Battery SOH (state of health) and SOC (state of charge) joint estimation method and device and storage medium

Abstract

The invention discloses a battery SOH and SOC joint estimation method, a computer device and a storage medium, and the method can achieve the joint training of an SOH estimation model and an SOC estimation model, enables the SOH estimation model to have the performance of estimating the SOH of a battery, enables the SOC estimation model to have the performance of estimating the SOC of the battery, and enables the SOH estimation model to have the performance of estimating the SOC of the battery. The SOH estimation model and the SOC estimation model can be used to estimate the SOH and the SOC of the battery; through performing hyper-parameter optimization on the trained SOH estimation model and the SOC estimation model, network parameters of the SOH estimation model and the SOC estimation model can be finer, so that the precision of estimating the SOH and the SOC of the battery by the SOH estimation model and the SOC estimation model is improved. The method is widely applied to the technical field of computers.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a battery SOH and SOC joint estimation method, a computer device and a storage medium. Background Art

[0002] Batteries are widely used in power supply and energy storage, transportation, and aerospace equipment. The battery's state of health (SOH) and state of charge (SOC) are two important indicators of the battery, which are closely related to the safety, stable operation and performance of the battery. Since SOH and SOC are indicators reflected by the interaction of substances inside the battery, current technology makes it difficult to analyze batteries on a large scale from the level of physical and chemical principles to obtain indicators such as SOH and SOC. Therefore, current battery-related technologies mainly estimate SOH and SOC based on the change laws of SOH and SOC. However, current battery-related technologies generally have disadvantages such as low estimation accuracy. Summary of the invention

[0003] In view of the technical problems of low estimation accuracy in estimating the SOH and SOC of a battery in current battery-related technologies, the present invention aims to provide a method, a computer device and a storage medium for jointly estimating the SOH and SOC of a battery.

[0004] In one aspect, an embodiment of the present invention includes a method for jointly estimating battery SOH and SOC, the method comprising the following steps:

[0005] Establishing SOH estimation model and SOC estimation model;

[0006] Acquire a first data set and a second data set; the first data set includes a first battery operating parameter and a battery health state collected from the battery during a charging phase of the battery, and the second data set includes a second battery operating parameter and a battery state of charge collected from the battery during a discharging phase of the battery;

[0007] Perform a first training process on the SOH estimation model using the first data set to obtain the trained SOH estimation model;

[0008] Performing a second training process on the SOC estimation model using the second data set to obtain the trained SOC estimation model;

[0009] Hyperparameter optimization is performed on the SOH estimation model and the SOC estimation model.

[0010] Furthermore, the establishing of the SOH estimation model and the SOC estimation model includes:

[0011] Build CNN modules, multi-scale convolutional attention modules, residual modules, and fully connected layers;

[0012] The CNN module, the multi-scale convolutional attention module, the residual module and the fully connected layer are connected in sequence to obtain the SOH estimation model.

[0013] Furthermore, the establishing of the SOH estimation model and the SOC estimation model includes:

[0014] Establish fully connected layers and multiple TCN modules;

[0015] Each of the TNN modules and the fully connected layer is connected in sequence to obtain the SOC estimation model.

[0016] Further, the using the first data set to perform a first training process on the SOH estimation model to obtain the trained SOH estimation model includes:

[0017] Dividing the first data set into a first training set and a first validation set; the first training set and the first validation set respectively include a plurality of the first battery operating parameters and a corresponding plurality of the battery health states;

[0018] Perform multiple rounds of first training processes on the SOH estimation model; in any round of the first training process, input the first battery operating parameter in the first training set into the SOH estimation model for processing, obtain the actual output of the SOH estimation model, take the battery health state in the first training set as the expected output, adjust the network parameters of the SOH estimation model according to the actual output and expected output of the SOH estimation model, input the first battery operating parameter in the first verification set into the SOH estimation model for processing, obtain the actual output of the SOH estimation model, take the battery health state in the first verification set as the expected output, and evaluate the SOH estimation model according to the actual output and expected output of the SOH estimation model;

[0019] Among the SOH estimation models trained in all rounds of the first training process, the SOH estimation model with the best evaluation is obtained as the trained SOH estimation model.

[0020] Further, performing a second training process on the SOC estimation model using the second data set to obtain the trained SOC estimation model includes:

[0021] Dividing the second data set into a second training set and a second validation set; the second training set and the second validation set respectively include a plurality of the second battery operating parameters and a corresponding plurality of the battery states of charge;

[0022] Perform multiple rounds of second training processes on the SOC estimation model; in any round of the second training process, input the actual output of the SOH estimation model in the same training stage and the second battery operating parameter in the second training set into the SOC estimation model for processing, obtain the actual output of the SOC estimation model, take the battery state of charge in the second training set as the expected output, adjust the network parameters of the SOC estimation model according to the actual output and expected output of the SOC estimation model, input the actual output of the SOH estimation model in the same training stage and the second battery operating parameter in the second verification set into the SOC estimation model for processing, obtain the actual output of the SOC estimation model, take the battery state of charge in the second verification set as the expected output, and evaluate the SOC estimation model according to the actual output and expected output of the SOC estimation model;

[0023] Among the SOC estimation models trained in all rounds of the second training process, the SOC estimation model with the best evaluation is obtained as the trained SOC estimation model.

[0024] Furthermore, the performing hyperparameter optimization on the SOH estimation model and the SOC estimation model includes:

[0025] Set the initial evaluation result value and the target evaluation result value;

[0026] Perform at least one round of optimization process; any round of the optimization process includes the following steps:

[0027] Randomly generate the first hyperparameter;

[0028] After configuring the estimation model using the first hyperparameter, performing a third training process on the estimation model to obtain a first evaluation result; wherein the estimation model is at least one of the SOH estimation model and the SOC estimation model;

[0029] Adjusting the first hyperparameter to obtain a second hyperparameter;

[0030] After configuring the estimation model using the second hyperparameter, performing a fourth training process on the estimation model to obtain a second evaluation result;

[0031] Using a simulated annealing algorithm, according to the first evaluation result, the second evaluation result and the evaluation result variable of the current round of optimization process, determine the optimal hyperparameter of the current round of optimization process; wherein, when the current round of optimization process is the first round of optimization process, the value of the evaluation result variable of the current round of optimization process is the value of the initial evaluation result;

[0032] Attenuate the evaluation result variables of this round of optimization process;

[0033] When the attenuated evaluation result variable of this round of optimization process is less than or equal to the target evaluation result value, hyperparameter optimization is performed with the optimal hyperparameter of this round of optimization process, and the entire optimization process is terminated. Otherwise, the next round of optimization process is executed.

[0034] Further, the using of the simulated annealing algorithm to determine the optimal hyperparameters of the current round of optimization process according to the first evaluation result, the second evaluation result and the evaluation result variables of the current round of optimization process includes:

[0035] Set the initial number of iterations, target number of iterations and first probability;

[0036] Perform at least one round of iteration process; any round of the iteration process includes the following steps:

[0037] When the second evaluation result is less than the first evaluation result, randomly determining the second hyperparameter as the optimal hyperparameter of this round of optimization process according to the first probability;

[0038] When the second evaluation result is greater than or equal to the first evaluation result, determine a second probability according to the first evaluation result and the second evaluation result and the evaluation result variable of this round of optimization process, obtain a random number, when the random number is less than the second probability, determine the second hyperparameter as the optimal hyperparameter of this round of optimization process, when the random number is greater than or equal to the second probability, determine the first hyperparameter as the optimal hyperparameter of this round of optimization process;

[0039] The variable of the number of iterations of the current iterative process is decayed; wherein, when the current iterative process is the first iterative process, the value of the variable of the number of iterations of the current iterative process is the initial number of iterations;

[0040] When the attenuated iteration number variable of the current iteration process is less than or equal to the target iteration number, all iteration processes are terminated, otherwise, the next iteration process is executed.

[0041] Furthermore, the battery SOH and SOC joint estimation method further includes:

[0042] During the charging stage of the battery, the third battery operating parameter is collected;

[0043] During the discharge phase of the battery, a fourth battery operating parameter is collected;

[0044] Inputting the third battery operating parameter into the SOH estimation model for processing to obtain an SOH estimation value;

[0045] The SOH estimation value and the fourth battery operating parameter are input into the SOC estimation model for processing to obtain an SOC estimation value.

[0046] On the other hand, an embodiment of the present invention also includes a computer device, including a memory and a processor, the memory is used to store at least one program, and the processor is used to load at least one program to execute the battery SOH and SOC joint estimation method in the embodiment.

[0047] On the other hand, an embodiment of the present invention further includes a computer-readable storage medium storing a program executable by a processor, and the program executable by the processor is used to execute the battery SOH and SOC joint estimation method in the embodiment when executed by the processor.

[0048] The beneficial effects of the present invention are: the battery SOH and SOC joint estimation method in the embodiment can realize the joint training of the SOH estimation model and the SOC estimation model, so that the SOH estimation model has the performance of estimating the SOH of the battery, and the SOC estimation model has the performance of estimating the SOC of the battery, so that the SOH estimation model and the SOC estimation model can be used to estimate the SOH and SOC of the battery; by performing hyperparameter optimization on the trained SOH estimation model and SOC estimation model, the network parameters of the SOH estimation model and the SOC estimation model can be made more refined, which is beneficial to improve the accuracy of the SOH estimation model and the SOC estimation model in estimating the SOH and SOC of the battery. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 Schematic diagram of the steps of jointly training the SOH estimation model and the SOC estimation model and optimizing the hyperparameters in the embodiment;

[0050] Figure 2 It is a schematic diagram of the principle of joint training and hyperparameter optimization of the SOH estimation model and the SOC estimation model in the embodiment;

[0051] Figure 3 Schematic diagram of the steps of using the SOH estimation model and the SOC estimation model to perform SOH and SOC estimation in the embodiment;

[0052] Figure 4 Schematic diagram of the principle of using the SOH estimation model and the SOC estimation model to perform SOH and SOC estimation in the embodiment. DETAILED DESCRIPTION

[0053] Terminology explanation:

[0054] Data-driven model: Data-driven means that the model adapts to the data, and the model changes according to different data. Typical data-driven models include deep data network models, etc.

[0055] Neural Network: A neural network is a computational model that simulates the way the human brain works. It is the basis of the fields of deep learning and machine learning. Neural networks consist of a large number of nodes (or "neurons") that are interconnected in the network and can process complex data inputs and perform various tasks such as classification, regression, pattern recognition, etc.

[0056] Battery Management System: A battery management system is an electronic device that can monitor and manage batteries. It collects and calculates parameters such as voltage, current, temperature, and SOC, and then controls the battery's charge and discharge process, thereby protecting the battery and improving its overall performance.

[0057] SOH: Sate of health is related to the remaining useful life (RUL) of the battery and is usually defined as: current capacity of the battery / nominal capacity.

[0058] SOC: State of charge is used to describe the current battery power, which can be described by the ratio of "remaining available capacity / maximum available capacity".

[0059] CNN: Convolutional Neural Network Features: Use convolutional layers to process data with spatial relationships, such as images. Widely used in image and video recognition, image classification, image segmentation, and natural language processing.

[0060] Residual Block: The residual block solves the gradient vanishing problem in deep neural networks by introducing "quick connections", so that deeper networks can be trained effectively. The core idea of ​​the residual block is to learn the residual function between input and output, rather than directly learning the mapping from input to output. This makes it easier to optimize the network.

[0061] Fully connected layer:

[0062] Attention Mechanism: Attention Mechanism is a commonly used technique in computer science and machine learning that can make the model more accurate and effective when processing sequence data. In traditional neural networks, the output of each neuron depends only on the output of all neurons in the previous layer, while in the attention mechanism, the output of each neuron not only depends on the output of all neurons in the previous layer, but can also be weighted according to different parts of the input data, that is, different weights are given to different parts. This allows the model to pay more attention to the key information in the input sequence, thereby improving the accuracy and efficiency of the model.

[0063] Channel attention module: The channel attention module assigns weights to each feature channel to indicate its importance. During the training process, the channel weights of unimportant features are suppressed, and the channel weights of highly relevant features are enhanced.

[0064] Spatial Attention Module: The spatial attention module places more emphasis on spatial location information and actively explores areas that are spatially related to a given task.

[0065] TCN: Temporal Convolutional Network, temporal convolutional neural network, the TCN model is a time series prediction model based on convolutional neural network, which has certain advantages over CNN in processing time series data.

[0066] Simulated annealing algorithm: The simulated annealing algorithm (SA) is derived from the principle of solid annealing and is a probability-based algorithm. The solid is heated to a sufficiently high temperature and then slowly cooled. When heated, the particles inside the solid become disordered as the temperature rises, the internal energy increases, and the molecules and atoms become more unstable. When slowly cooled, the particles gradually become ordered, the energy decreases, and the atoms become more stable. During the cooling (cooling) process, the solid reaches an equilibrium state at each temperature, and finally reaches the ground state at room temperature, and the internal energy is reduced to a minimum.

[0067] The simulated annealing algorithm starts from a relatively high initial temperature, and with the continuous decrease of the temperature parameter, it combines the probability jump characteristics to randomly search for the global optimal solution of the objective function in the solution space, that is, it can probabilistically jump out of the local optimal solution and eventually tend to the global optimal solution. The simulated annealing algorithm is an optimization algorithm of a serial structure that effectively avoids falling into the local minimum and eventually tends to the global optimal solution by giving the search process a time-varying probability jump that eventually tends to zero.

[0068] When estimating the SOH and SOC of a battery, the following problems are usually encountered:

[0069] When the battery's SOH and SOC are jointly estimated based on data-driven, the accuracy is low, especially in scenarios of fast charging and battery capacity regeneration, where it is more difficult to obtain high estimation accuracy. In addition, if the method of manually extracting SOH attenuation characteristics is used when the battery's SOH and SOC are jointly estimated based on data-driven, it is not only time-consuming and labor-intensive, but also ineffective in some cases. When SOH and SOC are jointly estimated, it is easy to ignore the hyperparameter optimization of the SOH estimation model and the SOC estimation model, thereby limiting the estimation accuracy of SOH and SOC.

[0070] In view of the above problems, this embodiment provides a method for jointly estimating battery SOH and SOC. Figure 1 , the battery SOH and SOC joint estimation method has the following steps:

[0071] P1. Establish SOH estimation model and SOC estimation model;

[0072] P2. Obtain the first data set and the second data set;

[0073] P3. Perform a first training process on the SOH estimation model using the first data set to obtain a trained SOH estimation model;

[0074] P4. Perform a second training process on the SOC estimation model using the second data set to obtain a trained SOC estimation model;

[0075] P5. Perform hyperparameter optimization on the SOH estimation model and the SOC estimation model.

[0076] Steps P1-P5 are steps for joint training and hyperparameter optimization of the SOH estimation model and the SOC estimation model. The principles of steps P1-P5 are as follows: Figure 2 shown.

[0077] In this embodiment, when executing step P1, that is, the step of establishing the SOH estimation model and the SOC estimation model, the following steps may be specifically performed:

[0078] P101. Establish CNN module, multi-scale convolutional attention module, residual module and fully connected layer;

[0079] P102. Connect the CNN module, multi-scale convolutional attention module, residual module and fully connected layer in sequence to obtain the SOH estimation model;

[0080] P103. Establish a fully connected layer and multiple TCN modules;

[0081] P104. Connect each TNN module and the fully connected layer in sequence to obtain the SOC estimation model.

[0082] Steps P101-P102 are steps for establishing a SOH estimation model. Figure 2 The established SOH estimation model is a hybrid neural network model that can automatically capture and enhance the highly correlated SOH decay features in the initial input charging data. Figure 2 ,The SOH estimation model consists of a convolutional neural network module CNN, a multi-scale convolutional attention module, a residual module and a fully connected layer, where the multi-scale convolutional attention module consists of a channel attention module and a spatial attention module, and the channel attention module contains a multi-scale information extraction part, a multi-scale information weight learning part and a multi-scale information weight fusion part. The multi-scale convolutional attention module first captures the features of the input battery charging data using convolution kernels of different sizes on the channel, and then further filters and enhances the features in the spatial dimension.

[0083] Steps P103-P104 are steps for establishing a SOC estimation model. Figure 2 ,The SOC estimation model consists of multiple temporal convolutional neural network modules TCN and ,fully connected layers.

[0084] When executing step P2, a battery (which can be a sample battery whose battery health status and battery charge status are stable and easy to test, and whose type can be a lithium-ion battery) is tested to obtain a first battery operating parameter and a battery health status to form a first data set, and a second battery operating parameter and a battery charge status to form a second data set.

[0085] Specifically, when executing step P2, the sample battery is subjected to multiple complete charge and discharge cycles; during the life cycle of the sample battery, the voltage, current and temperature data in all charge and discharge cycles are recorded.

[0086] For example, the first charging stage, the first discharging stage (the first charging stage and the first discharging stage constitute the first charging and discharging cycle, and so on), the second charging stage, the second discharging stage... the mth charging stage, the mth discharging stage, and so on, are carried out in sequence. m charging and discharging cycles.

[0087] Specifically, in the first charging stage, the voltage V cycle1 、Current I cycle1 and temperature T cycle1 As the first battery operating parameter obtained in the first charging stage, the battery health status SOH of the first charging stage is obtained by detecting the sample battery. cycle1 ; In the second charging stage, the voltage V cycle2 、Current I cycle2 and temperature T cycle2As the first battery operating parameter obtained in the second charging stage, the battery health status SOH of the second charging stage is obtained by detecting the sample battery. cycle2 ...In the mth charging stage, the voltage V is detected by testing the sample battery. cyclem 、Current I cyclem and temperature T cyclem As the first battery operating parameter obtained in the mth charging stage, the battery health status SOH of the mth charging stage is obtained by detecting the sample battery. cyclem The m first battery operating parameters and the m battery health states constitute a first data set.

[0088] Specifically, in the first discharge stage, the voltage V c ' ycle1 、Current I c ' ycle1 and temperature T c ' ycle1 As the second battery operating parameter obtained in the first discharge stage, the battery state of charge SOC of the first discharge stage is obtained by detecting the sample battery. cycle1 ; In the second discharge stage, the voltage V′ is detected on the sample battery cycle2 , current I′ cycle2 and temperature T′ cycle2 As the second battery operating parameter obtained in the second discharge stage, the battery state of charge SOC of the second discharge stage is obtained by detecting the sample battery. cycle2 ...In the mth discharge stage, the voltage V′ is detected by testing the sample battery. cyclem , current I′ cyclem and temperature T′ cyclem As the second battery operating parameter obtained in the mth discharge stage, the battery state of charge SOC of the mth discharge stage is obtained by detecting the sample battery. cyclem The m second battery operating parameters and the m battery charge states constitute a second data set.

[0089] In this embodiment, when executing step P2, preprocessing processes such as removing abnormal values, performing difference sampling in seconds, and standardization can be performed. In this embodiment, it is assumed that the obtained first data set and second data set are both preprocessed data.

[0090] In this embodiment, when executing step P3, that is, performing a first training process on the SOH estimation model using the first data set to obtain a trained SOH estimation model, the following steps may be specifically performed:

[0091] P301. Divide the first data set into a first training set and a first validation set;

[0092] P302. Perform multiple rounds of first training processes on the SOH estimation model; in any round of the first training process, input the first battery operating parameter in the first training set into the SOH estimation model for processing, obtain the actual output of the SOH estimation model, take the battery health state in the first training set as the expected output, adjust the network parameters of the SOH estimation model according to the actual output and expected output of the SOH estimation model, input the first battery operating parameter in the first verification set into the SOH estimation model for processing, obtain the actual output of the SOH estimation model, take the battery health state in the first verification set as the expected output, and evaluate the SOH estimation model according to the actual output and expected output of the SOH estimation model;

[0093] P303. Among the SOH estimation models trained in all rounds of the first training process, obtain the SOH estimation model with the best evaluation as the trained SOH estimation model.

[0094] In step P301, the first data set may be divided into a first training set and a first validation set at a ratio of 4:1. For example, the first battery operating parameter 1 (including the voltage V cycle1 、Current I cycle1 and temperature T cycle1 ) and battery health status SOH cycle1 , the first battery operating parameter 3 (including voltage V cycle3 、Current I cycle3 and temperature T cycle3 ) and battery health status SOH cycle3 , the first battery operating parameters 4 (including voltage V cycle4 、Current I cycle4 and temperature T cycle4 ) and battery health status SOH cycle4 ...The first battery operating parameter m (including voltage V cyclem 、Current I cyclem and temperature T cyclem ) and battery health status SOC cyclem wait The first battery operating parameters and The battery health status is divided into the first training set; the first battery operating parameter 2 includes the voltage V cycle2 、Current I cycle2 and temperature T cycle2 ) and battery health status SOH cycle2 ...The first battery operating parameter m-1 (including voltage V cyclem-1 、Current i cyclem-1 and temperature T cyclem-1 ) and battery health status SOH cyclem-1 wait The first battery operating parameters and The battery health states are divided into the first validation set.

[0095] In step P302, a plurality of rounds of first training processes are performed on the SOH estimation model, each round of the first training process uses a pair of first battery operating parameters and battery health status, A total of the first battery operating parameters and battery health status can be performed The first round of training process.

[0096] In this embodiment, taking one round of the first training process as an example, in this round of the first training process, the first battery operating parameter in the first training set (assuming that the voltage V cycle4 、Current I cycle4 and temperature T cycle4 ) is input into the SOH estimation model for processing to obtain the actual output y of the SOH estimation model cycle4 , using the battery health status SOH in the first training set cycle4 As the expected output, the actual output y of the model is estimated according to SOH cycle4 With the expected output SOH cycle4 , calculate the root mean square error RMSE cycle4 Or mean absolute error MAE cycle4 The error function value is equal, and the network parameters of the SOH estimation model are adjusted according to the error function value. Then the first battery operating parameter in the first validation set (assuming that the voltage V cycle2 、Current I cycle2 and temperature T cycle2 ) is input into the SOH estimation model for processing to obtain the actual output y of the SOH estimation model cycle2 , using the battery health status SOH in the first validation set cycle2 As the expected output, the actual output y of the model is estimated according to SOH cycle2 With the expected output SOH cycle2 , evaluate the SOH estimation model and obtain the evaluation value.

[0097] In step P302, multiple rounds of the first training process may be performed, and the total number of rounds of the first training process may be 50 rounds, and each round of the first training process has a corresponding evaluation value. When executing step P303, among the SOH estimation models trained in all rounds of the first training process, the SOH estimation model with the best evaluation (for example, the largest evaluation value) is obtained as the trained SOH estimation model obtained by executing step P3.

[0098] In this embodiment, when executing step P4, that is, performing a second training process on the SOC estimation model using the second data set to obtain a trained SOC estimation model, the following steps may be specifically performed:

[0099] P401. Divide the second data set into a second training set and a second validation set;

[0100] P402. Perform multiple rounds of second training processes on the SOC estimation model; in any round of the second training process, the actual output of the SOH estimation model in the same training stage and the second battery operating parameters in the second training set are input into the SOC estimation model for processing, the actual output of the SOC estimation model is obtained, the battery state of charge in the second training set is used as the expected output, and the network parameters of the SOC estimation model are adjusted according to the actual output and the expected output of the SOC estimation model, the actual output of the SOH estimation model in the same training stage and the second battery operating parameters in the second verification set are input into the SOC estimation model for processing, the actual output of the SOC estimation model is obtained, the battery state of charge in the second verification set is used as the expected output, and the SOC estimation model is evaluated according to the actual output and the expected output of the SOC estimation model;

[0101] P403. Among the SOC estimation models trained in all rounds of the second training process, obtain the SOC estimation model with the best evaluation as the trained SOC estimation model.

[0102] The principles of steps P401-P403 are similar to those of steps P301-P303.

[0103] In step P401, the second data set may be divided into a second training set and a second validation set at a ratio of 4:1. For example, the second battery operating parameter 1 (including the voltage V′) in the second data set cycle1 , current I′ cycle1 and temperature T′c ycle1 ) and battery state of charge SOC cycle1 , the second battery operating parameters 3 (including voltage V' cycle3 , current I′ cycle3 and temperature T′ cycle3 ) and battery state of charge SOC cycle3 , the second battery operating parameters 4 (including voltage V' cycle4 , current I′ cycle4 and temperature T′ cycle4 ) and battery state of charge SOC cycle4 ...The first battery operating parameter m (including voltage V' cyclem , current I′ cyclem and temperature T′ cyclem ) and battery state of charge SOCcyclem wait The second battery operating parameters and The battery state of charge is divided into the second training set; the second battery operating parameter 2 (including the voltage V′ cycle2 , current I′ cycle2 and temperature T′ cycle2 ) and battery state of charge SOC cycle2 ...The second battery operating parameter m-1 (including voltage V' cyclem-1 , current I′ cyclem-1 and temperature T′ cyclem-1 ) and battery state of charge SOC cyclem-1 wait The second battery operating parameters and The battery health status is divided into the second validation set.

[0104] In step P402, a plurality of rounds of second training processes are performed on the SOC estimation model, each round of the second training process uses a set of second battery operating parameters, battery state of charge, and actual output of the SOH estimation model in the same training stage. A total of the second battery operating parameters and battery health status can be performed The second round of training process.

[0105] In this embodiment, taking one round of the second training process as an example, in this round of the second training process, the second battery operating parameter in the second training set (assuming that the voltage V c ' ycle4 、Current I c ' ycle4 and temperature T c ' ycle4 ) and the actual output of the SOH estimation model (e.g., y obtained in step P302 cycle4 ) is input into the SOC estimation model for processing to obtain the actual output y′ of the SOC estimation model cycle4 , using the battery state of charge SOC in the second training set cycle4 As the expected output, the actual output y′ of the SOC estimation model is cycle4 With the expected output SOC cycle4 , calculate the root mean square error RMSE′ cycle4 Or mean absolute error MAE′ cycle4 The error function value is equal, and the network parameters of the SOC estimation model are adjusted according to the error function value, and then the second battery operating parameters in the second verification set (assuming that the voltage V′ is selected cycle2 , current I′ cycle2 and temperature T′ cycle2 ) is input into the SOC estimation model for processing to obtain the actual output y′ of the SOC estimation modelcycle2 , using the battery state of charge SOC in the second validation set cycle2 As the expected output, the actual output y′ of the SOC estimation model is cycle2 With the expected output SOC cycle2 , evaluate the SOC estimation model and obtain an evaluation value.

[0106] In step P402, multiple rounds of the second training process may be performed, and the total number of rounds of the second training process may be 50 rounds, and each round of the second training process has a corresponding evaluation value. When executing step P403, among the SOC estimation models trained in all rounds of the second training process, the SOC estimation model with the best evaluation (for example, the largest evaluation value) is obtained as the trained SOC estimation model obtained by executing step P4.

[0107] By executing steps P301-P303 and P401-P403, joint training of the SOH estimation model and the SOC estimation model can be achieved, so that the SOH estimation model has the performance of estimating the SOH of the battery, and the SOC estimation model has the performance of estimating the SOC of the battery, so that the SOH estimation model and the SOC estimation model can be used to estimate the SOH and SOC of the battery.

[0108] After executing steps P1-P4, execute step P5 to perform hyperparameter optimization on the trained SOH estimation model and SOC estimation model. By performing hyperparameter optimization, the network parameters of the SOH estimation model and the SOC estimation model can be made more refined, which is conducive to improving the accuracy of the SOH estimation model and the SOC estimation model in estimating the SOH and SOC of the battery.

[0109] In this embodiment, when executing step P302 or P402 to calculate the root mean square error or the mean absolute error, the formula used is:

[0110]

[0111] Among them, RMSE is the root mean square error, y′ i is the value of the i-th dimension in the expected output, y i is the value of the i-th dimension in the actual output, and n is the dimension of the expected output and the actual output.

[0112] In this embodiment, when executing step P5, that is, the step of optimizing the hyper parameters of the SOH estimation model and the SOC estimation model, the following steps may be specifically performed:

[0113] P501. Set the initial evaluation result value V0 and the target evaluation result value V f ;

[0114] P502. Perform at least one round of optimization process.

[0115] Any round of optimization process includes the following steps:

[0116] P50201. Randomly generate the first hyperparameter x;

[0117] P50202. After configuring the estimation model using the first hyperparameter x, perform a third training process on the estimation model to obtain a first evaluation result f(x);

[0118] P50203. Adjust the first hyperparameter x to obtain the second hyperparameter x′;

[0119] P50204. After configuring the estimation model using the second hyperparameter x′, perform a fourth training process on the estimation model to obtain a second evaluation result f(x′);

[0120] P50205. Use the simulated annealing algorithm to determine the optimal hyperparameters of the current round of optimization process based on the first evaluation result f(x) and the second evaluation result f(x′) and the evaluation result variables of the current round of optimization process;

[0121] P50206. Attenuate the evaluation result variable V of this round of optimization process;

[0122] P50207. When the attenuated evaluation result variable V of this round of optimization process is less than or equal to the target evaluation result value V f , perform hyperparameter optimization with the optimal hyperparameters of this round of optimization process, and end the execution of all optimization processes. Otherwise, execute the next round of optimization process.

[0123] In step P501, the initial evaluation result value V0 and the target evaluation result value V f are all fixed values, where the initial evaluation result value V0 can be a larger value, and the target evaluation result value V f Can be a smaller value.

[0124] In step P502, multiple rounds of optimization are performed, and each round of optimization performs steps P50201-P50207. One round of optimization is taken as an example for description.

[0125] In this embodiment, the optimization process in step P502 can be performed on both the SOH estimation model and the SOC estimation model. Therefore, in steps P50201-P50207, the SOH estimation model and the SOC estimation model are collectively referred to as estimation models. When the hyperparameters of the SOH estimation model or the SOC estimation model are actually optimized, the SOH estimation model or the SOC estimation model can be substituted into the estimation model in steps P50201-P50207.

[0126] In this embodiment, an evaluation result variable V is set, and the specific value of the evaluation result variable V changes with each round of optimization process. In the first round of optimization process, the evaluation result variable V is set to be the same as the initial evaluation result value V0, that is, V=V0.

[0127] In step P50201, a first hyperparameter x is randomly generated. Specifically, the type of the first hyperparameter x can be a learning rate, a number of network layers, a number of neurons, etc. The first hyperparameter x can be a vector composed of numerical values ​​of the learning rate, the number of network layers, the number of neurons, etc., and the specific numerical values ​​of the learning rate, the number of network layers, the number of neurons, etc. in the vector are randomly determined.

[0128] In step P50202, the estimation model is configured using the first hyperparameter x. Specifically, since the type of the first hyperparameter x in this embodiment is a learning rate, a number of network layers, a number of neurons, etc., when configuring, the values ​​of the learning rate, the number of network layers, the number of neurons, and other parameters of the estimation model are modified to the values ​​corresponding to the first hyperparameter x. After completing the configuration of the hyperparameters, a third training process is performed on the estimation model. Specifically, if the estimation model is a SOH estimation model, the third training process performed is the same as the first training process performed in step S3 (the training set used, etc. may be different); if the estimation model is a SOC estimation model, the third training process performed is the same as the second training process performed in step S4 (the training set used, etc. may be different).

[0129] Referring to the principles of step P302 and step P402, after the first training process or the second training process is completed, an evaluation value of the SOH estimation model or the SOC estimation model is obtained. Therefore, in step P50202, similarly, the SOH estimation model or the SOC estimation model will also obtain an evaluation value, which is called the first evaluation result f(x).

[0130] In step P50203, the first hyperparameter x is fine-tuned to obtain a second hyperparameter x′. The type of the second hyperparameter x′ is still the learning rate, the number of network layers, the number of neurons, etc., but the value is different from the first hyperparameter x. Specifically, the first hyperparameter x can be adjusted by growth and other adjustments to obtain the second hyperparameter x′.

[0131] In step P50204, based on the same principle as step P50202, the estimation model can be configured using the second hyperparameter x′, and the fourth training process is performed on the configured estimation model. Specifically, if the estimation model is an SOH estimation model, the fourth training process performed is the same as the first training process performed in step S3 (the training set used, etc. may be different); if the estimation model is an SOC estimation model, the fourth training process performed is the same as the second training process performed in step S4 (the training set used, etc. may be different).

[0132] In step P50204, based on the same principle as step P50202, the evaluation value of the SOH estimation model or the SOC estimation model, that is, the second evaluation result f(x′), can also be obtained.

[0133] In step P50205, the simulated annealing algorithm is used to determine the optimal hyperparameter x of the current round of optimization process based on the first evaluation result f(x) and the second evaluation result f(x′) and the evaluation result variables of the current round of optimization process. best .

[0134] In step P50206, the evaluation result variable V of this round of optimization process is decayed. Specifically, a fixed decay rate rate can be set, and the evaluation result variable V is multiplied by the decay rate rate to decay, that is, V=V×rate.

[0135] In step P50207, the attenuated evaluation result variable V of this round of optimization process is compared with the target evaluation result value V f For comparison. If V≤V f , then the optimal hyperparameter x of this round of optimization process best Perform hyperparameter optimization, end all optimization processes, and do not execute the next round of optimization; if V>V f , execute the next round of optimization process, that is, jump back to step P50201 and start re-executing a new round of steps P50201-P50207.

[0136] Specifically, in step P50207, the optimal hyperparameter x best When performing hyperparameter optimization, the hyperparameters of the estimated model, such as the learning rate, number of network layers, and number of neurons, can be set to the same value as the optimal hyperparameter x best same.

[0137] In this embodiment, when executing step P50205, that is, using the simulated annealing algorithm to determine the optimal hyperparameters of the current round of optimization process according to the first evaluation result, the second evaluation result and the evaluation result variables of the current round of optimization process, the following steps can be specifically performed:

[0138] P5020501. Set the initial number of iterations L0 and the target number of iterations L f and the first probability P1;

[0139] P5020502. Perform at least one round of iteration.

[0140] Any round of iteration process includes the following steps:

[0141] P502050201. When the second evaluation result f(x′) is less than the first evaluation result f(x), the second hyperparameter is randomly determined as the optimal hyperparameter of this round of optimization process according to the first probability P1;

[0142] P502050202. When the second evaluation result f(x′) is greater than or equal to the first evaluation result f(x), determine the second probability P2 according to the first evaluation result f(x) and the second evaluation result f(x′) and the evaluation result variable V of this round of optimization process, obtain a random number r, and when the random number r is less than the second probability P2, determine the second hyperparameter x′ as the optimal hyperparameter of this round of optimization process; when the random number r is greater than or equal to the second probability P2, determine the first hyperparameter x as the optimal hyperparameter of this round of optimization process;

[0143] P502050203. Attenuate the number of iterations variable L of this round of iteration process; wherein, when this round of iteration process is the first round of iteration process, the value of the number of iterations variable L of this round of iteration process is the initial number of iterations L0;

[0144] P502050204. When the attenuated number of iterations of this round of iterations is less than or equal to the target number of iterations L f , end the execution of all iterations, otherwise, execute the next round of iterations.

[0145] In step P5020501, the initial number of iterations L0 and the target number of iterations L f and the first probability P1 are fixed values, where the initial number of iterations L0 can be a larger value, and the target number of iterations L f Can be 0.

[0146] In step P5020502, multiple rounds of iterations are performed, and each round of iterations performs steps P502050201 to P502050204. One round of iterations is taken as an example for description.

[0147] In this embodiment, an iteration number variable L is set, and the specific value of the iteration number variable L changes with each round of iteration process. In the first round of iteration process, the iteration number variable L is set to be the same as the initial iteration number L0, that is, L=L0.

[0148] In steps P502050201-P502050202, the magnitude relationship between the first evaluation result f(x) and the second evaluation result f(x′) is determined. If the second evaluation result f(x′) is smaller than the first evaluation result f(x), that is, f(x′)<f(x), then step P502050201 is executed, and the second hyperparameter x′ is randomly determined as the optimal hyperparameter x in this round of optimization process according to the first probability P1. best , that is, it is possible that x best = x′ (the probability is the first probability P1), it is also possible that x best Take other values ​​(the probability is 1-P1); if the second evaluation result f(x′) is greater than or equal to the first evaluation result f(x), that is, f(x′)≥f(x), then execute step P502050202, and the second probability P2 can be calculated according to the following formula:

[0149]

[0150] Then generate a random number r, which satisfies 0<r<1, and determine the size relationship between the random number r and the second probability P2; if the random number r is less than the second probability P2, that is, r<P2, then the second hyperparameter x′ is determined as the optimal hyperparameter x in this round of optimization process. best , that is, x best =x′; if the random number r is greater than or equal to the second probability P2, that is, r≥P2, then the first hyperparameter x is determined as the optimal hyperparameter x for this round of optimization process best , that is, x best =x.

[0151] In step P502050203, the iteration number variable L of this round of iteration process is decayed. Specifically, 1 can be subtracted from the original iteration number variable L, that is, L=L-1.

[0152] In step P502050204, the attenuated iteration number variable L of this round of iteration process is compared with the target iteration number L f For comparison. If L≤L f (In this embodiment, the target number of iterations L f is 0, that is, L≤0), then the entire iteration process ends and the next iteration process is not performed; if L>L f (ie L>0), execute the next round of iteration process, that is, jump back to step P502050201 and start re-executing a new round of steps P502050201-P502050204.

[0153] In this embodiment, steps P502050201-P502050204 are steps nested in steps P50201-P50207. For example, in this embodiment, the iteration number variable L needs to decay from the initial iteration number L0 to 0, and needs to decay L0 times. That is, when each round of optimization process (i.e., steps P50201-P50207) is executed, L0 rounds of iteration process (i.e., steps P502050201-P502050204) are executed.

[0154] In this embodiment, by executing multiple rounds of optimization processes such as steps P50201-P50207, and executing multiple rounds of iterative processes such as steps P502050201-P502050204 in each round of optimization, it is possible to perform hyperparameter optimization of the SOH estimation model and the SOC estimation model using a simulated annealing algorithm, and the network parameters of the SOH estimation model and the SOC estimation model can be made more refined, which is beneficial to improving the accuracy of the SOH estimation model and the SOC estimation model in estimating the SOH and SOC of the battery.

[0155] In this embodiment, the SOH estimation model and the SOC estimation model that have been jointly trained and hyperparameter optimized in steps P1-P5 can be used to estimate the battery SOH and SOC. Figure 3 , you can perform the following steps:

[0156] S1. During the battery charging phase, the third battery operating parameters are collected;

[0157] S2. During the discharge phase of the battery, the fourth battery operating parameter is collected;

[0158] S3. Inputting the third battery operating parameter into the SOH estimation model for processing to obtain an SOH estimation value;

[0159] S4. Input the SOH estimation value and the fourth battery operating parameter into the SOC estimation model for processing to obtain the SOC estimation value.

[0160] Steps S1-S4 are steps for performing SOH and SOC estimation using the SOH estimation model and the SOC estimation model. The principles of steps S1-S4 are as follows: Figure 4 shown.

[0161] When executing step S1, the battery to be estimated (which may be a battery whose battery health status and battery charge status are unknown, and whose type may be a lithium-ion battery) is detected to obtain the third battery operating parameter and the fourth battery operating parameter.

[0162] Specifically, during a complete charging phase of the battery to be estimated, the voltage V″ of the battery to be estimated can be obtained by detecting the battery to be estimated. cycle , current I″ cycleand temperature T″ cycle As the third battery operating parameter; in a complete discharge phase of the battery to be estimated, the voltage V″′ is detected on the battery to be estimated cycle , current I″′ cycle and temperature T″′ cycle As the fourth battery operating parameter.

[0163] In this embodiment, when executing step S1, preprocessing processes such as removing abnormal values, performing difference sampling in seconds, and standardization can be performed. In this embodiment, it is assumed that the obtained third data set and fourth data set are both preprocessed data.

[0164] Reference Figure 4 In step S3, the third battery operating parameter is input into the SOH estimation model for processing. The SOH estimation model after joint training and hyperparameter optimization has the performance of processing the third battery operating parameter and outputting the SOH estimation value.

[0165] Reference Figure 4 In step S4, the SOH estimation value output by the SOH estimation model and the fourth battery operating parameter are input into the SOC estimation model for processing. The SOC estimation model after joint training and hyperparameter optimization has the performance of processing the fourth battery operating parameter and outputting the SOC estimation value.

[0166] Reference Figure 4 , the SOH estimation model and the SOC estimation model process the third battery operating parameter and the fourth battery operating parameter to obtain an SOH estimation value and an SOC estimation value, wherein the SOH estimation value represents an estimation of the SOH of the battery to be estimated, and the SOC estimation value represents an estimation of the SOC of the battery to be estimated.

[0167] In this embodiment, by executing steps S1-S4, the deep neural network structure of the SOH estimation model and the SOC estimation model can be used to automatically extract feature characteristics and automatically capture the attenuation characteristics of SOH and SOC in the input battery charging and discharging data; in addition, by using a hybrid neural network including a multi-scale convolutional attention module and a residual module as the SOH estimation model and the SOC estimation model, the SOH estimation accuracy and stability can be improved, thereby improving the accuracy of SOC estimation in the joint estimation process of SOH and SOC.

[0168] In summary, executing steps P1-P5 and S1-S4 has the following advantages:

[0169] 1. Use deep neural networks to automatically capture the SOH and SOC attenuation characteristics in lithium-ion battery charge and discharge data, and complete the estimation model building.

[0170] 2. The structure of the multi-scale convolutional attention module: It includes a channel attention module and a spatial attention module, where the channel attention module includes a multi-scale information extraction part, a multi-scale information weight learning part, and a multi-scale information weight fusion part.

[0171] 3. Use the SOH estimation hybrid neural network model structure consisting of a CNN module, a multi-scale convolutional attention module, a residual module, and a fully connected layer.

[0172] 4. Use the SOC estimation neural network model structure containing multiple TCN modules and fully connected layers.

[0173] 5. Use the simulated annealing optimization algorithm to tune the hyperparameters of the SOH estimation model and the SOC estimation model respectively.

[0174] 6. Both the SOH estimation part and the SOC estimation part are estimated through deep neural network modeling. At the same time, the SOH and SOC attenuation characteristics in the battery charging and discharging data are automatically extracted by utilizing the deep neural network's ability to automatically extract features.

[0175] 7. The SOH estimation model uses a hybrid model that includes a multi-scale convolutional attention module and a residual module, which makes the SOH estimation model more accurate and stable.

[0176] 8. The hyperparameters of the SOH estimation model and the SOC estimation model are tuned respectively through the simulated annealing optimization algorithm to further improve the accuracy of the SOH and SOC joint estimation model.

[0177] A computer program for executing the method for jointly estimating the battery SOH and SOC in this embodiment can be written and written into a computer device or storage medium. When the computer program is read out and run, the method for jointly estimating the battery SOH and SOC in this embodiment is executed, thereby achieving the same technical effect as the method for jointly estimating the battery SOH and SOC in the embodiment.

[0178] It should be noted that, unless otherwise specified, when a feature is referred to as being "fixed" or "connected" to another feature, it may be directly fixed or connected to the other feature, or it may be indirectly fixed or connected to the other feature. In addition, the descriptions of up, down, left, right, etc. used in the present disclosure are only relative to the relative positional relationship of the components of the present disclosure in the accompanying drawings. The singular forms of "a", "" and "the" used in the present disclosure are also intended to include the plural forms, unless the context clearly indicates other meanings. In addition, unless otherwise defined, all technical and scientific terms used in this embodiment have the same meaning as those generally understood by those skilled in the art. The terms used in the specification of this embodiment are only for describing specific embodiments and are not intended to limit the present invention. The term "and / or" used in this embodiment includes any combination of one or more related listed items.

[0179] It should be understood that, although the term first, second, third etc. may be adopted to describe various elements in the present disclosure, these elements should not be limited to these terms. These terms are only used to distinguish the same type of elements from each other. For example, without departing from the scope of the present disclosure, the first element may also be referred to as the second element, and similarly, the second element may also be referred to as the first element. The use of any and all examples or exemplary language ("for example", "such as" etc.) provided by the present embodiment is only intended to better illustrate embodiments of the present invention, and unless otherwise required, the scope of the present invention will not be limited.

[0180] It should be appreciated that embodiments of the present invention may be implemented or enforced by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. The method may be implemented in a computer program using standard programming techniques - including a non-transitory computer-readable storage medium configured with a computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner - according to the methods and drawings described in the specific embodiments. Each program may be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system. However, if desired, the program may be implemented in assembly or machine language. In any case, the language may be a compiled or interpreted language. In addition, the program may be run on a programmed dedicated integrated circuit for this purpose.

[0181] In addition, the operations of the process described in this embodiment may be performed in any suitable order, unless otherwise indicated in this embodiment or otherwise clearly contradicted by the context. The process described in this embodiment (or variations and / or combinations thereof) may be performed under the control of one or more computer systems configured with executable instructions, and may be implemented as a code (e.g., executable instructions, one or more computer programs, or one or more applications) executed on one or more processors in common, by hardware or a combination thereof. A computer program includes a plurality of instructions that may be executed by one or more processors.

[0182] Further, the method can be implemented in any type of computing platform that is operably connected to a suitable computer, including but not limited to a personal computer, a minicomputer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, etc. Various aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, an optical read and / or write storage medium, RAM, ROM, etc., so that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the process described herein. In addition, the machine-readable code, or a portion thereof, can be transmitted via a wired or wireless network. When such media includes instructions or programs that implement the above steps in conjunction with a microprocessor or other battery SOH and SOC joint estimator, the invention of this embodiment includes these and other different types of non-transitory computer-readable storage media. When programmed according to the methods and techniques of the present invention, the present invention also includes the computer itself.

[0183] The computer program can be applied to input data to perform the functions of the present embodiment, thereby converting the input data to generate output data stored in a non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the present invention, the converted data represents a physical and tangible object, including a specific visual depiction of the physical and tangible object produced on the display.

[0184] The above are only preferred embodiments of the present invention. The present invention is not limited to the above embodiments. As long as the technical effects of the present invention are achieved by the same means, any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the scope of protection of the present invention. Within the scope of protection of the present invention, its technical solutions and / or implementation methods may have various modifications and changes.

Claims

1. A method for jointly estimating battery SOH and SOC, characterized in that: The battery SOH and SOC joint estimation method comprises: Establishing SOH estimation model and SOC estimation model; Acquire a first data set and a second data set; the first data set includes a first battery operating parameter and a battery health state collected from the battery during a charging phase of the battery, and the second data set includes a second battery operating parameter and a battery state of charge collected from the battery during a discharging phase of the battery; Perform a first training process on the SOH estimation model using the first data set to obtain the trained SOH estimation model; Performing a second training process on the SOC estimation model using the second data set to obtain the trained SOC estimation model; Hyperparameter optimization is performed on the SOH estimation model and the SOC estimation model.

2. The battery SOH and SOC joint estimation method according to claim 1, characterized in that: The establishing of the SOH estimation model and the SOC estimation model comprises: Build CNN modules, multi-scale convolutional attention modules, residual modules, and fully connected layers; The CNN module, the multi-scale convolutional attention module, the residual module and the fully connected layer are connected in sequence to obtain the SOH estimation model.

3. The battery SOH and SOC joint estimation method according to claim 2, characterized in that: The establishing of the SOH estimation model and the SOC estimation model comprises: Establish fully connected layers and multiple TCN modules; Each of the TNN modules and the fully connected layer is connected in sequence to obtain the SOC estimation model.

4. The battery SOH and SOC joint estimation method according to claim 1, characterized in that: The step of performing a first training process on the SOH estimation model using the first data set to obtain the trained SOH estimation model includes: Dividing the first data set into a first training set and a first validation set; the first training set and the first validation set respectively include a plurality of the first battery operating parameters and a corresponding plurality of the battery health states; Perform multiple rounds of first training processes on the SOH estimation model; in any round of the first training process, input the first battery operating parameter in the first training set into the SOH estimation model for processing, obtain the actual output of the SOH estimation model, take the battery health state in the first training set as the expected output, adjust the network parameters of the SOH estimation model according to the actual output and expected output of the SOH estimation model, input the first battery operating parameter in the first verification set into the SOH estimation model for processing, obtain the actual output of the SOH estimation model, take the battery health state in the first verification set as the expected output, and evaluate the SOH estimation model according to the actual output and expected output of the SOH estimation model; Among the SOH estimation models trained in all rounds of the first training process, the SOH estimation model with the best evaluation is obtained as the trained SOH estimation model.

5. The battery SOH and SOC joint estimation method according to claim 4, characterized in that: The performing a second training process on the SOC estimation model using the second data set to obtain the trained SOC estimation model includes: Dividing the second data set into a second training set and a second validation set; the second training set and the second validation set respectively include a plurality of the second battery operating parameters and a corresponding plurality of the battery states of charge; Perform multiple rounds of second training processes on the SOC estimation model; in any round of the second training process, input the actual output of the SOH estimation model in the same training stage and the second battery operating parameter in the second training set into the SOC estimation model for processing, obtain the actual output of the SOC estimation model, take the battery state of charge in the second training set as the expected output, adjust the network parameters of the SOC estimation model according to the actual output and expected output of the SOC estimation model, input the actual output of the SOH estimation model in the same training stage and the second battery operating parameter in the second verification set into the SOC estimation model for processing, obtain the actual output of the SOC estimation model, take the battery state of charge in the second verification set as the expected output, and evaluate the SOC estimation model according to the actual output and expected output of the SOC estimation model; Among the SOC estimation models trained in all rounds of the second training process, the SOC estimation model with the best evaluation is obtained as the trained SOC estimation model.

6. The battery SOH and SOC joint estimation method according to claim 1, characterized in that: The performing hyperparameter optimization on the SOH estimation model and the SOC estimation model includes: Set the initial evaluation result value and the target evaluation result value; Perform at least one round of optimization process; any round of the optimization process includes the following steps: Randomly generate the first hyperparameter; After configuring the estimation model using the first hyperparameter, performing a third training process on the estimation model to obtain a first evaluation result; wherein the estimation model is at least one of the SOH estimation model and the SOC estimation model; Adjusting the first hyperparameter to obtain a second hyperparameter; After configuring the estimation model using the second hyperparameter, performing a fourth training process on the estimation model to obtain a second evaluation result; Using a simulated annealing algorithm, according to the first evaluation result, the second evaluation result and the evaluation result variable of the current round of optimization process, determine the optimal hyperparameter of the current round of optimization process; wherein, when the current round of optimization process is the first round of optimization process, the value of the evaluation result variable of the current round of optimization process is the value of the initial evaluation result; Attenuate the evaluation result variables of this round of optimization process; When the attenuated evaluation result variable of this round of optimization process is less than or equal to the target evaluation result value, hyperparameter optimization is performed with the optimal hyperparameter of this round of optimization process, and the entire optimization process is terminated. Otherwise, the next round of optimization process is executed.

7. The battery SOH and SOC joint estimation method according to claim 6, characterized in that: The using of the simulated annealing algorithm to determine the optimal hyperparameters of the current round of optimization process according to the first evaluation result, the second evaluation result and the evaluation result variables of the current round of optimization process comprises: Set the initial number of iterations, target number of iterations and first probability; Perform at least one round of iteration process; any round of the iteration process includes the following steps: When the second evaluation result is less than the first evaluation result, randomly determining the second hyperparameter as the optimal hyperparameter of this round of optimization process according to the first probability; When the second evaluation result is greater than or equal to the first evaluation result, determine a second probability according to the first evaluation result and the second evaluation result and the evaluation result variable of this round of optimization process, obtain a random number, when the random number is less than the second probability, determine the second hyperparameter as the optimal hyperparameter of this round of optimization process, when the random number is greater than or equal to the second probability, determine the first hyperparameter as the optimal hyperparameter of this round of optimization process; The variable of the number of iterations of the current iterative process is decayed; wherein, when the current iterative process is the first iterative process, the value of the variable of the number of iterations of the current iterative process is the initial number of iterations; When the attenuated iteration number variable of the current iteration process is less than or equal to the target iteration number, all iteration processes are terminated, otherwise, the next iteration process is executed.

8. The method for jointly estimating battery SOH and SOC according to any one of claims 1 to 7, characterized in that: The battery SOH and SOC joint estimation method further includes: During the charging stage of the battery, the third battery operating parameter is collected; During the discharge phase of the battery, a fourth battery operating parameter is collected; Inputting the third battery operating parameter into the SOH estimation model for processing to obtain an SOH estimation value; The SOH estimation value and the fourth battery operating parameter are input into the SOC estimation model for processing to obtain an SOC estimation value.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory is used to store at least one program, and the processor is used to load at least one program to execute the battery SOH and SOC joint estimation method according to any one of claims 1 to 8.

10. A computer-readable storage medium storing a program executable by a processor, characterized in that: The program executable by the processor is used to execute the battery SOH and SOC joint estimation method as described in any one of claims 1-8 when executed by the processor.

Images