A mechanism and data-driven fusion modeling method and system for a PEMFC

Abstract

The application provides a mechanism and data driven PEMFC fusion modeling method and system, relates to the fuel cell technical field, and is suitable for fuel cell modeling and parameter optimization.According to the working principle and experimental data of the PEMFC, a semi-mechanism and semi-data driven model is constructed; the parameter range of the to-be-identified parameters of the model is confirmed; the stack and experimental data are imported into the PEMFC model; the integrated error of the actual output voltage of the battery and the output voltage of the mechanism and data driven module is taken as an objective function, and a constraint condition is determined; an evolutionary programming algorithm is used to optimize the to-be-identified parameters, and an optimal solution is obtained; and the optimal control of the PEMFC is realized according to the optimal solution.Compared with the prior art, the application can realize more accurate model construction, can evaluate the iteration results of the last round in the evolution process, can adaptively adjust the evolution structure, can guide the population to evolve towards the optimal solution direction, can effectively improve the algorithm performance, and can realize accurate identification of the parameters of the PEMFC.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of fuel cells, and particularly relates to a PEMFC fusion modeling method based on mechanism and data driving. BACKGROUND

[0002] With the increasing demand for energy in society, the consumption of traditional energy is rapidly increasing, and a series of problems such as global warming caused by the use of traditional energy are gradually highlighted. Therefore, the development of efficient, clean, economic and sustainable energy has gradually become the focus of attention. Proton exchange membrane fuel cell (PEMFC) can utilize electrochemical reaction to directly convert chemical energy in fuel into electrical energy, and has high energy utilization rate, high efficiency, and good adaptability and controllability. With the wide application of PEMFC, the efficient modeling and optimal control of PEMFC gradually become a research hotspot, and the model construction and parameters are the key to control, so it is essential to construct a suitable PEMFC model and adopt an effective parameter identification method for optimal control implementation.

[0003] In the aspect of model construction, the commonly used PEMFC models include empirical model, mechanism model, semi-empirical model and data-driven model, although these models have certain effects in model construction, but still have limitations in accuracy and adaptability. In the aspect of parameter identification, the existing parameter identification methods include real-time adaptive parameter estimation method of nonlinear system, robust method based on gradient optimizer, etc. However, these methods may need complex calculation process, and may be sensitive to initial conditions, resulting in that the algorithm falls into local optimal solution. Therefore, part of the research begins to focus on heuristic algorithms, such as genetic algorithm, particle swarm algorithm, etc. Heuristic algorithm is based on intuition, experience and heuristic rules to solve the optimization and search of complex problems, is not sensitive to the selection of initial solution and parameter setting, and does not depend on the specific structure of the problem, and has strong global search ability. Although the existing methods have good performance, but there are still certain limitations in the adaptability of the search process.

[0004] Based on the above problems, it is necessary to propose a PEMFC fusion modeling method based on mechanism and data driving. SUMMARY

[0005] In view of the problems existing in the prior art, the application is proposed.

[0006] Therefore, the purpose of the application is to provide a modeling and parameter identification method of proton exchange membrane fuel cell, to realize efficient parameter identification by constructing a semi-mechanism and semi-data driving model, and combining evolutionary computing and programming tree idea.

[0007] To solve the above technical problems, the present application provides the following technical solutions: a PEMFC fusion modeling method based on mechanism and data driving, comprising: constructing a semi-mechanism semi-data driving model of PEMFC according to the working principle and experimental data of the proton exchange membrane fuel cell respectively; confirming the parameters to be identified in the PEMFC model and setting the range of the parameters to be identified; importing the stack data and experimental related data into the PEMFC model; taking the comprehensive error of the actual output voltage of the battery and the output voltage of the mechanism module and the data driving module as the objective function, and determining the constraint condition of the objective function; using the evolutionary programming algorithm to optimize the parameters to be identified, and obtaining the optimal parameters to be identified; and realizing the optimal control of the proton exchange membrane fuel cell according to the output optimal parameters to be identified.

[0008] As a preferred scheme of the PEMFC fusion modeling method based on mechanism and data driving, the semi-mechanism semi-data driving model of PEMFC comprises a mechanism module of the PEMFC model and a data driving module of the PEMFC model.

[0009] The mechanism module of the PEMFC model comprises modeling considering activation loss V act , ohmic loss V onmic and concentration difference loss V con , and the output voltage of the single cell is represented as:

[0010]

[0011] Wherein, ξ1, ξ2, ξ3, ξ4, ξ5, ξ6, ξ7 are the parameters to be optimized in the PEMFC model, represents the oxygen concentration in the cathode and the gas reaction interface, R C is the equivalent resistance of electron transfer, R M is the equivalent resistance of the proton exchange membrane, b is the parameter coefficient related to the state of the battery in the model, J is the actual current density, J max is the maximum current density, E n is the open circuit voltage of the PEMFC, T is the current gas temperature, i represents the battery current, exp() represents the exponential function with the natural logarithm e as the base, ln() represents the logarithmic function with the natural number e as the base, E n is the open circuit voltage of the PEMFC, and the calculation formula is represented as:

[0012]

[0013] Wherein, are the partial pressures of hydrogen and oxygen respectively;

[0014] The voltage of the PEMFC stack is equivalent to the voltage of n single cells, and the stack voltage VP is represented as:

[0015] V P = nV c

[0016] As a preferred scheme of the mechanism and data driven PEMFC fusion modeling method described in the application, the data driven module for establishing a PEMFC model comprises:

[0017] Collecting experimental data of the PEMFC under various operating conditions;

[0018] Cleaning the data, filling in missing values, normalizing or standardizing the data;

[0019] The to-be-identified parameters of the PEMFC model are input, and the output voltage of the data driven model is output;

[0020] An improved RNN model based on association mining is constructed as the data driven module, an association mining layer is added before the hidden layer, and the input x t and the output (a t-1 , b t-1 ) of the hidden layer at the previous moment are taken as the inputs of the association mining layer;

[0021] wherein a t-1 is the rule mined by the association mining layer at the previous moment, b t-1 is the output of the hidden layer at the previous moment, the association mining layer judges the association between the inputs and the output at the present moment according to the inputs, and the weights are distributed according to the association degree, and the association mining formula is represented as:

[0022] (a t , w t ) = ξ(x t , a t-1 , b t+1 )

[0023] wherein a t is the rule mined by the association mining layer at the present moment, w t is the weight distribution given by the association mining layer, and ξ() is the association mining layer function;

[0024] The hidden layer formula is represented as:

[0025] b t = h(U·x t + w t ·b t-1 )

[0026] wherein h() is the hidden layer function, U is the parameter matrix from the input layer to the association mining layer, and b t and b t-1are the outputs of the current time hidden layer and the last time hidden layer respectively, x t is the input of the current time;

[0027] The experimental data is used to train the model, the absolute error is used as the loss function, and the error between the model prediction value and the actual output value is minimized.

[0028] As a preferred scheme of the PEMFC fusion modeling method based on mechanism and data driving provided by the application, wherein the parameters to be identified in the PEMFC model are confirmed as ξ1, ξ2, ξ3, ξ4, ξ5, ξ6, ξ7, λ, R c , b, the range of the to-be-identified parameters is set as:

[0029]

[0030] min max

[0031] Wherein, ξ i is the optimized parameter ξ1, ξ2, ξ3, ξ4, ξ5, ξ6, ξ7 in the PEMFC model, and ξ i is the upper and lower bounds of ξ1, ξ2, ξ3, ξ4, ξ5, ξ6, ξ7, λ is the water content of the proton exchange membrane, and λ i is the upper and lower bounds of λ, R min is the electron transfer equivalent resistance, and R max is the upper and lower bounds of R C , and b is the parameter coefficient related to the state of the battery in the model, and b c is the upper and lower bounds of b. min ,R c max is the upper and lower bounds of R c , and b is the parameter coefficient related to the state of the battery in the model, and b min is the upper and lower bounds of b. max

[0032] As a preferred scheme of the PEMFC fusion modeling method based on mechanism and data driving provided by the application, wherein the constraint condition of the objective function is determined, including substituting the parameters after each iteration into the constructed mechanism module and the trained data driving module respectively, and obtaining the output voltages V M and V D of the two modules respectively.

[0033] The comprehensive error of the output voltages of the mechanism module and the data driving module and the actual output voltage is used as the objective function F f , and the function is expressed as:

[0034]

[0035] Wherein, N is the data amount, and i is the index, VR is the actual voltage;

[0036] The target function value is minimum when the error is minimum, and the parameters are the optimal solution.

[0037] As a preferred scheme of the mechanism and data-driven PEMFC fusion modeling method, the evolutionary programming algorithm optimization to be identified parameters includes:

[0038] Initialize parameters, initialize the population in a specified range;

[0039] Define the mutation strategy set, selection strategy set, scaling factor set and crossover factor set of the algorithm;

[0040] According to the target function F f Calculate the fitness s of the individual i , the evaluation formula is:

[0041]

[0042] Wherein R is the evaluation result, is the average value of the fitness, s max is the maximum value of the fitness, and NP is the population number;

[0043] Determine whether the number of iterations is met, if not, build an evolutionary programming tree for mutation, crossover, selection operation of the population, if met, end the iteration, and obtain the optimal parameter set.

[0044] As a preferred scheme of the mechanism and data-driven PEMFC fusion modeling method, the evolutionary programming tree includes a mutation selection layer, a mutation operation layer, a crossover layer and a selection layer.

[0045] The mutation selection layer selects a mutation strategy and a scaling factor value in the mutation strategy set and the scaling factor set respectively, and the selection method is determined by the result of the last iteration. If a better result appears, a strategy is randomly selected from the mutation strategy considering the optimal solution, and the scaling factor F is adaptively reduced, and the calculation formula is:

[0046]

[0047] If the last result does not change after iteration, a strategy is randomly selected from the mutation strategy without considering the optimal solution, and the scaling factor is adaptively increased, and the calculation formula is:

[0048]

[0049] Wherein F u is the upper limit of the scaling factor, and F lI is a lower bound of the scaling factor max I is a maximum number of iterations c I is a current iteration number

[0050] The mutation operation layer includes performing a mutation operation according to a selected mutation strategy and a scaling factor;

[0051] The crossover layer includes determining a value of a crossover factor, performing a crossover operation according to the determined value of the crossover factor, and the value of the crossover factor is determined by evaluation results of the population, and the calculation is expressed as:

[0052]

[0053] Wherein, CR is a crossover factor, CR u CR is an upper bound of the crossover factor l R is a lower bound of the crossover factor, and R is an evaluation result of the population

[0054] The selection layer includes randomly selecting a selection strategy in a selection strategy set, and performing a selection operation.

[0055] Another object of the present application is to provide a mechanism and data driven PEMFC fusion modeling system, by driving model construction module, a semi-mechanism and semi-data driven model combining working principle and experimental data is successfully established, which provides an accurate basis for parameter optimization; the parameter identification and definition module determines the range of the parameters to be optimized, ensuring the directionality and efficiency of the parameter optimization process; the data import and management module provides sufficient data support for model verification and optimization, enhancing the practicability and reliability of the model; the objective function design module provides a clear objective and boundary for the optimization process by defining a clear objective function and constraint condition, making the optimization process more targeted and effective; the evolutionary algorithm optimization module successfully determines the optimal parameter set by using advanced evolutionary programming algorithm, improving the prediction accuracy and control effect of the model; the optimal control execution module realizes accurate control of the proton exchange membrane fuel cell by using the optimal parameters, improves the performance and stability of the system, and lays a solid technical foundation for the wide application and commercialization of fuel cells.

[0056] The present application provides the following technical scheme: a mechanism and data driven PEMFC fusion modeling method, comprising: a driving model construction module, a parameter identification and definition module, a data import and management module, an objective function design module, an evolutionary algorithm optimization module, and an optimal control execution module;

[0057] The driving model construction module respectively constructs a semi-mechanism and semi-data driven model of the PEMFC according to the working principle and experimental data of the proton exchange membrane fuel cell;

[0058] The parameter identification and definition module identifies parameters to be identified in the PEMFC model and sets the range of the parameters to be identified;

[0059] The data import and management module imports stack data and experimental related data into the PEMFC model;

[0060] The objective function design module takes the comprehensive error of the actual output voltage of the battery and the output voltage of the mechanism module and the data-driven module as the objective function, and determines the constraint condition of the objective function;

[0061] The evolutionary algorithm optimization module uses evolutionary programming algorithm to optimize the parameters to be identified, and obtains the optimal parameters to be identified;

[0062] The optimal control execution module realizes the optimal control of the proton exchange membrane fuel cell according to the output optimal parameters to be identified.

[0063] A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of any one of the adaptive multi-camera capture and real-time free-view video rendering methods when executing the computer program.

[0064] A computer readable storage medium having a computer program stored thereon, characterized in that the computer program implements the steps of any one of the adaptive multi-camera capture and real-time free-view video rendering methods when executed by a processor.

[0065] The beneficial effects of the present application are: by combining mechanism model and data-driven model, the method can more accurately simulate the actual working state of PEMFC, and improve the prediction accuracy of the model. The semi-mechanism semi-data-driven model structure makes the method adapt to the operating characteristics of PEMFC under different working conditions, and has strong working condition adaptability. By optimizing the model parameters, the method can provide more effective control strategies for PEMFC, thereby improving the operating efficiency and stability of the fuel cell. It can monitor and adjust the operating state of PEMFC in real time, which helps to improve the stability and reliability of the system. By improving the operating efficiency and stability of PEMFC, the method can bring better economic benefits to users. It proposes new ideas and methods in the field of PEMFC modeling and control, which helps to promote the progress and development of related technologies. BRIEF DESCRIPTION OF DRAWINGS

[0066] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0067] Figure 1 A flow chart of the whole process of a PEMFC fusion modeling method based on mechanism and data driving of the present application.

[0068] Figure 2 A flow chart of the construction of a PEMFC model of a PEMFC fusion modeling method based on mechanism and data driving of the present application.

[0069] Figure 3 An improved RNN model based on association mining of a PEMFC fusion modeling method based on mechanism and data driving of the present application.

[0070] Figure 4 A flow chart of an evolutionary programming algorithm of a PEMFC fusion modeling method based on mechanism and data driving of the present application. DETAILED DESCRIPTION

[0071] In order to make the above objectives, features and advantages of the present application more apparent and comprehensible, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the protection scope of the present application.

[0072] Embodiment 1

[0073] Reference Figure 1 -- Figure 4 For the first embodiment of the present application, a PEMFC fusion modeling method based on mechanism and data driving is provided, as shown in Figure 1 , which comprises:

[0074] S1: respectively according to the working principle and experimental data of the proton exchange membrane fuel cell, a semi-mechanism semi-data driven model of the PEMFC is constructed.

[0075] It should be noted that the mechanism module and the data driven module for establishing the PEMFC model are as shown in Figure 2 .

[0076] Further, the mechanism module for establishing the PEMFC model is represented as:

[0077] It should be noted that in the process of generating electric energy by the fuel cell system, the internal battery will also affect the output voltage and cause consumption, in order to ensure that the established model simulates the actual system running state as much as possible, the activation loss V act , ohmic loss V onmic and concentration difference loss V con, the output voltage of the single cell is V c :

[0078] V c = E n -V act -V onmic -V con

[0079] wherein E n is the open circuit voltage of the PEMFC, which refers to the voltage when there is no current flowing through the cell, and the calculation formula is represented as:

[0080]

[0081] wherein T is the current gas temperature, ln() represents the logarithmic function with natural logarithm e as the base, are the partial pressures of hydrogen and oxygen respectively, and are represented as:

[0082]

[0083]

[0084] wherein RH a , RH c respectively represent the relative humidity of the steam in the anode and the cathode, P a , P c respectively represent the inlet pressure of the anode and the cathode, A represents the active area of the proton exchange membrane, and i represents the cell current, represents the saturation pressure of water vapor, and exp() represents the exponential function with natural logarithm e as the base, and the calculation formula is represented as:

[0085]

[0086] The activation loss refers to the fact that when the current passes through the electrode, the oxidation reaction and the reduction reaction respectively occurring in the anode and the cathode will affect the degree of electrification on the electrode, thereby causing deviation of the equilibrium value of the electrode potential, and the calculation formula of the activation loss is represented as:

[0087]

[0088] wherein ξ1, ξ2, ξ3, and ξ4 are the related parameters in the PEMFC model that need to be optimized, and cO2 represents the oxygen concentration in the cathode at the gas reaction interface, and the calculation formula is represented as:

[0089]

[0090] Ohmic loss refers to the electronic resistance generated by the current passing through the bipolar plate and the external bus, and the ion exchange membrane also has ion resistance to ion flow. These resistances will cause power loss. The calculation formula of Ohmic loss is expressed as:

[0091] V onmic = i(R M + R C )

[0092] Where R C is the equivalent resistance of electron transfer, R M is the equivalent resistance of proton exchange membrane, and is expressed as:

[0093]

[0094] Where l is the thickness of the proton exchange membrane, is the impedance that the hydrogen ion needs to overcome, and is expressed as:

[0095]

[0096] Where λ is the water content of the proton exchange membrane.

[0097] Concentration loss refers to the change in the concentration of the reactants at the anode and cathode, which affects the change in the voltage of the battery. The calculation formula of concentration loss is expressed as:

[0098]

[0099] Where ξ5, ξ6, and ξ7 are also part of the parameters related to the optimization of the PEMFC model, b is the parameter coefficient related to the state of the battery in the model, J is the actual current density, and J max is the maximum current density.

[0100] PEMFC stack is composed of multiple identical single cells connected in series, and the output voltage characteristics of the two are basically the same. Therefore, the voltage of the PEMFC stack is equivalent to the voltage of n single cells, and the stack voltage V P of the PEMFC can be expressed as:

[0101] V P = nV c

[0102] Further, the data-driven module of the PEMFC model is expressed as:

[0103] Collect experimental data of PEMFC under various operating conditions, including voltage output and parameters to be identified under different temperature, pressure, flow rate, etc.

[0104] Clean the data, fill in the missing values, normalize or standardize, etc. to ensure data quality;

[0105] The to-be-identified parameters of the PEMFC model are taken as inputs, and the output voltage of the data-driven model is taken as output;

[0106] As shown in Figure 3 , an improved RNN model based on association mining is constructed as a data-driven model, and an association mining layer is added before the hidden layer to determine the association between the inputs and the outputs before the current inputs according to the input, and the weights are allocated according to the association degree, and the association mining formula is represented as:

[0107] (a t ,w t )=ξ(x t ,a t-1 ,b t-1 )

[0108] Wherein, ξ() is the association mining layer function, x t is the input at the current time, a t , a t-1 is the rule mined by the association mining layer at the current time and the previous time, is the weight distribution given by the association mining layer, and b t is the output of the hidden layer at the previous time.

[0109] The hidden layer formula is represented as:

[0110] b t =h(U·x t +w t ·b t-1 )

[0111] Wherein, h() is the hidden layer function, U is the parameter matrix from the input layer to the association mining layer, b t and b t-1 are the outputs of the hidden layer at the current time and the previous time, respectively.

[0112] The improved RNN model constructed is a multiple-input single-output model, and therefore, the calculation formula of the final output Y is represented as:

[0113] b t =h(U·x t +w t ·b t-1 )

[0114] The experimental data is used to train the model to minimize the error between the predicted value of the model and the actual output value. The absolute error is taken as the loss function, and the calculation formula is represented as:

[0115]

[0116] Wherein, is the actual output value, and Y is the predicted value obtained by the model.

[0117] S2: confirming the parameters to be identified in the PEMFC model and setting the range of the parameters to be identified.

[0118] Further, confirming the parameters to be identified in the PEMFC model;

[0119] It should be noted that (ξ1, ξ2, ξ3, ξ4, ξ5, ξ6, ξ7, λ, R c , b) and the range of the parameters to be identified is represented as:

[0120]

[0121] wherein ξ i is the optimized parameter ξ1, ξ2, ξ3, ξ4, ξ5, ξ6, ξ7 in the PEMFC model, ξ i min , ξ i max are the upper and lower bounds of ξ1, ξ2, ξ3, ξ4, ξ5, ξ6, ξ7 respectively, λ is the water content of the proton exchange membrane, λ min , λ max are the upper and lower bounds of λ respectively, R c min , R c max are the upper and lower bounds of R c respectively, b min , b max are the upper and lower bounds of b respectively.

[0122] S3: importing the stack data and experimental related data into the PEMFC model.

[0123] It should be noted that the stack data and experimental related data are imported into the constructed PEMFC semi-mechanism semi-data driven model.

[0124] S4: taking the comprehensive error of the actual output voltage of the battery and the output voltage of the mechanism module and the data driven module as the objective function, and determining the constraint condition of the objective function;

[0125] It should be noted that the parameters (ξ1, ξ2, ξ3, ξ4, ξ5, ξ6, ξ7, λ, R c , b) after each iteration are respectively substituted into the constructed mechanism module and the trained data driven module, and the output voltages V M and V D of the two modules are respectively obtained;

[0126] The comprehensive error of the output voltages of the mechanism module and the data driven module and the actual output voltage is taken as the objective function, and the function is as follows:

[0127]

[0128] Wherein, N is the data amount, i is the index, V R is the actual voltage;

[0129] When the error is the smallest, the objective function value is the smallest, and the parameter at this time is considered as the optimal solution.

[0130] S5: As Figure 4 shown, the evolutionary programming algorithm is used to optimize the to-be-identified parameters, and the optimal to-be-identified parameters are obtained;

[0131] Further, the evolutionary programming algorithm optimization to-be-identified parameter expression is:

[0132] It should be noted that the initialization parameter is initialized in a specified range.

[0133] The mutation strategy set, selection strategy set, scaling factor set and crossover factor set of the algorithm are defined, wherein the mutation strategy set includes {DE / rand / 1, DE / rand / 2, DE / best / 1, DE / best / 2, DE / current-to-best / 1, DE / current-to-best / 2, DE / current-to-rand / 1, DE / current-to-rand / 2, DE / rand-to-best / 1, DE / rand-to-best / 2}, the selection strategy set includes {roulette selection, tournament selection, best individual selection, random selection}, the scaling factor set is a set of [0, 2], and the crossover factor set is a set of [0, 1].

[0134] According to the objective function F f , the fitness s i of the individual is calculated, the population quality is evaluated according to the fitness value of the individual, and the evaluation formula is represented as:

[0135]

[0136] Wherein, R is the evaluation result, is the average value of the fitness, s max is the maximum value of the fitness, and NP is the population quantity.

[0137] It is judged whether the iteration number is satisfied, if not, the following operation is performed, if satisfied, the iteration is ended, and the optimal parameter set is obtained.

[0138] According to the above set, the evolutionary programming tree is constructed, which is used for a series of operations such as mutation, crossover and selection on the population.

[0139] S6: According to the output optimal to-be-identified parameter, the optimal control of the proton exchange membrane fuel cell is realized.

[0140] Further, the evolutionary programming tree is constructed according to the mutation strategy set, the selection strategy set, the scaling factor set and the crossover factor set

[0141] It should be noted that the evolutionary programming tree levels are arranged from bottom to top as follows: the mutation selection layer, the mutation operation layer, the crossover layer and the selection layer, which are represented as follows:

[0142] Further, the mutation selection layer is represented as follows:

[0143] It should be noted that a mutation strategy and a scaling factor value are selected in the mutation strategy set and the scaling factor set respectively, and the selection method is determined by the result of the last iteration. If a better result appears in the last iteration, a strategy is randomly selected from the strategies considering the optimal solution, such as DE / best / 1, DE / best / 2, etc., and the scaling factor F is adaptively reduced, and the calculation formula is represented as follows:

[0144]

[0145] If the final result does not change after iteration, a strategy is randomly selected from the strategies not considering the optimal solution, and the scaling factor is adaptively increased, and the calculation formula is represented as follows:

[0146]

[0147] wherein F u is the upper bound of the scaling factor, F l is the lower bound of the scaling factor, I max is the maximum number of iterations, and I c is the current number of iterations.

[0148] Further, the mutation operation layer is represented as follows:

[0149] It should be noted that the mutation operation is performed according to the selected mutation strategy and the scaling factor.

[0150] Further, the crossover layer is represented as follows:

[0151] It should be noted that the value of the crossover factor is determined and the crossover operation is performed according to the determined value of the crossover factor. The value of the crossover factor is also determined by the evaluation result of the population, and the calculation formula is as follows:

[0152]

[0153] wherein CR is the crossover factor, CR u is the upper bound of the crossover factor, CR l is the lower bound of the crossover factor, and R is the evaluation result of the population.

[0154] Further, the selection layer is represented as follows:

[0155] It should be noted that one selection strategy is randomly selected in the selection strategy set and the selection operation is performed;

[0156] Embodiment 2

[0157] The following is an embodiment of the present application, in order to verify the beneficial effects of the present application, scientific demonstration is carried out through simulation experiment.

[0158] The following is an example of modeling and parameter identification based on the mechanism and data-driven PEMFC fusion modeling method provided by the present application:

[0159] Firstly, the mechanism module of the battery is constructed based on the basic mechanism of the proton exchange membrane fuel cell, and secondly, the data-driven module of the battery is constructed combined with experimental data, the experimental data is substituted into the improved RNN model constructed, the data-driven module is trained, and the absolute error is taken as the loss function, and the calculation formula is as follows:

[0160]

[0161] Wherein, Y is the actual output value, and Y is the predicted value obtained by the model.

[0162] Suppose when the value of AE is less than 0.1, the model training is completed.

[0163] The parameters to be identified are set as (ξ1, ξ2, ξ3, ξ4, ξ5, ξ6, ξ7, λ, R c , b), and the range of the parameters to be identified is assumed as follows:

[0164]

[0165] Evolution programming algorithm is used to optimize the to-be-identified parameters, in each iteration process, the to-be-identified parameters are input into the evolution programming tree, in the mutation selection layer, assuming that the last iteration appears better results, then the DE / best / 1 strategy is selected as the mutation strategy, and F l = 0.2, F u = 1.8, I max = 1000, the current iteration number I c = 10, the value of the scaling factor is calculated according to the formula .

[0166] Then enter the mutation operation layer, and perform mutation operation according to the selected mutation strategy and the scaling factor F:

[0167] v i = x best +F×(x r1 -x r2 )

[0168] Wherein, vi is a new individual to be generated, x best is the best individual in the current population, x r1 and x r2 are two different individuals randomly selected from the current population.

[0169] In the crossover layer, CR u = 0.9, CR l = 0.2, the crossover factor CR is calculated to perform the crossover operation. In the selection layer, the best individual strategy is selected to perform the selection operation.

[0170] After each iteration, the parameters are substituted into the mechanism module and the data-driven module, respectively, and the integrated error between the output voltages of the mechanism module and the data-driven module and the actual output voltage is taken as the objective function F f , F f According to the calculation of the fitness s i of the individual, the evaluation result R of the individual is calculated, and the next iteration is prepared. In the optimization process, it is assumed that the iteration number reaches 1000 times, the termination condition is met, the iteration is ended, and the optimal to-be-identified parameters are output, so as to realize the optimal control strategy of the proton exchange membrane fuel cell.

[0171] The PEMFC fusion modeling method based on mechanism and data driving provided by the application proposes a new semi-mechanism semi-data driving model, combines the ideas of the differential evolution algorithm and the genetic programming algorithm to construct an evolutionary programming tree, and judges the construction of the objective function according to the integrated error of the mechanism model and the data-driven model, so that the modeling and parameter identification of the proton exchange membrane fuel cell are effectively realized.

[0172] It should be noted that the above embodiments are only used to illustrate the technical solutions of the application and not to limit the application. Although the application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the application, and they should be covered in the scope of the claims of the application.

[0173] Example 3

[0174] The third embodiment of the application is different from the first two embodiments:

[0175] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the parts of the technical solutions that essentially contribute to the prior art or the parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0176] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered a list of executable instructions for implementing logic functions, and can be specifically embodied in any computer-readable medium for use by an instruction execution system, apparatus, or device, such as a computer-based system, a system including a processor, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions, or in conjunction with these instructions execution systems, apparatuses, or devices. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport programs for use by an instruction execution system, apparatus, or device, or in conjunction with these instruction execution systems, apparatuses, or devices.

[0177] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection having one or more wires (electrical devices), a portable computer diskette (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium can even be paper or other suitable medium on which the program can be printed, as the program can be electronically obtained, for example, by optical scanning of the paper or other medium, followed by editing, interpreting, or otherwise processing, if necessary, in other suitable ways to be electronically obtained, and then stored in the computer memory.

[0178] It should be understood that various parts of the present application can be realized in hardware, software, firmware, or a combination thereof. In the above embodiments, a plurality of steps or methods can be realized in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if realized in hardware, and as in another embodiment, it can be realized by any one or a combination of the following technologies known in the art: discrete logic circuit with logic gate circuit for implementing logic functions on data signals, application specific integrated circuit with suitable combination logic gate circuit, programmable gate array (PGA), field programmable gate array (FPGA), etc.

[0179] Embodiment 4

[0180] For an embodiment of the present application, a GASF-BP-based transmission line vibration signal pin fault detection system is provided, comprising:

[0181] A driving model construction module: a semi-mechanism semi-data driven model of the PEMFC is constructed according to the working principle and experimental data of the proton exchange membrane fuel cell respectively;

[0182] A parameter identification and definition module: the parameters to be identified in the PEMFC model are confirmed and the range of the parameters to be identified is set;

[0183] A data import and management module: the stack data and experimental related data are imported into the PEMFC model;

[0184] A target function design module: the integrated error of the actual output voltage of the battery and the output voltage of the mechanism module and the data driven module is taken as a target function, and the constraint condition of the target function is determined;

[0185] An evolutionary algorithm optimization module: the evolutionary programming algorithm is used to optimize the parameters to be identified, and the optimal parameters to be identified are obtained;

[0186] An optimal control execution module: the optimal control of the proton exchange membrane fuel cell is realized according to the output optimal parameters to be identified.

[0187] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limited. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. A mechanism and data-driven fusion modeling method for PEMFC, characterized in that: The application relates to a method for constructing a semi-mechanism semi-data driven model of a proton exchange membrane fuel cell (PEMFC) and implementing optimal control of the PEMFC. The method comprises the following steps: confirming parameters to be identified in the PEMFC model and setting a range of the parameters to be identified; importing stack data and experimental related data into the PEMFC model; taking the comprehensive error of the actual output voltage of the cell and the output voltage of the mechanism module and the data driven module as a target function, and determining the constraint condition of the target function; optimizing the parameters to be identified by using an evolutionary programming algorithm to obtain optimal parameters to be identified; implementing optimal control of the PEMFC according to the output optimal parameters to be identified. The method comprises the following steps: The mechanism module for establishing the PEMFC model includes modeling considering activation loss V act , ohmic loss V onmic , and concentration difference loss V con , and single cell output voltage V c is expressed as: wherein ξ1, ξ2, ξ3, ξ4, ξ5, ξ6, ξ7 are the parameters to be optimized in the PEMFC model, represents the oxygen concentration in the cathode at the gas reaction interface, R C is the equivalent resistance of electron transfer, R M is the equivalent resistance of proton exchange membrane, b is the parameter coefficient related to the state of the battery in the model, J is the actual current density, J max is the maximum current density, E n is the open circuit voltage of the PEMFC, T is the current gas temperature, i represents the battery current, exp() represents the exponential function with the base of natural logarithm e, ln() represents the logarithmic function with the base of natural logarithm e, E n is the open circuit voltage of the PEMFC, and the calculation formula is represented as: wherein P and P are the partial pressures of hydrogen and oxygen, respectively; equivalent the voltage of the PEMFC stack to the voltage of n single cells, and the stack voltage of the PEMFC is expressed as: V P = nV c The constraint condition of the determination of the target function comprises substituting the parameters after each iteration into the constructed mechanism module and the trained data-driven module respectively to obtain the output voltages V M and V D respectively The comprehensive error of the output voltage of the mechanism module and the data-driven module and the actual output voltage is taken as a target function F f , and the function is represented as: where N is the data amount, i is the index, V R is the actual voltage; when the error is minimum, the value of the target function is minimum, and the parameters are optimal solutions.

2. The mechanism and data-driven fusion modeling method for a PEMFC according to claim 1, characterized in that: The data driven module of the PEMFC model comprises the following steps: collecting experimental data of the PEMFC under various operating conditions; cleaning the data, filling in missing values, and normalizing or standardizing the data; An improved RNN model based on association mining is constructed as a data-driven module, an association mining layer is added before the hidden layer, the input x t and the output (a t-1 , b t-1 ) of the hidden layer at the last moment are taken as the input of the association mining layer. where a t-1 is the rule discovered by the correlation discovery layer at the previous time, b t-1 is the output of the hidden layer at the previous time, the correlation discovery layer determines the correlation between the current inputs and the outputs before them according to the input, and assigns weights according to the correlation degree. The correlation discovery formula is represented as: (a t ,w t ) = ξ(x t ,a t-1 ,b t+1 ) where a t is the rule discovered by the current time-related mining layer, w t is the weight distribution given by the correlation mining layer, and ξ() is the correlation mining layer function. taking the input of the parameters to be identified of the PEMFC model as the output of the voltage of the data driven model; b t = h(U · x t + w t · b t-1 ) where h() is a hidden layer function, U is a parameter matrix from the input layer to the associated mining layer, b t and b t-1 are the outputs of the current time hidden layer and the previous time hidden layer, respectively, and x t is the input at the current time. the formula of the hidden layer is expressed as:

3. The mechanism and data-driven fusion modeling method for a PEMFC according to claim 1, characterized in that: The identified parameters in the confirmed PEMFC model include ξ1, ξ2, ξ3, ξ4, ξ5, ξ6, ξ7, λ, and R c b, and the range of the identified parameters is set as: wherein ξ i are the parameters of the PEMFC model to be optimized ξ1, ξ2, ξ3, ξ4, ξ5, ξ6, ξ7, ξ i min , ξ i max are the upper and lower bounds of ξ1, ξ2, ξ3, ξ4, ξ5, ξ6, ξ7, respectively, λ is the water content of the proton exchange membrane, λ min , λ max are the upper and lower bounds of λ, respectively, R C is the equivalent resistance of electron transfer, R c min , R c max are the upper and lower bounds of R c , respectively, b is the parameter coefficient related to the state of the battery in the model, b min , b max are the upper and lower bounds of b, respectively.

4. The mechanism and data-driven fusion modeling method for a PEMFC according to claim 1, characterized in that: training the model by using experimental data, taking the absolute error as a loss function, and minimizing the error between the predicted value of the model and the actual output value. The method comprises the following steps: initializing the parameters, initializing the population in the specified range; According to the objective function F f The fitness s of the individual is calculated i The population quality is evaluated, and the evaluation formula is represented as wherein R is the evaluation result, s is the average of the fitness, max NP is the population size; defining a mutation strategy set, a selection strategy set, a scaling factor set and a crossover factor set of the algorithm; 5. The mechanism and data-driven fusion modeling method for a PEMFC according to claim 4, characterized in that: judging whether the iteration number is satisfied, if not, constructing an evolutionary programming tree for mutation, crossover and selection operations, if yes, ending the iteration to obtain the optimal parameter set. The method comprises the following steps: the mutation selection layer comprises selecting a mutation strategy and a scaling factor in the mutation strategy set and the scaling factor set respectively, the selection method is determined by the result of the last iteration, if a better result appears, a strategy is randomly selected from the mutation strategy set considering the optimal solution, and the scaling factor F is adaptively reduced, and the calculation formula is expressed as: where F u is the upper bound of the scaling factor, F l is the lower bound of the scaling factor, I max is the maximum number of iterations, I c is the current iteration number; if the last result does not change after iteration, a strategy is randomly selected from the mutation strategy set not considering the optimal solution, and the scaling factor is adaptively increased, and the calculation formula is expressed as: the mutation operation layer comprises executing a mutation operation according to the selected mutation strategy and the scaling factor; where CR is the crossover factor, CR u is the upper bound of the crossover factor, CR l is the lower bound of the crossover factor, and R is the population evaluation result. the crossover layer comprises determining the value of the crossover factor, executing a crossover operation according to the determined value of the crossover factor, and the value of the crossover factor is determined by the evaluation result of the population, and the calculation formula is expressed as:

6. A system based on the mechanism and data-driven PEMFC fusion modeling method according to any one of claims 1-5, characterized in that: the selection layer comprises randomly selecting a selection strategy in the selection strategy set, and executing a selection operation. The method comprises the following steps: a driving model construction module, a parameter identification and definition module, a data import and management module, a target function design module, an evolutionary algorithm optimization module and an optimal control execution module. The driving model construction module constructs a semi-mechanism semi-data driving model of the PEMFC according to working principles and experimental data of the proton exchange membrane fuel cell respectively; The parameter identification and definition module identifies parameters to be identified in the PEMFC model and sets a range of the parameters to be identified; The data import and management module imports stack data and experimental related data into the PEMFC model; The objective function design module takes a comprehensive error of actual output voltage of the cell and output voltage of the mechanism module and the data driving module as an objective function, and determines a constraint condition of the objective function; The evolutionary algorithm optimization module optimizes the parameters to be identified by using an evolutionary programming algorithm to obtain optimal parameters to be identified; The optimal control execution module realizes optimal control of the proton exchange membrane fuel cell according to the output optimal parameters to be identified. 7.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-6 when the computer program is executed by the processor. The processor executes the computer program to realize steps of the method in any one of claims 1 to 5.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize steps of the method in any one of claims 1 to 5.

Images