Battery mechanism model parameter identification method and system based on differential evolution algorithm

Abstract

The invention discloses a battery mechanism model parameter identification method and system based on a differential evolution algorithm, and the method comprises the steps: constructing an agent model training data set; training the neural network agent model to obtain a trained agent model; screening individuals to obtain an initial population; taking the trained proxy model as a target function of a differential evolution algorithm; inputting the initial population into a differential evolution algorithm to obtain an optimal population; inputting each group of data of the optimal population into an experiment system and a simulation system for evaluation to obtain a real voltage error corresponding to each group of data in the optimal population; whether the real voltage error is smaller than a set threshold value or not is judged, and if yes, the to-be-identified parameter corresponding to the minimum value of the real voltage error is directly output; and if not, setting the set proportion part in the evaluated optimal population as a dynamic data set, supplementing the dynamic data set into the proxy model training data set, and returning to the step of training the neural network proxy model.

Description

Technical Field

[0001] The present invention relates to the technical field of battery management, and in particular to a battery mechanism model parameter identification method and system based on a differential evolution algorithm. Background Art

[0002] In the field of battery system analysis and evaluation, battery system simulation is one of the important means to study battery performance and evaluate battery status. Therefore, it is crucial to study the model mechanism of the battery system. In addition, studying the battery mechanism model is of great significance for researchers to deeply understand the chemical and physical processes of the battery in order to optimize battery performance and extend its service life. However, for complex battery mechanism models, many parameters need to be identified. Therefore, it is very important to develop accurate and efficient battery mechanism model identification methods.

[0003] At present, the common methods for identifying battery mechanism model parameters are mainly the following:

[0004] Heuristic optimization algorithms such as genetic algorithm and particle swarm algorithm are used to identify the parameters of the mechanism model through experiments and model error minimization. This method is prone to fall into local optimality and has low convergence accuracy.

[0005] The least squares method is used to identify the parameters of the battery mechanism model by fitting the battery data and estimating the model parameters, thereby improving the prediction accuracy of the battery model. However, it is sensitive to noise and the data quality has a great impact on the results.

[0006] The Kalman filter method estimates the parameter values ​​at each moment using the known terminal voltage data through the state variables in the mathematical model, and then obtains the corresponding model parameter values. Its implementation is very complex.

[0007] Based on the defects of the above methods, there is an urgent need for a battery mechanism model parameter identification method with fast speed, low complexity and high convergence accuracy. Summary of the invention

[0008] In order to solve the deficiencies of the prior art, the present invention provides a battery mechanism model parameter identification method and system based on a differential evolution algorithm;

[0009] On the one hand, a battery mechanism model parameter identification method based on differential evolution algorithm is provided, including:

[0010] Based on each set of sampled parameters to be identified and the corresponding real errors between the experimental system and the simulation system, a training data set for the proxy model is constructed.

[0011] Using the proxy model training data set to train the neural network proxy model, and obtaining a trained proxy model;

[0012] Based on the agent model training data set, individuals are screened to obtain the initial population; the trained agent model is used as the objective function of the differential evolution algorithm; the initial population is input into the differential evolution algorithm to obtain the optimal population;

[0013] Each set of data of the optimal population is input into the experimental system and the simulation system for evaluation, and the actual voltage error corresponding to each set of data in the optimal population is obtained;

[0014] It is determined whether the true voltage error corresponding to each set of data in the optimal population is less than the set threshold. If so, the parameter to be identified corresponding to the minimum value of the true voltage error is directly output; if not, the set proportion of the optimal population after evaluation is set as the dynamic data set, and the dynamic data set is added to the proxy model training data set, and the training step of the neural network proxy model is returned.

[0015] On the other hand, a battery mechanism model parameter identification system based on differential evolution algorithm is provided, including:

[0016] A data set construction module is configured to: construct a proxy model training data set based on each group of sampled parameters to be identified and the corresponding real errors of the experimental system and the simulation system;

[0017] A training module is configured to: train the neural network proxy model using a proxy model training data set to obtain a trained proxy model;

[0018] The differential evolution module is configured to: screen individuals based on the agent model training data set to obtain an initial population; use the trained agent model as the objective function of the differential evolution algorithm; input the initial population into the differential evolution algorithm to obtain the optimal population;

[0019] The evaluation module is configured to: input each set of data of the optimal population into the experimental system and the simulation system for evaluation, and obtain the actual voltage error corresponding to each set of data in the optimal population;

[0020] The judgment module is configured to: judge whether the true voltage error corresponding to each group of data in the optimal population is less than a set threshold value. If so, directly output the to-be-identified parameter corresponding to the minimum value of the true voltage error; if not, set a set proportion of the optimal population after evaluation as a dynamic data set, add the dynamic data set to the proxy model training data set, and return to the training module.

[0021] On the other hand, there is also provided an electronic device, comprising:

[0022] a memory for non-transitory storage of computer-readable instructions; and

[0023] a processor for executing the computer readable instructions,

[0024] When the computer-readable instructions are executed by the processor, the method described in the first aspect is executed.

[0025] On the other hand, a storage medium is provided, which non-temporarily stores computer-readable instructions, wherein when the non-temporary computer-readable instructions are executed by a computer, the method described in the first aspect is executed.

[0026] On the other hand, a computer program product is provided, comprising a computer program, wherein the computer program is used to implement the method described in the first aspect when running on one or more processors.

[0027] The above technical solution has the following advantages or beneficial effects:

[0028] The present invention builds a neural network proxy model based on the correlation of identification parameters, effectively replacing the experimental system and the numerical simulation model. On this basis, the differential evolution algorithm for screening the initial population is used to optimize the parameters to be identified. The parameters to be identified each time are re-evaluated by the experimental system and the numerical simulation system, and then added to the training data set. The proxy model, the initial population and the optimization results are continuously updated to achieve parameter identification.

[0029] The present invention combines the correlation analysis of the parameters to be identified and constructs a three-layer neural network proxy model combined with a basis function model, which reduces the neural network training hyperparameters and reduces the complexity of the proxy model.

[0030] The present invention improves the initial population selection rule of the differential evolution algorithm, and selects better samples as the initial population in the data set of the training agent model, which greatly improves the optimization starting point of the optimization algorithm.

[0031] The present invention implements a logic closed loop by re-experimenting and simulating the optimal population of the optimization algorithm and substituting it back into the training data set, thereby continuously updating the proxy model, the initial population, and the optimal population until the parameters are identified, thereby improving the accuracy of the parameters to be identified. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0033] Figure 1 The flowchart of the proxy optimization method for battery mechanism model parameter identification described in the present invention;

[0034] Figure 2 A schematic diagram of a neural network model based on correlation analysis in the present invention;

[0035] Figure 3 A flow chart of the differential evolution algorithm for screening the initial population implemented in the present invention;

[0036] Figure 4 It is a flow chart of the cyclic update operation in the present invention. DETAILED DESCRIPTION

[0037] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0038] Embodiment 1

[0039] like Figure 1 As shown, this embodiment provides a battery mechanism model parameter identification method based on a differential evolution algorithm;

[0040] The battery mechanism model parameter identification method based on differential evolution algorithm includes:

[0041] S101: constructing a proxy model training data set based on each group of sampled parameters to be identified and the corresponding real errors of the experimental system and the simulation system;

[0042] S102: using the proxy model training data set to train the neural network proxy model to obtain a trained proxy model;

[0043] S103: Based on the agent model training data set, screen individuals to obtain an initial population; use the trained agent model as the objective function of the differential evolution algorithm; input the initial population into the differential evolution algorithm to obtain the optimal population;

[0044] S104: input each set of data of the optimal population into the experimental system and the simulation system for evaluation, and obtain the actual voltage error corresponding to each set of data in the optimal population;

[0045] S105: Determine whether the true voltage error corresponding to each set of data in the optimal population is less than the set threshold. If so, directly output the parameter to be identified corresponding to the minimum value of the true voltage error; if not, set the set proportion of the optimal population after evaluation as the dynamic data set, add the dynamic data set to the proxy model training data set, and return to S102.

[0046] It should be understood that the present invention continuously obtains the optimal population by cyclically updating the training data set, the proxy model and the initial population. When the true voltage error of the optimal individual in the optimal population meets the conditions, the optimization process is considered to be completed, and the optimal individual is the parameter to be identified.

[0047] Furthermore, the S101: constructing a proxy model training data set based on each group of sampled parameters to be identified and the corresponding real errors of the experimental system and the simulation system, includes:

[0048] (1-1): Construct a battery mechanism model of the target battery and obtain M parameters to be identified of the battery mechanism model;

[0049] (1-2): According to the Latin hypercube sampling rule, each parameter to be identified is divided into several intervals according to the value range of the parameter, and sampling is performed evenly in each interval to obtain several groups of sampled parameters to be identified, where each group of sampled parameters to be identified includes M types of parameters to be identified, and each type of parameter to be identified has a corresponding value;

[0050] (1-3): Under the set battery discharge conditions, a discharge experiment is performed on the target battery to obtain the reference voltage corresponding to each group of parameter sampling samples to be identified;

[0051] (1-4): Through numerical simulation, the numerical voltage corresponding to each group of parameter sampling samples to be identified is obtained;

[0052] (1-5): Calculate the difference between the reference voltage and the numerical voltage corresponding to each set of parameter sampling samples to be identified, and take the absolute difference as the error;

[0053] (1-6): Based on each set of sampled parameters to be identified and the corresponding errors, a proxy model training data set is constructed.

[0054] Furthermore, the (1-1): constructs an electrochemical mechanism model of the target battery, and obtains M parameters to be identified of the battery mechanism model, wherein the battery mechanism model to be identified is a classic P2D model, and the M parameters to be identified include physical parameters reflecting the battery itself, such as the radius of the electrode active particles, and parameters reflecting the electrochemical performance of the battery, such as the solid phase electronic conductivity.

[0055] Further, (1-2): according to the Latin hypercube sampling rule, each parameter to be identified is divided into several intervals according to the value range of the parameter, and sampling is performed uniformly in each interval to obtain several groups of sampled parameters to be identified, wherein each group of sampled parameters to be identified includes M types of parameters to be identified, and each type of parameter to be identified is provided with a corresponding value, including:

[0056] Assuming that the value range of the first parameter to be identified is 1 to 100, it is divided into 5 intervals, and the value ranges of the 5 intervals are 1 to 20, 21 to 40, 41 to 60, 61 to 80, and 81 to 100 respectively.

[0057] Among them, the value of each parameter to be identified is the M parameters to be identified of the battery mechanism model. Obtained by spatial Latin hypercube sampling.

[0058] Spatial Latin hypercube sampling refers to dividing the value range of each parameter to be identified into the same interval and performing uniform sampling in each interval to reduce the correlation between various parameters.

[0059] Furthermore, (1-3): under set battery discharge conditions, a discharge experiment is performed on the target battery to obtain a reference voltage corresponding to each group of sampled parameters to be identified; the discharge conditions include: constant current, discharge depth, discharge temperature and cut-off voltage.

[0060] Among them, the discharge experiment refers to connecting the battery to a load device and setting the battery discharge conditions to make it discharge, so as to test the battery's discharge performance and voltage changes during the discharge process.

[0061] The reference voltage u corresponding to each set of identification parameter sampling samples is obtained through battery discharge experiments to construct an accurate sample data set

[0062] Furthermore, (1-4) is as follows: by numerical simulation, a numerical voltage corresponding to each group of sampled parameters to be identified is obtained, wherein the numerical simulation specifically refers to simulating the battery discharge process using a computer based on the battery electrochemical model and according to the conditions of the discharge experiment to observe changes in the electrochemical performance of the battery.

[0063] The numerical voltage corresponding to each set of identification parameter sampling samples is obtained through parallel numerical simulation Constructing a simulation sample dataset

[0064] The battery discharge conditions include constant current I(t), discharge depth, temperature and cut-off voltage. The numerical simulation is consistent with the discharge experiment conditions. The numerical simulation solution methods include but are not limited to finite element and finite difference. The open circuit voltage at time T is recorded as the reference voltage u and the numerical voltage

[0065] Furthermore, the step (1-5): calculating the difference between the reference voltage and the numerical voltage corresponding to each group of sampled parameters to be identified, and taking the difference as the error e, comprises:

[0066]

[0067] Wherein, u represents the reference voltage corresponding to each set of sampled parameters to be identified obtained by the experimental system; Indicates the numerical simulation voltage corresponding to each group of parameter sampling samples to be identified.

[0068] Furthermore, (1-6): based on each set of sampled parameters to be identified and the corresponding errors, a proxy model training data set is constructed, wherein the proxy model training data set It includes: a number of samples, each sample includes: a group of parameters to be identified and corresponding errors, each group of parameters to be identified includes: M parameters to be identified, and each parameter to be identified is set with a corresponding value.

[0069] Furthermore, the step S102: using the proxy model training data set to train the proxy model to obtain a trained proxy model includes:

[0070] (2-1): Calculate the correlation coefficient between each parameter to be identified and the error in each group of parameter sampling samples to be identified, and determine whether the absolute value of the correlation coefficient is greater than the set threshold. If so, it means that the current parameter to be identified is a high-correlation parameter, otherwise, it means that the current parameter to be identified is a low-correlation parameter;

[0071] (2-2): If it is a highly correlated parameter, the current parameter to be identified is used as the input value of the linear basis function model, and the corresponding error is used as the output value of the linear basis function model to obtain the trained linear basis function model;

[0072] (2-3): If it is a low correlation parameter, the current parameter to be identified is used as the input value of the fully connected neural network, and the corresponding error is used as the output value of the fully connected neural network to obtain the trained fully connected neural network;

[0073] (2-4): Based on the trained linear basis function model and the trained fully connected neural network, a trained proxy model is obtained.

[0074] Furthermore, the (2-1) is to calculate the correlation coefficient between each parameter to be identified and the error in each group of sampled parameters to be identified, wherein the correlation coefficient is obtained by a Spearman correlation analysis algorithm.

[0075] The calculation formula of the Spearman correlation analysis algorithm is as follows:

[0076]

[0077] Among them, d i It represents the difference in the rank value of the i-th data pair, and n represents the number of data points.

[0078] According to the parameters to be identified Correlation analysis between the error e and the parameters is divided into high correlation parameters Parameters with low correlation And get the corresponding data set and The neural network proxy model is trained using two data sets. It is generally believed that the absolute value of the correlation coefficient |ρ|≥0.5 is highly correlated, otherwise it is lowly correlated.

[0079] Furthermore, (2-2): if it is a highly correlated parameter, the current parameter to be identified is used as the input value of the linear basis function model, and the corresponding error is used as the output value of the linear basis function model to obtain the trained linear basis function model, and the linear basis function model includes: a first input layer, a first hidden layer and a first output layer connected in sequence.

[0080] In order to take advantage of the strong overall correlation between high correlation parameters and output, a linear basis function model is used to characterize the mapping relationship between the two. The formula is as follows:

[0081]

[0082] Among them, the basis function θ(x i )You can choose Gaussian basis function, etc.

[0083] Based on the dataset Using genetic algorithm to identify the basic model parameters w i , where the loss function Loss in the identification process is defined as the output of the base model f(x) The root mean square error with e in the training data set is calculated as follows:

[0084]

[0085] Where n is the number of data points.

[0086] Furthermore, (2-3): if it is a low correlation parameter, the current parameter to be identified is used as the input value of the fully connected neural network, and the corresponding error is used as the output value of the fully connected neural network to obtain the trained fully connected neural network; wherein the fully connected neural network includes: a second input layer, a second hidden layer and a second output layer connected in sequence.

[0087] The mapping relationship between low-correlation parameters and output is represented by a fully connected neural network, and the nonlinear function in its hidden layer is set to the rectified linear unit ReLU, and the calculation formula is:

[0088] ReLU(x)=max{0,x}+min{0,x}*λ

[0089] Among them, x is the input value of the rectifier linear unit, λ is the preset weight parameter, max{0,x} and min{0,x} represent the maximum and minimum values ​​in {0,x} respectively.

[0090] With the basis function parameters unchanged, the training data set and The parameters of the fully connected neural network are obtained, and the training method of the fully connected neural network adopts the gradient descent method.

[0091] Furthermore, the (2-4): based on the trained linear basis function model and the trained fully connected neural network, a trained proxy model is obtained, including:

[0092] The first output layer and the second output layer are combined into a total output layer to obtain the trained proxy model.

[0093] It should be understood that Figure 2 As shown, the first input layer and the second input layer have a total of M nodes, and the first hidden layer has There is only one node related to There are N related nodes, each of the second hidden nodes is added with a bias Bias1, and the total output layer has one node; the input of the trained proxy model is the parameter to be identified The output is the voltage error e, which is the same as The relevant nodes are basis functions, and the second hidden layer is related to The relevant nodes are activation functions, the complexity of the proxy model is reduced, the training speed is improved, and it is helpful to avoid overfitting.

[0094] Furthermore, the S103: based on the agent model training data set, screening individuals to obtain an initial population includes:

[0095] The data in the agent model training data set are sorted in order from small to large voltage errors, and the top-ranked data sets are selected as the initial population; each data set includes: M parameters to be identified and the corresponding errors. The differential evolution algorithm aims to minimize the error e.

[0096] Furthermore, the S103: the optimal population includes: a plurality of groups of data, each group of data includes: M types of optimized parameters to be identified.

[0097] The objective function of the trained proxy model is implemented using the trained proxy model, with the parameters to be identified as optimization parameters and the minimization of the error e as the goal.

[0098] The data set for training the proxy model is arranged from small to large according to the error, and the better data is selected as the initial population, so as to improve the starting point of the evolutionary algorithm and increase the optimization speed, such as Figure 3 shown.

[0099] like Figure 3 As shown, the steps of differential evolution algorithm include:

[0100] (1): Screening the initial population;

[0101] (2): Determine whether the termination criteria are met. If so, output the optimal population and the optimal individual, and then end;

[0102] (3): If not, perform mutation and crossover operations, process boundary conditions, calculate the objective function, implement the objective function using the trained proxy model, perform selection operations, and return to (2).

[0103] Further, S105: if not, setting a set proportion of the optimal population after evaluation as a dynamic data set, and supplementing the dynamic data set to the proxy model training data set, including:

[0104] Output the optimal population from the differential evolution algorithm Conduct experimental system evaluation best and numerical simulation evaluation In order to improve the optimization speed, the numerical simulation process uses parallel computing or GPU heterogeneous computing;

[0105] Calculate the true voltage error of the output optimal population The individuals in the optimal population are arranged from small to large according to the true voltage error, and the individuals in the top-ranked set proportion are used as the dynamic data set, which is added to the training data set.

[0106] For the supplemented training data set, the basis function model and the neural network are retrained to obtain a proxy model with higher accuracy, and the data is reselected as the initial population of the differential evolution algorithm.

[0107] By cyclically updating the training data set, the proxy model, and the initial population, the optimal population is continuously obtained. When the voltage error of the optimal individual in the optimal population meets the conditions, the optimization process is considered to be over, and the parameters of this individual are identified, such as Figure 4 shown.

[0108] Set the end standard of the entire loop optimization, the calculation formula is:

[0109]

[0110] in, For parameter identification, J is the preset minimum voltage error.

[0111] Embodiment 2

[0112] This embodiment provides a battery mechanism model parameter identification system based on a differential evolution algorithm, including:

[0113] A data set construction module is configured to: construct a proxy model training data set based on each group of sampled parameters to be identified and the corresponding real errors of the experimental system and the simulation system;

[0114] A training module is configured to: train the neural network proxy model using a proxy model training data set to obtain a trained proxy model;

[0115] The differential evolution module is configured to: screen individuals based on the agent model training data set to obtain an initial population; use the trained agent model as the objective function of the differential evolution algorithm; input the initial population into the differential evolution algorithm to obtain the optimal population;

[0116] The evaluation module is configured to: input each set of data of the optimal population into the experimental system and the simulation system for evaluation, and obtain the actual voltage error corresponding to each set of data in the optimal population;

[0117] The judgment module is configured to: judge whether the true voltage error corresponding to each group of data in the optimal population is less than a set threshold value. If so, directly output the to-be-identified parameter corresponding to the minimum value of the true voltage error; if not, set a set proportion of the optimal population after evaluation as a dynamic data set, add the dynamic data set to the proxy model training data set, and return to the training module.

[0118] It should be noted that the above-mentioned data set construction module, training module, differential evolution module, evaluation module and judgment module correspond to steps S101 to S105 in Embodiment 1, and the examples and application scenarios implemented by the above-mentioned modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned Embodiment 1. It should be noted that the above-mentioned modules as part of the system can be executed in a computer system such as a set of computer executable instructions.

[0119] The description of each embodiment in the above embodiments has different emphases. For parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0120] The proposed system can be implemented in other ways. For example, the system embodiment described above is only illustrative, and the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0121] Embodiment 3

[0122] This embodiment also provides an electronic device, including: one or more processors, one or more memories, and one or more computer programs; wherein the processor is connected to the memory, and the one or more computer programs are stored in the memory. When the electronic device is running, the processor executes the one or more computer programs stored in the memory so that the electronic device executes the method described in the above embodiment one.

[0123] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0124] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0125] In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in a processor or an instruction in the form of software.

[0126] The method in the first embodiment can be directly embodied as a hardware processor, or a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it will not be described in detail here.

[0127] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0128] Embodiment 4

[0129] This embodiment further provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the method described in the first embodiment is completed.

[0130] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A battery mechanism model parameter identification method based on differential evolution algorithm, characterized in that: include: Based on each set of sampled parameters to be identified and the corresponding real errors of the experimental system and the simulation system, a training data set for the proxy model is constructed; Using the proxy model training data set to train the neural network proxy model, and obtaining a trained proxy model; Based on the agent model training data set, individuals are screened to obtain the initial population; The trained proxy model is used as the objective function of the differential evolution algorithm; the initial population is input into the differential evolution algorithm to obtain the optimal population; Each set of data of the optimal population is input into the experimental system and the simulation system for evaluation, and the actual voltage error corresponding to each set of data in the optimal population is obtained; It is determined whether the true voltage error corresponding to each set of data in the optimal population is less than the set threshold. If so, the parameter to be identified corresponding to the minimum value of the true voltage error is directly output; if not, the set proportion of the optimal population after evaluation is set as the dynamic data set, and the dynamic data set is added to the proxy model training data set, and the training step of the neural network proxy model is returned.

2. The battery mechanism model parameter identification method based on differential evolution algorithm as claimed in claim 1, characterized in that: Based on each set of sampled parameters to be identified and the corresponding real errors of the experimental system and the simulation system, a proxy model training data set is constructed, including: Construct a battery mechanism model of the target battery and obtain M parameters to be identified of the battery mechanism model; According to the Latin hypercube sampling rule, each parameter to be identified is divided into several intervals according to the value range of the parameter, and sampling is performed evenly in each interval to obtain several groups of sampled parameters to be identified, wherein each group of sampled parameters to be identified includes M types of parameters to be identified, and each type of parameter to be identified has a corresponding value; Under the set battery discharge conditions, a discharge experiment is performed on the target battery to obtain the reference voltage corresponding to each group of parameter sampling samples to be identified; Through numerical simulation, the numerical voltage corresponding to each group of sampled parameters to be identified is obtained; Calculate the difference between the reference voltage and the numerical voltage corresponding to each group of sampled parameters to be identified, and take the absolute difference as the error; Based on each set of sampled parameters to be identified and the corresponding errors, a proxy model training data set is constructed.

3. The battery mechanism model parameter identification method based on differential evolution algorithm as claimed in claim 1, characterized in that: The agent model training data set is used to train the neural network agent model to obtain the trained agent model, including: Calculate the correlation coefficient between each parameter to be identified and the error in each group of parameter sampling samples to be identified, and determine whether the absolute value of the correlation coefficient is greater than the set threshold. If so, it means that the current parameter to be identified is a high correlation parameter, otherwise, it means that the current parameter to be identified is a low correlation parameter; If it is a highly correlated parameter, the current parameter to be identified is used as the input value of the linear basis function model, and the corresponding error is used as the output value of the linear basis function model to obtain the trained linear basis function model; If it is a low correlation parameter, the current parameter to be identified is used as the input value of the fully connected neural network, and the corresponding error is used as the output value of the fully connected neural network to obtain the trained fully connected neural network; Based on the trained linear basis function model and the trained fully connected neural network, a trained proxy model is obtained.

4. The battery mechanism model parameter identification method based on differential evolution algorithm as claimed in claim 3, characterized in that: If it is a highly correlated parameter, the current parameter to be identified is used as an input value of the linear basis function model, and the corresponding error is used as an output value of the linear basis function model to obtain a trained linear basis function model, wherein the linear basis function model includes: a first input layer, a first hidden layer, and a first output layer connected in sequence; A linear basis function model is used to characterize the mapping relationship between high correlation parameters and output. The formula is as follows: Among them, the basis function θ(x i )Select Gaussian basis function; Based on the dataset Using genetic algorithm to identify the basic model parameters w i , where the loss function Loss in the identification process is defined as the output of the base model f(x) The root mean square error with e in the training data set is calculated as follows: Where n is the number of data points.

5. The battery mechanism model parameter identification method based on differential evolution algorithm as claimed in claim 3, characterized in that: If it is a low correlation parameter, the current parameter to be identified is used as the input value of the fully connected neural network, and the corresponding error is used as the output value of the fully connected neural network to obtain the trained fully connected neural network; wherein the fully connected neural network includes: a second input layer, a second hidden layer, and a second output layer connected in sequence; the mapping relationship between the low correlation parameter and the output is represented by a fully connected neural network, and the nonlinear function in the hidden layer is set to the rectified linear unit ReLU, and the calculation formula is: ReLU(x)=max{0,x}+min{0,x}*λ Where x is the input value of the rectifier linear unit, λ is the preset weight parameter, max{0,x} and min{0,x} represent the maximum and minimum values ​​in {0,x} respectively; With the basis function parameters unchanged, the training data set and The parameters of the fully connected neural network are obtained, and the training method of the fully connected neural network adopts the gradient descent method.

6. The battery mechanism model parameter identification method based on differential evolution algorithm as claimed in claim 1, characterized in that: Based on the agent model training data set, individuals are screened to obtain the initial population, including: The data in the proxy model training data set are sorted in ascending order of voltage error, and several groups of data with the highest sorting are selected as the initial population; wherein each group of data includes: M parameters to be identified and corresponding errors.

7. The battery mechanism model parameter identification method based on differential evolution algorithm as claimed in claim 1, characterized in that: The optimal population includes: a plurality of groups of data, each group of data includes: M optimized parameters to be identified; The objective function of the trained proxy model is implemented using the trained proxy model, with the parameters to be identified as optimization parameters and the minimization of the error e as the goal; If not, a set proportion of the optimal population after evaluation is set as a dynamic data set, and the dynamic data set is added to the agent model training data set; Output the optimal population from the differential evolution algorithm Conduct experimental system evaluation best and numerical simulation evaluation In order to improve the optimization speed, the numerical simulation process uses parallel computing or GPU heterogeneous computing; Calculate the true voltage error of the output optimal population The individuals in the optimal population are arranged from small to large according to the true voltage error, and the individuals in the top-ranked set proportion are used as the dynamic data set, which is added to the training data set.

8. The battery mechanism model parameter identification system based on differential evolution algorithm is characterized by: include: A data set construction module is configured to: construct a proxy model training data set based on each group of sampled parameters to be identified and the corresponding real errors of the experimental system and the simulation system; A training module is configured to: train the neural network proxy model using a proxy model training data set to obtain a trained proxy model; The differential evolution module is configured to: screen individuals based on the agent model training data set to obtain an initial population; use the trained agent model as the objective function of the differential evolution algorithm; input the initial population into the differential evolution algorithm to obtain the optimal population; The evaluation module is configured to: input each set of data of the optimal population into the experimental system and the simulation system for evaluation, and obtain the actual voltage error corresponding to each set of data in the optimal population; The judgment module is configured to: judge whether the true voltage error corresponding to each group of data in the optimal population is less than a set threshold value. If so, directly output the to-be-identified parameter corresponding to the minimum value of the true voltage error; if not, set a set proportion of the optimal population after evaluation as a dynamic data set, add the dynamic data set to the proxy model training data set, and return to the training module.

9. An electronic device, comprising: a memory for non-transitory storage of computer readable instructions; as well as a processor for executing the computer readable instructions, When the computer-readable instructions are executed by the processor, the method described in any one of claims 1 to 7 is executed.

10. A storage medium, characterized in that: The computer-readable instructions are non-transitory stored, wherein when the non-transitory computer-readable instructions are executed by a computer, the method according to any one of claims 1 to 7 is performed.

Images