Lithium battery health state prediction method, equipment and medium

Abstract

The invention relates to a lithium battery health state prediction method, which comprises the steps of obtaining a historical data sequence of a lithium battery health state from a health state database, and processing the historical data sequence to determine a historical data step length and a prediction step length; extracting a training set, a verification set and a test set from the historical data sequence according to the historical data step length and the prediction step length; constructing an SOH prediction model, and training the SOH prediction model according to the training set, the verification set and the test set; and according to an SOH prediction request of the lithium battery, obtaining a data sequence of a historical data step length from the health state database, inputting the data sequence into the SOH prediction model, and outputting a predicted SOH value. The invention also relates to a device and a medium. By analyzing the lithium battery health state historical data sequence, the historical data step length and the prediction step length related to the health state prediction model can be determined, the appropriate historical observation window length can be selected, and the future data length can be predicted, so that the accuracy and the reliability of the model are improved.

Description

Technical Field

[0001] The present application relates to the field of battery technology, and in particular to a method, device and medium for predicting the health status of a lithium battery. Background Art

[0002] The state of health (SOH) reflects the ability of a battery to store and supply energy relative to its initial conditions. It is usually determined based on the reduction in battery capacity or the increase in internal resistance. When the battery capacity drops to 80% of its initial value, or the internal resistance becomes 1.3 times its initial value, the fault threshold is reached. Lithium battery SOH prediction helps to understand the performance changes of the battery, detect potential faults in a timely manner, and take measures to extend the battery life and ensure the safe and reliable operation of new energy storage and power grids.

[0003] At present, there are many methods for predicting lithium battery SOH, such as physical model-based methods, data-driven methods, etc. However, data-driven methods usually require a large amount of random historical data to train the model, but in practical applications, due to the diversity of battery usage environments and working conditions, it is impossible to accurately obtain sufficient high-quality historical data to predict lithium battery SOH, resulting in low accuracy and poor adaptability in health status prediction. Summary of the invention

[0004] In order to solve the above technical problems, this application provides the following technical solutions:

[0005] On the one hand, the present application provides a method for predicting the health status of a lithium battery, the steps of which include:

[0006] Obtain a historical data sequence of the health status of the lithium battery from a health status database, and process the historical data sequence to determine a historical data step length and a prediction step length;

[0007] According to the historical data step size and the prediction step size, the training set, validation set and test set are extracted from the historical data sequence;

[0008] Construct a SOH prediction model and train the SOH prediction model based on the training set, validation set, and test set;

[0009] According to the SOH prediction request of the lithium battery, the data sequence of the historical data step is obtained from the health status database, the data sequence is input into the SOH prediction model, and the predicted SOH value is output.

[0010] In one embodiment, a historical data sequence of the health status of a lithium battery is obtained from a health status database, and the historical data sequence is processed to determine a historical data step length and a prediction step length, including:

[0011] From the health status database, obtain the health status values ​​at multiple evaluation time points and construct a historical data sequence;

[0012] Analyze the historical data sequence through the autocorrelation function to generate an autocorrelation image;

[0013] From the autocorrelation image, determine the time interval at which the autocorrelation coefficient is higher than a preset coefficient threshold, and determine the time interval as the historical data step length;

[0014] The prediction step size is determined based on the proportional relationship between the historical data step size and the prediction step size.

[0015] In one embodiment, generating an autocorrelation image includes: an X-axis of the autocorrelation image is the number of time lags, and a Y-axis is the autocorrelation coefficient.

[0016] In one embodiment, the autocorrelation function includes:

[0017] The formula for the autocorrelation coefficient at lag k is

[0018]

[0019] Where N is the length of the time series, x t is the time series value at time t, is the mean of the time series.

[0020] In one embodiment, according to the historical data step size and the prediction step size, a training set, a validation set, and a test set are extracted from the historical data sequence, including:

[0021] In the historical data sequence, the first 60% of the total cycle period of each battery sample is determined as the training set and validation set, and the remaining 40% of the data is used as the test set; among them, the training set accounts for 50% and the validation set accounts for 10%.

[0022] In one embodiment, extracting a training set from a historical data sequence includes:

[0023] The sliding window slides gradually on the data sequence of the sample data set, one unit at a time, and the data in the sliding window constitutes the training set.

[0024] In one implementation manner, constructing the SOH prediction model includes: selecting any one of the GRU algorithm, the Informer algorithm, and the DLinear algorithm to construct the SOH prediction model.

[0025] In one embodiment, training a SOH prediction model based on a training set, a validation set, and a test set includes:

[0026] The SOH prediction model architecture is trained according to the training set to obtain an initial SOH prediction model;

[0027] The initial SOH prediction model was verified and evaluated based on the validation set and test set to obtain a SOH prediction model that can be used for health status prediction.

[0028] On the other hand, the present application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the lithium battery health status prediction method provided in the above embodiment are implemented.

[0029] On the other hand, the present application also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the lithium battery health status prediction method provided in the above embodiment are implemented.

[0030] This application has at least the following beneficial effects:

[0031] In this application, by analyzing the historical data sequence of the health status of lithium batteries, the historical data step size and prediction step size related to the health status prediction model can be determined, which helps to select the appropriate historical observation window length and predict the length of future data, thereby improving the accuracy and reliability of the model. The technical solution of this application has good adaptability and scalability, and can be applied to lithium batteries of different types and specifications, as well as different application scenarios and systems. By continuously updating and improving the historical data sequence, the accuracy and adaptability of the prediction model can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A flowchart of a lithium battery health status prediction method provided in an embodiment of the present application.

[0033] Figure 2 A schematic diagram of an autocorrelation image provided by an embodiment of the present application.

[0034] Figure 3 A schematic diagram of forming a training set according to an embodiment of the present application.

[0035] Figure 4 A schematic diagram of a test set provided in accordance with an embodiment of the present application.

[0036] Figure 5 A schematic diagram of the process of training and predicting the GRU, Informer, and DLinear algorithms provided in one embodiment of the present application.

[0037] Reference numerals:

[0038] 1. Drying chamber; 2. Vacuum pump; 3. Fan; 4. Collection filter; 5. Material collector; 6. Vacuum filter; 7. Buffer chamber; 8. Discharge valve;

[0039] 11. Main cabin; 12. Insulation cover; 13. Air extraction port; 14. Feed port;

[0040] 71. Material storage compartment; 72. Maintenance compartment door;

[0041] 111, feeding end; 112, discharging end;

[0042] 200. Control device;

[0043] C01, cold source liquid inlet pipe; C02, cold source liquid outlet pipe;

[0044] H01, heat source liquid inlet pipe; H02, heat source liquid outlet pipe;

[0045] NP1, first negative pressure pipeline; NP2, second negative pressure pipeline; NP3, third negative pressure pipeline;

[0046] PP1, first positive pressure pipeline;

[0047] V1, pump body isolation valve; V2, cabin filter isolation valve; V3, air blowing control valve; V4, separation isolation valve; V5, discharge isolation valve; V6, hot liquid isolation valve;

[0048] P1, first pressure monitor; P2, second pressure monitor; P3, third pressure monitor. DETAILED DESCRIPTION

[0049] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present application, so the present application is not limited by the specific embodiments disclosed below.

[0050] In the description of the present application, it should be understood that if the terms "length", "width", "thickness", "up", "down", "vertical", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc. appear, the orientation or position relationship indicated by these terms is based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present application and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application.

[0051] In addition, if the terms "first", "second", and "third" appear, these terms are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first", "second", and "third" may explicitly or implicitly include at least one of the features. In the description of this application, if the term "plurality" appears, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0052] In this application, unless otherwise clearly specified and limited, if the terms "installed", "connected", "connected", "fixed" and the like appear, these terms should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integrated connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to the specific circumstances.

[0053] In the present application, unless otherwise clearly specified and limited, if there is a description that a first feature is "above" or "below" a second feature, etc., or similar descriptions appear, it may mean that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. Moreover, the first feature being "above", "above" and "above" the second feature may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. The first feature being "below", "below" and "below" the second feature may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is lower in level than the second feature.

[0054] It should be noted that if an element is referred to as being "fixed to" or "disposed on" or "set on" another element, it may be directly on the other element or there may be a central element. If an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. If any, the terms "vertical", "horizontal", "upper", "lower", "left", "right" and similar expressions used in this application are for illustrative purposes only and do not represent the only implementation method.

[0055] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0056] Reference Figure 1 As shown, in some embodiments of the present application, the present application provides a method for predicting the health status of a lithium battery, the steps of which include:

[0057] Step S1: Obtain a historical data sequence of the health status of the lithium battery, perform autocorrelation analysis on the historical data sequence, and determine the historical data step length and prediction step length related to the health status prediction model.

[0058] In this solution, the historical data step length refers to the length of the historical observation window, and the prediction step length refers to the length of the predicted future data. The historical data sequence is the health status value of the lithium battery in each cycle.

[0059] Wherein, step S1 specifically includes:

[0060] Step S11: Obtain health status values ​​at multiple evaluation time points from the health status database to construct a historical data sequence.

[0061] The health status value is obtained by inputting the health factor parameter value of the lithium battery into the lithium battery health assessment model.

[0062] Step S12: Analyze the historical data sequence through the autocorrelation function to generate an autocorrelation image, wherein the X-axis of the autocorrelation image is the number of time lags, and the Y-axis is the autocorrelation coefficient.

[0063] Among them, for the time series (historical data series) X = {x1, x2, ...x n}, its autocorrelation at lag k is as follows:

[0064]

[0065] Where N is the length of the time series, x t is the time series value (health status value) at time t, is the mean of the time series, and the autocorrelation coefficient r of lag k k The closer it is to 1, the more similar or periodic the time series is at that lag.

[0066] It should be noted that since the autocorrelation function (ACF) is used to measure the autocorrelation between a variable and its variables at different lag time points, the degree of correlation between the observed value (health status) at a given time point and the previous observed value (health status) can be determined.

[0067] Step S13: Determine the time interval in which the autocorrelation coefficient is higher than the preset coefficient threshold from the autocorrelation image, and determine the time interval as the historical data step length.

[0068] It should be noted that the present application has found that the larger the observation window, the lower the correlation between the observation value at a given time point and the previous observation value. Since smaller observation windows require more computing resources and processing time, larger windows cannot capture rapid changes and short-term fluctuations in the data in a timely manner. Therefore, the time interval in which the autocorrelation coefficient is higher than the preset coefficient threshold is selected.

[0069] Step S14: Determine the prediction step length according to the proportional relationship between the historical data step length and the prediction step length.

[0070] It should be noted that the proportional relationship can be set according to actual needs, usually a 2-fold relationship.

[0071] For example, after performing SOH autocorrelation analysis on a lithium battery, the autocorrelation graph is as follows: Figure 2 shown.

[0072] When the preset coefficient threshold is set to 0.8 and the proportional relationship is set to the historical data step length is twice the forecast step length, it can be observed that Figure 2 In the figure, when the time interval is 32, the autocorrelation of 0.9 can be maintained. When the time interval is 64, the autocorrelation is 0.8, and 0.8 corresponds to 64. Therefore, the historical observation window length of the health status prediction model is set to 64, and the predicted future data length is set to 36. That is, in the operation of the energy storage power station, if the sampling frequency is once every 10 days, the health status changes in the next 360 days are predicted.

[0073] Through the autocorrelation analysis of the lithium battery health status historical data series, the historical data step size and prediction step size related to the health status prediction model can be determined. This helps to select the appropriate historical observation window length and the length of predicted future data, thereby improving the accuracy and reliability of the model.

[0074] This technical solution has good adaptability and scalability, and can be applied to lithium batteries of different types and specifications, as well as different application scenarios and systems. By continuously updating and improving the historical data series, the accuracy and adaptability of the prediction model can be improved.

[0075] Step S2: Extract training set, validation set and test set from the historical data sequence according to the historical data step size and prediction step size.

[0076] Wherein, step S2 specifically includes:

[0077] Step S21: In the historical data sequence, determine the first 60% of the total cycle period of each battery sample as the training set and the validation set, and the remaining 40% of the data as the test set, of which the training set accounts for 50% and the validation set accounts for 10%.

[0078] Step S22: Divide the historical data sequence into fixed window sizes using a sliding window strategy.

[0079] like Figure 3 As shown, the data sequence of the sample data set is gradually slid, one unit at a time, to form a training set. For example, when the sample data set is divided, the historical data step corresponds to 1-64 (equivalent to one input of the prediction model), and the prediction step corresponds to 65-100 (equivalent to one output of the prediction model). Sliding one unit, the historical data step corresponds to 2-65 (equivalent to another input of the prediction model), and the prediction step corresponds to 66-101 (equivalent to another output of the prediction model), until 50% of the sample data set is extracted. It should be noted that if the length of 100 is not met in the end, this sample point will be discarded.

[0080] like Figure 4 As shown in the figure, in the prediction stage, the prediction step size is slid by units each time to form a test set. For example, when the data of the sample data set is divided starting from 40% of the data, the historical data step size corresponds to 1-64, and the prediction step size corresponds to 65-100. When sliding by 36 units, the historical data step size corresponds to 37-100, and the prediction step size corresponds to 101-136.

[0081] It should be noted that in order to enable the prediction steps to be spliced ​​together, the sliding window selects prediction step units, so that the prediction step splicing result (equivalent to multiple output data of the prediction model) is consistent with the data sequence in the test set, thereby improving the accuracy of the health status prediction model.

[0082] On the one hand, the prediction step size refers to the length of the predicted future data. Through the sliding window strategy, the window size can be adjusted according to different prediction step sizes, so as to extract training samples that are suitable for the prediction step size. This ensures that the model can be fully trained and verified under different prediction step sizes.

[0083] In order to enable the prediction steps to be spliced ​​together, the sliding window selects prediction step units, so that the prediction step splicing results (equivalent to multiple output data of the prediction model) are consistent with the data sequence in the test set, thereby improving the accuracy of the health status prediction model.

[0084] On the other hand, the sliding window strategy can generate multiple training samples on the historical data sequence, increasing the diversity of the data. This can avoid the model from overfitting to a specific time point or time period, and improve the generalization and robustness of the prediction model. The sliding window strategy can also better process time series data, capture dynamic features, increase data diversity, and improve the adaptability and accuracy of the prediction model.

[0085] Step S3: Train the SOH prediction model architecture according to the training set to obtain an initial SOH prediction model, verify and evaluate the initial SOH prediction model according to the validation set and the test set to obtain a SOH prediction model that can be used to predict the health status of lithium batteries.

[0086] As described in step S2 above, for the test set, the prediction starting point of the lithium battery SOH is set to 60% of the total cycle period, and the SOH value of the remaining 40% cycle period is used as the true value to compare with the prediction result, and then the SOH prediction model is established using the GRU algorithm or the Informer algorithm or the DLinear algorithm. After the SOH prediction model is established, the training set is input to train the SOH prediction model.

[0087] In a specific embodiment, the SOH prediction model is constructed by the GRU algorithm. The gated recurrent unit GRU is a variant of RNN, which is used to process sequence data. It has a gating mechanism that can help the model better capture long-term dependencies without being affected by the gradient vanishing problem, with fewer parameters and faster training speed. The basic formula for the forward propagation calculation of the GRU model at each time step is as follows: where σ represents the sigmoid function, W is the weight matrix of the model, b represents the adjustment coefficient of the model, the current time step is t, and the input is x t , the hidden state of the previous time step is h t-1 , the reset gate of the current time step is r t , the update gate is z t , the candidate hidden state is The hidden state at the current time step is h t .

[0088] Reset Gate t Controls the influence of the previous hidden state on the candidate hidden state.

[0089] r t =σ(W ir x t +b ir +h hr h t-1 +b hr )#

[0090] Update gate z t Controls the effect of the current input on the hidden state.

[0091] z t =σ(W iz x t +b iz +W hz h t-1 +b hz )#

[0092] The final hidden state h t is a weighted combination of the candidate hidden state at the current time step and the hidden state at the previous time step.

[0093]

[0094] Step S4: receiving a SOH prediction request for a lithium battery, obtaining an SOH data sequence of a historical data step from a health status database, inputting the SOH data sequence into an SOH prediction model, and obtaining a predicted SOH value of the prediction step.

[0095] In an actual usage scenario, the prediction effects of three time series prediction models, GRU, Informer, and DLinear, on battery samples with five different charging strategies.

[0096] 1) Among them, the number of training set and validation set calculations n1 is shown in the formula.

[0097] n1=N-historytimesteps-futuretimesteps#

[0098] The number of test set calculations n2 is shown in the formula.

[0099]

[0100] The number of cycles in the training set, validation set, and test set is shown in Table 1.

[0101] Table 1

[0102]

[0103]

[0104] 2) The process of building the Informer and DLinear models is as follows:

[0105] (I) The Informer model combines the Transformer's self-attention mechanism and encoder-decoder structure to process sequence data with long-term dependencies. The Informer model uses a probabilistic sparse self-attention mechanism to calculate the correlation of input variables in the input sequence, thereby capturing the long-term dependencies hidden in the sequence. The basic calculation formula is as follows: Q, K, V are the Query, Key, and Value matrices extracted from the input feature matrix, Softmax is the normalization equation, and d is the input feature dimension. is the filtered sparse matrix.

[0106]

[0107] The calculation formula for the process of convolutional layer distillation at time t of the Informer model encoder is as follows. Conv1d and MaxPool1d are used to reduce the amount of calculation and retain features with higher sparsity, so that the model pays more attention to the dominant attention score.

[0108]

[0109] The fully connected layer of the Informer model decoder outputs the final prediction result. For the vector, prior data and target placeholder sequence of the input decoder, 0 value is used for sequence placeholder to achieve one-step prediction of long sequences.

[0110]

[0111] (II) The DLinear model is optimized from the Transformer model. The DLinear model can reduce parameter adjustment and optimization, save time and reduce computational burden. It decomposes time series data into trend series X s and the remaining sequence X t , use a single-layer linear network to predict the two sequences obtained by decomposition, and output the results The prediction results H for two single-layer linear networks s and H t of superposition.

[0112] The basic calculation formula of the DLinear model is as follows:

[0113] H s =W s X s ∈R T×C , W s ∈R T×L #

[0114] H t =W t X t ∈R TXC , W t ∈R T×L #

[0115]

[0116] The input data is passed to the 2-layer GRU layer for forward calculation. The number of neurons is set to 64. The ReLU activation function is used for nonlinear mapping to obtain the output result. The Dropout layer is used to prevent overfitting, and MSE is used as the loss function. The optimizer uses the Adam optimization algorithm, and the learning rate is set to 0.005. The parameters of the Informer model and DLinear model use the default values.

[0117] The prediction starting point of the lithium battery SOH is set to 60% of the total cycle period, and the SOH value of the remaining 40% cycle period is used as the true value to compare with the prediction result.

[0118] 3) Use any of the GRU, Informer, and DLinear algorithms to establish a SOH prediction model, and calculate the two error indicators of MAE and RMSE of the prediction results to perform statistical analysis on the prediction accuracy.

[0119] Among them, the process of training and prediction using three algorithms, such as Figure 5 shown.

[0120] The error indicators MAE and RMSE are used to quantitatively analyze the estimation results of the three models. It can be obtained that on the five battery samples, the maximum RMSE and MAE results are 2.68% and 2.07% obtained by the Informer model, and the minimum RMSE and MAE results are 0.34% and 0.24% obtained by the GRU model. The two prediction error indicators of the three models are all within 3%, and they can capture the relationship between time series and track the aging trend of lithium batteries. They also have good generalization ability on battery samples with different charging and discharging strategies. The GRU model has the best error results in the battery samples. The prediction error RMSE and MAE of the GRU model and the DLinear model are both within 1%, and the difference in prediction performance between the two is small. According to the above data, the GRU model is preferred in this scheme to establish the SOH prediction model of the energy storage power station.

[0121] The number of charge and discharge cycles of each sample is different. The first 60% of the full life cycle is used as the training set and validation set. The training data volume of samples B1 and B2 is large and similar. The error results RMSE and MAE obtained by using the GRU model for prediction are 0.34%, 0.43% and 0.24%, 0.33% respectively. The prediction errors of these two samples are within 0.1%. The training data volume of B3, B4, and B5 is similar. The error results RMSE and MAE obtained by using the GRU model for prediction are 0.59%, 0.71%, 0.6% and 0.52%, 0.53%, 0.49% respectively. The RMSE error results of these three samples are within 0.15%, and the MAE error results are within 0.05%, indicating that the prediction errors of samples with similar training data volume are also similar. The prediction errors of B1 and B2 are smaller than those of B3, B4, and B5, indicating that the more training data, the better the learning effect of the model, the more accurate the prediction, and the better the evaluation index.

[0122] The GRU model is a deep learning model suitable for sequence data. It has the advantages of fast training speed, few parameters, and preventing gradient disappearance. It can reduce computing resources. Considering the large number of cells in the battery pack in the energy storage power station, it is urgent to establish an algorithm with low data requirements, strong generalization ability and small computational complexity. This application inputs the historical SOH value of the lithium battery into the GRU model to obtain the long-term prediction value of the battery SOH. The GRU model can use the historical data of SOH to more accurately predict the future SOH decay. The five selected battery samples have a large span of cycle times and different charging strategies. The prediction error of each sample can fully verify the generalization and accuracy of the proposed method.

[0123] In some embodiments of the present application, the present application also provides an electronic device, including a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps of the lithium battery health status prediction method provided in the above embodiment are implemented.

[0124] Furthermore, the present application also provides a computer-readable storage medium on which a computer management program is stored. When the computer management program is executed by a processor in an electronic device, the steps of the lithium battery health status prediction method provided in the above embodiment are implemented.

[0125] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0126] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, 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 generate 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 flowchart and / or block diagram. 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.

[0127] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate 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 A function specified in one or more boxes.

[0128] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0129] The foregoing description of specific exemplary embodiments of the present invention is for the purpose of illustration and demonstration. These descriptions are not intended to limit the present invention to the precise form disclosed, and it is clear that many changes and variations can be made based on the above teachings. The purpose of selecting and describing the exemplary embodiments is to explain the specific principles of the present invention and its practical application, so that those skilled in the art can realize and utilize various different exemplary embodiments of the present invention and various different selections and changes. The scope of the present invention is intended to be limited by the claims and their equivalents.

Claims

1. A method for predicting the health status of a lithium battery, characterized in that: The steps include: Obtain a historical data sequence of the health status of the lithium battery from a health status database, and process the historical data sequence to determine a historical data step length and a prediction step length; Extracting a training set, a validation set, and a test set from the historical data sequence according to the historical data step length and the prediction step length; Constructing a SOH prediction model, and training the SOH prediction model according to the training set, the validation set, and the test set; According to the SOH prediction request of the lithium battery, a data sequence of historical data steps is obtained from the health status database, the data sequence is input into the SOH prediction model, and the predicted SOH value is output.

2. The lithium battery health status prediction method according to claim 1, characterized in that: The method of obtaining a historical data sequence of the health status of the lithium battery from a health status database and processing the historical data sequence to determine a historical data step length and a prediction step length includes: From the health status database, obtain the health status values ​​at multiple evaluation time points and construct a historical data sequence; Analyze the historical data series through the autocorrelation function to generate an autocorrelation image; From the autocorrelation image, determine the time interval at which the autocorrelation coefficient is higher than a preset coefficient threshold, and determine the time interval as the historical data step length; The prediction step size is determined based on the proportional relationship between the historical data step size and the prediction step size.

3. The lithium battery health status prediction method according to claim 2, characterized in that: The generating of the autocorrelation image includes: the X-axis of the autocorrelation image is the number of time lags, and the Y-axis is the autocorrelation coefficient.

4. The lithium battery health status prediction method according to claim 3, characterized in that: The autocorrelation function comprises: The formula for the autocorrelation coefficient at lag k is Where N is the length of the time series, x t is the time series value at time t, is the mean of the time series.

5. The lithium battery health status prediction method according to claim 1, characterized in that: The extracting a training set, a validation set, and a test set from the historical data sequence according to the historical data step length and the prediction step length includes: In the historical data sequence, the first 60% of the total cycle period of each battery sample is determined as the training set and validation set, and the remaining 40% of the data is used as the test set; among them, the training set accounts for 50% and the validation set accounts for 10%.

6. The lithium battery health status prediction method according to claim 5, characterized in that: The step of extracting a training set from the historical data sequence comprises: The sliding window slides gradually on the data sequence of the sample data set, sliding one unit each time, and the data within the sliding window constitutes a training set.

7. The lithium battery health status prediction method according to claim 1, characterized in that: The constructing of the SOH prediction model includes: selecting any one of the GRU algorithm, the Informer algorithm, and the DLinear algorithm to construct the SOH prediction model.

8. The lithium battery health status prediction method according to claim 7, characterized in that: The step of training the SOH prediction model according to the training set, the validation set, and the test set includes: Training the SOH prediction model architecture according to the training set to obtain an initial SOH prediction model; The initial SOH prediction model is verified and evaluated according to the verification set and the test set to obtain a SOH prediction model that can be used for health status prediction.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the lithium battery health status prediction method described in any one of claims 1-8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the lithium battery health status prediction method described in any one of claims 1 to 8 are implemented.

Images