Battery health state estimation method, device, equipment, medium and program product

By extracting feature values ​​from battery charge and discharge historical data, building a multi-core limit learning machine model and optimizing parameters, the problem of narrow search range and low accuracy in battery health status estimation is solved, and higher prediction accuracy and faster iteration are achieved speed.

CN119064816APending Publication Date: 2024-12-03ZHEJIANG CHINT INSTR & METER

Patent Information

Application Number
CN202411548847.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-01
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing battery health status estimation methods have problems such as narrow search range, inability to guarantee accuracy and long iteration time for the region to search, especially in the prediction of the health status of lithium-ion batteries.

Method used

By extracting charge and discharge characteristic values ​​from multiple battery charge and discharge historical data, dividing the training set and test set, building a multi-core limit learning machine model, and optimizing model parameters using the sparrow algorithm, combining linear kernel functions and polynomial kernel functions, using the minimum two Multiplication is carried out iteratively to build a battery health status estimation model.

Benefits of technology

It improves the accuracy and prediction accuracy of battery health status estimation, expands the search range, shortens iteration time, and solves the problem of inability to guarantee accuracy and long-term search time.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119064816A_ABST
    Figure CN119064816A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of battery health state estimation, and discloses a battery health state estimation method, device and equipment, a medium and a program product, and the method comprises the steps: obtaining various types of battery charging and discharging historical data, and extracting a plurality of charging and discharging characteristic values from the various types of battery charging and discharging historical data, dividing the plurality of charge and discharge characteristic values into a charge and discharge characteristic value training set and a charge and discharge characteristic value test set; constructing a multi-kernel extreme learning machine model based on the charge and discharge characteristic value training set and a preset kernel function; optimizing the parameters of the multi-core extreme learning machine model by adopting a sparrow algorithm improved by various strategies to obtain optimized parameters; a sparrow algorithm improved by various strategies is adopted to construct a battery health state estimation model; and estimating the battery health state based on the charge and discharge characteristic value test set and the battery health state estimation model. The method solves the problems that in the battery state estimation process, the searching range is narrow, the precision cannot be guaranteed, and the parameter searching iteration time is long.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of battery health state estimation, in particular to the technical field of machine learning control of energy storage batteries, and specifically to a battery health state estimation method, device, equipment, medium and program product. Background Art

[0002] Lithium-ion batteries are widely used in energy storage power stations and industrial equipment because of their high power density, various shapes and sizes to meet the needs of different equipment, long life, recyclability and other advantages. However, the sharp decline in the capacity of aging batteries (lead-acid batteries, nickel-metal hydride batteries, nickel-cadmium batteries, flow batteries (vanadium batteries), sodium-sulfur batteries, lithium-ion batteries, lithium iron phosphate batteries, etc.) may cause the equipment to lose power at critical moments, especially in medical equipment and safety equipment, which will increase the possibility of safety accidents. In the process of ensuring the safety of energy storage equipment battery systems and extending their service life, the multiple safety hazards caused by battery aging have prompted the exploration and development of more intelligent battery management systems. In this context, accurate estimation of the health status of batteries such as lithium-ion batteries is particularly important. However, the internal reactions of the battery cannot be captured, and it is usually difficult to accurately estimate the health status of the battery. The state of health (SOH) of the battery reflects the aging of the battery, which can be defined as the ratio of the current maximum capacity to the initial capacity. The aging of the battery is directly manifested as an increase in the internal resistance of the battery. The general formula can be expressed as:

[0003] in, Q current Indicates the capacity of the current battery when fully charged, in Ah (ampere-hour); Q rated Indicates the initial rated capacity of the battery, in units of . Indicates the reference internal resistance; Indicates the internal resistance at the current temperature T and the number of cycles N; Indicates reference temperature The internal resistance of the new battery under Indicates reference temperature The internal resistance of the battery at the end of its life; is the temperature correction factor; is the cycle number correction factor; T is the current temperature; is the reference temperature; N is the number of charge and discharge cycles. As the battery ages, the internal resistance increases and its capacity decreases.

[0004] At present, the methods used for battery health state estimation can be mainly divided into two categories: one is the method based on physical models, and the other is the method based on data-driven. The method based on physical models requires the establishment of mathematical models of the physical or chemical behavior of internal degradation of the battery, including electrochemical models, which predict the battery degradation behavior by describing the kinetic process of internal electrochemical reactions; thermodynamic models, which consider the thermal effect and coupling effect of the battery at different temperatures; mechanical stress models, which study the influence of mechanical stress and strain inside the battery on the life; and health index models, which comprehensively consider multiple factors such as capacity decay, internal resistance increase and power reduction, and predict the overall health state of the battery through mathematical modeling. However, these methods usually require detailed information on the internal parameters of the battery, which is not easy to obtain. Data-driven methods do not rely on a detailed understanding of the physical and chemical processes inside the battery. Therefore, there is no need to construct complex equivalent circuits or electrochemical models in the modeling process, which reduces the dependence on the knowledge of complex chemical reactions inside the battery and helps to improve the generalization ability of the model. For example, extreme learning machine (ELM), multiple kernelized extreme learning machine (MKELM), neural network, etc. Although extreme learning machines have performed well in many applications, they also have some disadvantages, including sensitivity to parameter selection, limited generalization performance, lack of deep feature extraction capabilities, limited ability to handle large-scale data, and insufficient robustness to noise and outliers. In addition, extreme learning machine models have poor interpretability, are sensitive to data distribution, have high memory requirements, and random initialization affects performance. These disadvantages limit the use of extreme learning machines in some complex applications, and they need to be overcome by improving algorithms or combining other methods. Although multi-core extreme learning machines have performed well in dealing with nonlinear problems and improving prediction accuracy, they also have some disadvantages. Their complex model structure and combination of multiple kernel functions increase the difficulty of design and optimization, and parameter selection is difficult and requires a lot of experimental verification. Multi-core extreme learning machines have high requirements for computing resources and memory, long training time, and are prone to overfitting when processing high-dimensional data. Multi-core extreme learning machines have attracted widespread attention due to their good prediction accuracy and fast iteration performance in small sample data. During the prediction process, multi-core extreme learning machines need to optimize multiple kernel functions and their parameters. Commonly used optimization methods include grid search and other algorithms such as genetic algorithms and Bayesian optimization. However, the grid search method is very time-consuming and computationally demanding, while the genetic algorithm and Bayesian optimization may fall into local optimal solutions in some cases. Therefore, in general, there are still some limitations in using multi-core extreme learning machines to predict the health status of lithium-ion batteries. Summary of the invention

[0005] In view of this, the present invention provides a battery health state estimation method, device, equipment, medium and program product to solve the problems of narrow search range, unguaranteed accuracy and long parameter search iteration time in the battery state estimation process.

[0006] In a first aspect, the present invention provides a method for estimating a battery health state, the method comprising: Acquire multiple battery charge and discharge history data, extract multiple charge and discharge feature values ​​from the multiple battery charge and discharge history data, and divide the multiple charge and discharge feature values ​​into a charge and discharge feature value training set and a charge and discharge feature value test set; Construct a multi-core extreme learning machine model based on the charge and discharge characteristic value training set and the preset kernel function; The sparrow algorithm improved by multiple strategies is used to optimize the parameters of the multi-core extreme learning machine model, and the optimized multi-core extreme learning machine model parameters are obtained; A battery health status estimation model is constructed based on the charge and discharge characteristic value training set, optimized multi-core extreme learning machine model parameters, preset weight factors and the sparrow algorithm with multiple strategy improvements; The battery health state is estimated based on the charge and discharge characteristic value test set and the battery health state estimation model.

[0007] The battery health state estimation method provided by the present invention extracts multiple charge and discharge characteristic values ​​from multiple battery charge and discharge history data, divides the multiple charge and discharge characteristic values ​​into a charge and discharge characteristic value training set and a charge and discharge characteristic value test set to describe the uncertain characteristics of the battery health state; uses a preset kernel function to build a multi-core extreme learning machine model, uses a sparrow algorithm improved by multiple strategies to optimize the multi-core extreme learning machine model parameters, and obtains the optimized multi-core extreme learning machine model parameters; builds a battery health state estimation model based on the charge and discharge characteristic value training set, the optimized multi-core extreme learning machine model parameters, the preset weight factor and the sparrow algorithm improved by multiple strategies; estimates the battery health state based on the charge and discharge characteristic value test set and the battery health state estimation model. The present invention can improve the search range, operation speed and prediction accuracy of battery health state estimation using a multi-core extreme learning machine network through a multi-core extreme learning machine model and a sparrow algorithm improved by multiple strategies, and improve the accuracy of battery health state estimation. The problem of narrow search range, inability to guarantee accuracy and long parameter search iteration time in the battery state estimation process is solved.

[0008] In an optional implementation, extracting a plurality of charge and discharge characteristic values ​​from a plurality of battery charge and discharge history data includes: Extract constant current charging time, constant voltage rise time, IC curve peak, IC curve peak corresponding voltage, electrolyte resistance and charge transfer resistance of various batteries from various battery charge and discharge history data; Constant current charging time, constant voltage rise time, IC curve peak, IC curve peak corresponding voltage, electrolyte resistance and charge transfer resistance are used as multiple charging and discharging characteristic values.

[0009] The battery health state estimation method provided by the present invention constructs a training set and a test set to describe the uncertain characteristics of the battery health state by extracting constant current charging time, voltage corresponding to equal boost time, IC curve peak, voltage corresponding to IC curve peak, electrolyte resistance and charge transfer resistance from a variety of battery charge and discharge history data. These characteristic values ​​are closely related to the internal reaction of the battery, which helps to improve the prediction performance and applicability of the model and provides conditions for the subsequent construction of a multi-core extreme learning machine model.

[0010] In an optional implementation, the preset kernel function includes a linear kernel function and a polynomial kernel function; the multi-core extreme learning machine model includes a hidden layer and an output layer; The multi-core extreme learning machine model is constructed based on the charge and discharge characteristic value training set and the preset kernel function, including: The linear kernel function, the polynomial kernel function and the preset mixing coefficient are combined to obtain a combined kernel function; The weight coefficient of each charge and discharge feature value in the charge and discharge feature value training set is randomly generated in the hidden layer; In the output layer, the least squares method is used to iteratively calculate the combined kernel function, the weight coefficient of each charge and discharge eigenvalue, and the preset regularization coefficient to construct a multi-core extreme learning machine model.

[0011] The battery health state estimation method provided by the present invention uses a multi-core extreme learning machine model constructed by using linear kernel functions and polynomial kernel functions. Multiple kernel functions replace the random mapping in the traditional learning machine, and the least squares calculation method is used to replace the iterative calculation between the hidden layer and the output layer. The multi-core extreme learning machine better captures the nonlinear change characteristics of the battery health state through the combination of multiple kernel functions, thereby enhancing the expression ability and generalization performance of the multi-core extreme learning machine model.

[0012] In an optional implementation, the multi-core extreme learning machine model parameters include each charge and discharge eigenvalue, a linear kernel function, a polynomial kernel function, a weight coefficient, and a preset regularization coefficient; the multi-core extreme learning machine model parameters are optimized by using a sparrow algorithm improved by multiple strategies, and the optimized multi-core extreme learning machine model parameters include: The sparrow algorithm improved by multiple strategies is used to associate and fuse the regularization coefficient with the linear kernel function, the polynomial kernel function and each charge and discharge eigenvalue to obtain the first fusion parameter; The sparrow algorithm improved by multiple strategies is used to associate and fuse each charge and discharge characteristic value with the linear kernel function and the polynomial kernel function to obtain the second fusion parameter; The sparrow algorithm improved by multiple strategies is used to associate and fuse the weight coefficient with each charge and discharge characteristic value to obtain the third fusion parameter; The first fusion parameter, the second fusion parameter and the third fusion parameter are used as optimized multi-core extreme learning machine model parameters.

[0013] The battery health state estimation method provided by the present invention uses a sparrow search algorithm improved with multiple strategies to perform optimization operations such as data fusion on the kernel function parameters of a multi-core extreme learning machine, which can significantly improve the prediction accuracy of a subsequent battery health state estimation model.

[0014] In an optional embodiment, constructing a battery health state estimation model based on a charge and discharge characteristic value training set, optimized multi-core extreme learning machine model parameters, preset weight factors, and a sparrow algorithm improved by multiple strategies includes: Normalizing the charge and discharge characteristic values ​​in the charge and discharge characteristic value training set to obtain normalized charge and discharge characteristic values; The initial population of the sparrow algorithm under various improved strategies is randomly generated, and the optimal Latin hypercube algorithm is used to initialize the initial population to obtain the objective function; Based on the normalized charge and discharge eigenvalues ​​and preset weight factors, the sparrow algorithm improved with multiple strategies is used to iterate the objective function until the convergence condition is met and the iteration ends. The optimized multi-core extreme learning machine model parameters are assigned to the objective function obtained in the last iteration to obtain the battery health state estimation model.

[0015] The battery health state estimation method provided by the present invention, the sparrow search algorithm under multi-strategy improvement effectively avoids the problem of the sparrow search algorithm falling into the local optimum during the optimization process by introducing the optimal Latin hypercube sampling and the adjustment of the weight factor in the multi-core extreme learning machine model parameters, and further improves the adaptability and convergence speed of the multi-core extreme learning machine model. The optimal Latin hypercube algorithm is used to make the distribution of the initial population more uniform, and the coverage of the search space is improved. By dynamically adjusting the weight factor, the sparrow search algorithm under multi-strategy improvement achieves a balance between global search and local search, speeds up the optimization process, and reduces the iteration time.

[0016] In an optional implementation, the objective function is iterated based on the normalized charge and discharge characteristic values ​​and using a sparrow algorithm improved by multiple strategies until the convergence condition is met and the iteration ends, including: The normalized charge and discharge characteristic values ​​are divided into discoverers, participants and guards; Calculate the root mean square errors of the normalized charge and discharge characteristic values ​​corresponding to the discoverer, participant and guard respectively, and use the corresponding root mean square errors as the fitness value of the discoverer, the fitness value of the participant and the fitness value of the guard respectively; Based on the preset weight factor and using multiple strategies to improve the sparrow algorithm, the objective function is searched locally and globally, and the positions of the discoverer, participant and guard are updated according to the fitness value of the discoverer, the fitness value of the participant and the fitness value of the guard, until the sparrow algorithm improved by multiple strategies meets the convergence conditions and the iteration ends.

[0017] The battery health state estimation method provided by the present invention increases the search speed based on the dependence of the fitness value during iteration, because the position of the previous generation of discoverers is usually a relatively good search point, which helps to quickly find the global optimal solution or an area close to the global optimal solution. At the same time, a higher weight factor also expands the search range and improves the global search capability. In the later stage, as the number of iterations increases, the weight factor gradually decreases, reducing the algorithm's dependence on the position of the previous generation of discoverers. This decreasing weight factor helps the algorithm to jump out of the local optimal solution because it reduces the dependence on past information and increases the diversity of the search. By reducing the weight factor, the algorithm can explore new search areas more flexibly, thereby increasing the probability of finding the global optimal solution.

[0018] In an optional implementation, estimating the battery health state based on the charge and discharge characteristic value test set and the battery health state estimation model includes: Input the charge and discharge characteristic value test set into the battery health state estimation model to obtain the battery health state prediction data; The battery health state prediction data is denormalized, and the battery health state estimation result is obtained based on the denormalized prediction data.

[0019] The battery health state estimation method provided by the present invention inputs a charge and discharge characteristic value test set into a battery health state estimation model to obtain battery health state prediction data; performs denormalization processing on the battery health state prediction data, and obtains a battery health state estimation result based on the denormalized prediction data, which can significantly improve the effect of battery health state estimation and provide reliable control technology support for battery management and maintenance of electric vehicles and energy storage systems.

[0020] In a second aspect, the present invention provides a battery health status estimation device, the device comprising: A data acquisition and processing module is used to acquire multiple battery charge and discharge history data, extract multiple charge and discharge characteristic values ​​from the multiple battery charge and discharge history data, and divide the multiple charge and discharge characteristic values ​​into a charge and discharge characteristic value training set and a charge and discharge characteristic value test set; A multi-core extreme learning machine model building module is used to build a multi-core extreme learning machine model based on a charge and discharge characteristic value training set and a preset kernel function; A parameter optimization module is used to optimize the parameters of the multi-core extreme learning machine model by using the sparrow algorithm improved by multiple strategies to obtain the optimized multi-core extreme learning machine model parameters; The battery health state estimation model construction module is used to construct the battery health state estimation model based on the charge and discharge characteristic value training set, the optimized multi-core extreme learning machine model parameters, the preset weight factors and the sparrow algorithm under multiple strategy improvements; The battery health state estimation module is used to estimate the battery health state based on the charge and discharge characteristic value test set and the battery health state estimation model.

[0021] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, computer instructions are stored in the memory, and the processor executes the battery health status estimation method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0022] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the battery health status estimation method of the first aspect or any corresponding embodiment thereof.

[0023] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions, wherein the computer instructions are used to enable a computer to execute the battery health status estimation method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 is a flowchart of a battery health status estimation method according to an embodiment of the present invention; Figure 2 is a flow chart of another battery health status estimation method according to an embodiment of the present invention; Figure 3 is a flowchart of another battery health status estimation method according to an embodiment of the present invention; Figure 4 is a flowchart of another battery health status estimation method according to an embodiment of the present invention; Figure 5is a schematic diagram of a battery cycle charge and discharge curve according to an embodiment of the present invention; Figure 6 is a schematic diagram of an initial population after the optimal Latin hypercube algorithm is initialized according to an embodiment of the present invention; Figure 7 is a schematic diagram of randomly generating an initial population according to an embodiment of the present invention; Figure 8 is a comparison error diagram of various algorithms used for a B0006 battery according to an embodiment of the present invention; Fig. 9 is a comparison error diagram of various algorithms used for a B0007 battery according to an embodiment of the present invention; Fig.10 is a comparison error diagram of various algorithms used for the B00018 battery according to an embodiment of the present invention; Fig.11 is a structural block diagram of a battery health status estimation device according to an embodiment of the present invention; Fig.12 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0027] According to an embodiment of the present invention, an embodiment of a battery health status estimation method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0028] In this embodiment, a battery health status estimation method is provided, which can be used in computer terminals, such as central processing units, servers, etc. Figure 1 is a flow chart of a battery health status estimation method according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps: Step S101, acquiring multiple battery charge and discharge history data, extracting multiple charge and discharge feature values ​​from the multiple battery charge and discharge history data, and dividing the multiple charge and discharge feature values ​​into a charge and discharge feature value training set and a charge and discharge feature value test set.

[0029] Specifically, the battery in this embodiment includes but is not limited to lithium batteries. Taking lithium batteries as an example, the multiple lithium batteries are lithium batteries of different models, for example, lithium batteries of model B0006, model B0007 and model B0018. The multiple lithium battery charge and discharge history data include the following categories: 1. Basic charge and discharge parameter data: 1) Current data: ① Charging current: Record the current flowing into the battery during each charging process, including the constant current value in the constant current charging stage and the small current value in the possible trickle charging stage. For example, in a certain stage, the constant current charging is performed with a current of 1C (current of 1 times the rated capacity of the battery).

[0030] ②Discharge current: records the current flowing out during the battery discharge process. The discharge current will be different for different discharge scenarios (such as different load conditions), which is very important for evaluating the performance of the battery in different usage scenarios.

[0031] 2) Voltage data: ① Charging voltage: monitor the voltage changes of the battery during the charging process, including the voltage values ​​at different charging stages, such as the voltage rise at the constant current charging stage and the voltage stability at the constant voltage charging stage. Generally speaking, the battery charging cut-off voltage is an important parameter, and exceeding this voltage may cause damage to the battery.

[0032] ②Discharge voltage: record the voltage output of the battery during the discharge process. As the battery discharges, the voltage will gradually decrease. By analyzing the discharge voltage curve, you can understand the battery's discharge characteristics and remaining power.

[0033] 3) Time data: ① Charging time: Recording the time spent on each charge is important for evaluating the battery's charging speed and efficiency. It can also help determine whether the battery has charging abnormalities (such as charging time that is too long or too short).

[0034] ②Discharge time: Record the duration of each battery discharge. Combined with the discharge current and voltage data, the battery’s discharge capacity and energy output can be calculated.

[0035] 2. Capacity related data: 1) Charge and discharge capacity per cycle: records the amount of electricity stored and released by the battery during each charge and discharge process. This is a key indicator for evaluating battery performance and life. As the number of charge and discharge cycles increases, the charge and discharge capacity of the battery will gradually decay.

[0036] 2) Cumulative charge and discharge capacity: Statistics on the total charge and discharge capacity of the battery from the beginning of the test to the current moment. By analyzing the changes in the cumulative charge and discharge capacity, the overall usage and aging of the battery can be understood.

[0037] 3) Capacity retention rate: The ratio of the remaining capacity of the battery after each cycle to the initial capacity is calculated, which reflects the battery's ability to retain capacity after multiple charge and discharge cycles. For example, after 100 charge and discharge cycles, the battery's capacity retention rate is 80%, indicating that the battery's capacity has decreased by 20% compared to the initial state.

[0038] 3. Internal resistance related data: 1) DC internal resistance (DCIR): The DC resistance of the battery in different states is measured by applying specific current pulses during the battery charging and discharging process. The DC internal resistance will gradually increase with the use and aging of the battery, affecting the performance and efficiency of the battery.

[0039] 2) AC internal resistance (ACIR): Use AC signals to measure the internal resistance of the battery. AC internal resistance can reflect the changes in the electrochemical characteristics and electrode structure inside the battery, and is also important for analyzing the performance and health status of the battery.

[0040] 4. Temperature related data: 1) Battery surface temperature: During the charging and discharging process, the battery generates heat, causing the temperature to rise. Recording the temperature changes on the battery surface can help evaluate the thermal management performance of the battery and whether there is a risk of overheating.

[0041] 2) Ambient temperature: Record the temperature of the test environment, because the ambient temperature will affect the battery's charge and discharge performance and life. By analyzing the relationship between ambient temperature and battery performance, the battery's operating environment and conditions can be optimized.

[0042] 5. Energy efficiency related data: 1) Charging energy: records the electric energy consumed by the battery during the charging process, which is the product of charging current, charging voltage and charging time.

[0043] 2) Discharge energy: records the electrical energy released by the battery during the discharge process, which is the product of discharge current, discharge voltage and discharge time.

[0044] 3) Energy efficiency: Calculate the ratio of discharge energy to charge energy, which reflects the energy conversion efficiency of the battery during the charge and discharge process. The higher the energy efficiency, the better the battery performance and the smaller the energy loss.

[0045] 6. Other auxiliary data: 1) Battery model and batch information: record the model, production batch and other information of the test battery, so as to compare and analyze batteries of different models and batches, and understand the impact of production process and raw materials on battery performance.

[0046] 2) Test equipment information: record the model, parameters and other information of the charge and discharge test equipment used to ensure the accuracy and reliability of the test data, and also facilitate equipment troubleshooting when problems arise.

[0047] After calculation, multiple charging and discharging characteristic values ​​related to the internal chemical changes of the battery health status are extracted from the above-mentioned multiple battery charging and discharging history data, such as constant current charging time, voltage corresponding to equal boost time, IC curve peak, voltage corresponding to IC curve peak, electrolyte resistance and charge transfer resistance.

[0048] All the extracted charge and discharge characteristic values ​​are divided into a charge and discharge characteristic value training set and a charge and discharge characteristic value test set according to a certain ratio, for example, divided into a charge and discharge characteristic value training set and a charge and discharge characteristic value test set according to a ratio of 4:6.

[0049] Step S102: constructing a multi-core extreme learning machine model based on the charge and discharge characteristic value training set and a preset kernel function.

[0050] Specifically, the multi-core extreme learning machine model is an improved model based on the extreme learning machine. The extreme learning machine is a fast learning algorithm for a single hidden layer feedforward neural network. Its main feature is to randomly generate the weights and biases of the hidden layer, and then quickly calculate the weights of the output layer through the least squares method. On this basis, the multi-core extreme learning machine introduces multiple kernel functions to enhance the expressive power and generalization performance of the multi-core extreme learning machine model.

[0051] The kernel function is used to map the input data from a low-dimensional space to a high-dimensional space. By combining the global kernel and the local kernel, the model can simultaneously process the global trend and local details of the data. Therefore, when faced with complex nonlinear data, the generalization ability and prediction accuracy of the multi-core extreme learning machine model are greatly improved.

[0052] A multi-core extreme learning machine model can be constructed based on the charge and discharge characteristic value training set and the preset kernel function.

[0053] Step S103, using the sparrow algorithm improved with multiple strategies to optimize the multi-core extreme learning machine model parameters to obtain optimized multi-core extreme learning machine model parameters.

[0054] Specifically, the sparrow algorithm improved with multiple strategies is used to perform parameter data fusion and other optimization operations on the regularization coefficient, a linear kernel function, two polynomial kernel functions, and the weight size of the multi-core extreme learning machine model to obtain the optimized multi-core extreme learning machine model parameters.

[0055] Step S104, building a battery health state estimation model based on the charge and discharge characteristic value training set, the optimized multi-core extreme learning machine model parameters, the preset weight factors and the sparrow algorithm improved with multiple strategies.

[0056] Specifically, the sparrow algorithm with multiple strategy improvements is a process of finding the optimal solution to the problem by imitating the different division of labor among sparrows in the process of searching for food. In each iteration of the sparrow search algorithm, sparrows are usually divided into three types: discoverers, participants, and guards. During each iteration, the positions of discoverers, participants, and guards are updated. The specific operations are: generating and initializing the initial population, using the weight factor to balance the local and global search capabilities when the guard position is updated to finally obtain the optimal solution; bringing the optimal solution to multi-core extreme learning to obtain a battery health status estimation model.

[0057] Step S105 , estimating the battery health state based on the charge and discharge characteristic value test set and the battery health state estimation model.

[0058] Specifically, the charge and discharge characteristic value test set is input into the battery health state estimation model to obtain a prediction result, and the battery health state is estimated according to the prediction result.

[0059] The battery health state estimation method provided in this embodiment extracts multiple charge and discharge characteristic values ​​from multiple battery charge and discharge history data, divides the multiple charge and discharge characteristic values ​​into a charge and discharge characteristic value training set and a charge and discharge characteristic value test set to describe the uncertain characteristics of the battery health state; uses a preset kernel function to build a multi-core extreme learning machine model, and uses a sparrow algorithm improved by multiple strategies to optimize the multi-core extreme learning machine model parameters to obtain optimized multi-core extreme learning machine model parameters; builds a battery health state estimation model based on the charge and discharge characteristic value training set, the optimized multi-core extreme learning machine model parameters, the preset weight factor and the sparrow algorithm improved by multiple strategies; and estimates the battery health state based on the charge and discharge characteristic value test set and the battery health state estimation model. The present invention can improve the search range, operation speed and prediction accuracy (prediction results are shown in Table 1) of battery health state estimation using a multi-core extreme learning machine model and a sparrow algorithm improved by multiple strategies, and improve the accuracy of battery health state estimation. The problem of narrow search range, unguaranteed accuracy and long parameter search iteration time in the battery state estimation process is solved.

[0060] In this embodiment, a battery health status estimation method is provided, which can be used in computer terminals, such as central processing units, servers, etc. Figure 2 is a flow chart of a battery health status estimation method according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps: Step S201, acquiring multiple battery charge and discharge history data, extracting multiple charge and discharge feature values ​​from the multiple battery charge and discharge history data, and dividing the multiple charge and discharge feature values ​​into a charge and discharge feature value training set and a charge and discharge feature value test set.

[0061] Specifically, the above step S201 includes: Step S2011, extracting constant current charging time, constant voltage rise time, IC curve peak, IC curve peak corresponding voltage, electrolyte resistance and charge transfer resistance of various batteries from various battery charge and discharge history data.

[0062] Specifically, key characteristic values ​​are extracted from the battery charge and discharge data, including constant current charging time and constant voltage rise time. The battery cycle charge and discharge curve (such as Figure 5 As shown in the figure, the battery cycle charge and discharge curve is Kalman filtered to reduce noise interference, and the IC curve is constructed by differential method to read the peak value of the IC curve (the curve of the change of current with voltage during battery charge and discharge, incremental capacity, IC curve) and the voltage corresponding to the peak value of the IC curve. The electrolyte resistance and charge transfer resistance are read according to the EIS test (Electrochemical Impedance Spectroscopy, EIS).

[0063] The S2 Kalman filter is a method to remove noise from data, and its formula is as follows: (1); in, It's time k The state prediction is used to update the state. F k is the state transition matrix, B k is the control input matrix, u k is the control vector, P k|k-1 is the predicted state covariance, Q k is the process noise covariance.

[0064] In its update phase, new measurement data is used to correct the prediction and update the estimate of the system state. The calculation formula can be expressed as: (2); in, K k Its Kalman gain, an indicator that measures the relative weight of the predicted covariance and the measured covariance, is used to update the state estimate. H k is the measurement matrix, R k is the measurement noise covariance, I is the identity matrix.

[0065] (3); (4); in, I is the identity matrix, through which the uncertainty of the state estimate can be updated.

[0066] The difference method formula is expressed as: (5); in, is the rate of change of a variable Y at the jth moment; is the time derivative of a variable X at the jth moment; is the variable X at the jth moment; It is The variable X at the moment; M is the time sampling interval.

[0067] Step S2012, taking the constant current charging time, the constant voltage rise time, the IC curve peak, the voltage corresponding to the IC curve peak, the electrolyte resistance and the charge transfer resistance as a plurality of charge and discharge characteristic values.

[0068] It should be noted that, according to actual conditions, other charge and discharge characteristic values ​​may be extracted from the battery charge and discharge data. For example, battery capacity related data may be selected as charge and discharge characteristic values ​​to estimate the battery health status.

[0069] This embodiment only uses constant current charging time, constant voltage rise time, IC curve peak, IC curve peak corresponding voltage, electrolyte resistance and charge transfer resistance as examples of multiple charging and discharging characteristic values ​​for explanation without specific limitation.

[0070] Step S202: constructing a multi-core extreme learning machine model based on the charge and discharge characteristic value training set and a preset kernel function.

[0071] Specifically, the preset kernel function includes a linear kernel function and a polynomial kernel function; the multi-core extreme learning machine model includes a hidden layer and an output layer, and the above step S202 includes: Step S2021, combining the linear kernel function, the polynomial kernel function and the preset mixing coefficient to obtain a combined kernel function.

[0072] Specifically, the linear kernel function and the polynomial kernel function are expressed by the following formulas: (6); in, is a linear kernel function, is the polynomial kernel function, is the inner product of its input, Used to scale its inner product. Used to control the number of inner products, Represents the constant in the kernel function to prevent the model from overfitting or underfitting.

[0073] The combined kernel function is: (7); in, Represents the mixing coefficient, which can be used to adjust the mapping performance of the kernel function.

[0074] Step S2022, randomly generating a weight coefficient for each charge and discharge eigenvalue in the charge and discharge eigenvalue training set in the hidden layer.

[0075] Specifically, the size of the weight coefficient determines the importance of each charge and discharge characteristic value in the charge and discharge characteristic value training set in the multi-core extreme learning machine model. The weights of the hidden layer nodes are randomly generated or artificially given and do not need to be updated. The learning process only calculates the output weight coefficient.

[0076] Step S2023, in the output layer, the least squares method is used to iteratively calculate the combined kernel function, the weight coefficient of each charge and discharge eigenvalue, and the preset regularization coefficient to construct a multi-core extreme learning machine model.

[0077] Specifically, compared with the traditional kernel extreme learning machine algorithm, the multi-core extreme learning machine model uses multiple kernel functions to replace random mapping and uses the least squares method to replace the iterative calculation between the hidden layer and the output layer. The multi-core extreme learning machine model is expressed by the following formula: (8); Among them, X is the sample data, that is, each charge and discharge characteristic value; Output for multi-core extreme learning machine model; is the hidden layer output; The matrix proposed for the feature is, H=α1H1+α2H2, where H1 and H2 are kernel matrices generated by different kernel functions; is the preset regularization coefficient; is the identity matrix; is the target vector of the charge and discharge feature value training set; is the combined kernel function; is the kernel reconstruction matrix, which can be expressed as: , where D is the sample weight matrix, which is a diagonal matrix in which each element corresponds to the weight coefficient of a sample. Y is the target value of the training sample.

[0078] Step S203, using the sparrow algorithm improved with multiple strategies to optimize the multi-core extreme learning machine model parameters to obtain optimized multi-core extreme learning machine model parameters.

[0079] Specifically, the multi-core extreme learning machine model parameters include each charge and discharge characteristic value, a linear kernel function, a polynomial kernel function, a weight coefficient and a preset regularization coefficient; the above step S203 includes: Step S2031, using the sparrow algorithm improved by multiple strategies to associate and fuse the regularization coefficient with the linear kernel function, the polynomial kernel function and each charge and discharge eigenvalue, to obtain a first fusion parameter.

[0080] Specifically, the regularization coefficient is associated and fused with the linear kernel function and the polynomial kernel function respectively, specifically: the regularization coefficient is used to control the complexity and overfitting of the model. For the linear kernel function and polynomial kernel function of the multi-core extreme learning machine model, the regularization coefficient can adjust the degree of effect of the linear kernel function and the polynomial kernel function in the model. When the regularization coefficient is small, the linear kernel function and the polynomial kernel function have a greater impact on the model output, which may make the model pay more attention to the local characteristics of the data; when the regularization coefficient is large, it will have a certain inhibitory effect on the influence of the linear kernel function and the polynomial kernel function, and the model is more inclined to overall smooth fitting. For example, when processing battery capacity data with more noise, a larger regularization coefficient combined with appropriate linear kernel functions and polynomial kernel functions can reduce the impact of noise.

[0081] The regularization coefficient is associated with each charge and discharge characteristic value. Specifically, the regularization coefficient affects the model's capture of the changing trend of the charge and discharge characteristic values ​​in terms of constant current charging time, constant voltage rise time, IC curve peak, IC curve peak corresponding voltage, electrolyte resistance and charge transfer resistance. If the charge and discharge characteristic values ​​in the training set have large fluctuations, the appropriate regularization coefficient can allow the model to balance the fitting and generalization capabilities of the data during the learning process. For example, the charging current data has different change patterns in different cycles, and the regularization coefficient helps the model accurately grasp the changing rules and make accurate predictions.

[0082] Step S2032: using the sparrow algorithm improved by multiple strategies to associate and fuse each charge and discharge characteristic value with the linear kernel function and the polynomial kernel function respectively, to obtain a second fusion parameter.

[0083] Specifically, the linear kernel function determines the weight and role of the linear kernel function in the multi-core extreme learning machine. For data with a certain linear change trend such as constant current charging time and constant voltage rise time (in the process of battery aging, constant current charging time and constant voltage rise time may show a certain linear growth or decrease trend), the linear kernel function can capture this linear relationship well and integrate it into the model to improve the prediction accuracy.

[0084] Polynomial kernel function: The polynomial kernel function can introduce nonlinear characteristics to the model. For example, the change of battery capacity with the number of cycles is often not a simple linear relationship. The polynomial kernel function can better fit this nonlinear change. In the analysis of parameters such as the charge and discharge capacity of each cycle, the cumulative charge and discharge capacity, and the capacity retention rate, the polynomial kernel function can dig out the hidden nonlinear laws, thereby optimizing the model's evaluation of battery performance.

[0085] Step S2033: using the sparrow algorithm improved by multiple strategies to associate and fuse the weight coefficient with each charge and discharge characteristic value to obtain a third fusion parameter.

[0086] Specifically, the weight coefficient affects the electrical parameters: The weight determines the importance of each input parameter in the model. For electrical parameters such as constant current charging time, voltage corresponding to equal boost time, IC curve peak, voltage corresponding to IC curve peak, electrolyte resistance and charge transfer resistance, the distribution of weights will affect the model's contribution to the prediction or analysis of different voltage values. If in a certain application scenario, the influence of the voltage corresponding to the peak of the IC curve on the health status of the battery is more critical, the influence of the voltage corresponding to the peak of the IC curve can be increased through the optimized weight, so that the model can more accurately judge the battery status based on parameters such as the voltage corresponding to the peak of the IC curve.

[0087] The weight coefficient can also be integrated with temperature-related parameters: in terms of battery surface temperature and ambient temperature parameters, the weight size can adjust the role of temperature factors in the model. For example, in a high temperature environment, the battery performance changes significantly. The optimized weight can make the model pay more attention to temperature parameters, thereby more accurately predicting the changes in battery performance parameters such as charge and discharge capacity and internal resistance based on temperature changes.

[0088] Step S2034: using the first fusion parameter, the second fusion parameter, and the third fusion parameter as optimized multi-core extreme learning machine model parameters.

[0089] The overall effects of the optimized parameters of the sparrow algorithm under various strategy improvements include: Data fusion and model optimization: The parameters optimized by the sparrow algorithm can enable the multi-core extreme learning machine to better integrate various types of training set parameters. When processing EIS test data combined with charge and discharge cycle data, the optimized parameters can find the intrinsic connection between different data and improve the accuracy and reliability of the model's evaluation of battery performance.

[0090] Prediction and decision support: With optimized parameters, the model can more accurately predict the performance changes of the battery under different working conditions. For example, based on the current charging current, temperature and other parameters, combined with the optimized model parameters, the remaining capacity of the battery, the change trend of the DC internal resistance, etc. can be predicted, providing decision support for the use, maintenance and management of the battery.

[0091] Step S204, construct a battery health state estimation model based on the charge and discharge characteristic value training set, the optimized multi-core extreme learning machine model parameters, the preset weight factors and the sparrow algorithm under multiple strategy improvements. For details, please refer to Figure 1 Step S104 of the illustrated embodiment will not be described in detail here.

[0092] Step S205, estimate the battery health state based on the charge and discharge characteristic value test set and the battery health state estimation model. Figure 1 Step S105 of the illustrated embodiment will not be described in detail here.

[0093] The battery health state estimation method provided in this embodiment extracts constant current charging time, voltage corresponding to equal boost time, IC curve peak, voltage corresponding to IC curve peak from multiple battery charge and discharge history data, and uses electrolyte resistance and charge transfer resistance as charge and discharge characteristic values ​​to construct training sets and test sets to describe the uncertain characteristics of battery health state. These characteristic values ​​are closely related to the internal reaction of the battery, which helps to improve the prediction performance and applicability of the model, and provides conditions for the subsequent construction of a multi-core extreme learning machine model. The multi-core extreme learning machine model constructed by linear kernel function and polynomial kernel function, multiple kernel functions replace the random mapping in the traditional learning machine, and the least squares calculation method is used to replace the iterative calculation between the hidden layer and the output layer. The multi-core extreme learning machine better captures the nonlinear change characteristics of the battery health state through the combination of multiple kernel functions, thereby enhancing the expression ability and generalization performance of the multi-core extreme learning machine model. The sparrow search algorithm improved by multiple strategies is used to perform data fusion and other optimization operations on the kernel function parameters of the multi-core extreme learning machine, which can significantly improve the prediction accuracy of the subsequent battery health state estimation model.

[0094] In this embodiment, a battery health status estimation method is provided, which can be used in computer terminals, such as central processing units, servers, etc. Figure 3 is a flow chart of a battery health status estimation method according to an embodiment of the present invention. Figure 3 As shown, the process includes the following steps: Step S301, obtain multiple battery charge and discharge history data, extract multiple charge and discharge feature values ​​from the multiple battery charge and discharge history data, and divide the multiple charge and discharge feature values ​​into a charge and discharge feature value training set and a charge and discharge feature value test set. Figure 2 Step S201 of the illustrated embodiment will not be described in detail here.

[0095] Step S302: construct a multi-core extreme learning machine model based on the charge and discharge characteristic value training set and the preset kernel function. Figure 2 Step S202 of the illustrated embodiment will not be described in detail here.

[0096] Step S303, using the sparrow algorithm improved by multiple strategies to optimize the multi-core extreme learning machine model parameters, to obtain optimized multi-core extreme learning machine model parameters. Figure 2 Step S203 of the illustrated embodiment will not be described in detail here.

[0097] Step S304, building a battery health state estimation model based on the charge and discharge characteristic value training set, the optimized multi-core extreme learning machine model parameters, the preset weight factors and the sparrow algorithm under multiple strategy improvements.

[0098] The above step S304 includes: Step S3041 , normalizing the charge and discharge characteristic values ​​in the charge and discharge characteristic value training set to obtain normalized charge and discharge characteristic values.

[0099] Specifically, the normalization formula is: (9); in, is the charge and discharge characteristic value after normalization; It is the test data of the charge and discharge characteristic value; and are the minimum and maximum values ​​of the charge and discharge characteristic values ​​respectively.

[0100] Step S3042, randomly generate the initial population of the sparrow algorithm under various strategy improvements, and use the optimal Latin hypercube algorithm to initialize the initial population to obtain the objective function.

[0101] Specifically, the sparrow algorithm improved by multiple strategies is to imitate the different division of labor of the sparrow group in the process of finding food to find the optimal solution to the problem. First, m sparrows are randomly generated to represent a population. The formula is as follows: (10); like Figure 7As shown in , the randomly generated initial population of sparrows tends to cluster in space, which makes it easy to fall into the problem of local optimal solution and increase the time of optimization. Figure 6 As shown, the Latin hypercube algorithm can effectively solve this problem and make the initial population evenly distributed in space. Its objective function is as follows: (11); in, For the Dimension The value of samples; For the The lower bound of the dimension. b i For the The upper bound of the dimension. is the number of intervals in each dimension, which is also the total number of samples. After random arrangement The index of the interval, starting from 0. Is a function that generates random numbers in the interval [0,1) and is used to randomly select a point in the specified interval. Represents the arrangement of all dimensions The optimization is performed to maximize the following minimum distance expression. Indicates calculation of all sample point pairs The minimum distance between Make sure to compare the distances between different points. ‖xi-xj‖ is the Euclidean distance, which is used to measure any two sample points. and The distance between.

[0102] The solution steps can be divided into: Figure 6 and Figure 7 As shown in the figure, the optimal Latin hypercube algorithm initializes the population more uniformly. First, the optimal Latin hypercube sampling point L is used to define the objective function f(L), that is, the maximization of the minimum distance. The optimization algorithm is applied to the Latin hypercube sampling point L, and the sample position is updated to optimize the objective function f(L) iteratively until the convergence condition is met. Figure 6 and Figure 7 The points in are the sampling points of the two-dimensional vector. Figure 6 It means that when the optimal Latin hypercube is used to randomly generate sampling points, the sampling points generated by the population are evenly distributed, while when the optimal Latin hypercube is not used to randomly generate sampling points, the sampling points are unevenly distributed. The horizontal and vertical axes have no practical significance.

[0103] Step S3043, based on the normalized charge and discharge characteristic values ​​and the preset weight factors, the objective function is iterated by using the sparrow algorithm improved by multiple strategies until the convergence condition is met and the iteration ends, and the optimized multi-core extreme learning machine model parameters are assigned to the objective function obtained by the last iteration to obtain a battery health state estimation model.

[0104] In some optional implementations, in the above step S3043, the objective function is iterated based on the normalized charge and discharge characteristic value and the preset weight factor and the sparrow algorithm improved by multiple strategies until the convergence condition is met and the iteration ends, including: Step a1, dividing the normalized charge and discharge characteristic values ​​into discoverers, participants and guards.

[0105] Specifically, in each iteration of the sparrow search algorithm, the normalized charge and discharge characteristic values ​​are used as sparrows, which are usually divided into three types: discoverers, participants, and guards.

[0106] Step a2, respectively calculate the root mean square errors of the normalized charge and discharge characteristic values ​​corresponding to the discoverer, participant and guard, and use the corresponding root mean square errors as the fitness value of the discoverer, the fitness value of the participant and the fitness value of the guard.

[0107] Specifically, the expression of the root mean square error is as follows: (12); in, Represented as the first The model outputs the difference between the predicted value and the true value, and N is the length of the charge and discharge feature value training set.

[0108] Step a3, based on the preset weight factor and using the sparrow algorithm with multiple strategies to improve the objective function, local search and global search are performed on the objective function, and the positions of the discoverer, participant and guard are updated according to the fitness value of the discoverer, the fitness value of the participant and the fitness value of the guard, until the sparrow algorithm with multiple strategies improves meets the convergence condition and the iteration ends.

[0109] Specifically, during each iteration, the discoverer's position is updated as follows: (13); in, Indicates l+ In the 1st iteration i The location of the sparrow sample is located at j Weizhong; D max is the maximum number of iterations; α A random number in the interval (0,1); R is the guard value in [0,1];Q is a random number that follows a standard normal distribution; L for d A matrix whose dimensions are all 1; d is the dimension of the objective function; Z ST is the risk-free value.

[0110] The participant positions are updated as follows: (14); in, Indicates l+ The optimal position of the entire population in generation 1; Indicates l The worst position of the entire population after iterations; E Represents a 1×d matrix containing all 1s or -1s.

[0111] The guard position update part introduces a weight factor and smoothly adjusts the weight factor through the periodic change of the sine function. w , thereby improving the global search ability of the algorithm and avoiding falling into local optimality. The formula is as follows: (15); The position update after introducing the weight factor is as follows: (16); in, Indicates l The optimal population position after iterations; β is the step size for controlling the update position; K is a random number in the interval [-1,1]; f w Current worst fitness; f q Current optimal fitness; f i The fitness of the current guard; after improvement, the algorithm can be more dependent on the position of the previous generation of finders (also called scouts) in the early stages. This dependence increases the search speed, because the position of the previous generation of finders is usually a relatively good search point, which helps to quickly find the global optimal solution or an area close to the global optimal solution. At the same time, a higher weight factor also expands the search range and improves the global search capability. Later, as the number of iterations increases, the weight factor gradually decreases, reducing the algorithm's dependence on the position of the previous generation of finders. This decreasing weight factor helps the algorithm jump out of the local optimal solution because it reduces dependence on past information and increases the diversity of the search. By reducing the weight factor, the algorithm can explore new search areas more flexibly, thereby increasing the probability of finding the global optimal solution.

[0112] If the sparrow algorithm improved by multiple strategies meets the convergence conditions, the global optimal value is assigned to the regularization coefficient of the multi-core extreme learning machine model, a linear kernel parameter, two polynomial kernel parameters, and a weight value to obtain a battery health state estimation model; if the convergence conditions are not met, step S3043 is repeated until the convergence conditions are met.

[0113] The battery health state estimation method provided in this embodiment increases the search speed according to the dependence of the fitness value during iteration, because the position of the previous generation of discoverers is usually a relatively good search point, which helps to quickly find the global optimal solution or the area close to the global optimal solution. At the same time, a higher weight factor also expands the search range and improves the global search capability. In the later stage, as the number of iterations increases, the weight factor gradually decreases, reducing the algorithm's dependence on the position of the previous generation of discoverers. This decreasing weight factor helps the algorithm to jump out of the local optimal solution because it reduces the dependence on past information and increases the diversity of the search. By reducing the weight factor, the algorithm can explore new search areas more flexibly, thereby increasing the probability of finding the global optimal solution. The optimal Latin hypercube makes the initial sample points evenly distributed, avoids the problem of falling into the local optimal solution, and expands the search range for parameters. Weight factor: The adjustment of the weight factor helps the algorithm jump out of the problem of the local optimal solution. Through the decreasing weight factor, the algorithm gradually reduces its dependence on historical information, increases the diversity of the search, and thus increases the probability of finding the global optimal solution. The number of unnecessary iterations is reduced, the search efficiency is improved, and the convergence time of the model is shortened. Therefore, the present invention can improve the search range, operation speed and prediction accuracy (prediction results are shown in Table 1) of battery health state estimation using a multi-core extreme learning machine network, and improve the accuracy of battery health state estimation.

[0114] Step S305 , estimating the battery health state based on the charge and discharge characteristic value test set and the battery health state estimation model.

[0115] Specifically, the above step S305 includes: Step S3051, inputting the charge and discharge characteristic value test set into the battery health state estimation model to obtain battery health state prediction data.

[0116] Step S3052, performing denormalization processing on the battery health state prediction data, and obtaining a battery health state estimation result based on the denormalized prediction data.

[0117] Specifically, the anti-normalization formula is as follows: (17); in, is the normalized output prediction data; is the denormalized actual prediction data.

[0118] The battery health state estimation method provided in this embodiment, the sparrow search algorithm under multi-strategy improvement effectively avoids the problem of the sparrow search algorithm falling into the local optimum during the optimization process by introducing the optimal Latin hypercube sampling and the adjustment of the weight factor in the multi-core extreme learning machine model parameters, and further improves the adaptability and convergence speed of the multi-core extreme learning machine model. By using the optimal Latin hypercube algorithm, the distribution of the initial population is made more uniform, and the coverage of the search space is improved. By dynamically adjusting the weight factor, the sparrow search algorithm under multi-strategy improvement achieves a balance between global search and local search, speeds up the optimization process, and reduces the iteration time. The charge and discharge characteristic value test set is input into the battery health state estimation model to obtain the battery health state prediction data; the battery health state prediction data is denormalized, and the battery health state estimation result is obtained based on the denormalized prediction data, which can significantly improve the effect of battery health state estimation and provide reliable technical control support for battery management and maintenance of electric vehicles and energy storage systems.

[0119] As one or more specific application embodiments of the present invention, Figures 4 to 10 The battery health state estimation method provided by the present invention is further described in detail. This embodiment includes reading the internal chemical changes of the battery health state and the characteristic values ​​related thereto, including: constant current charging time, voltage corresponding to the equal boost time, IC curve peak, voltage corresponding to the IC curve peak, electrolyte resistance and charge transfer resistance as characteristic values, and establishing a LWSSA-MKELM model (L: Latin Hypercube Sampling optimal Latin hypercube sampling, W: Weight Factor weight factor, SSA: Sparrow Search Algorithm sparrow search algorithm, MKELM: Multi-Kernel Extreme Learning Machine multi-core extreme learning machine) at the same time. By training the LWSSA-MKELM model, the optimal battery health state estimation model is obtained, such as Figure 4 As shown, the specific steps include: Step S1: Figure 5It is the battery cycle charge and discharge curve, the horizontal axis is the number of cycles, and the vertical axis is the battery capacity change. This example uses the NASA lithium-ion battery data set (battery number B0006, battery number B0007, and battery number B0018) and is tested at room temperature in three different operating modes (charging, discharging, and impedance). Charging is performed at 1.5A in constant current mode until the battery voltage reaches 4.2V, and then continues in constant voltage mode until the charging current drops to 20mA. Discharging is performed at 2A in constant current mode until the battery voltage drops to 2.5V (B0006 battery), 2.2V (B0007 battery), and 2.5V (B0018 battery), respectively. Impedance measurements are performed by electrochemical impedance spectroscopy (EIS) frequency scanning from 0.1Hz to 5kHz. Repeated charge and discharge cycles accelerate the aging of the battery, and impedance measurements provide insights into changes in internal battery parameters as aging progresses. When the battery reaches the end-of-life standard, that is, the rated capacity drops by 30% (from 2Ahr to 1.4Ahr), the experiment stops and the battery charge and discharge history data obtained from the battery charge and discharge cycle test experiment is obtained; Step S2: Extract key characteristic values ​​from the battery's charge and discharge history data, including: constant current charging time, constant voltage rise time. Perform Kalman filtering on the battery's cyclic charge and discharge curve to reduce noise interference, and construct the IC curve by the differential method to read the IC curve peak and the IC curve corresponding voltage. At the same time, read the electrolyte resistance and charge transfer resistance according to the EIS test. Use the above six types of data as charge and discharge characteristic values, divide the charge and discharge characteristic values ​​into a charge and discharge characteristic value training set and a charge and discharge characteristic value test set in a ratio of 4:6, and then normalize the charge and discharge characteristic value training set and the charge and discharge characteristic value test set in the [0,1] interval.

[0120] The S2 Kalman filter is a method to remove noise from data, and its formula is as follows: (1); in, It's time k The state prediction is used to update the state. F k is the state transition matrix, B k is the control input matrix, u k is the control vector, P k|k-1 is the predicted state covariance, Q k is the process noise covariance.

[0121] In its update phase, new measurement data is used to correct the prediction and update the estimate of the system state. The calculation formula can be expressed as: (2); in, K k Its Kalman gain, an indicator that measures the relative weight of the predicted covariance and the measured covariance, is used to update the state estimate. H k is the measurement matrix, R k is the measurement noise covariance, I is the identity matrix.

[0122] (3); (4); in, I is the identity matrix, through which the uncertainty of the state estimate can be updated.

[0123] The difference method formula is expressed as: (5); in, is the rate of change of a variable Y at the jth moment; is the time derivative of a variable X at the jth moment; is the variable X at the jth moment; It is The variable X at the moment; M is the time sampling interval.

[0124] Step S3: construct a multi-kernel extreme learning machine model through linear kernel function and polynomial kernel function, and use the root mean square error as the fitness value.

[0125] Specifically, the multi-core extreme learning machine model is an improved model based on the extreme learning machine. The extreme learning machine is a fast learning algorithm for a single hidden layer feedforward neural network. Its main feature is to randomly generate the weights and biases of the hidden layer, and then quickly calculate the weights of the output layer through the least squares method. On this basis, the multi-core extreme learning machine introduces multiple kernel functions to enhance the expressiveness and generalization performance of the model. The specific contents are as follows: The kernel function is used to map the input data from low-dimensional space to high-dimensional space. By combining the global kernel and the local kernel, the model can simultaneously process the global trend and local details of the data, thereby greatly improving the generalization ability and prediction accuracy of the model when facing complex nonlinear data. The linear kernel function and the polynomial kernel function selected by the present invention are: (6); Where K linear (x,x′) is the linear kernel function, K poly(x, x′) is a polynomial kernel function, x·x´ is the inner product of its input, γ is used to scale its inner product, and d is used to control the number of inner products.

[0126] The combined kernel function is: (7); Where p represents the mixing coefficient, which can be used to adjust the mapping performance of the kernel function.

[0127] Compared with the traditional kernel extreme learning machine algorithm, the multi-core extreme learning machine uses multiple kernel functions to replace random mapping, and relies on the least squares method to replace the iterative calculation between the hidden layer and the output layer.

[0128] (8); Among them, X is the sample data, that is, each charge and discharge eigenvalue; f(x) is the output of the multi-core extreme learning machine model; h(x) is the hidden layer output; H is the feature proposed matrix; C is the regularization parameter; I is the unit matrix; T is the target vector of the charge and discharge eigenvalue training set; K(x,x N ) is the kernel function; W ELM Reconstruct the matrix for the kernel.

[0129] The root mean square error of the training set of the multi-core extreme learning machine is expressed as: (12); in, Represented as the first The difference between the predicted value and the true value, N is the length of the charge and discharge feature value training set.

[0130] S4: Use the sparrow algorithm improved by multiple strategies to optimize the parameters of the multi-core extreme learning machine model and obtain the optimized parameters.

[0131] Specifically, the method includes: using the sparrow algorithm improved by multiple strategies to associate and fuse the regularization coefficient with the linear kernel function, the polynomial kernel function and each charge and discharge eigenvalue, respectively, to obtain the first fusion parameter; using the sparrow algorithm improved by multiple strategies to associate and fuse each charge and discharge eigenvalue with the linear kernel function and the polynomial kernel function, respectively, to obtain the second fusion parameter; using the sparrow algorithm improved by multiple strategies to associate and fuse the weight coefficient with each charge and discharge eigenvalue, to obtain the third fusion parameter; and using the first fusion parameter, the second fusion parameter and the third fusion parameter as the optimized multi-core extreme learning machine model parameters.

[0132] S5: Figure 4Based on the multi-strategy improved sparrow algorithm to optimize the multi-core extreme learning machine flow chart, the optimized parameters are used to build a lithium-ion battery health status estimation model for the multi-strategy improved sparrow algorithm to optimize the extreme learning machine; The multi-strategy improvement imitates the different division of labor among sparrow groups in the process of searching for food to find the optimal solution to the problem. First, m sparrows are generated to represent a population.

[0133] (10); The randomly generated sparrow population is prone to clustering in space, which can easily lead to the problem of local optimal solution and increase the time of searching for the optimal solution. The optimal Latin hypercube algorithm can effectively solve this problem and make the initial population evenly distributed in space. The basic idea is as follows: (11); Among them, x ij is the value of the jth sample in the i-th dimension. i is the lower bound of the i-th dimension. b i is the upper bound of the i-th dimension. n is the number of intervals in each dimension, which is also the total number of samples. k is the index of the j-th interval after random permutation, starting from 0. rand() is a function that generates random numbers in the interval [0,1) and is used to randomly select a point in the specified interval. max represents the permutation of all dimensions. Optimize to maximize the following minimum distance expression. min means to calculate the minimum distance between all sample point pairs (i, j), i≠j ensures that the distances between different points are compared. ‖xi-xj‖ is the Euclidean distance, which is used to measure the distance between any two sample points xi and xj.

[0134] The solution steps can be divided into: Figure 6 and Figure 7 As shown in the figure, the optimal Latin hypercube algorithm population is initialized more uniformly. First, the Latin hypercube sampling point L is initialized to define the objective function f(L), that is, the maximization of the minimum distance. The optimization algorithm is applied to the Latin hypercube sampling point L, and the sample position is updated to optimize f(L) iteratively until the convergence condition is met.

[0135] Furthermore, in each iteration of the sparrow algorithm under multi-strategy improvement, sparrows are usually divided into three types: discoverers, participants, and guards. During each iteration, the discoverer's position update is described as follows: (13); in, Indicates l+ In the 1st iteration i The location of the sparrow sample is located at j Weizhong;D max is the maximum number of iterations; α A random number in the interval (0,1); R is the guard value in [0,1]; Q is a random number that follows a standard normal distribution; L for d A matrix whose dimensions are all 1; d is the dimension of the objective function; Z ST is the risk-free value.

[0136] The participant positions are updated as follows: (14); in, Indicates l+ The optimal position of the entire population in generation 1; Indicates l The worst position of the entire population after iterations; E Represents a 1×d matrix containing all 1s or -1s.

[0137] The weight factor is introduced in the guard position update part to smoothly adjust the weight factor w through the periodic change of the sine function, thereby improving the global search ability of the algorithm and avoiding falling into the local optimum. The formula is as follows: (15); The position update after introducing the weight factor is as follows: (16); in, Indicates l The optimal population position after iterations; β is the step size for controlling the update position; K is a random number in the interval [-1,1]; f w Current worst fitness; f q Current optimal fitness; f iThe fitness of the current guard; after improvement, the algorithm can be more dependent on the position of the previous generation of finders (scouts) in the early stages. This dependence increases the search speed, because the position of the previous generation of finders is usually a relatively good search point, which helps to quickly find the global optimal solution or an area close to the global optimal solution. At the same time, a higher weight factor also expands the search range and improves the global search capability. Later, as the number of iterations increases, the weight factor gradually decreases, reducing the algorithm's dependence on the position of the previous generation of finders. This decreasing weight factor helps the algorithm jump out of the local optimal solution because it reduces dependence on past information and increases the diversity of the search. By reducing the weight factor, the algorithm can explore new search areas more flexibly, thereby increasing the probability of finding the global optimal solution.

[0138] S6: The health status of lithium-ion batteries is estimated by using the obtained battery health status prediction model. The charge and discharge characteristic value test set includes the constant current charging time, constant voltage rise time, IC curve peak and its corresponding voltage, electrolyte resistance and charge transfer resistance of B0006 battery, B0007 battery and B00018 battery in the NASA data set.

[0139] The specific contents of step S6 include: Step S6.1: Create a multi-core extreme learning machine model from step S3, and normalize the charge and discharge characteristic value training set. The normalization formula is: (9); In the formula, is the charge and discharge characteristic value after normalization; It is the test data of the charge and discharge characteristic value; and are the minimum and maximum values ​​of the charge and discharge characteristic values ​​respectively.

[0140] Step S6.2: Initialize the population using the optimal Latin hypercube algorithm. Set the initial population size, the maximum number of evolutionary iterations, the ratio of discoverers, participants and guards, the guard value and the risk-free value.

[0141] Step S6.3: Select the root mean square error as the fitness function and sort the fitness value of each sparrow.

[0142] Step S6.4: Update the positions of the discoverer, participant and guard.

[0143] Step S6.5: Calculate the fitness value and update the sparrow position.

[0144] Step S6.6: If the sparrow algorithm meets the termination conditions, the global optimal value is assigned to the regularization coefficient of the multi-core extreme learning machine, a linear kernel parameter, two polynomial kernel parameters, and the weight value to obtain the optimal prediction model; if the algorithm requirements are not met, repeat steps S6.3, S6.4 and S6.5.

[0145] Step S7: Different battery test error results are as follows Figure 8 , Fig. 9 and Fig.10 As shown, the correlation coefficient R of the charge and discharge characteristic value test set is calculated 2 , RMSE (root mean square error), MAE (Mean Absolute Error) and MAPE (Mean Absolute Percentage Error). The specific calculation formulas are as follows: (18); Among them, y i is the true value, RSS represents the residual sum of squares of the charge and discharge eigenvalues ​​in the test set, TSS It represents the total sum of squares of the charge and discharge feature values ​​in the test set. R2 is used to evaluate the accuracy of the model. Its value is between 0 and 1. The closer it is to 1, the higher the reliability of the model.

[0146] In order to reflect the advantages of the present invention, the LWSSA-MKELM is compared with the ELM, LWSSA-KELM and SSA-MKELM models, and the prediction results are shown in Table 1 below.

[0147] Table 1

[0148] In Table 1, ELM (Extreme Learning Machine, ‌ELM) is the extreme learning machine model. LWSSA- KELM (Latin Hypercube Sampling, Weight Factor, Kernel Extreme Learning Machine). SSA-MKELM (Sparrow Search Algorithm Multi-Kernel Extreme Learning Machine).

[0149] By comparing the above data and Figure 8 , Fig. 9 and Fig.10It can be found that the LWSSA-MKELM model proposed in the present invention is significantly better than other similar models that have not been improved or have been partially improved. Specifically, in terms of battery health state estimation, the LWSSA-MKELM model significantly improves the prediction accuracy and calculation efficiency, while enhancing the stability and robustness of the model. These improvements enable the LWSSA-MKELM model to maintain excellent performance under different experimental conditions, providing a more reliable and efficient solution for battery health management and maintenance.

[0150] This embodiment provides a battery health state estimation method, which introduces the optimal Latin hypercube and weight factor improved sparrow search algorithm in the traditional multi-core extreme learning machine model. First, through constant current charging time, equal boost time corresponding voltage, IC curve peak, IC curve peak corresponding voltage, electrolyte resistance and charge transfer resistance, based on the above six points as characteristic values, a training set and a test set are constructed to describe the uncertain characteristics of the battery health state; secondly, a multi-core extreme learning machine constructed by a linear kernel function and a polynomial kernel function is used, and its root mean square error is used as the fitness function; then the optimal Latin hypercube is used to optimize the initial population position of the sparrow; then the population is initialized first, and the weight factor is used to balance its local and global search capabilities at the position of the early warning person to finally obtain its optimal parameters; the optimal parameters are brought to the multi-core extreme learning to obtain the optimal model. The present invention solves the problems of narrow search range, inability to guarantee accuracy and long parameter search iteration time in the process of battery state estimation, and can improve the computing speed and prediction accuracy of battery health state estimation using a multi-core extreme learning machine network.

[0151] In this embodiment, a battery health status estimation device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and will not be repeated here. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0152] This embodiment provides a battery health status estimation device, such as Fig.11 As shown, including: The data acquisition and processing module 1101 is used to acquire multiple battery charge and discharge history data, extract multiple charge and discharge feature values ​​from the multiple battery charge and discharge history data, and divide the multiple charge and discharge feature values ​​into a charge and discharge feature value training set and a charge and discharge feature value test set.

[0153] The multi-core extreme learning machine model construction module 1102 is used to construct a multi-core extreme learning machine model based on the charge and discharge characteristic value training set and a preset kernel function.

[0154] The parameter optimization module 1103 is used to optimize the multi-core extreme learning machine model parameters by using the sparrow algorithm improved with multiple strategies to obtain optimized multi-core extreme learning machine model parameters.

[0155] The battery health state estimation model construction module 1104 is used to construct a battery health state estimation model based on the charge and discharge characteristic value training set, the optimized multi-core extreme learning machine model parameters, the preset weight factors and the sparrow algorithm under multiple strategy improvements.

[0156] The battery health state estimation module 1105 is used to estimate the battery health state based on the charge and discharge characteristic value test set and the battery health state estimation model.

[0157] In some optional implementations, the data acquisition and processing module 1101 includes: The characteristic value extraction unit is used to extract the constant current charging time, constant voltage rise time, IC curve peak value, IC curve peak corresponding voltage, electrolyte resistance and charge transfer resistance of various batteries from various battery charge and discharge history data.

[0158] The first determining unit is used to use constant current charging time, constant voltage rise time, IC curve peak, IC curve peak corresponding voltage, electrolyte resistance and charge transfer resistance as multiple charging and discharging characteristic values.

[0159] In some optional implementations, the preset kernel function includes a linear kernel function and a polynomial kernel function; the multi-core extreme learning machine model includes a hidden layer and an output layer; the multi-core extreme learning machine model construction module 1102 includes: The combination unit is used to combine the linear kernel function, the polynomial kernel function and the preset mixing coefficient to obtain the combined kernel function.

[0160] The weight coefficient generating unit is used to randomly generate the weight coefficient of each charge and discharge characteristic value in the charge and discharge characteristic value training set in the hidden layer.

[0161] The multi-core extreme learning machine model building unit is used to iteratively calculate the combined kernel function, the weight coefficient of each charge and discharge eigenvalue and the preset regularization coefficient in the output layer using the least squares method to build a multi-core extreme learning machine model.

[0162] In some optional embodiments, the multi-core extreme learning machine model parameters include each charge and discharge eigenvalue, a linear kernel function, a polynomial kernel function, a weight coefficient, and a preset regularization coefficient; the parameter optimization module 1103 includes: The first fusion unit is used to associate and fuse the regularization coefficient with the linear kernel function, the polynomial kernel function and each charge and discharge eigenvalue by using the sparrow algorithm improved by multiple strategies to obtain the first fusion parameter.

[0163] The second fusion unit is used to associate and fuse each charge and discharge characteristic value with a linear kernel function and a polynomial kernel function by using a sparrow algorithm improved by multiple strategies to obtain a second fusion parameter.

[0164] The third fusion unit is used to associate and fuse the weight coefficient with each charge and discharge characteristic value by using the sparrow algorithm improved by multiple strategies to obtain a third fusion parameter.

[0165] The second determining unit is used to use the first fusion parameter, the second fusion parameter and the third fusion parameter as the optimized multi-core extreme learning machine model parameters.

[0166] In some optional implementations, the battery health status estimation model building module 1104 includes: The normalization unit is used to normalize the charge and discharge characteristic values ​​in the charge and discharge characteristic value training set to obtain normalized charge and discharge characteristic values.

[0167] The initialization unit is used to randomly generate the initial population of the sparrow algorithm under various strategy improvements, and use the optimal Latin hypercube algorithm to initialize the initial population to obtain the objective function.

[0168] The battery health state estimation model construction unit is used to iterate the objective function based on the normalized charge and discharge characteristic values ​​and preset weight factors and adopt the sparrow algorithm improved by multiple strategies until the convergence condition is met and the iteration ends. The optimized multi-core extreme learning machine model parameters are assigned to the objective function obtained in the last iteration to obtain the battery health state estimation model.

[0169] In some optional implementations, the battery health state estimation model building unit includes: The division subunit is used to divide the normalized charge and discharge characteristic values ​​into discoverers, participants and guards.

[0170] The fitness value calculation subunit is used to calculate the root mean square errors of the normalized charge and discharge characteristic values ​​corresponding to the discoverer, the participant and the guard, respectively, and use the corresponding root mean square errors as the fitness value of the discoverer, the fitness value of the participant and the fitness value of the guard, respectively.

[0171] The sparrow algorithm processing subunit is used to perform local and global searches on the objective function based on preset weight factors and using the sparrow algorithm improved by multiple strategies, and to update the positions of the discoverer, participant and guard according to the fitness value of the discoverer, the fitness value of the participant and the fitness value of the guard until the sparrow algorithm improved by multiple strategies meets the convergence conditions and the iteration ends.

[0172] In some optional implementations, the battery health status estimation module 1105 includes: The input unit is used to input the charge and discharge characteristic value test set into the battery health state estimation model to obtain battery health state prediction data.

[0173] The denormalization unit is used to perform denormalization processing on the battery health state prediction data, and obtain the battery health state estimation result based on the denormalized prediction data.

[0174] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0175] The battery health status estimation device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0176] The embodiment of the present invention also provides a computer device having the above Fig.11 A battery health status estimation device is shown.

[0177] See also Fig.12 , Fig.12 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Fig.12 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Fig.12 A processor 10 is taken as an example.

[0178] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0179] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0180] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0181] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0182] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Fig.12 The example of connecting through bus is taken in the following.

[0183] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device may be a touch screen.

[0184] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0185] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0186] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for estimating a battery health state, characterized in that: The method comprises: Acquire multiple battery charge and discharge history data, extract multiple charge and discharge characteristic values ​​from the multiple battery charge and discharge history data, and divide the multiple charge and discharge characteristic values ​​into a charge and discharge characteristic value training set and a charge and discharge characteristic value test set; Construct a multi-core extreme learning machine model based on the charge and discharge characteristic value training set and the preset kernel function; The multi-core extreme learning machine model parameters are optimized by using a sparrow algorithm improved by multiple strategies to obtain optimized multi-core extreme learning machine model parameters; Constructing a battery health state estimation model based on the charge and discharge characteristic value training set, the optimized multi-core extreme learning machine model parameters, the preset weight factors and the sparrow algorithm under multiple strategy improvements; The battery health state is estimated based on the charge and discharge characteristic value test set and the battery health state estimation model.

2. The method according to claim 1, characterized in that Extracting multiple charge and discharge characteristic values ​​from the multiple battery charge and discharge history data includes: Extracting constant current charging time, constant voltage rise time, IC curve peak, IC curve peak corresponding voltage, electrolyte resistance and charge transfer resistance of various batteries from the various battery charge and discharge history data; The constant current charging time, the constant voltage rise time, the IC curve peak, the voltage corresponding to the IC curve peak, the electrolyte resistance and the charge transfer resistance are used as multiple charge and discharge characteristic values.

3. The method according to claim 1, characterized in that The preset kernel function includes a linear kernel function and a polynomial kernel function; the multi-core extreme learning machine model includes a hidden layer and an output layer; The multi-core extreme learning machine model is constructed based on the charge and discharge characteristic value training set and the preset kernel function, including: The linear kernel function, the polynomial kernel function and the preset mixing coefficient are combined to obtain a combined kernel function; Randomly generate a weight coefficient for each charge and discharge characteristic value in the charge and discharge characteristic value training set in a hidden layer; In the output layer, the least squares method is used to iteratively calculate the combined kernel function, the weight coefficient of each charge and discharge eigenvalue and the preset regularization coefficient to construct a multi-core extreme learning machine model.

4. The method according to claim 3, characterized in that The multi-core extreme learning machine model parameters include each charge and discharge characteristic value, a linear kernel function, a polynomial kernel function, a weight coefficient and a preset regularization coefficient; The multi-core extreme learning machine model parameters are optimized by using the sparrow algorithm improved by multiple strategies, and the optimized multi-core extreme learning machine model parameters include: The sparrow algorithm improved by multiple strategies is used to associate and fuse the regularization coefficient with the linear kernel function, the polynomial kernel function and each charge and discharge eigenvalue to obtain the first fusion parameter; The sparrow algorithm improved by multiple strategies is used to associate and fuse each charge and discharge characteristic value with the linear kernel function and the polynomial kernel function to obtain the second fusion parameter; The sparrow algorithm improved by multiple strategies is used to associate and fuse the weight coefficient with each charge and discharge characteristic value to obtain the third fusion parameter; The first fusion parameter, the second fusion parameter and the third fusion parameter are used as optimized multi-core extreme learning machine model parameters.

5. The method according to claim 1, characterized in that The battery health state estimation model is constructed based on the charge and discharge characteristic value training set, the optimized multi-core extreme learning machine model parameters, the preset weight factors and the sparrow algorithm under multiple strategy improvements, including: Normalizing the charge and discharge characteristic values ​​in the charge and discharge characteristic value training set to obtain normalized charge and discharge characteristic values; The initial population of the sparrow algorithm under various improved strategies is randomly generated, and the optimal Latin hypercube algorithm is used to initialize the initial population to obtain the objective function; Based on the normalized charge and discharge eigenvalues ​​and preset weight factors, the objective function is iterated by using a sparrow algorithm improved by multiple strategies until the convergence condition is met and the iteration ends. The optimized multi-core extreme learning machine model parameters are assigned to the objective function obtained in the last iteration to obtain a battery health state estimation model.

6. The method according to claim 5, characterized in that The objective function is iterated based on the normalized charge and discharge characteristic value and the preset weight factor and the sparrow algorithm improved by using multiple strategies until the convergence condition is met and the iteration ends, including: The normalized charge and discharge characteristic values ​​are divided into discoverers, participants and guards; Calculate the root mean square errors of the normalized charge and discharge characteristic values ​​corresponding to the discoverer, participant and guard respectively, and use the corresponding root mean square errors as the fitness value of the discoverer, the fitness value of the participant and the fitness value of the guard respectively; Based on the preset weight factor and using multiple strategies to improve the sparrow algorithm, the objective function is searched locally and globally, and the positions of the discoverer, participant and guard are updated according to the fitness value of the discoverer, the fitness value of the participant and the fitness value of the guard, until the sparrow algorithm improved by multiple strategies meets the convergence conditions and the iteration ends.

7. The method according to claim 1, characterized in that Estimation of battery health status based on the charge and discharge characteristic value test set and battery health status estimation model includes: Inputting the charge and discharge characteristic value test set into a battery health state estimation model to obtain battery health state prediction data; The battery health state prediction data is denormalized, and a battery health state estimation result is obtained based on the denormalized prediction data.

8. A battery health status estimation device, characterized in that: The device comprises: A data acquisition and processing module, used to acquire multiple battery charge and discharge history data, extract multiple charge and discharge characteristic values ​​from the multiple battery charge and discharge history data, and divide the multiple charge and discharge characteristic values ​​into a charge and discharge characteristic value training set and a charge and discharge characteristic value test set; A multi-core extreme learning machine model building module is used to build a multi-core extreme learning machine model based on a charge and discharge characteristic value training set and a preset kernel function; A parameter optimization module, used to optimize the multi-core extreme learning machine model parameters by using a sparrow algorithm improved by multiple strategies to obtain optimized multi-core extreme learning machine model parameters; A battery health state estimation model construction module is used to construct a battery health state estimation model based on the charge and discharge characteristic value training set, the optimized multi-core extreme learning machine model parameters, the preset weight factors and the sparrow algorithm under multiple strategy improvements; The battery health state estimation module is used to estimate the battery health state based on the charge and discharge characteristic value test set and the battery health state estimation model.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the battery health status estimation method according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the battery health state estimation method according to any one of claims 1 to 7.

11. A computer program product, characterized in that The method comprises computer instructions, wherein the computer instructions are used to cause a computer to execute the battery health state estimation method according to any one of claims 1 to 7.

Citation Information

Patent Citations

  • Short-term load prediction method for optimizing Elman neural network based on sparrow search algorithm

    CN113689029A

  • Wind power prediction method for optimizing kernel extreme learning machine based on sparrow search algorithm

    CN114065610A

  • Wind power composite prediction method based on PCA-BP-SSA algorithm

    CN115496187A

  • Lithium ion battery health state prediction method optimized by improved sparrow algorithm

    CN115498283A

  • Method for training prediction model for time series data prediction and related equipment

    CN115545151A

Cited By

  • Battery charging protection method and system based on intelligent control and storage medium

    CN120016657A

  • Battery health state assessment method and system, storage medium and computer

    CN120195575A

  • Battery examination and vehicle network interaction control method based on SSA and Autoformer

    CN120454232A

  • Battery physical examination and vehicle-network interaction control method based on SSA and Autoformer

    CN120454232B

  • State estimation method, device and equipment for energy storage battery and medium

    CN120468679A