Battery remaining capacity estimation method and device, storage medium and computer equipment

Abstract

According to the battery remaining capacity estimation method provided by the invention, a plurality of working conditions of battery operation and a plurality of initial sliding window length values corresponding to an adaptive extended Kalman filtering algorithm are determined; under each initial sliding window length value, calculating an average absolute error value of the battery remaining capacity of the adaptive extended Kalman filtering algorithm under each working condition; selecting a target sliding window length value from the initial sliding window length values according to the average absolute error values; and when it is determined that the length value of the target sliding window has the temperature migration capability in the adaptive extended Kalman filtering algorithm, calculating the remaining capacity of the battery by using the length value of the target sliding window. The method effectively solves the problems that in the prior art, the sliding window length parameter lacks an effective optimization method, and estimation precision is reduced due to working condition and temperature changes, and the precision and stability of SOC estimation are remarkably improved.

Description

Technical Field

[0001] The present application relates to the technical field of lithium-ion battery SOC estimation, and in particular to a battery remaining power estimation method, device, storage medium and computer equipment. Background Art

[0002] In recent years, as the application scope of lithium-ion batteries continues to expand, accurately estimating their state of charge (SOC) has become increasingly important. Model-based methods are widely used because of their simple calculations and easy implementation. They usually use equivalent circuit models combined with adaptive extended Kalman filter (AEKF) algorithms to achieve accurate estimation of SOC. However, when using the AEKF algorithm to estimate SOC, changes in operating conditions or temperature will affect the estimation accuracy. In order to ensure the estimation accuracy of the algorithm, the key parameter of the sliding window length needs to be adjusted frequently. In addition, the value of the sliding window length also has a significant impact on the estimation accuracy of the algorithm.

[0003] At present, in the process of estimating SOC using the AEKF algorithm, there is a lack of effective optimization methods and related characteristics research for the sliding window length parameter. If the parameter is set unreasonably, it will have a negative impact on the estimation accuracy of the algorithm. In addition, when the working conditions or temperature change, the parameter needs to be reset to ensure the estimation accuracy of the algorithm, which will undoubtedly bring many inconveniences to the research and application of the algorithm. In summary, the optimization research on the sliding window length parameter in the AEKF algorithm is of great significance, which helps to improve the accuracy and stability of SOC estimation and provide more reliable technical support for the management and application of lithium-ion batteries. Summary of the invention

[0004] The purpose of the present application is to solve at least one of the above-mentioned technical deficiencies, especially the technical deficiencies in the prior art of lacking an effective optimization method for sliding window length parameters and related characteristic research.

[0005] In a first aspect, the present application provides a method for estimating the remaining battery power, the method comprising:

[0006] Determine multiple operating conditions of the battery and multiple initial sliding window length values ​​corresponding to the adaptive extended Kalman filter algorithm;

[0007] Under each initial sliding window length value, the mean absolute error value of the remaining battery power of the adaptive extended Kalman filter algorithm in each working condition is calculated;

[0008] According to each mean absolute error value, a target sliding window length value is selected from each initial sliding window length value;

[0009] When it is determined that in the adaptive extended Kalman filter algorithm, the target sliding window length value has temperature migration capability, the target sliding window length value is used to calculate the remaining battery power.

[0010] In one embodiment, the step of selecting a target sliding window length value from each initial sliding window length value according to each mean absolute error value includes:

[0011] Calculate the summary value of each mean absolute error value, and calculate the ratio of each mean absolute error value to the summary value;

[0012] For each initial sliding window length value, calculate the cumulative value of the corresponding ratios of each mean absolute error value under the initial sliding window length value;

[0013] Among the accumulated values, the initial sliding window length value corresponding to the accumulated value with the smallest value is used as the target sliding window length value.

[0014] In one embodiment, the step of calculating the summary value of each mean absolute error value includes:

[0015] Cross-tabulate the initial sliding window length values ​​and the average absolute error values ​​under various working conditions to obtain a contingency table;

[0016] In a contingency table, summary values ​​are obtained by calculating the sum of the column sums or the sum of the row sums.

[0017] In one embodiment, for each initial sliding window length value, the step of calculating the cumulative value of the corresponding ratios of the mean absolute error values ​​under the initial sliding window length value includes:

[0018] Cross-tabulate the ratios of each initial sliding window length value and each working condition to obtain a probability distribution table;

[0019] In the probability distribution table, calculate the cumulative value of the corresponding ratio of each working condition under each initial sliding window length value.

[0020] In one embodiment, the step of determining that the target sliding window length value has temperature migration capability in the adaptive extended Kalman filter algorithm includes:

[0021] According to the target sliding window length value, construct at least two sample matrices corresponding to different temperatures;

[0022] After calculating the correlation matrix according to each sample matrix, the eigenvalue of the correlation structure matrix is ​​calculated according to the correlation matrix;

[0023] Set the correlation hypothesis and irrelevance hypothesis based on the eigenvalues ​​and construct the test statistic;

[0024] The statistical value is calculated according to the test statistic and compared with a preset critical value. If the statistical value exceeds the critical value, it is determined that the target sliding window length value is transferable at different temperatures.

[0025] In one embodiment, the correlation matrix is ​​as follows:

[0026]

[0027] The correlation structure matrix is ​​shown below:

[0028]

[0029] in, and Represent the sample covariance matrix of each sample matrix, and represents the cross covariance matrix between each sample matrix, represents the correlation structure matrix, Represents the mean absolute error value corresponding to a temperature, Represents the mean absolute error value corresponding to another temperature.

[0030] In one embodiment, the correlation assumption and the irrelevance assumption are as follows:

[0031]

[0032] The test statistic is as follows:

[0033]

[0034] in, , represents the smaller value in each sample matrix dimension, represents a sample matrix dimension, represents another sample matrix dimension, represents the eigenvalue, Represents a statistic used to measure the correlation between two sets of variables. represents the irrelevance assumption, represents the correlation hypothesis, represents another statistic used for hypothesis testing, represents the number of variables in a sample matrix, represents the number of variables in another sample matrix, represents the eigenvalue number of the correlation structure matrix, Indicates the level of inspection.

[0035] In a second aspect, the present application provides a battery remaining power estimation device, the device comprising:

[0036] An initial sliding window length value determination module is used to determine multiple initial sliding window length values ​​corresponding to multiple operating conditions of the battery and the adaptive extended Kalman filter algorithm;

[0037] A mean absolute error value calculation module is used to calculate the mean absolute error value of the remaining battery power of the adaptive extended Kalman filter algorithm in each working condition under each initial sliding window length value;

[0038] A target sliding window length value selection module is used to select a target sliding window length value from each initial sliding window length value according to each mean absolute error value;

[0039] The temperature migration capability verification module is used to calculate the remaining battery power using the target sliding window length value after determining that the target sliding window length value has the temperature migration capability in the adaptive extended Kalman filter algorithm.

[0040] In a third aspect, the present application provides a storage medium: the storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the battery remaining power estimation method as described in any of the above embodiments.

[0041] In a fourth aspect, the present application provides a computer device, comprising: one or more processors, and a memory;

[0042] The memory stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the steps of the remaining battery power estimation method in any of the above embodiments are performed.

[0043] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0044] The battery remaining capacity estimation method provided in the present application provides diversified basic data for the optimization of the sliding window length parameters by determining multiple operating conditions of the battery and multiple initial sliding window length values ​​corresponding to the adaptive extended Kalman filter algorithm; under each initial sliding window length value, the mean absolute error value of the remaining battery capacity of the adaptive extended Kalman filter algorithm in each operating condition is calculated, and the performance of different sliding window length values ​​is evaluated by quantizing the error, so that the sliding window length value with the best performance under different operating conditions can be accurately screened out; according to each mean absolute error value, a target sliding window length value is selected from each initial sliding window length value, and this process not only realizes the optimization of the sliding window length parameters, but also effectively solves the problem of reduced estimation accuracy caused by changes in operating conditions, ensuring that a higher SOC estimation accuracy can be obtained under different operating conditions; when it is determined that the target sliding window length value has temperature migration capability, it is used to calculate the remaining battery capacity, further solving the influence of temperature changes on the estimation accuracy, so that the method can maintain stable SOC estimation performance under different temperature environments. In summary, this method effectively solves the problem in the prior art of lack of effective optimization method for sliding window length parameters and the problem of decreased estimation accuracy caused by operating conditions and temperature changes, significantly improves the accuracy and stability of SOC estimation, and provides more reliable technical support for the management and application of lithium-ion batteries. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0046] Figure 1 A flowchart of a method for estimating the remaining battery power provided in an embodiment of the present application;

[0047] Figure 2 An example diagram of the SOC under different operating conditions estimated by the AEKF algorithm provided in the embodiment of the present application at 25°C;

[0048] Figure 3 This is an example diagram of the SOC under different working conditions estimated by the AEKF algorithm provided in the embodiment of the present application at 45°C;

[0049] Figure 4 A schematic diagram of the structure of a battery remaining power estimation device provided in an embodiment of the present application;

[0050] Figure 5 A schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0052] The present application provides a method for estimating the remaining battery power. The following embodiments are described by taking the method applied to a computer device as an example. It can be understood that the computer device can be any device with data processing functions, including but not limited to a single server, a server cluster, a personal laptop computer, a desktop computer, etc. Figure 1 As shown, the method may include the following steps:

[0053] S101: Determine multiple operating conditions of the battery and multiple initial sliding window length values ​​corresponding to the adaptive extended Kalman filter algorithm.

[0054] Among them, multiple operating conditions of the battery refer to the operating conditions of the battery under different working conditions. These operating conditions cover various complex situations that the battery may encounter during actual use, such as constant current charging, constant current discharging, dynamic stress test conditions, federal urban cycle conditions, highway cycle conditions, and pulse charge and discharge conditions. Each operating condition has specific requirements and change rules for battery parameters such as current, voltage, and temperature, thereby fully reflecting the performance of the battery in different usage scenarios.

[0055] The Adaptive Extended Kalman Filter (AEKF) algorithm is a recursive algorithm for state estimation, which is particularly suitable for dealing with state estimation problems in nonlinear systems. In the AEKF algorithm, the sliding window length is a key parameter that determines the amount of historical data considered in the estimation process. The initial sliding window length value refers to the initial value of the sliding window length set when the algorithm starts running. These different initial values ​​constitute multiple sets of initial sliding window length values. For example, depending on the type of battery, application scenario, and empirical data, the initial sliding window length value can be set to different values ​​such as 50, 100, 150, 200, etc. These values ​​will affect the algorithm's estimation accuracy and convergence speed of the remaining battery capacity (SOC).

[0056] In this step, for determining multiple operating conditions of the battery, first, the data acquisition system can be used to monitor various parameters of the battery in the actual operation process in real time, including current, voltage, temperature, etc. Through long-term experiments and actual application data collection, a large amount of battery operation data in different scenarios is accumulated. For example, in the actual driving process of electric vehicles, the battery current and voltage changes of the vehicle under different road conditions such as urban roads, highways, mountain roads, and corresponding ambient temperature and other information are recorded. Secondly, based on the collected data, combined with the battery usage specifications and industry standards, the battery operation conditions can be classified and defined. For example, according to the magnitude and change law of the current, the constant current charging condition is defined as the operating state of the battery when the charging current remains constant (such as 1A); the dynamic stress test (DST) condition is defined as the state of the battery operating under a specific current pulse sequence (such as the charge and discharge current that changes at a certain time interval and amplitude). Finally, in a laboratory environment, the battery test equipment can be used to simulate various defined conditions and perform charge and discharge experiments on the battery. By repeating multiple experiments, the accuracy and representativeness of the defined conditions are verified to ensure that these conditions can fully cover the situations that the battery may encounter in actual applications. For example, when simulating constant current discharge conditions, different discharge currents (such as 0.5C, 1C, 2C, etc.) and termination voltages are set to observe changes in battery parameters such as voltage and capacity to verify the stability and consistency of battery performance under this condition.

[0057] In order to determine the multiple initial sliding window length values ​​corresponding to the adaptive extended Kalman filter algorithm, we can refer to the existing battery management technology literature, industry standards and previous research results to understand the commonly used sliding window length ranges for different battery types and application scenarios. For example, for the application of lithium-ion batteries in electric vehicles, some studies have shown that the initial sliding window length between 100 and 200 can achieve better SOC estimation results. A series of experimental schemes can also be designed to set multiple initial sliding window length values ​​for testing for different battery types and working conditions. For example, for a certain lithium iron phosphate battery, under constant current discharge conditions, the initial sliding window length is set to 50, 100, 150, 200, 250, etc., and multiple charge and discharge cycle experiments are performed under each length value, and the corresponding SOC estimation error and other performance indicators are recorded. The performance of the AEKF algorithm under different initial sliding window length values ​​can also be evaluated by analyzing the experimental data, including SOC estimation accuracy, algorithm convergence speed, and robustness to noise. According to the evaluation results, the initial sliding window length values ​​with good comprehensive performance under various working conditions are selected as candidate sets. For example, it is found that when the initial sliding window length is 150, the mean absolute error of SOC estimation is small under constant current discharge conditions and dynamic stress test conditions, and the algorithm converges quickly, so this value can be included in the candidate set.

[0058] It can be understood that the purpose of determining multiple operating conditions of the battery is to comprehensively consider various situations that the battery may encounter in actual use and ensure that the designed estimation method has wide applicability and robustness. Only by fully understanding the operating characteristics of the battery under different operating conditions can sufficient data support be provided for subsequent parameter optimization. The purpose of determining multiple initial sliding window length values ​​corresponding to the adaptive extended Kalman filter algorithm is to find the most suitable sliding window length through parameter optimization and improve the estimation accuracy and stability of the algorithm.

[0059] S102: Under each initial sliding window length value, calculating the mean absolute error value of the remaining battery power of the adaptive extended Kalman filter algorithm in each operating condition.

[0060] Among them, the mean absolute error value is an important indicator for measuring the accuracy of the adaptive extended Kalman filter (AEKF) algorithm in estimating the remaining battery capacity (SOC). It is obtained by calculating the absolute value of the error between the SOC value estimated by the algorithm and the actual SOC value of the battery under specific working conditions, and then taking the average of these absolute error values. For example, under constant current discharge conditions, after multiple charge and discharge cycle experiments, a set of SOC estimation error data is obtained. After calculating the absolute value of these data and averaging them, the mean absolute error value under this condition is obtained. The smaller the value, the higher the estimation accuracy of the algorithm under this condition.

[0061] In this step, after determining multiple initial sliding window length values, subsequent calculation operations need to be performed for each value. For example, the four initial sliding window length values ​​of 50, 100, 150, and 200 have been determined, so the next step is to process these four values ​​separately. First, set the sliding window length parameter in the AEKF algorithm to the first value 50 to ensure that the algorithm uses a sliding window of this length to process data in subsequent calculations; then, keep other algorithm parameters unchanged to ensure the accuracy of the experiment and avoid interference with the results caused by changes in other parameters; then, start the AEKF algorithm and start running it at this sliding window length, preparing to estimate the SOC of the battery under different operating conditions.

[0062] Before starting the calculation, you need to ensure that you have collected enough operating data of the battery under different working conditions, including current, voltage, temperature, and the corresponding actual SOC value. These data can be collected by simulating various working conditions in a laboratory environment through experimental equipment, or they can be obtained from the battery management system in actual applications. For example, for constant current charging conditions, record the current, voltage, and temperature data of the battery at regular time intervals (such as 1 minute) during the process of charging from 0% SOC to 100% SOC, and measure the real SOC value through high-precision equipment as a reference standard. For each working condition, the collected data is input into the AEKF algorithm with the initial sliding window length set in chronological order. The algorithm estimates the SOC value of the battery in real time based on the input data. During the estimation process, it is necessary to record the estimated SOC value and the corresponding actual SOC value at each time point. For example, under the dynamic stress test (DST) condition, as the current pulsates, the AEKF algorithm continuously updates the battery model parameters and outputs the corresponding SOC estimation value, and the computer synchronously stores these estimation values ​​and the actual values ​​obtained by the benchmark method.

[0063] After obtaining the estimated SOC value and the actual SOC value, the error between the two is calculated point by point, that is, the actual SOC value minus the estimated SOC value, to obtain the error sequence. Then, take the absolute value of each error value in the error sequence to obtain the absolute error sequence. For example, in a certain pulse charging process, the estimated SOC value is 30%, and the actual SOC value is 32%, then the error is 2%, and the absolute error is 2%; if the next estimated SOC value is 35%, and the actual value is 33%, the error is -2%, and the absolute error is still 2%. Add all the values ​​in the obtained absolute error sequence, and then divide them by the length of the sequence, that is, the number of data points, to obtain the average absolute error value corresponding to the initial sliding window length value under this working condition. For example, under a certain initial sliding window length, after calculating 100 data points, the sum of the absolute error sequence obtained is 150%, and the average absolute error value is 1.5%. This process needs to be repeated for each working condition to obtain the average absolute error value of the algorithm in all working conditions under this initial sliding window length.

[0064] It is understandable that only by actually running the algorithm and obtaining the estimated value can we compare and analyze it with the true value, and then evaluate the performance of the algorithm under different working conditions. The purpose of calculating the mean absolute error is to quantify the error between the estimated value and the true value. By taking the average value of the absolute error, the overall estimation accuracy of the algorithm can be more intuitively reflected. Performing these operations at each initial sliding window length value can perform a comprehensive performance evaluation for different parameter settings, thereby providing a reliable basis for the subsequent selection of the optimal sliding window length value.

[0065] S103: Selecting a target sliding window length value from each initial sliding window length value according to each mean absolute error value.

[0066] Among them, the target sliding window length value refers to the optimal sliding window length value selected from multiple initial sliding window length values ​​according to certain selection strategies and standards. This value can enable the AEKF algorithm to achieve the best accuracy and stability in SOC estimation under different battery operating conditions. For example, among the mean absolute error values ​​corresponding to multiple initial sliding window length values, the sliding window length corresponding to the value with the smallest error is found, which is the target sliding window length value, which enables the algorithm to output high-precision SOC estimation results under various operating conditions.

[0067] In this step, before starting to select the target sliding window length value, the mean absolute error values ​​under each initial sliding window length value that have been calculated can be sorted and preprocessed. First, these data are arranged in the order of the size of the initial sliding window length value to facilitate subsequent analysis and comparison. For example, the mean absolute error values ​​corresponding to the initial sliding window length values ​​of 50, 100, 150, and 200 are recorded in the corresponding positions to form a data table. At the same time, the data is detected and eliminated for outliers to ensure the accuracy and reliability of the data. For example, if a mean absolute error value deviates significantly from the normal range, it may be caused by accidental errors or equipment failures during the experiment, and it needs to be eliminated and the data needs to be re-acquired.

[0068] In order to select the target sliding window length value from multiple initial sliding window length values, a suitable selection strategy can be determined. Common selection strategies include the minimum error criterion, the comprehensive performance evaluation method, etc. The minimum error criterion directly selects the initial sliding window length value corresponding to the minimum mean absolute error value as the target value; the comprehensive performance evaluation rule not only considers the mean absolute error value, but also combines other factors such as the convergence speed of the algorithm, the computational complexity, etc., and determines the optimal target sliding window length value through weighted scoring. For example, in some application scenarios with high real-time requirements, in addition to pursuing high estimation accuracy, it is also necessary to consider the running speed of the algorithm. At this time, it is more appropriate to use the comprehensive performance evaluation method.

[0069] According to the determined selection strategy, the target sliding window length value is selected from each initial sliding window length value. If the minimum error criterion is adopted, the average absolute error values ​​corresponding to all initial sliding window length values ​​are traversed, the smallest error value is found, and then the corresponding sliding window length value is recorded as the target value. For example, when the initial sliding window length value is 100, the average absolute error value is 1.2%, and the error is greater than this value at other length values, then 100 is the target sliding window length value. If the comprehensive performance evaluation method is adopted, it is necessary to build an evaluation model to quantify and score each performance indicator under each initial sliding window length value, and then calculate the total score, and finally select the sliding window length value with the highest total score as the target value. For example, assuming that the average absolute error value accounts for 60% of the weight, the algorithm convergence speed accounts for 30% of the weight, and the computational complexity accounts for 10% of the weight, the computer device will calculate the comprehensive score of each initial sliding window length value based on these weights and the corresponding indicator values, and the one with the highest score is the target sliding window length value.

[0070] It can be understood that the purpose of selecting the target sliding window length value is to determine the optimal parameter that can be stably used and has high estimation accuracy under different working conditions from a large number of candidate values, thereby effectively solving the problem of the lack of effective optimization methods for sliding window length parameters in the prior art and avoiding estimation errors caused by unreasonable parameter settings. By accurately screening the target sliding window length value, the accuracy and stability of the battery remaining power estimation can be significantly improved, the impact caused by changes in working conditions can be reduced, and more reliable technical support can be provided for the management and application of lithium-ion batteries, thereby improving the performance and safety of the entire battery system.

[0071] S104: When it is determined in the adaptive extended Kalman filter algorithm that the target sliding window length value has temperature migration capability, the target sliding window length value is used to calculate the remaining battery power.

[0072] Among them, temperature migration capability refers to the target sliding window length value being able to maintain its stability and adaptability to the accuracy of battery remaining capacity (SOC) estimation under different temperature environments. Specifically, when the battery temperature changes, such as from 25°C to 40°C, the target sliding window length value with temperature migration capability enables the adaptive extended Kalman filter (AEKF) algorithm to still accurately estimate the SOC under new temperature conditions without the need to readjust the parameter. This capability ensures the reliability and consistency of the algorithm under different temperature scenarios, which is crucial for battery management in complex environments.

[0073] In this step, the battery can be charged and discharged under different temperature environments to collect its operating data under various working conditions. For example, under low temperature (-20°C), normal temperature (25°C) and high temperature (50°C), the battery is subjected to constant current charging and discharging, dynamic stress testing and other working conditions experiments, and the current, voltage, temperature and actual SOC value and other data are recorded. Under each temperature condition, the AEKF algorithm is set using the target sliding window length value, and the algorithm is run to estimate the battery SOC. The average absolute error between the SOC value estimated by the algorithm and the actual SOC value at different temperatures is calculated. For example, at -20°C, after multiple charge and discharge cycles, a set of SOC estimation error data is obtained, and its average absolute error value is calculated. Compare the average absolute error values ​​of the target sliding window length values ​​at different temperatures. If the error values ​​remain at a low level and the variation is small at different temperatures, it means that the target sliding window length value has good temperature migration capability. For example, if the mean absolute error values ​​are 1.3%, 1.1%, and 1.5% at -20°C, 25°C, and 50°C, respectively, the value is considered to be temperature transferable.

[0074] After determining that the target sliding window length value has temperature transferability, the sliding window length parameter of the AEKF algorithm is set to the target sliding window length value. At the same time, according to the type and characteristics of the battery, other algorithm parameters such as process noise covariance and measurement noise covariance are initialized. For example, for a certain ternary lithium-ion battery, the sliding window length is set to the previously determined target value of 120, and the process noise covariance is initialized to 0.01, the measurement noise covariance is initialized to 0.02, etc. During the actual operation of the battery, the current, voltage, and temperature data are collected in real time by sensors and input into the AEKF algorithm. These data are preprocessed, such as filtering and noise reduction, to improve the quality and reliability of the data. For example, a digital filter is used to smooth the collected voltage signal to remove high-frequency noise interference. The AEKF algorithm estimates the SOC of the battery in real time based on the real-time input data and the set parameters. During the estimation process, the algorithm continuously updates the internal state and model parameters of the battery to adapt to the dynamic changes of the battery under different working conditions and temperatures. The estimated SOC value at each time point is recorded and can be displayed or stored as needed for subsequent use by the battery management system. For example, during the driving of an electric vehicle, the estimated SOC value of the battery is displayed in real time to provide the driver with accurate power information.

[0075] It can be understood that by verifying its applicability under different temperature conditions, it can be ensured that the selected parameters have wide adaptability and stability in practical applications. Calculating the remaining battery capacity using a target sliding window length value with temperature migration capability can avoid estimation errors caused by temperature changes and improve the accuracy and reliability of the estimation. It solves the problem in the prior art that the sliding window length parameter needs to be frequently adjusted when the temperature changes, reducing the inconvenience and extra workload caused by temperature changes. At the same time, by using a verified target sliding window length value, the accuracy and stability of the battery remaining capacity estimation can be significantly improved, providing more reliable technical support for the management and application of lithium-ion batteries, thereby improving the performance and safety of the entire battery system.

[0076] In the above embodiment, by determining multiple operating conditions of the battery and multiple initial sliding window length values ​​corresponding to the adaptive extended Kalman filter algorithm, a variety of basic data are provided for the optimization of the sliding window length parameters; under each initial sliding window length value, the mean absolute error value of the remaining battery power of the adaptive extended Kalman filter algorithm in each operating condition is calculated, and the performance of different sliding window length values ​​is evaluated by quantizing the error, so that the sliding window length value with the best performance under different operating conditions can be accurately screened out; according to each mean absolute error value, the target sliding window length value is selected from each initial sliding window length value, and this process not only realizes the optimization of the sliding window length parameters, but also effectively solves the problem of reduced estimation accuracy caused by changes in operating conditions, ensuring that a higher SOC estimation accuracy can be obtained under different operating conditions; when it is determined that the target sliding window length value has temperature migration capability, it is used to calculate the remaining battery power, further solving the influence of temperature changes on the estimation accuracy, so that the method can maintain stable SOC estimation performance under different temperature environments. In summary, this method effectively solves the problem in the prior art of lack of effective optimization method for sliding window length parameters and the problem of decreased estimation accuracy caused by operating conditions and temperature changes, significantly improves the accuracy and stability of SOC estimation, and provides more reliable technical support for the management and application of lithium-ion batteries.

[0077] In one embodiment, the step of selecting a target sliding window length value from each initial sliding window length value according to each mean absolute error value includes:

[0078] Calculate the summary value of each mean absolute error value, and calculate the ratio of each mean absolute error value to the summary value;

[0079] For each initial sliding window length value, calculate the cumulative value of the corresponding ratios of each mean absolute error value under the initial sliding window length value;

[0080] Among the accumulated values, the initial sliding window length value corresponding to the accumulated value with the smallest value is used as the target sliding window length value.

[0081] Among them, the summary value refers to the sum of the average absolute error values ​​under each working condition, which reflects the overall error level of the AEKF algorithm in all working conditions under each initial sliding window length value. The ratio refers to the relative proportional relationship between each average absolute error value and the summary value, which is used to measure the proportion of the average absolute error value of each working condition in the overall error under a certain initial sliding window length value. By calculating the ratio, the relative size of the error under different working conditions and the distribution of the error of each working condition under different initial sliding window length values ​​can be more intuitively compared. The cumulative value refers to the sum of the corresponding ratios of the average absolute error values ​​under each working condition for each initial sliding window length value, which comprehensively reflects the error distribution of the initial sliding window length value under all working conditions.

[0082] Specifically, first, make sure that the mean absolute error values ​​of the AEKF algorithm under different initial sliding window length values ​​under various working conditions have been obtained. These data can be obtained through experiments or simulations and stored in a data table for subsequent calculations. For example, for initial sliding window length values ​​of 50, 100, 150, 200, and working conditions including constant current charging, constant current discharge, dynamic stress testing, etc., the mean absolute error values ​​under each combination have been recorded. Add all the mean absolute error values ​​to get a summary value. Divide the mean absolute error value under each working condition by the corresponding summary value to get the ratio of the mean absolute error value to the summary value under that condition.

[0083] The ratio data of each working condition under each initial sliding window length value are sorted together for the convenience of cumulative calculation. For example, for the initial sliding window length values ​​of 50, 100, 150, and 200, their ratios under each working condition are listed respectively. Add the ratios of all working conditions under each initial sliding window length value to obtain the cumulative value corresponding to the length value. For example, for the initial sliding window length value of 100, its ratios under three working conditions are 5.5%, 1.9%, and 2.6%, respectively, and the cumulative value is 10%. Compare the cumulative values ​​corresponding to all initial sliding window length values ​​to find the cumulative value with the smallest value. For example, for the cumulative values ​​calculated above, the cumulative value with the smallest value is 2%, corresponding to the initial sliding window length value of 200. The initial sliding window length value corresponding to the minimum cumulative value is determined as the target sliding window length value. For example, in this example, the target sliding window length value is 200. This value indicates that among all the initial sliding window length values, when the sliding window length is set to 200, the overall error distribution of the AEKF algorithm under various working conditions is the most ideal, that is, the cumulative value of the corresponding ratio of the mean absolute error value under various working conditions is the smallest, indicating that this sliding window length value can enable the algorithm to achieve the best comprehensive performance under different working conditions.

[0084] In this embodiment, the calculated summary value can fully reflect the overall error level of the algorithm under all initial sliding window length values ​​in all working conditions, providing a basis for the subsequent ratio calculation; the ratio calculation can quantify the proportion of each working condition in the overall error, which is helpful for a more detailed analysis of the performance of the algorithm under different working conditions; the calculation of the cumulative value integrates the ratio information of each working condition, providing a scientific basis for screening among multiple initial sliding window length values. By selecting the target sliding window length value corresponding to the minimum cumulative value, it can ensure that the AEKF algorithm can achieve the best comprehensive performance under different working conditions, effectively reduce the SOC estimation error, improve the reliability and intelligence level of the battery management system, and reduce the inaccurate estimation problem caused by unreasonable selection of sliding window length, thereby extending the battery life and improving the overall performance and user experience of battery-using equipment.

[0085] In one embodiment, the step of calculating the summary value of each mean absolute error value includes:

[0086] Cross-tabulate the initial sliding window length values ​​and the average absolute error values ​​under various working conditions to obtain a contingency table;

[0087] In a contingency table, summary values ​​are obtained by calculating the sum of the column sums or the sum of the row sums.

[0088] Among them, the contingency table is a table used to show the relationship between two or more categorical variables. In this embodiment, the rows of the contingency table can represent different initial sliding window length values, and the columns can represent various battery operating conditions. Each cell in the table contains the mean absolute error value under the corresponding row (initial sliding window length value) and column (operating condition) combination. For example, if the initial sliding window length value has four options of 50, 100, 150, and 200, and the battery operating conditions have three types of constant current charging, constant current discharge, and dynamic stress testing, then the contingency table will be a table with 4 rows and 3 columns, and each cell records the mean absolute error value under the corresponding sliding window length and operating condition.

[0089] The total value of the column sum refers to the sum of the data of each column in the contingency table, and then the total value obtained by summing up these column sums, which reflects the overall level of the mean absolute error value of a specific working condition under all initial sliding window length values. The total value of the row sum refers to the sum of the data of each row in the contingency table, and then the total value obtained by summing up these row sums, which reflects the overall level of the mean absolute error value of a specific initial sliding window length value under all working conditions.

[0090] Specifically, first, make sure that the mean absolute error values ​​for each initial sliding window length value under different operating conditions have been obtained. These data are usually obtained through experiments or simulations and stored in a database or data file. For example, for initial sliding window length values ​​of 50, 100, 150, 200, and operating conditions including constant current charging, constant current discharging, dynamic stress testing, etc., the mean absolute error value for each combination has been recorded. Arrange these data in the form of a two-dimensional table, where the rows represent the initial sliding window length values, the columns represent the operating conditions, and the cells are filled with the corresponding mean absolute error values. Use data processing software or programming languages ​​such as Python, MATLAB, etc. to create a contingency table.

[0091] For each column in the contingency table, add up all the mean absolute error values ​​under the column to get the sum of the column. Repeat this process to calculate the column sums of all working condition columns. Add up the column sums of all columns to get the total value of the column sums. Alternatively, for each row in the contingency table, add up all the mean absolute error values ​​under the row to get the sum of the row. Repeat this process to calculate the row sums of all rows with the initial sliding window length value. Add up the row sums of all rows to get the total value of the row sums.

[0092] In this embodiment, constructing a contingency table can systematically organize and display the average absolute error values ​​under different combinations of initial sliding window length values ​​and working conditions, making the data relationship clear and providing an intuitive basis for subsequent calculations and analysis. Calculating the sum of the column sum or row sum can quantitatively summarize the entire data set from different dimensions. The sum of the column sum reflects the overall error level of each working condition under all initial sliding window length values, while the sum of the row sum reflects the overall error level of each initial sliding window length value under all working conditions.

[0093] In one embodiment, for each initial sliding window length value, the step of calculating the cumulative value of the corresponding ratios of the mean absolute error values ​​under the initial sliding window length value includes:

[0094] Cross-tabulate the ratios of each initial sliding window length value and each working condition to obtain a probability distribution table;

[0095] In the probability distribution table, calculate the cumulative value of the corresponding ratio of each working condition under each initial sliding window length value.

[0096] Among them, the probability distribution table is a table used to show the distribution of ratios of different initial sliding window length values ​​under various operating conditions. In this embodiment, the rows of the table represent different initial sliding window length values, the columns represent various battery operating conditions, and each cell in the table contains the ratio under the corresponding row (initial sliding window length value) and column (operating condition) combination.

[0097] Specifically, first, ensure that the ratios of the initial sliding window length values ​​under different working conditions have been obtained. These ratios are calculated by dividing the average absolute error value under each working condition by the corresponding summary value. For example, for an initial sliding window length value of 100, the average absolute error value under the constant current charging condition is 1.2%, and the summary value is 4.7%, then the ratio is 25.5%. Arrange these ratio data into a two-dimensional table, where the rows represent the initial sliding window length values, the columns represent the working conditions, and the cells are filled with the corresponding ratios. Use data processing software or programming language to create a probability distribution table. Group the data in the probability distribution table according to the initial sliding window length value, and ensure that all the working condition ratio data under each initial sliding window length value are together for easy cumulative calculation. For example, for initial sliding window length values ​​of 50, 100, 150, and 200, list their ratios under each working condition respectively. For each initial sliding window length value, add the ratios of all working conditions under that length value to obtain the cumulative value corresponding to that length value.

[0098] In this embodiment, constructing a probability distribution table can systematically organize and display the distribution of ratios under different combinations of initial sliding window length values ​​and working conditions, making the data relationship clear and providing an intuitive basis for subsequent calculations and analysis. Calculating the cumulative value of the corresponding ratios of each working condition under each initial sliding window length value can comprehensively reflect the overall performance of the sliding window length value under all working conditions. By comparing the cumulative values ​​of different initial sliding window length values, the target sliding window length value that can achieve the best comprehensive performance under various working conditions can be screened out, thereby improving the accuracy and stability of the remaining battery power estimation, optimizing the performance of the battery management system, extending the battery life, and improving the overall performance and user experience of the device using the battery.

[0099] In one embodiment, the step of determining that the target sliding window length value has temperature migration capability in the adaptive extended Kalman filter algorithm includes:

[0100] According to the target sliding window length value, construct at least two sample matrices corresponding to different temperatures;

[0101] After calculating the correlation matrix according to each sample matrix, the eigenvalue of the correlation structure matrix is ​​calculated according to the correlation matrix;

[0102] Set the correlation hypothesis and irrelevance hypothesis based on the eigenvalues ​​and construct the test statistic;

[0103] The statistical value is calculated according to the test statistic and compared with a preset critical value. If the statistical value exceeds the critical value, it is determined that the target sliding window length value is transferable at different temperatures.

[0104] Among them, the sample matrix refers to the matrix form in which the operating data of the battery at different temperatures are organized according to certain rules. The correlation matrix is ​​a matrix used to describe the correlation relationship between different features in the sample matrix. Each element in the matrix represents the correlation coefficient between two features, and the value range is between [-1,1]. The closer the absolute value is to 1, the stronger the correlation is; the closer it is to 0, the weaker the correlation is. The eigenvalue of the correlation structure matrix refers to the value extracted from the correlation matrix that can reflect the structural characteristics of the matrix. The size of the eigenvalue represents the degree of dispersion or importance of the data in the direction of the corresponding eigenvector in the correlation matrix. The correlation hypothesis refers to the assumption that there is a correlation relationship between sample data at different temperatures, that is, temperature changes will not completely change the inherent correlation structure of battery operation data. The irrelevance hypothesis is the opposite, assuming that there is no correlation between sample data at different temperatures, that is, temperature changes cause the correlation structure of battery operation data to completely change. These two assumptions are the basis for statistical testing, and the test statistic is used to determine which assumption is more in line with the actual situation. The test statistic is a value calculated based on the sample data, which is used to measure the degree of difference between the observed data and the hypothesis. In the scenario of judging whether the target sliding window length value is temperature transferable, the test statistic is calculated based on the correlation matrix and eigenvalues, and is used to compare the differences in the correlation structure of sample data at different temperatures. If the value of the test statistic exceeds the pre-set critical value, it means that the difference between the observed data and the irrelevance hypothesis is large enough, so that the irrelevance hypothesis can be rejected and the correlation hypothesis can be accepted, that is, the target sliding window length value is considered to be transferable at different temperatures. The critical value is a pre-set threshold value used to compare with the value of the test statistic to determine whether to reject the null hypothesis, that is, the irrelevance hypothesis. The setting of the critical value is usually based on statistical distribution theory and the selected significance level, such as 0.05, 0.01, etc. Different significance levels correspond to different critical values, reflecting the tolerance for the risk of misjudgment. For example, when the significance level is 0.05, for a specific statistical test method, the corresponding critical value can be obtained by looking up a table or calculating.

[0105] Specifically, the battery is charged and discharged under different temperature environments, and its operating data under various working conditions are collected. These data are based on current, voltage, temperature, and the SOC value and actual SOC value estimated by the AEKF algorithm using the target sliding window length value. These data are stored according to the experimental sequence and temperature classification to prepare for the subsequent construction of the sample matrix. Features related to SOC estimation, such as current, voltage, temperature, and SOC estimation error, are selected from the collected data. For the experimental data at each temperature, samples are extracted at fixed time intervals or number of cycles, and each sample contains the above selected feature values. The sample data at each temperature are arranged in the selected feature order to form a matrix.

[0106] For each sample matrix corresponding to a temperature, the correlation coefficients between different features are calculated to construct a correlation matrix. For example, for the sample matrix at 25°C, the correlation coefficients between current and voltage, current and temperature, current and SOC estimation error, voltage and temperature, etc. are calculated and filled into the corresponding positions of the correlation matrix. Using linear algebra methods, the correlation matrix is ​​decomposed by eigenvalue to obtain the eigenvalues ​​of the correlation structure matrix. Eigenvalue decomposition can be implemented through mathematical libraries in computer devices or related functions in programming languages. For example, in Python, the linalg.eig function of the NumPy library can be used to input the correlation matrix and output the eigenvalues ​​and corresponding eigenvectors. The eigenvalues ​​are arranged in order from large to small, indicating the contribution of different feature combinations in the correlation matrix.

[0107] Based on the eigenvalues ​​of the correlation structure matrix, the correlation hypothesis and the irrelevance hypothesis are set. The correlation hypothesis assumes that the sample data at different temperatures have the same correlation structure, that is, the eigenvalues ​​have similar distributions at different temperatures; the irrelevance hypothesis assumes that the correlation structures of the sample data at different temperatures are different, that is, there are significant differences in the distribution of eigenvalues. According to the distribution characteristics of the eigenvalues, select the appropriate statistical test method to construct the test statistic. Common methods include Hotelling's T square test, Bartlett's sphericity test, etc.

[0108] Substitute the correlation matrix corresponding to the sample matrix at different temperatures into the calculation formula of the test statistic to obtain the specific statistical value. For example, for the two sample matrices at 25℃ and 40℃, calculate their correlation matrices respectively, and then substitute them into the formula of Bartlett's sphericity test. According to the selected statistical test method and the set significance level, consult the statistical distribution table or use the statistical function in the computer device to calculate the corresponding critical value. Compare the calculated statistical value with the critical value. If the statistical value exceeds the critical value, it means that at the current significance level, the irrelevance hypothesis is rejected and the correlation hypothesis is accepted, that is, the target sliding window length value is considered to be transferable at different temperatures; otherwise, it is considered to be not transferable.

[0109] In this embodiment, constructing sample matrices corresponding to different temperatures can systematically organize and present the operating data of the battery under various temperature conditions, providing a basis for subsequent correlation analysis. Calculating the eigenvalues ​​of the correlation matrix and the correlation structure matrix can quantitatively describe the correlation between different features in the data and their structural characteristics, revealing the inherent correlation pattern of the data. Setting the correlation hypothesis and the irrelevance hypothesis and constructing the test statistic is the theoretical basis for statistical testing. Through a scientific hypothesis framework and statistical methods, the similarities and differences of data correlation at different temperatures can be objectively judged. Finally, by comparing the test statistic value with the critical value, it is possible to clearly conclude whether the target sliding window length value has temperature transferability. The execution of this series of steps can effectively screen out the target sliding window length value that can maintain high estimation accuracy over a wide temperature range, ensure the reliability and consistency of the adaptive extended Kalman filter (AEKF) algorithm under different temperature conditions, improve the intelligence level of the battery management system, enhance the adaptability and competitiveness of equipment using batteries in complex temperature environments, and also help to extend the service life of the battery and improve its safety.

[0110] In one embodiment, the correlation matrix is ​​as follows:

[0111]

[0112] The correlation structure matrix is ​​shown below:

[0113]

[0114] in, and Represent the sample covariance matrix of each sample matrix, and represents the cross covariance matrix between each sample matrix, represents the correlation structure matrix, Represents the mean absolute error value corresponding to a temperature, Represents the mean absolute error value corresponding to another temperature.

[0115] In this embodiment, by calculating the correlation matrix and the correlation structure matrix, the correlation structure between the battery operation data at different temperatures can be quantified. Based on the eigenvalues ​​of the correlation structure matrix, hypothesis testing can be performed to determine whether the target sliding window length value is transferable at different temperatures. This process helps to improve the accuracy and stability of the battery remaining power estimation, ensure the reliability and consistency of the adaptive extended Kalman filter algorithm under different temperature conditions, thereby optimizing the performance of the battery management system, extending the battery life, and improving the adaptability and overall performance of the equipment using the battery in complex temperature environments.

[0116] In one embodiment, the correlation assumption and the irrelevance assumption are as follows:

[0117]

[0118] The test statistic is as follows:

[0119]

[0120] in, , represents the smaller value in each sample matrix dimension, represents a sample matrix dimension, represents another sample matrix dimension, represents the eigenvalue, Represents a statistic used to measure the correlation between two sets of variables. represents the irrelevance assumption, represents the correlation hypothesis, represents another statistic used for hypothesis testing, represents the number of variables in a sample matrix, represents the number of variables in another sample matrix, represents the eigenvalue number of the correlation structure matrix, Indicates the level of inspection.

[0121] In this embodiment, by setting the correlation hypothesis and the irrelevance hypothesis and calculating the test statistic, it is possible to scientifically determine whether the target sliding window length value is transferable at different temperatures. This process can quantify the correlation difference between sample data at different temperatures, provide a specific numerical basis for hypothesis testing, and objectively determine the applicability of the target sliding window length value at different temperatures. This helps to improve the accuracy and stability of the remaining battery power estimation, ensure the reliability and consistency of the adaptive extended Kalman filter algorithm under different temperature conditions, optimize the performance of the battery management system, extend the battery life, and improve the adaptability and overall performance of devices using batteries in complex temperature environments.

[0122] To facilitate understanding of the solution of the present application, specific examples are provided below for illustration.

[0123] Step 1: First, set the sliding window length value in the algorithm to different parameter values, and obtain the algorithm to estimate the MAE of SOC under different working conditions and temperatures under different sliding window length values.

[0124] Setting of sliding window length parameters: The sliding window length value range is obtained in the set S={1, 2, ..., n}, where n is the maximum sliding window length value.

[0125] Obtaining the algorithm MAE: traverse all elements in the set S, set the sliding window length parameters in the AEKF algorithm to the values ​​in the above set, then estimate the SOC of the battery under different working conditions, and calculate the MAE corresponding to the algorithm.

[0126] In this example, the value of n is set to 120, so the set S = {1, 2, ..., 120}, which contains a total of 120 different values. Four different working conditions are adopted, namely Bei Jing Dynamic Stress Test (BJDST), Dynamic Stress Test (DST), Federal Urban Driving Schedule (FUDS) and Supplemental Federal Test Procedure-US06 (US06), and the experimental temperature is 25°C. After setting the sliding window length in the AEKF algorithm to the values ​​in the set S, the SOC of the lithium-ion battery under different working conditions is estimated, and the MAE of the algorithm is calculated.

[0127] Step 2: Using MAE as the analysis data, use contingency table analysis to select the optimized sliding window length parameter value.

[0128] Fill in the obtained MAE value into the contingency table, as shown in the table:

[0129]

[0130] Among them, factor a represents the setting range of the sliding window length parameter, and state b represents different working conditions. The solution formula for different column sums in each column is as follows:

[0131]

[0132] The solution formula for each row and the sum of different rows is as follows:

[0133]

[0134] In the table, the sum e... is the sum of all row and column sum values. The calculation formula is as follows:

[0135]

[0136] In practical applications, each element in the contingency table is the frequency corresponding to different states and factors. Here, the MAE value corresponding to the algorithm is used as an element in the table for discussion when the sliding window length value in the algorithm is set to different numerical conditions to estimate the SOC under different working conditions. Now process the data in the table and divide each data in the table by the total value e. At this time, the sum of all elements in the table is 1, so each element in the contingency table can be regarded as a probability distribution table, as shown in the following table:

[0137]

[0138] According to the above probability distribution table, construct a set M, with columns and elements as elements in the set M, that is, M={P 1. ,P 2. , …, P n.}. Now find the smallest element from the set M, that is, the element with the smallest probability value. At this time, the element corresponds to an a i (i=1,…,n), this value is the sliding window length to be selected. The smaller the probability value, the smaller the MAE value of the algorithm, that is, the higher the estimation accuracy of the algorithm.

[0139] In this embodiment, the probability of finding the minimum value in the set M is 0.006523, which corresponds to a 101 , so the optimized sliding window length parameter value is 101. After setting this parameter in the AEKF algorithm to 101, the SOC is estimated. The results and error curves of the algorithm estimating the SOC under different working conditions are as follows Figure 2 shown.

[0140] Step 3: Use canonical correlation analysis combined with hypothesis testing to verify that the sliding window length parameter in the AEKF algorithm has temperature migration capabilities.

[0141] The main purpose is to solve the typical correlation coefficient of two sets of vectors. Then construct statistics and perform hypothesis tests to verify that the two sets of random vectors have significant correlation.

[0142] (1) Construct sample matrix:

[0143]

[0144] Among them, x ij (i=1,…,n; j=1,…,m) is the MAE of the algorithm at a certain temperature, y ij (i=1,…,n; j=1,…,m) is the MAE of the algorithm at another different temperature. The rows and columns in the matrices X and Y represent the sliding window length values ​​and different working conditions, respectively. Note that each element in the matrix is ​​obtained by standardizing the original data.

[0145] (2) Calculate the sample correlation matrix R and each sub-matrix in the matrix R:

[0146]

[0147]

[0148] Defining the Matrix , and then calculate the eigenvalues ​​of matrix A. After obtaining the eigenvalues ​​of matrix A, perform a hypothesis test to verify whether X and Y have a significant correlation.

[0149] (3) Propose a hypothesis:

[0150]

[0151] Construct the test statistic:

[0152]

[0153] in .like , then reject the original hypothesis, and it is believed that X and Y are correlated. This proves that the sliding window length in the AEKF algorithm has temperature migration capability, that is, under a certain temperature condition, after optimizing the sliding window length of the AEKF algorithm, when the temperature changes, it is no longer necessary to re-optimize the parameter value, thereby reducing the complexity of the algorithm application.

[0154] In this example, r=4, and the statistic T k and By calculation, we can get T k =393.7, obviously, , so X and Y have significant correlation. Now set the sliding window length parameter in the AEKF algorithm to the value optimized at 25℃, that is, set to 101. Then, in this case, use the algorithm to directly estimate the SOC under different working conditions at 45℃. The algorithm's estimation results and error curves are shown in the figure. Figure 3 shown.

[0155] In this example, at 25°C, four different operating conditions are used as experimental conditions, and the optimized sliding window length value is 101. At this time, the results and error curves of SOC estimation using the AEKF algorithm are as follows: Figure 2 shown.

[0156] In this embodiment, T k =393.7, significantly greater than , so it can be proved that the sliding window length in the AEKF algorithm has temperature migration capability. Then the sliding window length in the AEKF algorithm is set to the parameter value optimized at 25°C, that is, the sliding window length is set to 101, which is directly used to estimate the SOC of lithium-ion batteries under different working conditions at 45°C. The estimation results and error curves are shown in Figure 3 shown.

[0157] The following is a description of a battery remaining capacity estimation device provided in an embodiment of the present application. The battery remaining capacity estimation device described below and the battery remaining capacity estimation method described above can be referred to in correspondence with each other. Figure 4 As shown, the present application provides a battery remaining power estimation device, the device comprising:

[0158] An initial sliding window length value determination module 201 is used to determine multiple initial sliding window length values ​​corresponding to multiple operating conditions of the battery and the adaptive extended Kalman filter algorithm;

[0159] A mean absolute error value calculation module 202 is used to calculate the mean absolute error value of the remaining battery power of the adaptive extended Kalman filter algorithm in each working condition under each initial sliding window length value;

[0160] A target sliding window length value selection module 203 is used to select a target sliding window length value from each initial sliding window length value according to each mean absolute error value;

[0161] The temperature migration capability verification module 204 is used to calculate the remaining battery power using the target sliding window length value after determining that the target sliding window length value has the temperature migration capability in the adaptive extended Kalman filter algorithm.

[0162] In one embodiment, the target sliding window length value selection module 203 includes:

[0163] a ratio calculation unit, used to calculate a summary value of each mean absolute error value, and calculate a ratio of each mean absolute error value to the summary value;

[0164] A cumulative value calculation unit, used for calculating, for each initial sliding window length value, a cumulative value of the corresponding ratios of the mean absolute error values ​​under the initial sliding window length value;

[0165] The target sliding window length value selection unit is used to select, among the accumulated values, the initial sliding window length value corresponding to the accumulated value with the smallest value as .

[0166] In one embodiment, the ratio calculation unit includes:

[0167] A contingency table determination subunit is used to cross-tabulate each initial sliding window length value and the average absolute error value under each working condition to obtain a contingency table;

[0168] The summary value calculation unit is used to obtain the summary value by calculating the sum of each column sum or the sum of each row sum in the contingency table.

[0169] In one embodiment, the accumulated value calculation unit includes:

[0170] A probability distribution table determination subunit is used to cross-tabulate the ratios of each initial sliding window length value and each working condition to obtain a probability distribution table;

[0171] The cumulative value calculation subunit is used to calculate the cumulative value of the corresponding ratio of each working condition under each initial sliding window length value in the probability distribution table.

[0172] In one embodiment, the temperature transfer capability verification module 204 includes:

[0173] A sample matrix construction unit, used to construct at least two sample matrices corresponding to different temperatures according to a target sliding window length value;

[0174] An eigenvalue calculation unit is used to calculate the eigenvalue of the correlation structure matrix according to the correlation matrix after calculating the correlation matrix according to each sample matrix;

[0175] A test statistic construction unit, used to set a correlation hypothesis and an irrelevance hypothesis according to a characteristic value, and to construct a test statistic;

[0176] The temperature transferability verification unit is used to calculate the statistical value according to the test statistic and compare it with the preset critical value. If the statistical value exceeds the critical value, it is determined that the target sliding window length value is transferable at different temperatures.

[0177] In one embodiment, the correlation matrix is ​​as follows:

[0178]

[0179] The correlation structure matrix is ​​shown below:

[0180]

[0181] in, and Represent the sample covariance matrix of each sample matrix, and represents the cross covariance matrix between each sample matrix, represents the correlation structure matrix, Represents the mean absolute error value corresponding to a temperature, Represents the mean absolute error value corresponding to another temperature.

[0182] In one embodiment, the correlation assumption and the irrelevance assumption are as follows:

[0183]

[0184] The test statistic is as follows:

[0185]

[0186] in, , represents the smaller value in each sample matrix dimension, represents a sample matrix dimension, represents another sample matrix dimension, represents the eigenvalue, Represents a statistic used to measure the correlation between two sets of variables. represents the irrelevance assumption, represents the correlation hypothesis, represents another statistic used for hypothesis testing, represents the number of variables in a sample matrix, represents the number of variables in another sample matrix, represents the eigenvalue number of the correlation structure matrix, Indicates the level of inspection.

[0187] In one embodiment, the present application also provides a storage medium, which stores computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the battery remaining power estimation method as described in any of the above embodiments.

[0188] In one embodiment, the present application also provides a computer device having computer-readable instructions stored therein. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the battery remaining power estimation method as described in any of the above embodiments.

[0189] Indicatively, Figure 5 As shown, Figure 5 This is a schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. The computer device 300 may be provided as a server. Figure 5The computer device 300 includes a processing component 302, which further includes one or more processors, and a memory resource represented by a memory 301, for storing instructions executable by the processing component 302, such as an application. The application stored in the memory 301 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 302 is configured to execute instructions to perform the battery remaining power estimation method of any of the above embodiments.

[0190] The computer device 300 may further include a power supply component 303 configured to perform power management of the computer device 300, a wired or wireless network interface 304 configured to connect the computer device 300 to a network, and an input / output (I / O) interface 305. The computer device 300 may operate based on an operating system stored in the memory 301, such as Windows Server TM, Mac OS X TM, Unix TM, Linux TM, Free BSD TM, or the like.

[0191] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0192] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the term "include", "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not clearly listed, or also includes elements inherent to such process, method, article or equipment. In the absence of more restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or equipment including the elements. Herein, "one", "one", "said", "the" and "it" may also include plural forms, unless the context clearly indicates another way. A plurality refers to at least two cases, such as 2, 3, 5 or 8, etc. "And / or" includes any and all combinations of the relevant listed items.

[0193] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can refer to each other.

[0194] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for estimating remaining battery capacity, characterized in that: The method comprises: Determine multiple operating conditions of the battery and multiple initial sliding window length values ​​corresponding to the adaptive extended Kalman filter algorithm; Under each of the initial sliding window length values, calculating the mean absolute error value of the remaining battery power of the adaptive extended Kalman filter algorithm in each of the operating conditions; According to each of the mean absolute error values, selecting a target sliding window length value from each of the initial sliding window length values; When it is determined that in the adaptive extended Kalman filter algorithm, the target sliding window length value has temperature migration capability, the target sliding window length value is used to calculate the remaining battery power.

2. The method for estimating the remaining battery capacity according to claim 1, characterized in that: The step of selecting a target sliding window length value from each of the initial sliding window length values ​​according to each of the mean absolute error values ​​comprises: Calculating a summary value of each of the mean absolute error values, and calculating a ratio of each of the mean absolute error values ​​to the summary value; For each of the initial sliding window length values, calculate the cumulative value of the corresponding ratios of the mean absolute error values ​​under the initial sliding window length value; Among the accumulated values, the initial sliding window length value corresponding to the accumulated value with the smallest value is used as the target sliding window length value.

3. The method for estimating the remaining battery capacity according to claim 2, characterized in that: The step of calculating the summary value of each mean absolute error value comprises: Cross-tabulating the initial sliding window length values ​​and the average absolute error values ​​under the working conditions to obtain a contingency table; In the contingency table, the summary value is obtained by calculating the sum of the sums of each column or the sum of each row.

4. The method for estimating the remaining battery capacity according to claim 2, characterized in that: The step of calculating, for each of the initial sliding window length values, a cumulative value of the corresponding ratios of the mean absolute error values ​​under the initial sliding window length value comprises: Cross-tabulate the ratios of the initial sliding window length values ​​and the working conditions to obtain a probability distribution table; In the probability distribution table, the cumulative value of the corresponding ratios of each operating condition under each initial sliding window length value is calculated.

5. The method for estimating the remaining battery capacity according to claim 1, characterized in that: The step of determining that, in the adaptive extended Kalman filter algorithm, the target sliding window length value has temperature migration capability comprises: According to the target sliding window length value, construct at least two sample matrices corresponding to different temperatures; After calculating the correlation matrix according to each of the sample matrices, the eigenvalue of the correlation structure matrix is ​​calculated according to the correlation matrix; Setting a correlation hypothesis and an irrelevance hypothesis according to the eigenvalues, and constructing a test statistic; A statistical value is calculated according to the test statistic and compared with a preset critical value. If the statistical value exceeds the critical value, it is determined that the target sliding window length value is transferable at different temperatures.

6. The method for estimating the remaining battery capacity according to claim 5, characterized in that: The correlation matrix is ​​as follows: The correlation structure matrix is ​​shown below: in, and Respectively represent the sample covariance matrix of each of the sample matrices, and represents the cross covariance matrix between each of the sample matrices, represents the correlation structure matrix, Represents the mean absolute error value corresponding to a temperature, Represents the mean absolute error value corresponding to another temperature.

7. The method for estimating the remaining battery capacity according to claim 5, characterized in that: The correlation assumptions and non-correlation assumptions are as follows: The test statistic is as follows: in, , represents the smaller value in each dimension of the sample matrix, represents a sample matrix dimension, represents another sample matrix dimension, represents the eigenvalue, Represents a statistic used to measure the correlation between two sets of variables. represents the irrelevance assumption, represents the correlation hypothesis, represents another statistic used for hypothesis testing, represents the number of variables in a sample matrix, represents the number of variables in another sample matrix, represents the eigenvalue number of the correlation structure matrix, Indicates the level of inspection.

8. A battery remaining capacity estimation device, characterized in that: The device comprises: An initial sliding window length value determination module is used to determine multiple initial sliding window length values ​​corresponding to multiple operating conditions of the battery and the adaptive extended Kalman filter algorithm; A mean absolute error value calculation module is used to calculate the mean absolute error value of the remaining battery power of the adaptive extended Kalman filter algorithm in each of the working conditions under each of the initial sliding window length values; A target sliding window length value selection module, used for selecting a target sliding window length value from each of the initial sliding window length values ​​according to each of the mean absolute error values; The temperature migration capability verification module is used to calculate the remaining battery power using the target sliding window length value after determining that the target sliding window length value has the temperature migration capability in the adaptive extended Kalman filter algorithm.

9. A storage medium, characterized in that: The storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the battery remaining power estimation method as described in any one of claims 1 to 7.

10. A computer device, characterized in that: include: one or more processors, and memory; The memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the battery remaining power estimation method as described in any one of claims 1 to 7 are performed.

Images