neural network prediction model is constructed based on the multiple sets of data sets, specifically including: For the electrolytic cell to be predicted, the parameter values ​​of the input variables are changed equidistantly within the variation range of each input variable, and only one parameter value of an input variable is changed each time, and a set of output variable data is correspondingly generated through experimental or simulation operations; multiple parameter values ​​of each input variable are cross-changed, thereby correspondingly generating multiple sets of output data to form a data set; The data set is divided into a training set and a test set according to a preset division ratio, wherein the training set is used as a sample for predictive modeling training of the BP neural network prediction model, and the test set is used as a sample for predictive model testing of the BP neural network prediction model.

6. A method for evaluating and predicting the performance of electrolytic cell operating parameters according to claim 5, characterized in that: The structural parameters of the BP neural network prediction model include the number of input layer nodes, the number of output layer nodes, the number of hidden layer nodes, the link weight from the input layer to the hidden layer, the hidden layer weight, the link weight from the hidden layer to the output layer, and the output layer weight; the link bias between the input layer and the hidden layer, the hidden layer bias, the link bias between the hidden layer and the output layer, and the output layer bias.

7. A method for evaluating and predicting the performance of electrolytic cell operating parameters according to claim 6, characterized in that: The number of input layer nodes is the number of input variables, the number of output layer nodes is the number of output variables, and the number of hidden layer nodes is determined as follows: Among them, P is the number of hidden layer nodes, m is the number of input layer nodes, n is the number of output layer nodes, and o is the grid accuracy prediction coefficient.

8. The method for evaluating and predicting the performance of electrolytic cell operating parameters according to claim 6, characterized in that: The structural parameters of the BP neural network prediction model are optimized by using a genetic algorithm to obtain an optimized BP neural network prediction model, which specifically includes: The link weights from the input layer to the hidden layer, the hidden layer weights, the link weights from the hidden layer to the output layer, the output layer weights, the link biases between the input layer and the hidden layer, the hidden layer biases, the link biases between the hidden layer and the output layer, and the output layer biases in the structural parameters of the BP neural network prediction model are encoded into a chromosome in the genetic algorithm, and the error function of the BP neural network prediction model is used as the fitness function of the genetic algorithm. Each chromosome gene represents the value of a weight parameter or a bias parameter; wherein the error functions corresponding to different output variables are the same; In the genetic algorithm, selection, crossover, mutation and generation of new populations are carried out in sequence, and the fitness corresponding to the new population is iteratively calculated until the maximum evolutionary generation is reached or the fitness function is less than the preset error threshold, and the structural parameters of the optimized BP neural network prediction model are obtained. Based on the structural parameters of the optimized BP neural network prediction model, the optimized BP neural network prediction model is obtained.

9. A method for evaluating and predicting the performance of electrolytic cell operating parameters according to claim 8, characterized in that: After obtaining the optimized BP neural network prediction model based on the structural parameters of the optimized BP neural network prediction model, it also includes: The error verification index corresponding to each output variable of the optimized BP neural network prediction model is calculated through the test set, and the error verification of the optimized BP neural network prediction model is performed. If any error verification index fails, the structural parameters of the BP neural network prediction model are updated again until all error verification indexes pass, and the optimized BP neural network prediction model that passes the error verification is obtained. Based on the optimized BP neural network prediction model that passes the error verification, the external operating parameters and internal structural parameters of the current electrolytic cell are predicted. Among them, the error verification indicators include MSE, MAE, RMSE, MAPE and R 2 ; The calculation method of the error verification index corresponding to each output variable of the optimized BP neural network prediction model is as follows: Where N is the number of samples in the test set; X i represents the predicted value of each output variable of the optimized BP neural network prediction model corresponding to the i-th sample; Y i represents the actual simulated value of the output variable corresponding to the i-th sample; Represents the average value of the actual simulated value of the output variable corresponding to all samples.

10. A method for evaluating and predicting the performance of electrolytic cell operating parameters, characterized in that: include: A determination module is provided, wherein external operating parameters and internal structural parameters of the electrolyzer are used as input variables; Based on the input variables and the variation range of each input variable, determining the performance evaluation index of the electrolyzer, and taking the performance evaluation index as the output variable; wherein the electrolyzer is a proton exchange membrane electrolyzer; A construction module is used to obtain multiple sets of data sets based on input variables and output variables, and a BP neural network prediction model is constructed based on the multiple sets of data sets; The optimization module uses a genetic algorithm to optimize the structural parameters of the BP neural network prediction model to obtain an optimized BP neural network prediction model; The prediction module obtains the current external operating parameters and internal structural parameters of the electrolytic cell, and performs prediction based on the current external operating parameters, internal structural parameters and the optimized BP neural network prediction model to obtain the performance prediction result of the electrolytic cell.

Images