Lithium battery health state prediction method, apparatus and device, and storage medium

Abstract

The embodiment of the invention provides a lithium battery health state prediction method and device, equipment and a storage medium. The method comprises the following steps: acquiring monitoring data and corresponding SOH (state of health) data of a lithium battery in a cyclic charging and discharging process; performing dimensionality reduction on the monitoring data of the lithium battery in the cyclic charging and discharging process by using a PCA algorithm to obtain dimensionality-reduced monitoring data; constructing a data set by taking the dimension-reduced monitoring data as sample data and taking the SOH data corresponding to the dimension-reduced monitoring data as label data; training the GRU model by using the data set, and performing hyper-parameter optimization by using an SSA algorithm during training to obtain a lithium battery health state prediction model; and predicting SOH data of the specified lithium battery by using the lithium battery health state prediction model. In this way, the accuracy of lithium battery health state prediction can be improved, and the calculation burden can be reduced.

Description

Technical Field

[0001] The present invention relates to the field of lithium battery technology, and in particular to a lithium battery health status prediction method, device, equipment and storage medium. Background Art

[0002] In today's rapidly developing new energy field, lithium batteries, as a highly efficient and environmentally friendly energy storage solution, have been widely used in many occasions such as electric vehicles, mobile devices, and large-scale energy storage systems. With the increasing application of lithium batteries, safety accidents caused by performance degradation due to their charge and discharge cycles occasionally occur. The state of health (SOH) of a lithium battery is defined as the ratio of the actual capacity of the current cycle to the initial rated capacity, which can reflect the current battery's ability to store electrical energy relative to an unused battery. Generally, when the actual capacity of the battery drops to 70% to 80% of the initial capacity, it can be determined that it can no longer be used, that is, the end of life (End of Life, EOL). Therefore, accurately predicting the SOH of a lithium battery is of great significance for extending battery life, improving system reliability, and reducing maintenance costs.

[0003] Traditional lithium battery health status prediction methods often rely on empirical models and mechanism models, requiring researchers to have a strong theoretical knowledge background. At the same time, these methods have problems such as complex modeling, low accuracy and poor generalization. With the rise of technologies such as deep learning, its powerful feature extraction capabilities and ability to model complex nonlinear relationships provide a new solution for lithium battery SOH prediction. Therefore, data-driven lithium battery health status prediction has become a hot research topic.

[0004] However, the aging of lithium battery performance is a complex physical and chemical process affected by multiple factors. These factors are coupled with each other and are nonlinear and uncertain, which makes the prediction of lithium battery health status less accurate and has a heavy computational burden. Summary of the invention

[0005] In a first aspect, an embodiment of the present invention provides a method for predicting the health status of a lithium battery, the method comprising:

[0006] Obtain monitoring data of the lithium battery cycle charge and discharge process and its corresponding SOH data;

[0007] The principal component analysis (PCA) algorithm is used to reduce the dimension of the monitoring data of the lithium battery cycle charge and discharge process to obtain the reduced dimension monitoring data;

[0008] The reduced-dimensional monitoring data is used as sample data, and the SOH data corresponding to the reduced-dimensional monitoring data is used as label data to construct a data set;

[0009] The Gated Recurrent Unit (GRU) model was trained using the dataset, and the Sparrow Search Algorithm (SSA) algorithm was used to optimize hyperparameters during training to obtain a lithium battery health status prediction model.

[0010] Use the lithium battery health status prediction model to predict the SOH data of a specified lithium battery.

[0011] In some implementations of the first aspect, the PCA algorithm is used to reduce the dimension of the monitoring data of the lithium battery cyclic charge and discharge process to obtain the reduced-dimensional monitoring data, including:

[0012] Preprocessing the monitoring data of the lithium battery cycle charge and discharge process, where the preprocessing includes: missing value filling, outlier removal, and standardization;

[0013] The preprocessed monitoring data is reduced in dimension using the PCA algorithm to obtain reduced-dimensional monitoring data. In some implementations of the first aspect, the preprocessed monitoring data is reduced in dimension using the PCA algorithm to obtain reduced-dimensional monitoring data, including:

[0014] Calculate the covariance matrix of the preprocessed monitoring data, and calculate multiple eigenvectors and their corresponding eigenvalues ​​according to the covariance matrix;

[0015] Sort each eigenvector according to the eigenvalue from large to small, calculate the cumulative contribution rate according to the eigenvalue of the eigenvector in the sorted order, until the cumulative contribution rate is greater than or equal to the preset threshold, and select multiple eigenvectors used in the current cumulative contribution rate calculation to generate a principal component matrix;

[0016] The preprocessed monitoring data is projected into the space corresponding to the principal component matrix to obtain the reduced-dimensional monitoring data.

[0017] In some implementations of the first aspect, a GRU model is trained using a data set, and an SSA algorithm is used during training to perform hyperparameter optimization to obtain a lithium battery health status prediction model, including:

[0018] Divide the data set into training set and test set according to the preset ratio;

[0019] The GRU model is trained using the training set, and the SSA algorithm is used to optimize hyperparameters during training. The performance of the trained GRU model is evaluated using the test set. If the trained GRU model passes the performance evaluation, it is used as a lithium battery health status prediction model.

[0020] In some implementations of the first aspect, the GRU model includes a sequence input layer, a GRU layer with multiple hidden units, a fully connected layer and an output layer; the GRU layer is used to capture the temporal correlation between features, and uses a reset gate and an update gate to work together. The reset gate can selectively ignore the state of the previous moment, and the update gate generates a value from 0 to 1 to update the unit state, determining how much historical information is retained.

[0021] In some implementations of the first aspect, using the SSA algorithm to perform hyperparameter optimization during training includes:

[0022] Step 1: Determine the range of hyperparameter selection;

[0023] Step 2: Generate multiple initial hyperparameter combinations, that is, initialize the positions of multiple sparrows, and then initialize the sparrow population;

[0024] Step 3: Calculate the fitness value of each hyperparameter combination;

[0025] Step 4: Update the hyperparameter combination according to the fitness value, that is, update the position of the sparrow, and then update the sparrow population;

[0026] Step 5: Repeat steps 3-4 until the stopping condition is reached and output the optimal hyperparameter combination.

[0027] In some implementations of the first aspect, using a lithium battery health status prediction model to predict SOH data of a lithium battery includes:

[0028] The lithium battery health status prediction model is used to process the dimensionality reduction monitoring data of the specified lithium battery to predict the SOH data of the specified lithium battery.

[0029] In a second aspect, an embodiment of the present invention provides a lithium battery health status prediction device, the device comprising:

[0030] An acquisition module is used to acquire monitoring data of the lithium battery cycle charge and discharge process and its corresponding SOH data;

[0031] A dimension reduction module is used to reduce the dimension of the monitoring data of the lithium battery cycle charge and discharge process by using the PCA algorithm to obtain reduced-dimensional monitoring data;

[0032] A construction module is used to construct a data set using the reduced-dimensionality monitoring data as sample data and the SOH data corresponding to the reduced-dimensionality monitoring data as label data;

[0033] The training module is used to train the GRU model using the data set and use the SSA algorithm to optimize the hyperparameters during training to obtain a lithium battery health status prediction model;

[0034] The prediction module is used to predict the SOH data of a specified lithium battery using a lithium battery health status prediction model.

[0035] In a third aspect, an embodiment of the present invention provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method described above.

[0036] In a fourth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium storing computer instructions, where the computer instructions are used to enable a computer to execute the method described above.

[0037] In an embodiment of the present invention, the PCA algorithm can be used to first reduce the dimension of the original charge and discharge cycle monitoring data of the lithium battery, retain the key information related to the capacity degradation process of the lithium battery, and remove irrelevant noise. Then, based on the reduced-dimensional monitoring data and its corresponding SOH data, a GRU model combined with the SSA algorithm is trained to obtain a lithium battery health status prediction model, and then the model is used to predict the health status of the lithium battery to improve the accuracy of the prediction and reduce the computational burden.

[0038] It should be understood that the contents described in the summary of the invention are not intended to limit the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The above and other features, advantages and aspects of the embodiments of the present invention will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. The accompanying drawings are used to better understand the present invention and do not constitute a limitation of the present invention. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0040] Figure 1 A flowchart of a lithium battery health status prediction method provided by an embodiment of the present invention;

[0041] Figure 2 This is a schematic diagram comparing the prediction effects of lithium battery health status using different models;

[0042] Figure 3 A schematic diagram of a lithium battery health state prediction effect of a lithium battery health state prediction method;

[0043] Figure 4 A structural diagram of a lithium battery health status prediction device provided by an embodiment of the present invention;

[0044] Figure 5 The figure is a structural diagram of an exemplary electronic device capable of implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] In addition, the term "and / or" in the present invention is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present invention generally indicates that the associated objects before and after are in an "or" relationship.

[0047] In order to solve the technical problems arising from the background technology, the embodiments of the present invention provide a method, device, equipment and storage medium for predicting the health status of a lithium battery. Specifically, the monitoring data of the lithium battery's cyclic charge and discharge process and its corresponding SOH data are obtained; the PCA algorithm is used to reduce the dimension of the monitoring data of the lithium battery's cyclic charge and discharge process to obtain the reduced-dimensional monitoring data; the reduced-dimensional monitoring data is used as sample data, and the SOH data corresponding to the reduced-dimensional monitoring data is used as label data to construct a data set; the data set is used to train the GRU model, and the SSA algorithm is used to optimize the hyperparameters during the training to obtain a lithium battery health status prediction model; the lithium battery health status prediction model is used to predict the SOH data of a specified lithium battery.

[0048] In this way, the PCA algorithm can be used to reduce the dimension of the original charge and discharge cycle monitoring data of the lithium battery, retain the key information related to the capacity degradation process of the lithium battery, and remove irrelevant noise. Then, the GRU model combined with the SSA algorithm is trained based on the reduced-dimensional monitoring data and its corresponding SOH data to obtain a lithium battery health status prediction model, and then the model is used to predict the health status of the lithium battery to improve the accuracy of the prediction and reduce the computational burden.

[0049] In conjunction with the accompanying drawings, a lithium battery health status prediction method, device, equipment and storage medium provided by the embodiments of the present invention are described in detail through specific embodiments.

[0050] Figure 1 A flowchart of a lithium battery health status prediction method provided by an embodiment of the present invention is as follows Figure 1 As shown, the lithium battery health status prediction method 100 may include:

[0051] S110, obtaining monitoring data of the lithium battery's cyclic charge and discharge process and its corresponding SOH data.

[0052] That is, the monitoring data of the lithium battery and its corresponding SOH data are obtained during each charge and discharge cycle.

[0053] The acquired monitoring data may include data of variables such as battery internal resistance, maximum temperature and charging time, which are represented in the following matrix form:

[0054]

[0055] In the formula, X represents the acquired monitoring data, m represents the number of variables in the monitoring data; n represents the total number of lithium battery charge and discharge cycles; element x ij It represents the monitoring value of the jth variable during the i-th charge and discharge cycle of the lithium battery, where i = 1, 2, ..., n, j = 1, 2, ..., m.

[0056] The acquired SOH data can be calculated based on the discharge capacity data. The specific calculation formula is as follows:

[0057]

[0058] In the formula, SOH(i) represents the SOH data of the lithium battery during the i-th cycle charge and discharge; C i It represents the actual capacity of the lithium battery during the i-th cycle charge and discharge, and is equal to Q Discharge (i) refers to the capacity of the lithium battery during the i-th charge and discharge cycle; C0 represents the initial rated capacity of the lithium battery; i = 1, 2, ..., n, represents the number of charge and discharge cycles of the lithium battery.

[0059] S120, using a PCA algorithm to reduce the dimension of the monitoring data of the lithium battery cyclic charge and discharge process to obtain reduced-dimensional monitoring data.

[0060] In some embodiments, the monitoring data of the lithium battery cyclic charge and discharge process may be preprocessed first, and then the preprocessed monitoring data may be reduced in dimension using a PCA algorithm to obtain reduced-dimensional monitoring data.

[0061] Among them, preprocessing can include: missing value completion, outlier removal, standardization, etc.

[0062] Specifically, missing value completion and outlier elimination can prevent missing values ​​and outliers from affecting the selection of health indicators and the accuracy of lithium battery health status prediction results; standardization can eliminate dimensional differences and prevent the prediction results from being dominated by variables with larger orders of magnitude, while alleviating the slowdown in model convergence caused by large differences in the orders of magnitude of monitoring data.

[0063] It is worth noting that the PCA algorithm described here is used to reduce the dimension of the pre-processed monitoring data to obtain the reduced-dimensional monitoring data, which may include the following steps:

[0064] Step 1: Calculate the covariance matrix of the preprocessed monitoring data, and calculate multiple eigenvectors and their corresponding eigenvalues ​​according to the covariance matrix.

[0065] Step 2: Sort each eigenvector according to the eigenvalue from large to small, and calculate the cumulative contribution rate according to the eigenvalue of the eigenvector in the sorted order, until the cumulative contribution rate is greater than or equal to the preset threshold, and select the k eigenvectors used in the current cumulative contribution rate calculation to generate the principal component matrix. For example, the principal component matrix can be expressed as: V = [F1, F2, ..., F k ].

[0066] Step 3: Project the preprocessed monitoring data into the space corresponding to the principal component matrix to obtain reduced-dimensional monitoring data. In other words, multiply the preprocessed monitoring data by the principal component matrix to obtain reduced-dimensional monitoring data. Exemplarily, the above process can be expressed using the following formula:

[0067]

[0068] Where X2 represents the reduced dimension monitoring data in matrix form; X represents the preprocessed monitoring data; and V represents the principal component matrix.

[0069] S130, using the reduced-dimensionality monitoring data as sample data and the SOH data corresponding to the reduced-dimensionality monitoring data as label data to construct a data set.

[0070] That is, each row of elements in the dimension-reduced monitoring data matrix is ​​regarded as sample data, and the SOH data of the number of charge and discharge cycles to which each row of elements belongs is regarded as label data.

[0071] S140, using the data set to train the GRU model, and using the SSA algorithm to optimize hyperparameters during training to obtain a lithium battery health status prediction model.

[0072] In some embodiments, the data set can be divided into a training set and a test set according to a preset ratio (for example, 7:3), and the training set data is normalized before being input into the model. The training set is used to train a pre-built GRU model, and the SSA algorithm is used to optimize hyperparameters during training. The test set is used to evaluate the performance of the trained GRU model. If the trained GRU model passes the performance evaluation, it is used as a lithium battery health status prediction model.

[0073] The GRU model built here includes a sequence input layer, a GRU layer with multiple hidden units, a fully connected layer and an output layer; the GRU layer is used to capture the temporal association between important features, and uses a reset gate and an update gate to work together. The reset gate can selectively ignore the state of the previous moment, while the update gate generates a value from 0 to 1 to update the unit state and determine how much historical information is retained. For example, the GRU model can be expressed using the following formula:

[0074] r t =σ(W r ·[h t―1 ,x t ]+b r );

[0075] z t =σ(W z ·[h t―1 ,x t ]+b z );

[0076]

[0077] In the formula, x t Represents the single charge and discharge cycle monitoring data at the current moment; r t ,z t Respectively represent the outputs of the reset gate and update gate in the GRU layer; h t-1 , h t Respectively represent the hidden layer state at the previous moment, the temporary and current hidden layer output; W r ,W r , Represents the weight matrices of the reset gate, update gate and hidden state respectively; b r ,b z , They represent the bias terms of the reset gate, update gate and hidden state respectively; σ() represents the sigmoid activation function; tan h() represents the tanh activation function.

[0078] Furthermore, the GRU model built here selects the Adam gradient descent algorithm for parameter update and sets the model's initial hyperparameters, including the maximum number of iterations N. Iteration , batch size Bach_size, initial learning rate R Initial , learning rate reduction factor F Drop , the number of hidden units in the GRU layer C Hide wait.

[0079] It is worth noting that the hyperparameter optimization using the SSA algorithm during training described here can include the following steps:

[0080] Step 1: Determine the range of hyperparameter selection.

[0081] Step 2: Generate multiple initial hyperparameter combinations, that is, initialize the positions of multiple sparrows, and then initialize the sparrow population.

[0082] Step 3: Calculate the fitness value of each hyperparameter combination.

[0083] Step 4: Update the hyperparameter combination according to the fitness value, that is, update the position of the sparrow, and then update the sparrow population.

[0084] Step 5: Repeat steps 3-4 until the stopping condition is reached and output the optimal hyperparameter combination.

[0085] In order to further understand the technical principle of using the SSA algorithm to optimize super parameters, the following is a detailed explanation with an example:

[0086] In the solvable space, initialize a sparrow population, including the position vector X Position With the initial fitness value F X , the mathematical expression is:

[0087]

[0088] In the formula, d represents the dimension of the variable to be optimized; n represents the total number of sparrows; and f represents the fitness value.

[0089] With each iteration, the position of individual sparrows changes accordingly. In the SSA algorithm, the sparrow population is divided into explorers and followers. Explorers have higher energy and are responsible for searching the space for food and providing location and orientation information, while followers obtain food based on the information. In addition, during the entire search process, some sparrows will act as scouts to monitor whether there is any danger and issue warnings.

[0090] The mathematical expression of explorer position iteration is:

[0091]

[0092] In the formula, denotes the position of the i-th sparrow in the j-th dimension in the t-th and t+1-th iterations respectively; α denotes a random number between 0 and 1 (excluding 0 and 1); t max represents the maximum number of iterations; R2 represents the warning value between 0 and 1; δ st represents a safe value between 0.5 and 1; Q represents a random number that follows a normal distribution; L represents a 1×d matrix whose elements are all 1.

[0093] The mathematical expression for follower position iteration is:

[0094]

[0095] In the formula, X worst Respectively represent the best position and the global worst position of the explorer in the t+1th iteration; A + =A T (AA T ) ―1 Represents a 1-by-d matrix whose elements are all -1 or +1.

[0096] The mathematical expression of the scout position iteration is:

[0097]

[0098] In the formula, represents the global optimal position after t iterations; β represents the compensation parameter; K represents a random number between -1 and 1; f i ,f g ,f w They represent the fitness value of the current scout individual and the optimal and worst fitness values ​​respectively; ε represents a constant that makes the denominator non-zero.

[0099] S150, predicting SOH data of a specified lithium battery using a lithium battery health status prediction model.

[0100] In some embodiments, a lithium battery health status prediction model may be used to process the dimensionality reduction monitoring data of a specified lithium battery to predict the SOH data of the specified lithium battery.

[0101] In summary, according to the embodiments of the present invention, at least the following technical effects are achieved:

[0102] (1) Aiming at the capacity degradation process of lithium batteries with complex coupled factors, the present invention proposes a lithium battery health status prediction method based on PCA-SSA-GRU, which can reduce the dimension of the original charge and discharge cycle monitoring data of the battery, retain the key information related to the capacity degradation process, remove irrelevant noise, improve the prediction accuracy, and reduce the computational cost.

[0103] (2) In view of the fact that the traditional single data-driven method has high requirements on the integrity and quantity of historical data and is often only applicable to situations under specific charging conditions, the present invention proposes a streamlined SSA-GRU network model, and the hyperparameters are continuously optimized during the model training process. It can not only process the charge and discharge cycle data of lithium batteries under the same charging protocol, but also predict the health status of lithium batteries under different charging protocols, thereby improving generalization and accuracy.

[0104] The following is a detailed description of the lithium battery health status prediction method 100 provided in the embodiment of the present invention in conjunction with a specific embodiment, as shown below:

[0105] This embodiment will use the Toyota Research Institute (TRI) dataset. The TRI dataset was released by MIT and Stanford University in 2019 in conjunction with the Toyota Research Institute. It contains 124 lithium iron phosphate / graphite battery data cycled under 72 different charging conditions. Compared with the battery dataset in which the charging and discharging process is under constant current and constant voltage test conditions, it is very valuable for studying the degradation process of batteries under dynamic working conditions. This dataset has been used by many researchers. The experimental battery of the TRI dataset is a lithium iron phosphate / graphite battery produced by A123 Systems, with a nominal capacity of 1.1Ah and a nominal voltage of 3.3V. They were all cycled on a 48-channel Arbin LBT constant potentiostat in a forced convection temperature chamber set at 30°C, with a cycle life ranging from 150 to 2300 times. The TRI data set contains three batches of battery data (hereinafter referred to as Batch 1, Batch 2 and Batch 3). Each batch contains approximately 48 batteries, corresponding to three groups of experiments, where each battery adopts a one-step or two-step charging protocol in the format of "C1(Q1)-C2", which means: charging with the constant current charging protocol of C1 to reach Q1, i.e. the current switching state of charge (SOC), and the charging protocol is converted to C2. In addition, the discharge protocol of the batteries is 4C. Each batch of data contains variables within the cycle: including variables such as internal resistance, temperature, charge and discharge capacity, and variables for each cycle: including variables such as maximum / minimum temperature and charging time. Based on this TRI data set, the lithium battery health status prediction method 100 may include the following steps:

[0106] (1) Collect the lithium battery cycle charge and discharge monitoring data and its corresponding SOH data of "Batch 3" in the TRI dataset.

[0107] According to the experimental notes, the computer automatically restarted during the experiments of Batch 1 and Batch 2, resulting in partial missing and capacity mutations in the experimental data of the two batches. At the same time, due to errors in the electronic connection of some channels and the detachment of thermocouples, some data curves were noisy. Therefore, the data of Batch 1 and Batch 2 cannot be used here. Therefore, the lithium battery cycle charge and discharge monitoring data of "Batch 3" in the TRI data set is collected as the monitoring data of the lithium battery cycle charge and discharge process, which is represented by the following matrix form:

[0108]

[0109] In the formula, X represents the acquired monitoring data, m represents the number of variables in the monitoring data; n represents the total number of lithium battery charge and discharge cycles; element x ij It represents the monitoring value of the jth variable during the i-th charge and discharge cycle of the lithium battery, where i = 1, 2, ..., n, j = 1, 2, ..., m.

[0110] The SOH data here is the SOH data of 8 lithium batteries, which is calculated based on the discharge capacity data. The specific calculation formula is as follows:

[0111]

[0112] In the formula, SOH(i) represents the SOH data of the lithium battery during the i-th cycle charge and discharge; C i It represents the actual capacity of the lithium battery during the i-th cycle charge and discharge, and is equal to Q Discharge (i) refers to the capacity of the lithium battery during the i-th charge and discharge cycle; C0 represents the initial rated capacity of the lithium battery; i = 1, 2, ..., n, represents the number of charge and discharge cycles of the lithium battery.

[0113] (2) Preprocess the monitoring data.

[0114] In order to verify that the method of the present invention can be applied to the construction of a health indicator system for batteries under different charging protocols, 8 batteries with completely different charging protocols in "Batch 3" were selected as research objects. Since there are no missing values ​​in the experimental data of the 8 batteries, only standardization is performed. Data standardization is one of the key steps in data preprocessing and lithium battery health status detection. It can eliminate dimensional differences, improve model performance, and perform more accurate analysis when multiple variables exist. In this embodiment, the Z-Score standardization method is adopted, and its function expression is as follows: for the sequence x1, x2,…, x n To transform:

[0115] here

[0116] Therefore, the new sequence y1,y2,…,y n has a mean of 0 and a variance of 1 and is dimensionless.

[0117] (3) Use the PCA algorithm to reduce the dimension of the preprocessed monitoring data to obtain reduced-dimensional monitoring data.

[0118] The covariance matrix of the preprocessed monitoring data is calculated, and a plurality of eigenvectors and their corresponding eigenvalues ​​are calculated according to the covariance matrix.

[0119] Sort each eigenvector from large to small according to the eigenvalue, calculate the cumulative contribution rate according to the eigenvalue of the eigenvector in the sorted order, until the cumulative contribution rate is greater than or equal to the preset threshold, and select the k eigenvectors used in the current cumulative contribution rate calculation to generate the principal component matrix. For example, the principal component matrix can be expressed as: V = [F1, F2, ..., F k ].

[0120] The preprocessed monitoring data is projected into the space corresponding to the principal component matrix to obtain the reduced-dimensional monitoring data. In other words, the preprocessed monitoring data is multiplied by the principal component matrix to obtain the reduced-dimensional monitoring data. Exemplarily, the above process can be expressed using the following formula:

[0121]

[0122] In the formula, X2 represents the reduced dimension monitoring data in matrix form; X represents the preprocessed monitoring data; and V represents the principal component matrix. Taking the first battery as an example, the reduced dimension monitoring data X2 with a dimension of 771×2 is obtained.

[0123] (4) The reduced-dimensionality monitoring data is used as sample data and the SOH data corresponding to the reduced-dimensionality monitoring data is used as label data to construct a dataset.

[0124] (5) The GRU model is trained using the dataset, and the SSA algorithm is used to optimize hyperparameters during training to obtain a lithium battery health status prediction model.

[0125] The data set is divided into a training set and a test set according to a 7:3 ratio. The training set data is normalized before being input into the model. The pre-built GRU model is trained using the training set, and the SSA algorithm is used to optimize hyperparameters during training. The performance of the trained GRU model is evaluated using the test set. If the trained GRU model passes the performance evaluation, it will be used as a lithium battery health status prediction model.

[0126] The GRU model built here includes a sequence input layer with an input dimension of 3, a GRU layer with multiple hidden units, a fully connected layer, and a 1-dimensional output layer; the GRU layer is used to capture the temporal association between important features, and uses a reset gate and an update gate to work together. The reset gate can selectively ignore the state of the previous moment, while the update gate generates a value from 0 to 1 to update the unit state and determine how much historical information is retained. For example, the GRU model can be expressed using the following formula:

[0127] r t =σ(W r ·[h t―1 ,x t ]+br );

[0128] z t =σ(W z ·[h t―1 ,x t ]+b z );

[0129]

[0130] In the formula, x t Represents the single charge and discharge cycle monitoring data at the current moment; r t ,z t Respectively represent the outputs of the reset gate and update gate in the GRU layer; h t-1 , h t Respectively represent the hidden layer state at the previous moment, the temporary and current hidden layer output; W r ,W z , Represents the weight matrices of the reset gate, update gate and hidden state respectively; b r ,b z , They represent the bias terms of the reset gate, update gate and hidden state respectively; σ() represents the sigmoid activation function; tan h() represents the tanh activation function.

[0131] Furthermore, the GRU model built here selects the Adam gradient descent algorithm for parameter update and sets the model's initial hyperparameters, including the maximum number of iterations N. Iteration , batch size Bach_size, initial learning rate R Initial , learning rate reduction factor F Drop , the number of hidden units in the GRU layer C Hide Etc. Taking the first battery as an example, including the maximum number of iterations N Iteration =200, learning rate reduction factor F Drop =0.2, and the batch size Bach_size and initial learning rate R Initial , the number of hidden units in the GRU layer C Hide The three hyperparameters that are optimized as the training process changes are set in the subsequent steps.

[0132] Use Sparrow Search Algorithm (SSA) to configure the batch size Bach_size and initial learning rate R in the GRU network model Initial , the number of hidden units in the GRU layer C Hide The value curves of the three hyperparameters are [64,28], [0.001,0.01], and [64,256] respectively.

[0133] The hyperparameter optimization described here may include the following steps:

[0134] In the solvable space, a sparrow population is initialized to a size of 50×3, i.e., n=50, d=3, and explorers are set to account for 20% of the total population, and the rest are followers. The safety threshold is set to 0.8, and 20% of individuals are randomly selected from the population as scouts to warn of danger. In this embodiment, the prediction error RMSE value of the PCA-SSA-GRU model test set is used as the fitness value. After all fitness values ​​are calculated and saved, the optimal value of the individual in the initial population and the global optimal value are obtained. Its position vector X Position With the initial fitness value F X , the mathematical expression is:

[0135]

[0136] In the formula, d represents the dimension of the variable to be optimized; n represents the total number of sparrows; and f represents the fitness value.

[0137] With each iteration, the position of individual sparrows changes accordingly. In the SSA algorithm, the sparrow population is divided into explorers and followers. Explorers have higher energy and are responsible for searching the space for food and providing location and orientation information, while followers obtain food based on the information. In addition, during the entire search process, some sparrows will act as scouts to monitor whether there is any danger and issue warnings.

[0138] The mathematical expression of explorer position iteration is:

[0139]

[0140] In the formula, denotes the position of the i-th sparrow in the j-th dimension in the t-th and t+1-th iterations respectively; α denotes a random number between 0 and 1 (excluding 0 and 1); t max represents the maximum number of iterations; R2 represents the warning value between 0 and 1; δ st represents a safe value between 0.5 and 1; Q represents a random number that follows a normal distribution; L represents a 1×d matrix whose elements are all 1.

[0141] The mathematical expression for follower position iteration is:

[0142]

[0143] In the formula, X worst Respectively represent the best position and the global worst position of the explorer in the t+1th iteration; A + =A T (AA T )―1 Represents a 1-by-d matrix whose elements are all -1 or +1.

[0144] The mathematical expression of the scout position iteration is:

[0145]

[0146] In the formula, represents the global optimal position after t iterations; β represents the compensation parameter; K represents a random number between -1 and 1; f i ,f g ,f w They represent the fitness value of the current scout individual and the optimal and worst fitness values ​​respectively; ε represents a constant that makes the denominator non-zero.

[0147] When the PCA-SSA-GRU model is trained and the termination condition is met, the optimal hyperparameter combination and the optimal GRU model parameters are output. Taking the first battery as an example, the optimal hyperparameter combination is batch size Bach_size = 65, initial learning rate R Initial =0.009854, number of hidden units in the GRU layer C Hide =123.

[0148] (6) Use the lithium battery health status prediction model to predict the SOH data of the specified lithium battery.

[0149] To summarize, the embodiments of the present invention address the shortcomings of existing lithium battery health status prediction methods, take batteries under 8 different non-constant current and constant voltage charging conditions as research objects, and use the "dimensionality reduction" algorithm to extract key information reflecting changes in the actual health status of the battery from multiple mutually coupled factors, thereby reducing the computational cost. In addition, the GRU network model suitable for a variety of charging conditions is established by combining limited historical data.

[0150] Figure 2The following is a schematic diagram comparing the prediction effects of lithium battery health status of different models. As the number of cycles increases, the CNN model has a significant mutation in the later stage of prediction. In contrast, the LSTM model and the GRU model are more robust in SOH prediction. However, after zooming in on the part with the number of cycles from 700 to 770, it can be found that the prediction curve of the GRU model is more consistent with the actual curve. The GRU model is relatively simple in structure and has fewer parameters, which enables the GRU model to train and converge faster when facing a smaller data set while maintaining good generalization ability. Although the LSTM model is more complex in design and has more parameters, it can theoretically capture longer-term time dependencies. However, in this case, when the data volume is small, the LSTM model may face the risk of overfitting, resulting in its prediction accuracy being inferior to that of the GRU model. In addition, the training process of the LSTM model is more time-consuming and converges more slowly.

[0151] Figure 3 The schematic diagram of the lithium battery health status prediction effect of the lithium battery health status prediction method of this embodiment (taking the first battery as an example), Table 1 is a table of SOH prediction results of 8 batteries in the GRU model, including the optimal hyperparameter combination and prediction evaluation index RMSE, MAE values, RMSE and MAE orders of magnitude are both within 10 -2 , with higher accuracy.

[0152] Table 1

[0153]

[0154] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described order of actions, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0155] The above is an introduction to a method embodiment. The following is a further explanation of the solution of the present invention through an apparatus embodiment.

[0156] Figure 4 A structural diagram of a lithium battery health status prediction device provided by an embodiment of the present invention, such as Figure 4 As shown, the lithium battery health status prediction device 400 may include:

[0157] The acquisition module 410 is used to acquire monitoring data of the lithium battery's cyclic charge and discharge process and its corresponding SOH data.

[0158] The dimension reduction module 420 is used to reduce the dimension of the monitoring data of the lithium battery cycle charge and discharge process by using the PCA algorithm to obtain reduced-dimensional monitoring data.

[0159] The construction module 430 is used to construct a data set using the reduced-dimensionality monitoring data as sample data and the SOH data corresponding to the reduced-dimensionality monitoring data as label data.

[0160] The training module 440 is used to train the GRU model using the data set and use the SSA algorithm to optimize the hyperparameters during the training to obtain a lithium battery health status prediction model.

[0161] The prediction module 450 is used to predict the SOH data of a specified lithium battery using a lithium battery health status prediction model.

[0162] Understandably, Figure 4 Each module / unit in the lithium battery health status prediction device 400 shown has the function of realizing Figure 1 The functions of the various steps in the lithium battery health status prediction method 100 shown and the ability to achieve their corresponding technical effects are not described here for the sake of brevity.

[0163] Figure 5 5 is a block diagram of an exemplary electronic device capable of implementing an embodiment of the present invention. Electronic device 500 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device 500 may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown in the present invention, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required in the present invention.

[0164] like Figure 5 As shown, the electronic device 500 may include a computing unit 501, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the electronic device 500 can also be stored. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0165] Multiple components in the electronic device 500 are connected to the I / O interface 505, including: an input unit 506, such as a keyboard, a mouse, etc.; an output unit 507, such as various types of displays, speakers, etc.; a storage unit 508, such as a disk, an optical disk, etc.; and a communication unit 509, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 509 allows the electronic device 500 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0166] The computing unit 501 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 501 performs the various methods and processes described above, such as method 100. For example, in some embodiments, the method 100 may be implemented as a computer program product, including a computer program, which is tangibly contained in a computer-readable medium, such as a storage unit 508. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 500 via ROM 502 and / or communication unit 509. When the computer program is loaded into RAM 503 and executed by the computing unit 501, one or more steps of the method 100 described above may be performed. Alternatively, in other embodiments, the computing unit 501 may be configured to perform the method 100 in any other appropriate manner (e.g., by means of firmware).

[0167] The various embodiments described above in the present invention can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs, which may be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general programmable processor, which may receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0168] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0169] In the context of the present invention, a computer-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. A computer-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a computer-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0170] It should be noted that the present invention also provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute method 100 and achieve the corresponding technical effect achieved by executing the method in an embodiment of the present invention. For the sake of concise description, they will not be repeated here.

[0171] In addition, the present invention also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the method 100 is implemented.

[0172] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution disclosed in the present invention can be achieved, and the present invention is not limited here.

[0173] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting the health status of a lithium battery, characterized in that: The method comprises: Obtain monitoring data of the lithium battery cycle charge and discharge process and its corresponding SOH data; The PCA algorithm is used to reduce the dimension of the monitoring data of the lithium battery cycle charge and discharge process to obtain the reduced dimension monitoring data; The reduced-dimensionality monitoring data is used as sample data, and the SOH data corresponding to the reduced-dimensionality monitoring data is used as label data to construct a data set; The GRU model is trained using the data set, and the SSA algorithm is used to optimize hyperparameters during training to obtain a lithium battery health status prediction model; The lithium battery health status prediction model is used to predict the SOH data of a specified lithium battery.

2. The method according to claim 1, characterized in that The PCA algorithm is used to reduce the dimension of the monitoring data of the lithium battery cycle charge and discharge process to obtain the reduced dimension monitoring data, including: Preprocessing the monitoring data of the lithium battery cycle charge and discharge process, wherein the preprocessing includes: missing value filling, outlier removal, and standardization; The PCA algorithm is used to reduce the dimension of the preprocessed monitoring data to obtain reduced-dimensional monitoring data.

3. The method according to claim 2, characterized in that The PCA algorithm is used to reduce the dimension of the pre-processed monitoring data to obtain the reduced-dimensional monitoring data, including: Calculating a covariance matrix of the preprocessed monitoring data, and calculating a plurality of eigenvectors and their corresponding eigenvalues ​​according to the covariance matrix; Sort each eigenvector according to the eigenvalue from large to small, calculate the cumulative contribution rate according to the eigenvalue of the eigenvector in the sorted order, until the cumulative contribution rate is greater than or equal to the preset threshold, and select multiple eigenvectors used in the current cumulative contribution rate calculation to generate a principal component matrix; The preprocessed monitoring data is projected into the space corresponding to the principal component matrix to obtain dimension-reduced monitoring data.

4. The method according to claim 1, characterized in that: The GRU model is trained using the data set, and the SSA algorithm is used to optimize the hyperparameters during the training to obtain a lithium battery health status prediction model, including: Dividing the data set into a training set and a test set according to a preset ratio; The GRU model is trained using the training set, and the SSA algorithm is used to optimize hyperparameters during training. The performance of the trained GRU model is evaluated using the test set. If the trained GRU model passes the performance evaluation, it is used as a lithium battery health status prediction model.

5. The method according to claim 4, characterized in that The GRU model includes a sequence input layer, a GRU layer with multiple hidden units, a fully connected layer and an output layer; the GRU layer is used to capture the temporal association between features, and uses a reset gate and an update gate to work together. The reset gate can selectively ignore the state of the previous moment, and the update gate generates a value from 0 to 1 to update the unit state, determining how much historical information is retained.

6. The method according to claim 4, characterized in that The SSA algorithm is used to optimize hyperparameters during training, including: Step 1: Determine the range of hyperparameter selection; Step 2: Generate multiple initial hyperparameter combinations, that is, initialize the positions of multiple sparrows, and then initialize the sparrow population; Step 3: Calculate the fitness value of each hyperparameter combination; Step 4: Update the hyperparameter combination according to the fitness value, that is, update the position of the sparrow, and then update the sparrow population; Step 5: Repeat steps 3-4 until the stopping condition is reached and output the optimal hyperparameter combination.

7. The method according to claim 1, characterized in that The method of using the lithium battery health status prediction model to predict the SOH data of the lithium battery includes: The lithium battery health status prediction model is used to process the dimension reduction monitoring data of the specified lithium battery to predict the SOH data of the specified lithium battery.

8. A lithium battery health status prediction device, characterized in that: The device comprises: An acquisition module is used to acquire monitoring data of the lithium battery cycle charge and discharge process and its corresponding SOH data; A dimension reduction module is used to reduce the dimension of the monitoring data of the lithium battery cycle charge and discharge process by using the PCA algorithm to obtain reduced-dimensional monitoring data; A construction module, used to construct a data set using the reduced-dimensionality monitoring data as sample data and the SOH data corresponding to the reduced-dimensionality monitoring data as label data; A training module, used to train the GRU model using the data set, and use the SSA algorithm to optimize hyperparameters during training to obtain a lithium battery health status prediction model; The prediction module is used to predict the SOH data of a specified lithium battery using the lithium battery health status prediction model.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable a computer to execute the method according to any one of claims 1 to 7.

Images