Method, device and equipment for predicting remaining service life of battery and storage medium

Abstract

The invention relates to the technical field of battery management, in particular to a battery remaining service life prediction method, device and equipment and a storage medium, and aims to solve the problem of how to construct an applicable neural network model based on the correlation among capacity, SOH and RUL to accurately and quickly predict the RUL. In order to achieve the purpose, the method comprises the steps of predicting a capacity value through a trained capacity prediction model based on a battery data sequence, predicting an SOH value through a trained SOH prediction model based on the battery data sequence and the predicted capacity value, and predicting an RUL value through a trained RUL prediction model based on the battery data sequence, the predicted capacity value and the predicted SOH value. The capacity prediction model is constructed based on the Bi LSTM network, so that forward and backward data information can be considered at the same time when the battery data sequence is processed, and the prediction accuracy is remarkably improved; and the prediction results of the capacity and the SOH are used for training the RUL prediction model, so that the correlation among model parameters can be improved, and the robustness of the whole model is improved.

Description

Technical Field

[0001] The present application relates to the field of battery management technology, and specifically to a method, device, equipment and storage medium for predicting remaining battery life. Background Art

[0002] As the current mainstream energy storage product, lithium batteries have been widely used in daily life and industrial production such as machinery and equipment, transportation vehicles, and power grid energy storage systems. In battery management, the decision of the battery management system (BMS) mainly depends on the prediction of battery status such as the state of charge (SOC) and the state of health (SOH). At the same time, considering that battery aging may cause safety accidents, the prediction of the remaining useful life (RUL) of the battery is equally important. The RUL can help users determine whether the current battery can work in a safe state in the future.

[0003] At present, although the prediction method based on data-driven can obtain relatively accurate prediction results, the algorithms usually adopted by this method, such as convolutional neural network, BP neural network, support vector machine, etc., often have complex network structure, many parameters, long training time, gradient disappearance / explosion and other problems; and the existing prediction method based on data-driven, mostly predicts a single target, rarely considers using the predicted results of a certain category to train and / or predict the results of other categories, that is, ignores the correlation between capacity, SOC, SOH, RUL and other data. Therefore, how to build a practical neural network model based on the correlation between capacity, SOH, and RUL, and use the predicted results of a certain category to train and / or predict the results of other categories, so as to achieve accurate and rapid prediction of RUL, has become an urgent problem to be solved.

[0004] Accordingly, the art needs a new battery remaining service life prediction solution to solve the above problems. Summary of the invention

[0005] In order to overcome the above-mentioned defects, the present application is proposed to solve or at least partially solve the technical problem of how to construct a practical neural network model based on the correlation between capacity, SOH and RUL to accurately and quickly predict RUL.

[0006] In a first aspect, a method for predicting remaining battery life is provided, the method comprising: Acquire a first number of battery data groups to form a battery data sequence, wherein the battery data groups include voltage data, current data, and temperature data; Based on the battery data sequence, predicting the capacity value of the battery by using a trained capacity prediction model; Based on the battery data sequence and the predicted capacity value, predicting the SOH value of the battery through a trained SOH prediction model; Based on the battery data sequence, the predicted capacity value and the predicted SOH value, the RUL value of the battery is predicted by using a trained RUL prediction model.

[0007] In a technical solution of the above-mentioned method for predicting the remaining service life of a battery, the method further includes: Constructing the capacity prediction model based on the BiLSTM network; Constructing the SOH prediction model based on the LSTM network; The RUL prediction model is constructed based on the LSTM network.

[0008] In a technical solution of the above-mentioned method for predicting the remaining service life of a battery, the method further includes: Constructing the capacity prediction model based on at least one first feature extraction layer, a first fully connected layer and a first output layer, wherein the first feature extraction layer is constructed based on a BiLSTM network; Constructing the SOH prediction model based on at least one second feature extraction layer, a second fully connected layer and a second output layer, wherein the second feature extraction layer is constructed based on an LSTM network; The RUL prediction model is constructed based on at least one third feature extraction layer, a third fully connected layer and a third output layer, wherein the third feature extraction layer is constructed based on an LSTM network.

[0009] In a technical solution of the above-mentioned battery remaining service life prediction method, the battery remaining service life prediction model constructed based on the capacity prediction model, the SOH prediction model and the RUL prediction model also includes a first input reconstruction unit and a second input reconstruction unit, and the method also includes: Inputting the battery data sequence and the predicted capacity value into the first input reconstruction unit for reconstruction to obtain a first reconstructed battery data sequence, wherein the first reconstructed battery data sequence is input data of the SOH prediction model; The battery data sequence, the predicted capacity value and the predicted SOH value are input into the second input reconstruction unit for reconstruction to obtain a second reconstructed battery data sequence, wherein the second reconstructed battery data sequence is input data of the RUL prediction model.

[0010] In a technical solution of the above-mentioned method for predicting the remaining service life of a battery, the method further includes: Adding the predicted capacity value to each of the battery data groups in the battery data sequence to obtain the first reconstructed battery data sequence, wherein the first reconstructed battery data group in the first reconstructed battery data sequence includes the voltage data, the current data, the temperature data and the predicted capacity value; The predicted capacity value and the predicted SOH value are added to each of the battery data groups in the battery data sequence to obtain the second reconstructed battery data sequence, wherein the second reconstructed battery data group in the second reconstructed battery data sequence includes the voltage data, the current data, the temperature data, the predicted capacity value and the predicted SOH value.

[0011] In a technical solution of the above-mentioned method for predicting the remaining service life of a battery, the method further includes: Divide the sample data set based on the preset sliding window and sliding step size to obtain the training data set; Based on the training data set, training the capacity prediction model to obtain the trained capacity prediction model; Based on the trained capacity prediction model, obtaining a predicted capacity training data set corresponding to the training data set; Based on the training data set and the predicted capacity training data set, construct a first reconstruction training data set; Based on the first reconstructed training data set, training the SOH prediction model to obtain the trained SOH prediction model; Based on the trained SOH prediction model, obtaining a predicted SOH training data set corresponding to the training data set; constructing a second reconstructed training data set based on the training data set, the predicted capacity training data set and the predicted SOH training data set; Based on the second reconstructed training data set, the RUL prediction model is trained to obtain the trained RUL prediction model.

[0012] In a technical solution of the above-mentioned method for predicting the remaining service life of a battery, the method further includes: The first number is set equal to a window size of the sliding window.

[0013] In a second aspect, a battery remaining service life prediction device is provided, the device comprising: A data acquisition module, the data acquisition module is configured to acquire a first number of battery data groups to form a battery data sequence, wherein the battery data group includes voltage data, current data and temperature data; A remaining useful life prediction module, wherein the remaining useful life prediction module is configured to perform the following operations: Based on the battery data sequence, predicting the capacity value of the battery by using a trained capacity prediction model; Based on the battery data sequence and the predicted capacity value, predicting the SOH value of the battery through a trained SOH prediction model; Based on the battery data sequence, the predicted capacity value and the predicted SOH value, the RUL value of the battery is predicted by using a trained RUL prediction model.

[0014] In a third aspect, a computer device is provided, comprising a processor and a storage device, wherein the storage device is suitable for storing a plurality of program codes, and the program codes are suitable for being loaded and run by the processor to execute the battery remaining service life prediction method described in any one of the above technical solutions.

[0015] In a fourth aspect, a storage medium is provided, in which a plurality of program codes are stored, wherein the program codes are suitable for being loaded and run by a processor to execute the battery remaining service life prediction method described in any one of the above technical solutions.

[0016] The above one or more technical solutions of the present application have at least one or more of the following beneficial effects: in implementing the battery remaining service life prediction technical solution provided by the present application, the BiLSTM network model is used in the capacity prediction link, so that the capacity prediction model can consider both forward and backward data information when processing time series data (battery data sequence), thereby more accurately capturing the contextual dependencies in the sequence and significantly improving the prediction accuracy; the capacity prediction results and the SOH value prediction results are used for the training of the RUL prediction model, which can improve the correlation between the model parameters in the battery remaining service life prediction model, increase the robustness of the overall model, and make the battery remaining service life prediction model more suitable for actual prediction scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The disclosure of the present application will become more easily understood with reference to the accompanying drawings. It is easy for those skilled in the art to understand that these drawings are only for illustrative purposes and are not intended to limit the scope of protection of the present application.

[0018] Figure 1 It is a schematic diagram of the main steps of a method for predicting remaining battery service life according to an embodiment of the present application.

[0019] Figure 2 It is a structural schematic diagram of a battery remaining service life prediction model according to an embodiment of the present application.

[0020] Figure 3 It is a detailed structural diagram of a battery remaining service life prediction model according to an embodiment of the present application.

[0021] Figure 4 It is a schematic diagram of the main steps of a battery remaining service life prediction model training method according to an embodiment of the present application.

[0022] Figure 5 It is an example diagram of a comparison curve of the training loss and the validation loss of a capacity prediction model according to an embodiment of the present application, and a comparison curve of the true value and the predicted value.

[0023] Figure 6 It is an example diagram of a comparison curve of the training loss and the validation loss of the SOH prediction model according to an embodiment of the present application, and a comparison curve of the true value and the predicted value.

[0024] Figure 7 It is an example diagram of a comparison curve of the training loss and the validation loss of the RUL prediction model according to an embodiment of the present application, and a comparison curve of the true value and the predicted value.

[0025] Figure 8 is a schematic diagram of a battery remaining service life prediction device according to an embodiment of the present application.

[0026] Fig. 9 is a schematic diagram of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution of the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0028] Those skilled in the art should understand that these implementations are only used to explain the technical principles of the present application and are not intended to limit the protection scope of the present application. Those skilled in the art can make adjustments to them as needed to adapt to specific application scenarios.

[0029] See attached Figure 1 , Figure 1 Schematic diagram of the main steps of the method for predicting the remaining service life of a battery in an embodiment of the present application. The main steps of the method for predicting the remaining service life of a battery in an embodiment of the present application include: Step S101: Acquire a first number of battery data groups to form a battery data sequence; Step S102: Based on the battery data sequence, predict the capacity value of the battery through the trained capacity prediction model; Step S103: Based on the battery data sequence and the predicted capacity value, the SOH value of the battery is predicted by using the trained SOH prediction model; Step S104: Based on the battery data sequence, the predicted capacity value and the predicted SOH value, the RUL value of the battery is predicted by using the trained RUL prediction model.

[0030] In step S101, during the charging process or the discharging process, according to the time sequence, a first number of battery data groups can be obtained through the BMS of the battery; or a first number of battery data groups can be directly obtained through a voltage sensor, a current sensor, and a temperature sensor, wherein the battery data group includes voltage data (voltage value), current data (current value), and temperature data (temperature value) of the battery. In other words, a first number of voltage data, a first number of current data, and a first number of temperature data are obtained.

[0031] According to the first number of battery data groups arranged in time order, a battery data sequence is obtained. As an example, the first number of voltage data is v1, v2, ..., v n ,…,v N , the first number of current data is i1, i2, ..., i n ,…,i N , the first number of temperature data is T1, T2, ..., T n ,…,T N , where v n 、i n and T n represents the battery data acquired at the nth moment, where N is the first number. The time sequence is usually in the order from the front to the back, for example, the moment corresponding to v1 is earlier than the moment corresponding to v2. The battery data sequence according to the time sequence is expressed as: {{v1,i1,T1},{v2,i2,T2},…,{v n ,i n ,T n},…,{v N ,i N ,T N}} It can be written as {X1,X2,…,X n ,…,X N}, where the battery data set X n = {v n ,i n ,T n}, n = 1, ..., N, that is, each n in the battery data sequence corresponds to a historical moment, and each historical moment corresponds to a set of voltage data, current data and temperature data.

[0032] It should be noted that, for the charging and discharging processes of the battery, the current has a direction, and the current direction is indicated by "+" and "-". For example, -I indicates the discharge current, and +I indicates the charge current.

[0033] Continue reading Figure 2 , combined with Figure 2 illustrate Figure 1 The main steps of the battery remaining service life prediction method in . Figure 2 It is a structural schematic diagram of a battery remaining service life prediction model according to an embodiment of the present application.

[0034] like Figure 2 As shown, the battery remaining service life prediction model includes a capacity prediction model 1, a SOH prediction model 2, a RUL prediction model 3, a first input reconstruction unit 4 and a first input reconstruction unit 5.

[0035] In step S102, {X1, X2, …, X n ,…,X N} is the input data of the battery remaining service life prediction model. First, the battery data sequence {X1, X2, …, X n ,…,X N The trained capacity prediction model 1 is input in sequence, and the capacity value of the battery is predicted by the trained capacity prediction model 1 to obtain the predicted capacity value C.

[0036] It should be noted that the capacity value described in this application refers to the maximum capacity value of the battery corresponding to the current charging process or discharging process, that is, the current maximum capacity value. Using the current maximum capacity value to predict the SOH value can make the predicted SOH value more accurate; at the same time, the current maximum capacity value can be further used to calculate the SOC value, partially eliminating the impact of the SOH value change on the SOC calculation.

[0037] In step S103, the battery data sequence {X1, X2, ..., X n ,…,X N} and the predicted capacity value C are all input into the first input reconstruction model 4 for reconstruction to obtain a first reconstructed battery data sequence. The first reconstructed battery data sequence is both the output data of the first input reconstruction model 4 and the input data of the SOH prediction model.

[0038] Considering that the capacity value of the battery can be considered to be the same during a charge and discharge process, each battery data group X n = {v n ,i n ,T n The predicted capacity values ​​C corresponding to the battery data are all the same. The predicted capacity values ​​C are added to each battery data group of the battery data sequence to obtain a first reconstructed battery data sequence.

[0039] The first reconstructed battery data group in the first reconstructed battery data sequence includes voltage data, current data, temperature data and predicted capacity value. The first reconstructed battery data sequence can be expressed as: {{v1,i1,T1,C},{v2,i2,T2,C},…,{v n ,i n ,T n ,C},…,{v N ,i N ,T N ,C}} Denoted as {X′1,X′2,…,X′ n ,…,X′ N}, where the first reconstructed battery data set X′ n = {v n ,i n ,T n ,C}, n=1,…,N.

[0040] The first reconstructed battery data sequence {X′1, X′2, …, X′ n ,…,X′ N The trained SOH prediction model 2 is input in sequence, and the SOH value of the battery is predicted by the trained SOH prediction model 2 to obtain the predicted SOH value H.

[0041] In step S104, the battery data sequence {X1, X2, ..., X n ,…,X N}, the predicted capacity value C and the predicted SOH value H are all input into the second input reconstruction model 5 to obtain the second reconstructed battery data sequence. The second reconstructed battery data sequence is both the output data of the second input reconstruction model 5 and the input data of the RUL prediction model.

[0042] Also considering that in a charge and discharge process, the battery capacity and battery SOH value can be considered the same, so each group of battery parameters X n = {v n ,i n ,T n The predicted capacity value C and SOH value H corresponding to} are the same. The predicted capacity value C and the predicted SOH value H are added to each battery data group of the battery data sequence to obtain a second reconstructed battery data sequence.

[0043] The second reconstructed battery data group in the second reconstructed battery data sequence includes voltage data, current data, temperature data, predicted capacity value and predicted SOH value. The second reconstructed battery data sequence can be expressed as: {{v1,i1,T1,C,H},{v2,i2,T2,C,H},…,{v n ,i n ,T n ,C,H},…,{v N ,i N ,T N ,C,H}} Denoted as {X″1,X″2,…,X″ n ,…,X″ N}, where the second reconstructed battery data set X″ n = {v n ,i n ,T n ,C,H}, n=1,…,N.

[0044] The second reconstructed battery data sequence {X″1, X″2, …, X″ n ,…,X″ N} Input the trained RUL prediction model 3 in sequence, and use the trained RUL prediction model 3 to predict the RUL value of the battery to obtain the battery data sequence {X1, X2, …, X n ,…,X N}The predicted RUL value corresponding to .

[0045] Continue reading Figure 3 , attached Figure 3 is a detailed structural diagram of a battery remaining service life prediction model according to an embodiment of the present application, wherein: Figure 3 (a) is a detailed structural diagram of the capacity prediction model. Figure 3 (b) is a detailed structural diagram of the SOH prediction model. Figure 3 (c) is a detailed structural diagram of the RUL prediction model.

[0046] See also Figure 3 (a) A capacity prediction model is constructed based on a Bidirectional Long Short-Term Memory (BiLSTM) network. Specifically, the capacity prediction model includes at least one first feature extraction layer, a first fully connected layer, and a first output layer, wherein the first feature extraction layer is constructed based on the BiLSTM network.

[0047] In this application, the BiLSTM network model is used in the capacity prediction link, so that the capacity prediction model can simultaneously consider forward and backward data information when processing time series data (battery data sequence), thereby more accurately capturing the contextual dependencies in the sequence and significantly improving the prediction accuracy.

[0048] Extract the battery data sequence {X1,X2,…,X n ,…,X N}, the number of BiLSTM neurons in each first feature extraction layer is the second number, that is, the output dimension of each first feature extraction layer is 1*the second number, where the second number is less than or equal to the first number.

[0049] The first fully connected layer is used to fuse the second quantity of feature data to obtain a third quantity of feature data after dimension compression, that is, the output dimension of the first fully connected layer is 1*the third quantity, where the third quantity is smaller than the second quantity.

[0050] The first output layer is used to perform a weighted sum operation on the third amount of feature data to obtain a predicted capacity value C corresponding to the battery data sequence.

[0051] See also Figure 3 (b) A SOH prediction model is constructed based on an LSTM network (Long Short-Term Memory). Specifically, the SOH prediction model includes at least one second feature extraction layer, a second fully connected layer, and a second output layer, wherein the second feature extraction layer is constructed based on the LSTM network.

[0052] Through the LSTM network, the first reconstructed battery data sequence {X′1, X′2, …, X′ n ,…,X′ N}, the number of LSTM neurons in each second feature extraction layer is the fourth number, that is, the output dimension of each second feature extraction layer is 1*fourth number. In an embodiment of the present application, the fourth number is equal to the second number.

[0053] The second fully connected layer is used to fuse the feature data of the fourth quantity to obtain the fifth quantity feature data after dimension compression, that is, the output dimension of the second fully connected layer is 1*fifth quantity, where the fifth quantity is smaller than the fourth quantity.

[0054] The second output layer is used to perform a weighted sum operation on the fifth number of feature data to obtain a predicted SOH value H corresponding to the battery data sequence.

[0055] The RUL prediction model is constructed based on the LSTM network. Specifically, the RUL prediction model includes at least one third feature extraction layer, a third fully connected layer and a third output layer, wherein the third feature extraction layer is constructed based on the LSTM network.

[0056] Through the LSTM network, the second reconstructed battery data sequence {X″1,X″2,…,X″ n ,…,X″N}, the number of LSTM neurons in each third feature extraction layer is the sixth number, that is, the output dimension of each third feature extraction layer is 1*sixth number. In an embodiment of the present application, the sixth number is equal to the second number.

[0057] The third fully connected layer is used to fuse the feature data of the sixth quantity to obtain the feature data of the seventh quantity after dimension compression, that is, the output dimension of the third fully connected layer is 1*seventh quantity, where the seventh quantity is smaller than the sixth quantity.

[0058] The third output layer is used to perform a weighted sum operation on the seventh number of feature data to obtain a predicted RUL value corresponding to the battery data sequence.

[0059] Continue reading Figure 4 , Figure 4 Schematic diagram of the main steps of the battery remaining service life prediction model training method in the embodiment of the present application. The main steps of the battery remaining service life prediction model training method in the embodiment of the present application include: Step S401: Divide the sample data set based on a preset sliding window and sliding step size to obtain a training data set; Step S402: training a capacity prediction model based on a training data set to obtain a trained capacity prediction model; Step S403: Based on the trained capacity prediction model, obtain a predicted capacity training data set corresponding to the training data set; Step S404: constructing a first reconstructed training data set based on the training data set and the predicted capacity training data set; Step S405: training a SOH prediction model based on the first reconstructed training data set to obtain a trained SOH prediction model; Step S406: Based on the trained SOH prediction model, obtain a predicted SOH training data set corresponding to the training data set; Step S407: constructing a second reconstructed training data set based on the training data set, the predicted capacity training data set and the predicted SOH training data set; Step S408: Based on the second reconstructed training data set, train the RUL prediction model to obtain a trained RUL prediction model.

[0060] Considering that the battery remaining service life prediction model includes three relatively independent sub-models in structure (capacity prediction model, SOH prediction model and RUL prediction model), but these three sub-models have certain dependencies in data (SOH prediction model depends on the output data of capacity prediction model, and RUL prediction model depends on the output data of both capacity prediction model and SOH prediction model). Therefore, based on the data processing flow in the battery remaining service life prediction model and the data dependency between the three sub-models, when the battery remaining service life prediction model is constructed, a training strategy is adopted to train the three sub-models independently in sequence, that is, first train the capacity prediction model, then train the SOH prediction model, and finally train the RUL prediction model.

[0061] In the sample data set, the voltage data are v1,v2,…,v m ,…, the current data is i1,i2,…,i m ,…, the temperature data is T1, T2,…, T m ,…, where {v m ,i m ,T m} represents the mth sample data group. And the sample data set also includes information such as the number of cycles, charge or discharge identification, capacity value, SOH value, RUL value, etc. corresponding to the mth group of sample data.

[0062] In an embodiment of the present application, the preset window size of the sliding window is 100, and the sliding step size is 50. Accordingly, when the trained battery remaining service life prediction model is used, the first number needs to be set to the same as the window size of the sliding window, that is, the first number is 100.

[0063] In step S401, the sample data set is divided based on the sliding window (100) and the sliding step size (50) to obtain a training data set, that is, each training data set includes 100 sample data sets.

[0064] As an example, the first training data group of the training data set obtained by dividing the sample data set according to the sliding window (100) is: {{v1,i1,T1},{v2,i2,T2},…,{v 100 ,i 100 ,T 100}}, the second training data set obtained by dividing the sample data set according to the sliding window (W=100) and the sliding step size (S=50) is: {{v 51 ,i 51 ,T 51},{v 52 ,i 52 ,T 52},…,{v 150 ,i 150,T 150}}, and so on, we get the training data set {Y1,Y2,…,Y p ,…,Y P}, where the pth training data set can be expressed as Y p ={{v (p-1)*S+1 ,i (p-1)*S+1 ,T (p-1)*S+1},…,{v (p-1)*S+W ,i (p-1)*S+W ,T (p-1)*S+W}},p=1,…,P.

[0065] It should be noted that, since the capacity value, SOH value and RUL value of the battery can be considered as a fixed value during each charging or discharging process, accordingly, when dividing the sample data set according to the sliding window and the sliding step, it is necessary to make the data in the sliding window be sample data in the same charging or discharging process. Accordingly, each training data group corresponds to a set of capacity value, SOH value and RUL value. As an example, the pth training data group Y p The 100 sample data groups in the sample data set all correspond to the same group capacity value C in the sample data set p 、SOH value H p and RUL value RUL p , that is, the pth training data set Y p With {C p ,H p ,RUL p}. Furthermore, we can get the training data set {Y1,Y2,…,Y p ,…,Y P} corresponding to the capacity value sample data set {C1,C2,…,C p ,…,C P}、SOH value sample data set {H1,H2,…,H p ,…,H P} and RUL value sample data set {RUL1, RUL2, …, RUL p ,…,RUL P}.

[0066] In step S402, the training data set {Y1, Y2, ..., Y p ,…,Y P} as the input data of the capacity prediction model, and based on the capacity value sample data set {C1,C2,…,C p ,…,C P}, train the capacity prediction model to obtain a trained capacity prediction model.

[0067] In step S403, the training data set {Y1, Y2, ..., Y p ,…,Y P} is used as the input data of the trained capacity prediction model to predict the predicted capacity training data set {C′1,C′2,…,C′ p ,…,C′ P}.

[0068] In step S404, based on the training data set {Y1, Y2, ..., Y p ,…,Y P} and the prediction capacity training dataset {C′1,C′2,…,C′ p ,…,C′ P}, construct the first reconstruction training data set {Y′1,Y′2,…,Y′ p ,…,Y′ P}.

[0069] Specifically, C′ p Add to Y p In each sample data set in , the pth reconstructed training data set in the first reconstructed training data set is obtained: Y′ p ={{v (p-1)*S+1 ,i (p-1)*S+1 ,T (p-1)*S+1 ,C′ p},…,{v (p-1)*S+W ,i (p-1)*S+W ,T (p-1)*S+W ,C′ p}}, p=1,…,P.

[0070] In step S405, the first reconstructed training data set {Y′1, Y′2, …, Y′ p ,…,Y′ P} as the input data of the SOH prediction model, and based on the SOH value sample data set {H1,H2,…,H p ,…,H P}, train the SOH prediction model to obtain a trained capacity prediction model.

[0071] In step S406, the first reconstructed training data set {Y′1, Y′2, …, Y′ p ,…,Y′ P} is used as the input data of the trained SOH prediction model, and the training data set {Y1,Y2,…,Y p ,…,Y P} corresponding to the predicted SOH training dataset {H′1,H′2,…,H′ p ,…,H′ P}.

[0072] In step S407, based on the training data set {Y1, Y2, ..., Y p ,…,Y P}、Prediction capacity training data set {C′1,C′2,…,C′ p ,…,C′ P} and predict SOH training data set {H′1,H′2,…,H′ p ,…,H′ P}, construct the second reconstruction training data set {Y″1,Y″2,…,y″ p ,…,Y″ P}.

[0073] Specifically, C′ p and H′ p Add to Y p In each sample data set in, obtain the pth reconstructed training data set in the second reconstructed training data set: Y″ p ={{v (p-1)*S+1 ,i (p-1)*S+1 ,T (p-1)*S+1 ,C′ p ,H′ p},…,{v (p-1)*S+W ,i (p-1)*S+W ,T (p-1)*S+W ,C′ p ,H′ p}}, p=1,…,P.

[0074] In this application, the capacity prediction results and the SOH value prediction results are used for the training of the RUL prediction model, which can improve the correlation between the model parameters in the battery remaining service life prediction model, increase the robustness of the overall model, and make the battery remaining service life prediction model more suitable for actual prediction scenarios.

[0075] In another embodiment, the first reconstructed training data set {Y′1, Y′2, …, Y′ p ,…,Y′ P} and predict SOH training data set {H′1,H′2,…,H′ p ,…,H′ P}Construct the second reconstruction training data set {Y″1,Y″2,…,Y″ p ,…,Y″ P}.

[0076] In step S408, the second reconstructed training data set {Y″1, Y″2, …, Y″ p ,…,Y″ P} is the input data of the RUL prediction model, and is based on the RUL value sample data set {RUL1, RUL2, …, RULp ,…,RUL P}, train and obtain the trained RUL prediction model.

[0077] In another embodiment, during the RUL prediction model training process, a preset proportion of LSTM neurons can be randomly closed so that the RUL prediction model does not rely on a specific part of neurons during training, thereby improving the generalization ability of the model and reducing the risk of overfitting of the model.

[0078] It should be noted that the training methods and processes of the capacity prediction model constructed based on the BiLSTM network, the SOH prediction model constructed based on the LSTM network, and the RUL prediction model constructed based on the LSTM network are all well known to technical personnel in this field. The general training methods of the BiLSTM network and the LSTM network can be referred to. The specific training methods and processes of the above three prediction models will not be repeated here.

[0079] In one embodiment of the present application, the window size of the sliding window used to obtain the training data set is 100, and the sliding step size is 50; accordingly, the number of battery data groups in the battery data sequence serving as input data for the battery remaining service life prediction model is also 100 (the first number).

[0080] The capacity prediction model includes three first feature extraction layers constructed by BiLSTM networks, and the number of BiLSTM neurons in each first feature extraction layer is 100 (the second quantity), and the dimension of the output data of the first fully connected layer is 1*50 (the third quantity).

[0081] The SOH prediction model includes 3 second feature extraction layers constructed by LSTM networks, and the number of LSTM neurons in each second feature extraction layer is 100 (the fourth quantity), and the dimension of the output data of the second fully connected layer is 1*50 (the fifth quantity).

[0082] The RUL prediction model also includes 3 layers of the third feature extraction layer constructed by the LSTM network, and the number of LSTM neurons in each layer of the third feature extraction layer is 100 (the sixth quantity), and the dimension of the output data of the third fully connected layer is 1*50 (the seventh quantity).

[0083] A lithium-ion battery with a battery model of 18650 and a rated capacity of 2Ah is selected as the processing object of the method of this application, and the lithium-ion battery data set disclosed by NASA is used as the sample data set. Among them, 80% of the sample data groups in the sample data set are used to generate the training data set, and the remaining 20% ​​of the sample data groups are used to generate the test data set.

[0084] It should be noted that the method for generating the test data set is the same as that for generating the training data set. Both use a sliding window with a window size of 100 and a sliding step size of 50 to divide the sample data groups. For the specific division method, please refer to the aforementioned method for generating the training data set, which will not be repeated here.

[0085] Reference Figure 4 The battery remaining service life prediction model training method shown uses a training data set to train a capacity prediction model, a SOH prediction model, and a RUL prediction model; during the training process, a test data set is used to verify each model, and the model parameters are optimized to obtain a trained battery remaining service life prediction model (including a trained capacity prediction model, a trained SOH prediction model, and a trained RUL prediction model).

[0086] Reference Figure 1 The battery remaining service life prediction method shown uses a test data set to obtain the predicted value of the RUL value (and the predicted values ​​of the capacity value and the SOH value) through a trained battery remaining service life prediction model, and evaluates the trained capacity prediction model, SOH prediction model and RUL prediction model.

[0087] The comparison curves of the training loss (Train Loss) and validation loss (Validation Loss) of the capacity prediction model are as follows: Figure 5 As shown in (a), the comparison curve of the capacity true value (True Values) and predicted value (Predicted Values) of the capacity prediction model is as follows Figure 5 (b) as shown.

[0088] The comparison curves of the training loss (Train Loss) and validation loss (Validation Loss) of the SOH prediction model are as follows: Figure 6 As shown in (a), the comparison curve of the capacity true value (True Values) and predicted value (Predicted Values) of the capacity prediction model is as follows Figure 6 (b) as shown.

[0089] The comparison curves of the training loss (Train Loss) and validation loss (Validation Loss) of the RUL prediction model are as follows: Figure 7 As shown in (a), the comparison curve of the capacity true value (True Values) and predicted value (Predicted Values) of the capacity prediction model is as follows Figure 7 (b) as shown.

[0090] from Figure 5 (a) Figure 6 (a) and Figure 7(a) It can be seen that with the increase in the number of training epochs, the training loss value and the verification loss value decrease rapidly, which means that the battery remaining service life prediction model of the present application has high training efficiency and excellent training effect.

[0091] from Figure 5 (b) Figure 6 (b) and Figure 7 (b) It can be seen that the true value and the predicted value fit well, which means that the battery remaining service life prediction model of the present application has the advantage of high accuracy.

[0092] It should be noted that although the method of the present application is described using 18650 lithium-ion batteries as an example, those skilled in the art will appreciate that the method of the present application is also applicable to lithium-ion batteries of other specifications and models, as well as other types of energy storage batteries, such as nickel batteries, sodium-sulfur batteries, etc.

[0093] When the method of the present application is applied to batteries of different specifications, models and / or types, it is necessary to use the sample data set of the battery to retrain the battery remaining service life prediction model to obtain a trained battery remaining service life prediction model suitable for the battery. The sample data set can be a public sample data set, a sample data set provided by a battery manufacturer, or a sample data set obtained by the user through laboratory calibration and actual measurement.

[0094] It should be noted that in the embodiment of the present application, although a capacity prediction model is constructed based on a BiLSTM network, a SOH prediction model is constructed based on an LSTM network, and a RUL prediction model is constructed based on an LSTM network. However, those skilled in the art will understand that in order to achieve the effect of the present application, other network models such as a CNN network and a Transformer network may be used to construct one or more of the prediction models. As an example, a capacity prediction model is constructed based on a Transformer network, a SOH prediction model is constructed based on a CNN network, and a RUL prediction model is constructed based on a Transformer network; for another example, a capacity prediction model is constructed based on a Transformer network, a SOH prediction model is constructed based on an LSTM network, and a RUL prediction model is constructed based on a BiLSTM network, etc. Those skilled in the art may select a suitable network model to implement the technical solution of the present application according to actual conditions.

[0095] In addition, the number of the first feature extraction layer n in the capacity prediction model, the number of the second feature extraction layer m in the SOH prediction model, and the number of the third feature extraction layer p in the RUL prediction model, as well as the dimensions of the output data of the feature extraction layer and the fully connected layer of each prediction model, those skilled in the art can also select the appropriate number of feature extraction layers of each prediction model, as well as the dimensions of the output data of the feature extraction layer and the fully connected layer of each prediction model according to actual conditions. These adjusted solutions are equivalent to the technical solutions described in this application, and therefore will also fall within the protection scope of this application.

[0096] Furthermore, the present application also provides a device for predicting the remaining service life of a battery.

[0097] See attached Figure 8 , Figure 8 Schematic diagram of a battery remaining service life prediction device according to an embodiment of the present application. Figure 8 As shown, the battery remaining service life prediction device in the embodiment of the present application includes a data acquisition module 81 and a remaining service life prediction module 82. The data acquisition module 81 is configured to perform the operation in step S101. The remaining service life prediction module 82 includes Figure 2 The battery remaining service life prediction model shown in the figure includes a capacity prediction model 1, a SOH prediction model 2, a RUL prediction model 3, a first input reconstruction unit 4 and a first input reconstruction unit 5, wherein the detailed structures of the capacity prediction model 1, the SOH prediction model 2 and the RUL prediction model 3 are as shown in Figure 3 The remaining useful life prediction module 82 is configured to perform the operations in step S102 to step S104. For specific technical details, please refer to the method section of the embodiment of the present application.

[0098] Furthermore, the present application also provides a computer device.

[0099] See attached Fig. 9 , Fig. 9 is a schematic diagram of a computer device according to an embodiment of the present application. Fig. 9 As shown, the computer device in the embodiment of the present application mainly includes a storage device 91 and a processor 92. The storage device 91 can be configured to store a program for executing the battery remaining service life prediction method of the above method embodiment, and the processor 92 can be configured to execute the program in the storage device, which includes but is not limited to the program for executing the battery remaining service life prediction method of the above method embodiment. For ease of explanation, only the part related to the embodiment of the present application is shown. For specific technical details not disclosed, please refer to the method part of the embodiment of the present application.

[0100] In the embodiment of the present application, the computer device may be a control device device including various electronic devices. In some possible implementations, the computer device may include multiple storage devices 91 and multiple processors 92. The program for executing the battery remaining service life prediction method of the above method embodiment can be divided into multiple subprograms, and each subprogram can be loaded and run by multiple processors to execute different steps of the battery remaining service life prediction method of the above method embodiment. Specifically, each subprogram can be stored in different storage devices 91, and each processor 92 can be configured to execute the program in one or more storage devices 91 to jointly implement the battery remaining service life prediction method of the above method embodiment, that is, each processor 92 executes different steps of the battery remaining service life prediction method of the above method embodiment, to jointly implement the battery remaining service life prediction method of the above method embodiment.

[0101] The above-mentioned multiple processors 92 may be processors deployed on the same device. For example, the above-mentioned computer device may be a high-performance device composed of multiple processors, and the above-mentioned multiple processors 92 may be processors configured on the high-performance device. In addition, the above-mentioned multiple processors 92 may also be processors deployed on different devices. For example, the above-mentioned computer device may be a server cluster, and the above-mentioned multiple processors 92 may be processors on different servers in the server cluster.

[0102] Furthermore, the present application also provides a storage medium.

[0103] In a storage medium embodiment according to the present application, the storage medium may be configured to store a program for executing the battery remaining service life prediction method of the above method embodiment, and the program may be loaded and run by the processor to implement the above battery remaining service life prediction method. For ease of explanation, only the parts related to the embodiment of the present application are shown. For specific technical details not disclosed, please refer to the method part of the embodiment of the present application. The storage medium may be a storage device formed by various electronic devices. As an example, the storage medium in the embodiment of the present application is a non-temporary storage medium.

[0104] Those skilled in the art should be able to realize that, although the various steps in the above embodiments are described in a specific order, those skilled in the art can understand that, in order to achieve the effect of the present application, different steps do not have to be executed in such an order, and they can be executed simultaneously (in parallel) or in other orders. These adjusted schemes are equivalent technical schemes to the technical schemes described in the present application, and therefore will also fall within the scope of protection of the present application.

[0105] Those skilled in the art should be able to appreciate that the method steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of electronic hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0106] It should be noted that the terms "first", "second", "third", "fourth", etc. in the specification and claims of the present invention and the above-mentioned drawings are only used to distinguish similar objects, rather than to describe or indicate a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein.

[0107] In addition, in the description of the present application, the term "A and / or B" represents all possible combinations of A and B, such as only A, only B, or A and B.

[0108] So far, the technical solution of the present application has been described in conjunction with an embodiment shown in the accompanying drawings, but it is easy for those skilled in the art to understand that the protection scope of the present application is obviously not limited to these specific embodiments. Without departing from the principles of the present application, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present application.

Claims

1. A method for predicting remaining battery life, characterized in that: The method comprises: Acquire a first number of battery data groups to form a battery data sequence, wherein the battery data groups include voltage data, current data, and temperature data; Based on the battery data sequence, predicting the capacity value of the battery by using a trained capacity prediction model; Based on the battery data sequence and the predicted capacity value, predicting the SOH value of the battery through a trained SOH prediction model; Based on the battery data sequence, the predicted capacity value and the predicted SOH value, the RUL value of the battery is predicted by using a trained RUL prediction model.

2. The method for predicting the remaining service life of a battery according to claim 1, characterized in that: The method further comprises: Constructing the capacity prediction model based on the Bi LSTM network; Constructing the SOH prediction model based on the LSTM network; The RUL prediction model is constructed based on the LSTM network.

3. The method for predicting the remaining service life of a battery according to claim 1 or 2, characterized in that: The method further comprises: Constructing the capacity prediction model based on at least one first feature extraction layer, a first fully connected layer and a first output layer, wherein the first feature extraction layer is constructed based on a Bi LSTM network; Constructing the SOH prediction model based on at least one second feature extraction layer, a second fully connected layer and a second output layer, wherein the second feature extraction layer is constructed based on an LSTM network; The RUL prediction model is constructed based on at least one third feature extraction layer, a third fully connected layer and a third output layer, wherein the third feature extraction layer is constructed based on an LSTM network.

4. The method for predicting the remaining service life of a battery according to claim 1 or 2, characterized in that: The battery remaining service life prediction model constructed based on the capacity prediction model, the SOH prediction model and the RUL prediction model further includes a first input reconstruction unit and a second input reconstruction unit, and the method further includes: Inputting the battery data sequence and the predicted capacity value into the first input reconstruction unit for reconstruction to obtain a first reconstructed battery data sequence, wherein the first reconstructed battery data sequence is input data of the SOH prediction model; The battery data sequence, the predicted capacity value and the predicted SOH value are input into the second input reconstruction unit for reconstruction to obtain a second reconstructed battery data sequence, wherein the second reconstructed battery data sequence is input data of the RUL prediction model.

5. The method for predicting the remaining service life of a battery according to claim 4, characterized in that: The method further comprises: Adding the predicted capacity value to each of the battery data groups in the battery data sequence to obtain the first reconstructed battery data sequence, wherein the first reconstructed battery data group in the first reconstructed battery data sequence includes the voltage data, the current data, the temperature data and the predicted capacity value; The predicted capacity value and the predicted SOH value are added to each of the battery data groups in the battery data sequence to obtain the second reconstructed battery data sequence, wherein the second reconstructed battery data group in the second reconstructed battery data sequence includes the voltage data, the current data, the temperature data, the predicted capacity value and the predicted SOH value.

6. The method for predicting the remaining service life of a battery according to claim 1, characterized in that: The method further comprises: Divide the sample data set based on the preset sliding window and sliding step size to obtain the training data set; Based on the training data set, training the capacity prediction model to obtain the trained capacity prediction model; Based on the trained capacity prediction model, obtaining a predicted capacity training data set corresponding to the training data set; Based on the training data set and the predicted capacity training data set, construct a first reconstructed training data set; Based on the first reconstructed training data set, training the SOH prediction model to obtain the trained SOH prediction model; Based on the trained SOH prediction model, obtaining a predicted SOH training data set corresponding to the training data set; constructing a second reconstructed training data set based on the training data set, the predicted capacity training data set and the predicted SOH training data set; Based on the second reconstructed training data set, the RUL prediction model is trained to obtain the trained RUL prediction model.

7. The method for predicting the remaining service life of a battery according to claim 6, characterized in that: The method further comprises: The first number is set equal to a window size of the sliding window.

8. A battery remaining service life prediction device, characterized in that: The device comprises: A data acquisition module, the data acquisition module is configured to acquire a first number of battery data groups to form a battery data sequence, wherein the battery data group includes voltage data, current data and temperature data; A remaining useful life prediction module, wherein the remaining useful life prediction module is configured to perform the following operations: Based on the battery data sequence, predicting the capacity value of the battery by using a trained capacity prediction model; Based on the battery data sequence and the predicted capacity value, predicting the SOH value of the battery through a trained SOH prediction model; Based on the battery data sequence, the predicted capacity value and the predicted SOH value, the RUL value of the battery is predicted by using a trained RUL prediction model.

9. A computer device comprising a processor and a storage device, wherein the storage device is suitable for storing a plurality of program codes, characterized in that: The program code is suitable for being loaded and run by the processor to execute the battery remaining service life prediction method according to any one of claims 1 to 7.

10. A storage medium storing a plurality of program codes, characterized in that: The program code is suitable for being loaded and run by a processor to execute the method for predicting the remaining service life of a battery according to any one of claims 1 to 7.

Images