A data-driven lithium battery health status assessment method

Abstract

The present invention provides a data-driven lithium battery health status assessment method, which selects data from a battery aging data set as features for feature preprocessing to obtain a typical feature extraction interval, and combines the feature data of the typical feature extraction interval with the actual battery SOH to obtain a data set for each typical feature extraction interval. The training battery is divided and trained using a neural network algorithm for its data set of different typical feature extraction intervals to obtain a corresponding battery SOH assessment model, and the final model is used for lithium battery health status assessment. The above method requires that the lithium battery meet two conditions: the voltage is less than 4.0V during charging and the voltage value changes by more than 0.2V. On this basis, features can be extracted from any voltage range of the lithium battery charging voltage, alleviating the situation where the lithium battery feature extraction conditions are harsh and single; compared with deep learning, this method does not require complex parameter calculations and is more suitable for current battery management systems.

Description

Technical Field

[0001] The present invention relates to the field of battery management, and in particular to a data-driven lithium battery health status assessment method. Background Art

[0002] Lithium-ion batteries, with their advantages of high energy density, high nominal voltage, low self-discharge, and long service life, have been widely used in numerous industrial fields. However, during use, battery performance can gradually and irreversibly degrade due to corrosion of the positive and negative electrode materials, electrolyte denaturation, and aging and short-circuiting of the positive and negative separators. When lithium-ion battery degradation reaches a certain level, it will no longer be able to maintain normal operation of the equipment and may even lead to serious consequences. Therefore, to prevent accidents and improve the operational stability of battery systems, accurate assessment of the lithium-ion battery's state of health (SOH) is particularly important.

[0003] The state of health reflects the battery's ability to store and supply energy / power relative to its current state at the beginning. The ratio of the battery's current capacity to the rated capacity is usually used to characterize the state of health. On the one hand, in the existing research on lithium battery SOH evaluation, the development and application of intelligent algorithms have greatly improved the estimation accuracy. However, these algorithms represented by deep learning are relatively complex and require more computing resources, which brings greater computing pressure to the battery management system (BMS) based on microcontrollers. On the other hand, most algorithms extract features from charging data in a fixed interval. However, in the actual application of batteries, the battery's charging start voltage and end voltage are not fixed, and the battery charging voltage interval does not necessarily meet the voltage interval required by the algorithm. Therefore, the estimation accuracy is difficult to guarantee. In view of the problem that the lithium battery feature extraction conditions are harsh and single, and the evaluation algorithm is relatively complex, which leads to limited online use of the battery SOH evaluation model, the present invention studies the battery SOH evaluation method in BMS. Summary of the Invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a data-driven lithium battery health status assessment method.

[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0006] A data-driven lithium battery health status assessment method includes the following steps:

[0007] S1. Obtain the actual battery SOH and the charging voltage curve under constant current charging conditions from the battery aging dataset, and select the duration of the constant charging voltage interval as the feature;

[0008] S2. Divide the complete charging voltage interval in the charging voltage curve into several sub-voltage intervals using sliding windows of different sizes, calculate the charging duration of each battery cycle in different sub-voltage intervals, obtain characteristic data of each sub-voltage interval, and obtain a typical feature extraction interval after preprocessing;

[0009] S3. Combining the normalized characteristic data of the typical feature extraction interval with the actual SOH of the battery to obtain a data set of each typical feature extraction interval, and arranging the data set of the typical feature extraction interval using a nonlinear autoregressive structure NARX to obtain a data set of a nonlinear autoregressive structure of each typical feature extraction interval;

[0010] S4. Repeat the above steps S2-S3 for different batteries and divide them into training batteries and test batteries. Train and optimize the data sets of different typical feature extraction intervals of the training batteries to obtain a trained model.

[0011] S5. Use the trained model obtained in S4 to evaluate the health status of the lithium battery.

[0012] Furthermore, the features in S1 are represented as follows:

[0013]

[0014] Among them, T F Characterized by Is charged to the lower limit of the voltage range V min The moment of time, Is charged to the upper limit of the voltage range V max The moment of time.

[0015] Furthermore, the S2 specifically includes the following steps:

[0016] S21. Divide the 3.7V-4.2V charging voltage range into several sub-voltage ranges using sliding windows of different sizes, calculate the charging duration of each battery cycle in different sub-voltage ranges, and obtain characteristic data for each sub-voltage range;

[0017] S22, dividing the characteristic data of each sub-voltage interval by its maximum value and normalizing it to the interval [0, 1];

[0018] S23. Calculate the Pearson coefficient between the characteristic data of different sub-voltage intervals and the actual SOH of the battery, and fix the sub-voltage interval corresponding to the maximum absolute value of the Pearson coefficient and other sub-voltage intervals belonging to the same sliding window as typical feature extraction intervals. In particular, the sub-voltage interval with a Pearson coefficient less than 0.8 is not used as a typical feature extraction interval.

[0019] Furthermore, step S3 includes the following steps:

[0020] S31, combining the normalized feature data extracted from the typical feature extraction intervals with the battery SOH to obtain a data set D for each typical feature extraction interval;

[0021] S32. Taking a data sample of a typical feature extraction interval as an example, the data set D of the interval is arranged in a nonlinear autoregressive structure NARX by the following formula, with the current feature data and historical data as feature vectors and the current battery SOH as the target vector, to obtain the data set D of the nonlinear autoregressive structure NARX of the interval: N .

[0022] Furthermore, the target vector of the current battery SOH in S32 is expressed as:

[0023] SOH * (t) = f(T F (t), T F (t-1),…,T F (tz), SOH′(t-1),…, SOH′(tz))

[0024] Among them, SOH * (t) is the target vector, which represents the output of the battery SOH evaluation model at the t-th cycle. f represents the nonlinear mapping relationship between the target vector and the feature vector. T F (t) represents the characteristic data collected by the BMS in the t-th cycle, z represents the delay step of the characteristic vector, and SOH′(tz) represents the health status of the battery in the tz-th cycle.

[0025] Furthermore, step S4 includes the following steps:

[0026] S41, applying the above feature data extraction method and step S3 to different batteries, and dividing them into training batteries and test batteries;

[0027] S42. Dataset D of different typical feature extraction intervals for training battery N , using the edRVFL algorithm respectively, the battery SOH evaluation model of different typical feature extraction intervals is obtained;

[0028] S43. Use the test battery to test the battery SOH evaluation model in different typical feature extraction intervals, and continuously optimize the number of hidden layers L, the number of hidden layer neurons N and the activation function g of the hidden layer neurons of each model so that the root mean square error meets the requirements on the test battery. After the optimization is completed, the final battery SOH evaluation model in different typical feature extraction intervals is obtained.

[0029] Furthermore, the calculation formula of the root mean square error in S43 can be expressed as:

[0030]

[0031] Among them, SOH test Represents the estimation result of the battery SOH evaluation model on the test battery, SOH real It represents the actual SOH of the tested battery, n is the total number of cycles of the battery aging test, and i is the i-th cycle of the battery aging test.

[0032] The present invention has the following beneficial effects:

[0033] The beneficial effects of the present invention are:

[0034] 1. A data-driven lithium battery health status assessment method is proposed. This method only requires that the lithium battery meets the V min Less than 4.0V and voltage value change greater than 0.2V. On this basis, features can be extracted from any voltage range of the lithium battery charging voltage, alleviating the situation where the lithium battery feature extraction conditions are harsh and single. Compared with deep learning, this method does not require complex parameter calculations and is more suitable for current battery management systems.

[0035] 2. The use of nonlinear autoregressive structure NARX increases the feature dimension of the edRVFL algorithm input, so that edRVFL not only learns the nonlinear mapping relationship between the current cycle target vector and the feature vector, but also learns the nonlinear mapping relationship between the current cycle target vector and the historical cycle feature vector and the historical battery SOH vector. Therefore, by introducing historical information, the battery SOH evaluation model improves the estimation accuracy of the current cycle battery health status. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 The figure is a flow chart of the data-driven lithium battery health status assessment method of the present invention.

[0037] Figure 2 This is the feature extraction method used in the embodiment of the present invention.

[0038] Figure 3 This is the training and testing process of the embodiment of the present invention.

[0039] Figure 4 Schematic diagram of the neural network structure of the edRVFL algorithm in the embodiment.

[0040] Figure 5These are the evaluation results of the corresponding models of batteries in different voltage ranges according to the embodiments of the present invention, wherein (a) is the estimation result of the model corresponding to the voltage range of 3.70V-3.85V for CS2-36 batteries, (b) is the estimation result of the model corresponding to the voltage range of 3.75V-3.90V for CS2-36 batteries, (c) is the estimation result of the model corresponding to the voltage range of 3.80V-3.95V for CS2-36 batteries, (d) is the estimation result of the model corresponding to the voltage range of 3.85V-4.00V for CS2-36 batteries, (e) is the estimation result of the model corresponding to the voltage range of 3.90V-4.05V for CS2-36 batteries, (f) is the estimation result of the model corresponding to the voltage range of 3.95V-4.10V for CS2-36 batteries, and (g) is the estimation result of the model corresponding to the voltage range of 4.00V-4.15V for CS2-36 batteries. DETAILED DESCRIPTION

[0041] The specific embodiments of the present invention are described below to facilitate understanding of the present invention by those skilled in the art. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

[0042] A data-driven lithium battery health status assessment method, such as Figure 1 As shown, the following steps are included:

[0043] S1. Obtain the actual SOH of the battery and the charging voltage curve under constant current charging conditions from the battery aging data set, such as Figure 2 As shown, the duration of the equal charging voltage interval is selected as the feature, and the feature is defined as T F , the expression is:

[0044]

[0045] Where, Is charged to the lower limit of the voltage range V min The moment of time, Is charged to the upper limit of the voltage range V max the moment of time;

[0046] S2. Divide the complete charging voltage interval in the charging voltage curve into several sub-voltage intervals using sliding windows of different sizes, calculate the charging duration of each battery cycle in different sub-voltage intervals, obtain characteristic data of each sub-voltage interval, normalize these characteristic data respectively, and use the Pearson correlation analysis method to calculate characteristic data with a strong correlation with the actual SOH of the battery. The sub-voltage interval where the characteristic data is located and other sub-voltage intervals belonging to the same sliding window are fixed as typical feature extraction intervals. Specifically, the following steps are included:

[0047] S21. Divide the charging voltage range of 3.7V-4.2V into several sub-voltage ranges using sliding windows of different sizes (1. The size of the sliding window is a multiple of 0.05, i.e., 0.05V, 0.1V, 0.15V...0.45V, 0.5V; 2. The V of adjacent sub-voltage ranges under the same sliding window is min The difference is 0.05V), and the charging duration of each cycle of the battery in different sub-voltage intervals is calculated to obtain characteristic data of each sub-voltage interval;

[0048] S22, dividing the characteristic data of each sub-voltage interval by its maximum value and normalizing it to the interval [0, 1];

[0049] S23. Calculate the Pearson coefficient between the characteristics of different sub-voltage intervals and the actual SOH of the battery, and fix the sub-voltage interval corresponding to the maximum absolute value of the Pearson coefficient and other sub-voltage intervals belonging to the same sliding window as typical feature extraction intervals. In particular, the sub-voltage interval with a Pearson coefficient less than 0.8 is not used as a typical feature extraction interval.

[0050] S3. Combine the normalized characteristic data of the typical feature extraction interval with the actual SOH of the battery to obtain the data set D of each typical feature extraction interval. Arrange the data set D with the nonlinear autoregressive structure NARX to obtain the data set D of the nonlinear autoregressive structure NARX of each typical feature extraction interval. N ;

[0051] Specifically, this embodiment includes the following steps:

[0052] S31, combining the normalized feature data extracted from the typical feature extraction intervals with the battery SOH to obtain a data set D for each typical feature extraction interval;

[0053] S32. Taking a data sample of a typical feature extraction interval as an example, the data set D of the interval is arranged in a nonlinear autoregressive structure NARX by the following formula, with the current feature data and historical data as feature vectors and the current battery SOH as the target vector, to obtain the data set D of the nonlinear autoregressive structure NARX of the interval: N ;

[0054] SOH * (t) = f(T F (t), T F (t-1),…,T F (tz), SOH′(t-1),…, SOH′(tz))

[0055] Where SOH * (t) is the target vector, which represents the output of the battery SOH evaluation model at the t-th cycle. f represents the nonlinear mapping relationship between the target vector and the feature vector. T F (t) represents the characteristic data collected by the BMS in the t-th cycle, z represents the delay step of the characteristic vector, and SOH′(tz) represents the health status of the battery in the tz-th cycle.

[0056] S33: The data samples of other typical feature extraction intervals are also obtained by using step S302 to obtain the corresponding data set D N .

[0057] S4, apply the above feature data extraction method and step S3 to different batteries, and divide them into training batteries and test batteries, and extract the data set D of different typical feature extraction intervals of the training batteries. N Using the ensemble deep Random Vector Functional Link (edRVFL) neural network algorithm, we trained battery SOH assessment models corresponding to different typical feature extraction intervals and tested them on test batteries. We optimized the number of hidden layers, the number of hidden layer neurons, and the activation function of each model, including the following steps:

[0058] S41, applying the above feature data extraction method and step S3 to different batteries, and dividing them into training batteries and test batteries;

[0059] S42. Dataset D of different typical feature extraction intervals for training battery N , respectively using the edRVFL algorithm to obtain the battery SOH evaluation model for different typical feature extraction intervals. The neural network structure of the edRVFL algorithm is as follows Figure 4 As shown, the principle of the algorithm is:

[0060] Define the output H of the first hidden layer1 , which can be expressed as

[0061]

[0062] Where, Represents the input data of the model input layer, where n represents the number of input data samples, d represents the feature dimension, and W 1 Represents the weight vector between the input layer and the first hidden layer, B 1 Represents the bias vector of the first hidden layer. The output of the deeper lth hidden layer can be expressed as:

[0063]

[0064] Where, W l Represents the weight vector between the lth hidden layer and the input layer and the l-1th hidden layer, B l Represents the bias vector of the lth hidden layer, where W 1 、B 1 、W l 、B l They are randomly generated but remain unchanged after the model training is completed.

[0065] Define dataset D N The target vector in is T, Define the output weight vector between the lth hidden layer and the output layer as β l , Then β l The calculation formula is:

[0066]

[0067] Where, Indicates [H l ,X], the output of the lth layer can be expressed as:

[0068]

[0069] The final output of the battery SOH evaluation model can be expressed as

[0070]

[0071] Save the number of hidden layers L, the number of hidden layer neurons N, the activation function type g, the weights W between the hidden layer and the input layer, and the hidden layer 1 、W l , hidden layer neuron bias B 1 、B l , the output weight β between the hidden layer and the output layerl The battery SOH evaluation model is obtained.

[0072] S43. Use the test battery to test the battery SOH evaluation model in different typical feature extraction intervals, and continuously optimize the number of hidden layers L, the number of hidden layer neurons N, and the activation function g of the hidden layer neurons of each model to make the root mean square error (RMSE) meet the requirements on the test battery. After the optimization is completed, the final battery SOH evaluation model for different typical feature extraction intervals is obtained. The training and testing flow chart is as follows: Figure 3 As shown, the calculation formula of RMSE can be expressed as:

[0073]

[0074] Where SOH test Represents the estimation result of the battery SOH evaluation model on the test battery, SOH real It represents the actual SOH of the tested battery, n is the total number of cycles of the battery aging test, and i is the i-th cycle of the battery aging test.

[0075] Due to the existence of the NARX structure, it is necessary to know the characteristic data of the battery from the 1st to the zth cycle and the actual SOH. Therefore, in the test phase, the data set D of different typical feature extraction intervals of the test battery is N The actual SOH of the battery after the zth cycle is replaced by the estimated value of the battery SOH evaluation model, and the rest remain unchanged. The output of the test battery SOH evaluation model can be expressed as:

[0076]

[0077] The edRVFL algorithm combines RVFL with deep learning and ensemble learning in a special way, achieving accurate assessment of battery SOH without significantly increasing model parameters. In addition, since most parameters in the model are randomly generated, only the output weight vector between the hidden layer and the output layer needs to be calculated, which greatly reduces the computational complexity of the model.

[0078] S5. The final model is used to evaluate the health status of lithium batteries.

[0079] This example uses four battery aging data sets from the University of Maryland as an example for analysis. These four batteries are numbered CS2-35, CS2-36, CS2-37, and CS2-38. Cycling experiments were performed at room temperature using the same standard charge and discharge protocol. The standard charge protocol includes two charging steps: constant current (CC) charging and constant voltage (CV) charging. The CC step of each charging cycle is charged at a rate of 0.5C (0.55A) until the voltage reaches the charge cutoff voltage (4.2V). After that, the battery voltage is maintained at 4.2V, and CV charging is performed until the charging current reaches the cutoff current (0.05A). The discharge protocol is to discharge the battery at a CC state of 1C until the battery voltage reaches 2.7V.

[0080] like Figure 1 As shown in FIG, a data-driven lithium battery health status assessment method provided by the present invention, the specific steps are as follows:

[0081] Step 1: Process and analyze the aging data of four lithium batteries to obtain the actual SOH of the battery and the charging voltage curve under constant current charging conditions, and select T F As a feature;

[0082] Step 2: Divide the 3.7V-4.2V charging voltage range into several sub-voltage ranges using sliding windows of different sizes (1. The size of the sliding window is a multiple of 0.05, i.e. 0.05V, 0.1V, 0.15V...0.45V, 0.5V; 2. The V values ​​of adjacent sub-voltage ranges under the same sliding window are min The difference is 0.05V), and the charging duration of each cycle of the battery in different sub-voltage intervals is calculated to obtain characteristic data of each sub-voltage interval;

[0083] The characteristic data of each sub-voltage interval is divided by its maximum value and normalized to the interval [0,1];

[0084] Calculate the Pearson coefficient between the characteristic data of the four battery sub-voltage intervals and the actual SOH of the battery. The sub-voltage interval corresponding to the maximum absolute value of the Pearson coefficient and other sub-voltage intervals belonging to the same sliding window are fixed as typical feature extraction intervals. In particular, the sub-voltage interval with a Pearson coefficient less than 0.8 is not used as a typical feature extraction interval.

[0085] A comprehensive analysis of the Pearson correlation between the characteristic data of the four batteries and the actual SOH of the batteries was conducted, and the seven voltage intervals of 3.70V-3.85V, 3.75V-3.90V, 3.80V-3.95V, 3.85V-4.00V, 3.90V-4.05V, 3.95V-4.10V, and 4.00V-4.15V were selected as typical feature extraction intervals.

[0086] Step 3: Combine the normalized feature data extracted from the seven typical feature extraction intervals of the four batteries with the actual SOH of the batteries to obtain a data set D for each typical feature extraction interval of each battery;

[0087] The data set D of each typical feature extraction interval of each battery is arranged in a nonlinear autoregressive structure NARX to obtain the data set D of the nonlinear autoregressive structure NARX N ,According to the subsequent model test results, the z of the nonlinear autoregressive structure is set to 30;

[0088] Step 4: Select CS2-35 from the four batteries as the test battery and the remaining three batteries as the training batteries; use the edRVFL algorithm in Matlab to extract the data set D of the seven typical features of the training battery. N The battery SOH evaluation models of different intervals are trained separately; in the testing phase, the data sets D of different typical feature extraction intervals of the test battery are used to extract the battery SOH evaluation models of different intervals. N The actual SOH of the battery after the 30th cycle is replaced by the estimated value of the battery SOH evaluation model; the number of hidden layers L, the number of hidden layer neurons N, and the activation function g of the hidden layer neurons of each model are continuously optimized so that the RMSE between the estimated value of the battery SOH evaluation model and the actual SOH of the battery meets the requirements;

[0089] Step 5: Use CS2-36 as the test battery and the remaining batteries as the training batteries, and repeat the model training and testing parts in step 4;

[0090] Step 6: Use CS2-37 as the test battery and the remaining batteries as the training batteries, and repeat the model training and testing parts in step 4;

[0091] Step 7: Use CS2-38 as the test battery and the remaining batteries as the training batteries, and repeat the model training and testing parts in step 4;

[0092] The parameter settings of the battery SOH evaluation model corresponding to the 7 typical feature extraction intervals are finally determined as shown in Table 1:

[0093] Table 1 Parameter settings of different models

[0094]

[0095] Taking the CS2-36 battery as an example, the estimated results of the battery SOH evaluation model corresponding to each typical feature extraction interval of the battery are as follows: Figure 5 (a)-(g)

[0096] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0097] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0098] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0099] Specific embodiments are used in the present invention to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core ideas. At the same time, for those skilled in the art, according to the ideas of the present invention, there may be changes in the specific implementation methods and application scopes. In summary, the contents of this specification should not be understood as limiting the present invention.

[0100] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and it should be understood that the scope of protection of the present invention is not limited to such specific descriptions and embodiments. Those skilled in the art can make various other specific variations and combinations based on the technical teachings disclosed in the present invention without departing from the essence of the present invention, and such variations and combinations are still within the scope of protection of the present invention.

Claims

1. A data-driven lithium battery health status assessment method, characterized in that: The steps include: S1. Obtain the actual battery SOH and the charging voltage curve under constant current charging conditions from the battery aging dataset, and select the duration of the constant charging voltage interval as the feature; S2. Divide the complete charging voltage interval in the charging voltage curve into several sub-voltage intervals using sliding windows of different sizes, calculate the charging duration of each battery cycle in different sub-voltage intervals, obtain characteristic data of each sub-voltage interval, and obtain a typical feature extraction interval after preprocessing; S3. Combining the normalized characteristic data of the typical feature extraction interval with the actual SOH of the battery to obtain a data set of each typical feature extraction interval, and arranging the data set of the typical feature extraction interval using a nonlinear autoregressive structure NARX to obtain a data set of a nonlinear autoregressive structure of each typical feature extraction interval, specifically comprising the following steps: S31. Combine the normalized feature data extracted from the typical feature extraction interval with the battery SOH to obtain a data set for each typical feature extraction interval. ; S32. Taking a data sample of a typical feature extraction interval as an example, the data set of this interval is Arrange the data in the nonlinear autoregressive structure NARX, make the current feature data and historical data as feature vectors, and the current battery SOH as the target vector, and get the data set of the nonlinear autoregressive structure NARX in this interval. ; S4. Repeat the above steps S2-S3 for different batteries and divide them into training batteries and test batteries. Train and optimize the data sets of different typical feature extraction intervals of the training batteries to obtain a trained model. Specifically, the steps include the following: S41, applying the above feature data extraction method and step S3 to different batteries, and dividing them into training batteries and test batteries; S42. Datasets for different typical feature extraction intervals of training batteries , using the edRVFL algorithm respectively, the battery SOH evaluation model of different typical feature extraction intervals is obtained; S43. Use test batteries to test battery SOH evaluation models in different typical feature extraction intervals, and continuously optimize the number of hidden layers of each model. , the number of hidden layer neurons and the activation function of the hidden layer neurons , so that the root mean square error meets the requirements on the test battery, and the final battery SOH evaluation model for different typical feature extraction intervals is obtained after the optimization; S5. Use the trained model obtained in S4 to evaluate the health status of the lithium battery.

2. The data-driven lithium battery health status assessment method according to claim 1, characterized in that: The features in S1 are expressed as: in, Characterized by Is charged to the lower limit of the voltage range The moment of time, Is charged to the upper limit of the voltage range The moment of time.

3. The data-driven lithium battery health status assessment method according to claim 1, characterized in that: The S2 specifically includes the following steps: S21. Divide the charging voltage range of 3.7V-4.2V into multiple sub-voltage ranges using sliding windows of different sizes, calculate the charging duration of each battery cycle in different sub-voltage ranges, and obtain characteristic data for each sub-voltage range; S22, dividing the characteristic data of each sub-voltage interval by its maximum value and normalizing it to the interval [0, 1]; S23. Calculate the Pearson coefficient between the characteristic data of different sub-voltage intervals and the actual SOH of the battery, and fix the sub-voltage interval corresponding to the maximum absolute value of the Pearson coefficient and other sub-voltage intervals belonging to the same sliding window as typical feature extraction intervals.

4. The data-driven lithium battery health status assessment method according to claim 1, characterized in that: The target vector of the current battery SOH in S32 is expressed as: in, is the target vector, indicating the The output of the secondary cycle battery SOH assessment model, Represents the nonlinear mapping relationship between the target vector and the feature vector, Indicates the The characteristic data collected by the BMS in the next cycle, represents the delay step of the eigenvector, Indicates the The health status of the battery after the cycle.

5. The data-driven lithium battery health status assessment method according to claim 1, characterized in that: The calculation formula of the root mean square error in S43 can be expressed as: in, Represents the estimation result of the battery SOH evaluation model on the test battery, Indicates the actual SOH of the tested battery, is the total number of cycles of the battery aging experiment, is the i-th cycle of the battery aging experiment.

Images