Lithium ion battery life prediction model construction method and lithium ion battery remaining service life prediction method

Abstract

The invention relates to a lithium ion battery life prediction model construction method and a lithium ion battery remaining service life prediction method. The method comprises the following steps: acquiring an original network model; the original network model comprises an original inner-layer network and an outer-layer network; the outer network is used for predicting the remaining service life of the lithium ion battery; fusing the mechanism information of the lithium ion battery with an original inner-layer network to obtain an inner-layer network and a to-be-trained life prediction model composed of the inner-layer network and an outer-layer network; the inner-layer network is used for evaluating the battery aging condition of the lithium ion battery; constructing a to-be-trained group-driven model according to the plurality of to-be-trained life prediction models; and according to the sample data of the lithium ion battery and the residual service life label corresponding to the sample data, carrying out synchronous training on each to-be-trained service life prediction model in the group driving model to obtain a trained target group driving model. The method can improve the prediction precision of the remaining service life of the lithium ion battery.

Description

Technical Field

[0001] The present application relates to the technical field of lithium-ion batteries, and in particular to a method for constructing a lithium-ion battery life prediction model, a method, device, computer equipment, storage medium and computer program product for predicting the remaining service life of a lithium-ion battery. Background Art

[0002] Accurately predicting the remaining service life of lithium-ion batteries has an important impact on the safe, stable and efficient operation of lithium-ion batteries and battery systems. Among them, the prediction of the remaining service life of lithium-ion batteries based on machine learning has always been a research hotspot in this field. At present, the life prediction methods of lithium-ion batteries are still mainly based on pure battery material performance methods or pure machine learning methods, which cannot comprehensively consider various factors for comprehensive prediction, resulting in difficulty in further improving the prediction accuracy.

[0003] In addition, in the specific application process of machine learning algorithms, even for a machine learning model with constant hyperparameters, the prediction accuracy of the model after training is not constant. How to find the optimal result with the highest accuracy in multiple trainings with changes in accuracy is also a problem worth exploring. Summary of the invention

[0004] Based on this, it is necessary to provide a lithium-ion battery life prediction model construction method, lithium-ion battery remaining service life prediction method, device, computer equipment, computer-readable storage medium and computer program product that can improve the prediction accuracy of the remaining service life of lithium-ion batteries in response to the above-mentioned technical problems.

[0005] In a first aspect, the present application provides a method for constructing a lithium-ion battery life prediction model. The method comprises:

[0006] Acquire an original network model; the original network model includes an original inner network and an outer network; the outer network is used to predict the remaining service life of the lithium-ion battery;

[0007] The mechanism information of the lithium-ion battery is integrated with the original inner network to obtain an inner network and a life prediction model to be trained composed of the inner network and the outer network; the inner network is used to evaluate the battery aging of the lithium-ion battery;

[0008] Constructing a group driving model to be trained according to the plurality of life prediction models to be trained;

[0009] According to the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data, each of the life prediction models to be trained in the group drive model is synchronously trained to obtain a trained target group drive model.

[0010] In one embodiment, the mechanism information of the lithium-ion battery is merged with the original inner network to obtain an inner network, including:

[0011] Obtaining the mechanism information according to the positive and negative electrode solid phase diffusion information in the electrochemical model of the lithium ion battery;

[0012] Determining initial conditions and boundary conditions corresponding to the mechanism information;

[0013] The mechanism information, the initial conditions and the boundary conditions are fused with the loss function of the original inner network to obtain the inner network.

[0014] In one embodiment, according to the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data, each of the life prediction models to be trained in the group drive model is synchronously trained to obtain a trained target group drive model, including:

[0015] Initializing the hyperparameters of each life prediction model to be trained based on different random seeds to obtain the model parameters of each life prediction model to be trained;

[0016] Inputting the sample data into each of the life prediction models to be trained, respectively, to obtain a plurality of predicted remaining service lives;

[0017] According to the life difference between each predicted remaining service life and the remaining service life label, the model parameters of each life prediction model to be trained are iteratively updated to obtain the trained target group driving model.

[0018] In one embodiment, before synchronously training each of the life prediction models to be trained in the group drive model according to the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data to obtain the trained target group drive model, it also includes:

[0019] Obtaining a cycle life of the lithium-ion battery according to the maximum available capacity of the lithium-ion battery;

[0020] Obtaining battery operation data and battery material data of the lithium-ion battery under any charge and discharge cycle as the sample data; the sample data includes voltage data and current data of the lithium-ion battery; the battery material data includes solid phase particle radius;

[0021] According to the difference between the cycle life and the number of cycles corresponding to the sample data, a remaining service life label corresponding to the sample data is obtained.

[0022] In a second aspect, the present application provides a method for predicting the remaining service life of a lithium-ion battery. The method comprises:

[0023] Inputting the charging time, current data and solid phase particle radius of the lithium-ion battery into each trained life prediction model in the target group driving model, respectively, to obtain the predicted remaining service life of the lithium-ion battery output by each trained life prediction model; the target group driving model is obtained according to the steps of the method described in the first aspect;

[0024] According to the predicted remaining service life, a target life prediction model that meets a preset model performance condition is selected from the plurality of trained life prediction models included in the target group driving model;

[0025] The target remaining service life of the lithium-ion battery is obtained according to the predicted remaining service life output by the target service life prediction model.

[0026] In one embodiment, the estimated remaining useful life is obtained by:

[0027] Using the inner network in the trained life prediction model, based on the charging time, the current data and the solid phase particle radius, the battery aging state of the lithium-ion battery is evaluated to obtain battery aging information of the lithium-ion battery;

[0028] The voltage data, the current data and the battery aging information of the lithium-ion battery are input into the outer network of the trained life prediction model to obtain the predicted remaining service life of the lithium-ion battery.

[0029] In one embodiment, according to the predicted remaining service life, a target life prediction model that meets the preset model performance conditions is selected from the plurality of trained life prediction models, including:

[0030] Performing average processing and standard deviation processing on all the output predicted remaining useful lives to obtain a mean value and a standard deviation of the predicted remaining useful lives;

[0031] According to the mean and the standard deviation, a target life prediction model with a value less than a quantile threshold is screened out from the plurality of trained life prediction models; the quantile threshold is set according to the mean and the standard deviation.

[0032] In a third aspect, the present application also provides a device for constructing a lithium-ion battery life prediction model. The device comprises:

[0033] A life prediction module, used to input the charging time, current data and solid phase particle radius of the lithium-ion battery into each trained life prediction model in the target group driving model, respectively, to obtain the predicted remaining service life of the lithium-ion battery output by each trained life prediction model; the target group driving model is obtained according to the steps of the method described in the first aspect;

[0034] A performance screening module, configured to screen out a target life prediction model that meets a preset model performance condition from a plurality of the trained life prediction models included in the target group driving model according to the predicted remaining service life;

[0035] The life determination module is used to obtain the target remaining service life of the lithium-ion battery according to the predicted remaining service life output by the target life prediction model.

[0036] In a fourth aspect, the present application also provides a device for predicting the remaining service life of a lithium-ion battery. The device comprises:

[0037] A life prediction module, used to input the charging time, current data and solid phase particle radius of the lithium-ion battery into each trained life prediction model in the target group driving model, respectively, to obtain the predicted remaining service life of the lithium-ion battery output by each trained life prediction model; the target group driving model is obtained according to the steps of the method described in the first aspect;

[0038] A performance screening module, configured to screen out a target life prediction model that meets a preset model performance condition from a plurality of the trained life prediction models included in the target group driving model according to the predicted remaining service life;

[0039] The life determination module is used to obtain the target remaining service life of the lithium-ion battery according to the predicted remaining service life output by the target life prediction model.

[0040] In a fifth aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0041] Acquire an original network model; the original network model includes an original inner network and an outer network; the outer network is used to predict the remaining service life of the lithium-ion battery;

[0042] The mechanism information of the lithium-ion battery is integrated with the original inner network to obtain an inner network and a life prediction model to be trained composed of the inner network and the outer network; the inner network is used to evaluate the battery aging of the lithium-ion battery;

[0043] Constructing a group driving model to be trained according to the plurality of life prediction models to be trained;

[0044] According to the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data, each of the life prediction models to be trained in the group drive model is synchronously trained to obtain a trained target group drive model.

[0045] In a sixth aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0046] Inputting the charging time, current data and solid phase particle radius of the lithium-ion battery into each trained life prediction model in the target group driving model, respectively, to obtain the predicted remaining service life of the lithium-ion battery output by each trained life prediction model; the target group driving model is obtained according to the steps of the method described in the first aspect;

[0047] According to the predicted remaining service life, a target life prediction model that meets a preset model performance condition is selected from the plurality of trained life prediction models included in the target group driving model;

[0048] The target remaining service life of the lithium-ion battery is obtained according to the predicted remaining service life output by the target service life prediction model.

[0049] In a seventh aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0050] Acquire an original network model; the original network model includes an original inner network and an outer network; the outer network is used to predict the remaining service life of the lithium-ion battery;

[0051] The mechanism information of the lithium-ion battery is integrated with the original inner network to obtain an inner network and a life prediction model to be trained composed of the inner network and the outer network; the inner network is used to evaluate the battery aging of the lithium-ion battery;

[0052] Constructing a group driving model to be trained according to the plurality of life prediction models to be trained;

[0053] According to the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data, each of the life prediction models to be trained in the group drive model is synchronously trained to obtain a trained target group drive model.

[0054] In an eighth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0055] Inputting the charging time, current data and solid phase particle radius of the lithium-ion battery into each trained life prediction model in the target group driving model, respectively, to obtain the predicted remaining service life of the lithium-ion battery output by each trained life prediction model; the target group driving model is obtained according to the steps of the method described in the first aspect;

[0056] According to the predicted remaining service life, a target life prediction model that meets a preset model performance condition is selected from the plurality of trained life prediction models included in the target group driving model;

[0057] The target remaining service life of the lithium-ion battery is obtained according to the predicted remaining service life output by the target service life prediction model.

[0058] In a ninth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0059] Acquire an original network model; the original network model includes an original inner network and an outer network; the outer network is used to predict the remaining service life of the lithium-ion battery;

[0060] The mechanism information of the lithium-ion battery is integrated with the original inner network to obtain an inner network and a life prediction model to be trained composed of the inner network and the outer network; the inner network is used to evaluate the battery aging of the lithium-ion battery;

[0061] Constructing a group driving model to be trained according to the plurality of life prediction models to be trained;

[0062] According to the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data, each of the life prediction models to be trained in the group drive model is synchronously trained to obtain a trained target group drive model.

[0063] In a tenth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0064] Inputting the charging time, current data and solid phase particle radius of the lithium-ion battery into each trained life prediction model in the target group driving model, respectively, to obtain the predicted remaining service life of the lithium-ion battery output by each trained life prediction model; the target group driving model is obtained according to the steps of the method described in the first aspect;

[0065] According to the predicted remaining service life, a target life prediction model that meets a preset model performance condition is selected from the plurality of trained life prediction models included in the target group driving model;

[0066] The target remaining service life of the lithium-ion battery is obtained according to the predicted remaining service life output by the target service life prediction model.

[0067] The above-mentioned lithium-ion battery life prediction model construction method, lithium-ion battery remaining service life prediction method, device, computer equipment, storage medium and computer program product obtain the original network model; the original network model includes an original inner network and an outer network; the outer network is used to predict the remaining service life of the lithium-ion battery; the mechanism information of the lithium-ion battery is integrated with the original inner network to obtain the inner network, and the life prediction model to be trained composed of the inner network and the outer network; the inner network is used to evaluate the battery aging of the lithium-ion battery; according to multiple life prediction models to be trained, a group drive model to be trained is constructed; according to the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data, each life prediction model to be trained in the group drive model is synchronously trained to obtain a trained target group drive model. By adopting this method, the model is optimized by introducing the mechanism information of lithium-ion batteries, so that the optimized life prediction model can analyze the material performance state inside the lithium-ion battery in a deeper level, thereby improving the life prediction accuracy of the life prediction model; a group-driven model is also built through multiple life prediction models, so that the group-driven model comprehensively considers the prediction results of multiple life prediction models during training and application, which can avoid the error of the target remaining service life outputted in the final output due to the decrease in the prediction accuracy of a single life prediction model, so that the group-driven model can better adapt to the working conditions and environmental changes of different batteries during the prediction process, greatly improving the prediction accuracy of the remaining service life of lithium-ion batteries. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 A schematic diagram of a process for constructing a lithium-ion battery life prediction model in one embodiment;

[0069] Figure 2 is a schematic diagram of the structure of an inner network in one embodiment;

[0070] Figure 3 A schematic diagram of a training process and a testing process of a group drive model in one embodiment;

[0071] Figure 4 is a test result of a life prediction model in a target group driving model in an embodiment;

[0072] Figure 5 A schematic diagram of a flow chart of a method for predicting the remaining service life of a lithium-ion battery in one embodiment;

[0073] Figure 6 is a network structure diagram of an outer network in one embodiment;

[0074] Figure 7 A schematic diagram of a life prediction model in which a root mean square error reaches and does not reach a quantile threshold in one embodiment;

[0075] Figure 8 A structural block diagram of a device for constructing a life prediction model for a lithium-ion battery in one embodiment;

[0076] Fig. 9 A structural block diagram of a device for predicting the remaining useful life of a lithium-ion battery in one embodiment;

[0077] Fig.10 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0078] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0079] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0080] In one embodiment, Figure 1 As shown, a method for constructing a lithium-ion battery life prediction model is provided. This embodiment uses the method applied to a server as an example. It can be understood that the method can also be applied to a terminal, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0081] Step S101, obtaining an original network model; the original network model includes an original inner network and an outer network; the outer network is used to predict the remaining service life of the lithium-ion battery.

[0082] Among them, lithium-ion batteries mainly rely on the movement of lithium ions between the positive and negative electrodes to work.

[0083] Specifically, the server can use a neural network model to construct an original network model, for example, using a fully connected network as the original network model, which includes a conventional original inner network for feature extraction and an outer network for predicting the remaining service life of a lithium-ion battery.

[0084] Step S102, the mechanism information of the lithium-ion battery is integrated with the original inner network to obtain the inner network and the life prediction model to be trained composed of the inner network and the outer network; the inner network is used to evaluate the battery aging of the lithium-ion battery.

[0085] Specifically, in order to more comprehensively analyze the performance of lithium-ion batteries, the server can integrate the mechanism information of lithium-ion batteries, the initial conditions of the mechanism information, and the boundary conditions into the loss function of the original inner network, so that the inner network obtained after integration has the ability to evaluate the battery aging of lithium-ion batteries. For ease of distinction, the server can set the model that integrates the mechanism information as the life prediction model to be trained. In actual applications, the inner network has a total of 3 fully connected layers, each containing 64 neurons, and the outer network has a total of 3 fully connected layers, each containing 64 neurons.

[0086] Step S103, constructing a group drive model to be trained according to a plurality of life prediction models to be trained.

[0087] Specifically, the server generates multiple (for example, 100) independent life prediction models to be trained, and introduces a group-driven strategy to build the generated multiple life prediction models to be trained into a model group, and the server obtains a group-driven model to be trained. It should be noted that each life prediction model to be trained in the group-driven model is independent of each other, and all life prediction models to be trained have the same hyperparameters, but the initialization values ​​of the hyperparameters can be different, and all life prediction models to be trained are trained and predicted synchronously. Among them, the hyperparameters include learning rate, number of training times and optimizer, and the specific values ​​of the hyperparameters are shown in Table 1.

[0088] Table 1

[0089]

[0090] Step S104, according to the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data, each life prediction model to be trained in the group drive model is synchronously trained to obtain a trained target group drive model.

[0091] Specifically, the server obtains sample data of lithium-ion batteries and the remaining service life labels corresponding to the sample data, and divides the sample data into a training set and a test set; the training set is used to synchronously train all the life prediction models to be trained in the group drive model, and after reaching the preset maximum number of training times, the server obtains the trained target group drive model. The ratio of the training set to the test set can be set to 4:1.

[0092] In the above-mentioned lithium-ion battery life prediction model construction method, an original network model is obtained; the original network model includes an original inner network and an outer network; the outer network is used to predict the remaining service life of the lithium-ion battery; the mechanism information of the lithium-ion battery is integrated with the original inner network to obtain an inner network, and a life prediction model to be trained composed of the inner network and the outer network; the inner network is used to evaluate the battery aging of the lithium-ion battery; based on multiple life prediction models to be trained, a group drive model to be trained is constructed; based on the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data, each life prediction model to be trained in the group drive model is synchronously trained to obtain a trained target group drive model. By adopting this method, the model is optimized by introducing the mechanism information of lithium-ion batteries, so that the optimized life prediction model can analyze the material performance state inside the lithium-ion battery in a deeper level, thereby improving the life prediction accuracy of the life prediction model; a group-driven model is also built through multiple life prediction models, so that the group-driven model comprehensively considers the prediction results of multiple life prediction models during training and application, which can avoid the error of the target remaining service life outputted in the final output due to the decrease in the prediction accuracy of a single life prediction model, so that the group-driven model can better adapt to the working conditions and environmental changes of different batteries during the prediction process, greatly improving the prediction accuracy of the remaining service life of lithium-ion batteries.

[0093] In one embodiment, the above step S102, fusing the mechanism information of the lithium-ion battery with the original inner network to obtain the inner network, specifically includes the following contents: obtaining the mechanism information according to the positive and negative electrode solid phase diffusion information in the electrochemical model of the lithium-ion battery; determining the initial conditions and boundary conditions corresponding to the mechanism information; fusing the mechanism information, initial conditions and boundary conditions with the loss function of the original inner network to obtain the inner network.

[0094] Among them, the mechanism information of lithium-ion batteries refers to the information about the battery charging and discharging operation and electrochemical reaction mechanism, such as the migration of lithium ions, the transfer of charges and chemical reactions.

[0095] Specifically, the server can select the positive and negative electrode solid phase diffusion equations in the lithium-ion battery electrochemical model as the mechanism information to be incorporated, and the specific form is as follows:

[0096]

[0097] In the formula, c s,j (r, t) represents the solid phase concentration, r represents the radius of the solid phase spherical particle, D s,j is the solid phase diffusion coefficient of lithium ions, j=n represents the negative electrode, j=p represents the positive electrode

[0098] The initial conditions corresponding to this equation are:

[0099]

[0100] In the formula, c s,0 is the initial concentration of lithium ions in the solid phase of the electrode.

[0101] The boundary conditions are:

[0102]

[0103] In the formula, R s,j is the maximum radius of the positive and negative electrode solid particles, I(t) is the current, A is the electrode surface area, L j is the thickness of the positive and negative electrodes, F is a constant, a j is the specific surface area of ​​the positive and negative electrode solid particles.

[0104] The formula for solving the actual electrode solid phase lithium ion concentration is:

[0105]

[0106] Where C s,j (t) is the actual solid phase lithium ion concentration of the electrode, SOC(t) is the battery state of charge at time t, is the concentration ratio of negative electrode lithium ions in the fully charged state, is the concentration ratio of negative electrode lithium ions in the fully discharged state.

[0107] Furthermore, the specific values ​​of the parameters involved in the above mechanism information, initial conditions or boundary conditions are shown in Table 2. + Indicates the thickness of the positive electrode, L - Represents the thickness of the negative electrode, R + Represents the maximum radius of the positive electrode solid phase particle, R - Indicates the maximum radius of the negative electrode solid phase particles, a + represents the specific surface area of ​​the positive electrode solid phase particles, a - Represents the specific surface area of ​​the negative electrode solid particles.

[0108] Table 2

[0109]

[0110] Further, Figure 2 The server is based on the mechanism information L 扩散方差 , initial condition L 初始条件 and boundary conditions L 边界条件 , get the loss function to be fused Loss phy , the loss function to be fused is combined with the original loss function Loss of the original inner network data Fusion, such as Loss phy +Loss data , then the server obtains the target loss function Loss of the inner network, that is, Loss=Loss phy +Loss data .

[0111] In this embodiment, by integrating the positive and negative electrode solid-phase diffusion equations representing the mechanism information and their corresponding initial conditions and boundary conditions into the loss function of the original inner network, the inner machine that integrates the mechanism information can analyze the aging of lithium-ion batteries in more depth, which is beneficial to improving the life prediction accuracy of the life prediction model.

[0112] In one embodiment, the above step S104, based on the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data, synchronously trains each life prediction model to be trained in the group driving model to obtain a trained target group driving model, which specifically includes the following contents: based on different random seeds, the hyperparameters of each life prediction model to be trained are initialized to obtain the model parameters of each life prediction model to be trained; the sample data are input into each life prediction model to be trained to obtain multiple predicted remaining service lives; according to the life difference between each predicted remaining service life and the remaining service life label, the model parameters of each life prediction model to be trained are iteratively updated to obtain the trained target group driving model.

[0113] Figure 3 The figure is a schematic diagram of the training and testing process of the swarm-driven model. Specifically, 1) Determine the input and output of the model: For the inner network, the charging time, current, and solid phase particle radius of the lithium-ion battery are used as input data, and the solid phase concentration c of the lithium-ion battery is used as input data. s,j (r, t) and the solid phase diffusion coefficient D s,j As output data; for the outer network, the voltage data, current data and solid phase concentration c of the lithium-ion battery under any charge and discharge cycle s,j(r, t) is used as input data, and the predicted remaining service life corresponding to the cycle is used as output data. 2) Determine the hyperparameters of the model: The server can use different random seeds to initialize the hyperparameters of each life prediction model to be trained in the group-driven model, and obtain the model parameters of each life prediction model to be trained. 3) Determine the training set and test set of the model: The server divides the sample data and the remaining service life label corresponding to the sample data into a training set and a test set; the group-driven model is trained using the test set, for example, the sample data belonging to the test set is used as training data, and is input into each life prediction model to be trained in the group-driven model, respectively, to obtain the predicted remaining service life output by each life prediction model to be trained; according to the life difference between each predicted remaining service life and the remaining service life label, the model parameters of each life prediction model to be trained are iteratively updated, and after the set training termination condition is reached, the server obtains the trained target group-driven model.

[0114] Furthermore, the server may also use the sample data belonging to the test set as test data to test the prediction accuracy of the target group driving model. Figure 4 This is a schematic diagram of the test results of the life prediction model in the target group driving model. Figure 4 It can be seen that the life prediction model of this embodiment has a predicted value that is very close to the actual value, and can achieve high-precision prediction of the remaining service life of the lithium-ion battery.

[0115] In this embodiment, based on different random seeds, the hyperparameters of each life prediction model to be trained are initialized to obtain the model parameters of each life prediction model to be trained, so as to improve the diversity of the trained life prediction models; using the training data, iterative training of each life prediction model to be trained in the group drive model is realized, so that the trained target group drive model has the ability to predict the remaining service life of the lithium-ion battery.

[0116] In one embodiment, in the above step S104, each life prediction model to be trained in the group drive model is synchronously trained according to the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data, and before the trained target group drive model is obtained, it also includes: obtaining the cycle life of the lithium-ion battery according to the maximum available capacity of the lithium-ion battery; obtaining the battery operation data and battery material data of the lithium-ion battery under any charge and discharge cycle and taking them as sample data; the sample data includes the voltage data and current data of the lithium-ion battery; the battery material data includes the radius of the solid phase particle; and obtaining the remaining service life label corresponding to the sample data according to the difference between the cycle life and the number of cycles corresponding to the sample data.

[0117] The battery operation data refers to data describing the operation status of the lithium-ion battery during the charging and discharging process. For example, the battery operation data may be voltage data and current data.

[0118] Among them, battery material data refers to data describing the physical and electrochemical materials of lithium-ion batteries. For example, battery material data can be the maximum available capacity and the radius of solid phase particles.

[0119] Specifically, the server can pre-collect the battery operation data and battery material data of the lithium-ion battery as sample data; for example, the voltage data, current data, state of charge (SOC), maximum available capacity and solid phase particle radius of the lithium-ion battery in the charging state in each charge and discharge cycle are collected. The server multiplies the maximum available capacity of the lithium-ion battery by 0.8 to obtain the life threshold of the lithium-ion battery, and uses the number of cycles corresponding to the life threshold as the cycle life of the lithium-ion battery, that is, the cycle life of the lithium-ion battery is calculated according to the 80% life threshold of the lithium-ion battery. The server extracts the voltage data, current data and solid phase particle radius of the target step of the lithium-ion battery under any charge and discharge cycle, and calculates the remaining service life label corresponding to the charge and discharge cycle; the server calculates the difference between the cycle life of the lithium-ion battery and the number of cycles corresponding to the charge and discharge cycle, and obtains the remaining service life of the lithium-ion battery under the charge and discharge cycle, and then the server sets the remaining service life as the remaining service life label corresponding to the sample data.

[0120] In this embodiment, based on the battery operation data and battery material data of the lithium-ion battery under any charge and discharge cycle, the sample data and the remaining service life label corresponding to the sample data are reasonably obtained, laying the foundation for subsequent model training and model testing.

[0121] In one embodiment, Figure 5 As shown, a method for predicting the remaining service life of a lithium-ion battery is provided. This embodiment uses the method applied to a server as an example for illustration. It can be understood that the method can also be applied to a terminal, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0122] Step S501, the charging time, current data and solid phase particle radius of the lithium-ion battery are respectively input into each trained life prediction model in the target group driving model to obtain the predicted remaining service life of the lithium-ion battery output by each trained life prediction model; the target group driving model is obtained according to the steps of the above-mentioned lithium-ion battery life prediction model construction method.

[0123] Specifically, the server inputs the charging time, current data and solid phase particle radius of the lithium-ion battery into each trained life prediction model in the target group driving model, so as to perform battery aging status evaluation and remaining service life prediction processing on the lithium-ion battery in turn through each trained life prediction model, and finally obtains the predicted remaining service life of the lithium-ion battery output by each trained life prediction model.

[0124] Step S502 , based on the predicted remaining service life, a target life prediction model that meets a preset model performance condition is selected from a plurality of trained life prediction models included in the target group driving model.

[0125] Specifically, after obtaining the predicted remaining service life output by all trained life prediction models, the server can also perform model performance evaluation on all life prediction models to obtain performance evaluation results for each life prediction model; based on the performance evaluation results, models with poor performance in the target group driving model are excluded, and the life prediction model in the target group driving model that meets the preset model performance conditions is set as the target life prediction model.

[0126] Step S503, obtaining the target remaining service life of the lithium-ion battery according to the predicted remaining service life output by the target service life prediction model.

[0127] Specifically, the server averages the predicted remaining service life output by all target life prediction models, and uses the average predicted remaining service life obtained by the average processing as the target remaining service life of the lithium-ion battery.

[0128] In the above-mentioned method for predicting the remaining service life of lithium-ion batteries, the charging time, current data and solid phase particle radius of the lithium-ion battery are respectively input into each trained life prediction model in the target group driving model to obtain the predicted remaining service life of the lithium-ion battery output by each trained life prediction model; according to the predicted remaining service life, a target life prediction model that meets the preset model performance conditions is screened out from the multiple trained life prediction models included in the target group driving model; according to the predicted remaining service life output by the target life prediction model, the target remaining service life of the lithium-ion battery is obtained. With this method, while using the target group driving model to predict the remaining service life of the lithium-ion battery, the preset model performance conditions are also used to exclude the predicted remaining service life output by the life prediction model with poor performance, thereby further improving the prediction accuracy of the remaining service life of the lithium-ion battery.

[0129] In one embodiment, the predicted remaining service life is obtained in the following manner: through the inner network in the trained life prediction model, the battery aging status of the lithium-ion battery is evaluated based on the charging time, current data and solid phase particle radius to obtain the battery aging information of the lithium-ion battery; the voltage data, current data and battery aging information of the lithium-ion battery are input into the outer network of the trained life prediction model to obtain the predicted remaining service life of the lithium-ion battery.

[0130] For each trained life prediction model in the target group driving model, the server first inputs the charging time, current data and solid phase particle radius into the inner network of each life prediction model, so as to evaluate the battery aging status of the lithium-ion battery through the inner network, and output the solid phase diffusion coefficient and solid phase concentration reflecting the battery aging information of the lithium-ion battery. The server inputs the charging time and current data of the lithium-ion battery, as well as the solid phase diffusion coefficient output by the inner network, into the outer network of each life prediction model, so as to predict the remaining service life of the lithium-ion battery through the outer network, and finally output the predicted remaining service life of the lithium-ion battery. Figure 6 This is the network structure diagram of the outer network.

[0131] The solid phase diffusion coefficient refers to the diffusion rate of lithium ions in the electrode material, and the solid phase concentration refers to the concentration of lithium ions in the electrode material.

[0132] In this embodiment, the inner network and the outer network in the trained life prediction model comprehensively consider the battery aging state and the operating state of the lithium-ion battery, thereby effectively improving the accuracy of the predicted remaining service life.

[0133] In one embodiment, the above step S502, based on the predicted remaining service life, selects a target life prediction model that meets the preset model performance conditions from multiple trained life prediction models included in the target group driving model, specifically including the following contents: performing averaging and standard deviation processing on all output predicted remaining service lives to obtain the mean and standard deviation of the predicted remaining service life; based on the mean and standard deviation, selecting a target life prediction model that is less than a quantile threshold from multiple trained life prediction models; the quantile threshold is set based on the mean and standard deviation.

[0134] The server can use the mean and standard deviation of the predicted values ​​of the service life prediction model to collaboratively evaluate the prediction effect of each life prediction model. Specifically, the server performs average processing and standard deviation processing on the predicted remaining service life output by all life prediction models to obtain the mean and standard deviation of the predicted remaining service life. The server can determine the root mean square error of each life prediction model based on the mean and standard deviation, use the quartile of the root mean square error as the quantile threshold, exclude life prediction models with a root mean square error greater than the quantile threshold, and use the life prediction models that are not excluded as the target life prediction models. Figure 7 Schematic diagram of the life prediction model when the root mean square error reaches and does not reach the quantile threshold, as shown in Figure 7 As shown in the figure, among the 100 life prediction models generated, the life prediction models whose root mean square error of the prediction results is greater than the quantile threshold (for example, 5%) are effectively eliminated ( Figure 7 The label in is "before selection"), and the life prediction models with a root mean square error of the prediction results less than or equal to 5% are retained ( Figure 7 The label in is "after selection"), which can effectively improve the overall prediction accuracy.

[0135] In this embodiment, the mean and standard deviation of the predicted remaining useful life output by all life prediction models are used to exclude the output results of the life prediction models with poor prediction performance, and only the output results of the target life prediction model whose mean and standard deviation reach the quantile threshold are retained, thereby avoiding the influence of poor quality output results on the prediction accuracy and greatly improving the accuracy of the target remaining useful life.

[0136] In order to more clearly illustrate the method for constructing a lithium-ion battery life prediction model provided by the embodiment of the present disclosure, the following is a specific example of the method for constructing a lithium-ion battery life prediction model. Another method for constructing a lithium-ion battery life prediction model is provided, which can be applied to Figure 1 The server in the system includes the following:

[0137] First, based on the operating data of lithium-ion batteries, data extraction and data preprocessing operations are performed. Then, in order to address the problem that traditional life prediction methods cannot take into account both mechanism and machine learning, the mechanism information of lithium-ion batteries is introduced into the original network model. For example, the positive and negative electrode solid phase diffusion equations of the lithium-ion battery electrochemical model can be introduced into the neural network loss function. In addition, in order to address the problem that the output results of machine learning models are prone to volatility, a group-driven strategy model is also introduced. In the constructed large-scale independent life prediction model group, the prediction effect of each model is collaboratively evaluated through the mean and standard deviation of the model prediction value, and the models with poor performance are excluded using the quartile as the threshold. Finally, the output results of the models that are not excluded are averaged as the final life prediction result.

[0138] In this embodiment, the model is optimized by introducing the mechanism information of lithium-ion batteries, so that the optimized life prediction model can analyze the material performance state inside the lithium-ion battery in a deeper level, thereby improving the life prediction accuracy of the life prediction model; a group-driven model is also built through multiple life prediction models, so that the group-driven model comprehensively considers the prediction results of multiple life prediction models during training and application, which can avoid the error of the target remaining service life outputted in the final output due to the decrease in the prediction accuracy of a single life prediction model, so that the group-driven model can better adapt to the working conditions and environmental changes of different batteries during the prediction process, thereby greatly improving the prediction accuracy of the remaining service life of lithium-ion batteries.

[0139] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0140] Based on the same inventive concept, the embodiment of the present application also provides a lithium-ion battery life prediction model construction device for implementing the above-mentioned lithium-ion battery life prediction model construction method. The implementation solution provided by the device to solve the problem is similar to the implementation solution recorded in the above-mentioned method, so the specific limitations in one or more lithium-ion battery life prediction model construction device embodiments provided below can refer to the limitations of the lithium-ion battery life prediction model construction method above, and will not be repeated here.

[0141] In one embodiment, Figure 8 As shown, a lithium-ion battery life prediction model construction device 800 is provided, including: a model acquisition module 801, a model optimization module 802, a model construction module 803 and a model training module 804, wherein:

[0142] The model acquisition module 801 is used to acquire the original network model; the original network model includes an original inner network and an outer network; the outer network is used to predict the remaining service life of the lithium-ion battery.

[0143] The model optimization module 802 is used to fuse the mechanism information of the lithium-ion battery with the original inner network to obtain the inner network and the life prediction model to be trained composed of the inner network and the outer network; the inner network is used to evaluate the battery aging of the lithium-ion battery.

[0144] The model building module 803 is used to build a group driving model to be trained based on multiple life prediction models to be trained.

[0145] The model training module 804 is used to synchronously train each life prediction model to be trained in the group drive model according to the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data, so as to obtain a trained target group drive model.

[0146] In one embodiment, the model optimization module 802 is also used to obtain mechanism information based on the positive and negative electrode solid phase diffusion information in the electrochemical model of the lithium-ion battery; determine the initial conditions and boundary conditions corresponding to the mechanism information; and fuse the mechanism information, initial conditions and boundary conditions with the loss function of the original inner network to obtain the inner network.

[0147] In one embodiment, the model training module 804 is also used to initialize the hyperparameters of each life prediction model to be trained based on different random seeds to obtain the model parameters of each life prediction model to be trained; input the sample data into each life prediction model to be trained to obtain multiple predicted remaining service lives; and iteratively update the model parameters of each life prediction model to be trained according to the life difference between each predicted remaining service life and the remaining service life label to obtain a trained target group driving model.

[0148] In one embodiment, the lithium-ion battery life prediction model building device 800 also includes a label setting module, which is used to obtain the cycle life of the lithium-ion battery based on the maximum available capacity of the lithium-ion battery; obtain the battery operation data and battery material data of the lithium-ion battery under any charge and discharge cycle and use them as sample data; the sample data includes the voltage data and current data of the lithium-ion battery; the battery material data includes the radius of the solid phase particles; and obtain the remaining service life label corresponding to the sample data based on the difference between the cycle life and the number of cycles corresponding to the sample data.

[0149] In one embodiment, Fig. 9 As shown, a lithium-ion battery remaining service life prediction device 900 is provided, comprising: a service life prediction module 901, a performance screening module 902 and a service life determination module 903, wherein:

[0150] The life prediction module 901 is used to input the charging time, current data and solid phase particle radius of the lithium-ion battery into each trained life prediction model in the target group driving model, respectively, to obtain the predicted remaining service life of the lithium-ion battery output by each trained life prediction model; the target group driving model is obtained according to the steps of the above-mentioned lithium-ion battery life prediction model construction method.

[0151] The performance screening module 902 is used to screen out a target life prediction model that meets a preset model performance condition from a plurality of trained life prediction models included in the target group driving model according to the predicted remaining service life.

[0152] The life determination module 903 is used to obtain the target remaining service life of the lithium-ion battery according to the predicted remaining service life output by the target life prediction model.

[0153] In one embodiment, the lithium-ion battery remaining service life prediction device 900 also includes a result prediction module, which is used to evaluate the battery aging status of the lithium-ion battery based on the charging time, current data and solid phase particle radius through the inner network in the trained life prediction model to obtain the battery aging information of the lithium-ion battery; the voltage data, current data and battery aging information of the lithium-ion battery are input into the outer network of the trained life prediction model to obtain the predicted remaining service life of the lithium-ion battery.

[0154] In one embodiment, the performance screening module 902 is also used to perform averaging and standard deviation processing on all output predicted remaining useful lives to obtain a mean and standard deviation of the predicted remaining useful lives; based on the mean and standard deviation, a target life prediction model with a value less than a quantile threshold is screened out from a plurality of trained life prediction models; the quantile threshold is obtained based on the mean and standard deviation.

[0155] Each module in the above-mentioned lithium-ion battery life prediction model construction device and lithium-ion battery remaining service life prediction device can be implemented in whole or in part by software, hardware and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0156] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.10As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data of lithium-ion batteries. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for constructing a lithium-ion battery life prediction model and / or a method for predicting the remaining service life of a lithium-ion battery is implemented.

[0157] Those skilled in the art will understand that Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0158] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0159] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0160] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0161] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0162] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0163] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the present application. It should be noted that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for constructing a lithium-ion battery life prediction model, characterized in that: The method comprises: Acquire an original network model; the original network model includes an original inner network and an outer network; the outer network is used to predict the remaining service life of the lithium-ion battery; The mechanism information of the lithium-ion battery is integrated with the original inner network to obtain an inner network and a life prediction model to be trained composed of the inner network and the outer network; the inner network is used to evaluate the battery aging of the lithium-ion battery; Constructing a group driving model to be trained according to the plurality of life prediction models to be trained; According to the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data, each of the life prediction models to be trained in the group drive model is synchronously trained to obtain a trained target group drive model.

2. The method according to claim 1, characterized in that The step of fusing the mechanism information of the lithium-ion battery with the original inner network to obtain an inner network includes: Obtaining the mechanism information according to the positive and negative electrode solid phase diffusion information in the electrochemical model of the lithium ion battery; Determining initial conditions and boundary conditions corresponding to the mechanism information; The mechanism information, the initial conditions and the boundary conditions are fused with the loss function of the original inner network to obtain the inner network.

3. The method according to claim 1, characterized in that The step of synchronously training each of the life prediction models to be trained in the group drive model according to the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data to obtain a trained target group drive model includes: Initializing the hyperparameters of each life prediction model to be trained based on different random seeds to obtain the model parameters of each life prediction model to be trained; Inputting the sample data into each of the life prediction models to be trained, respectively, to obtain a plurality of predicted remaining service lives; According to the life difference between each predicted remaining service life and the remaining service life label, the model parameters of each life prediction model to be trained are iteratively updated to obtain the trained target group driving model.

4. The method according to any one of claims 1 to 3, characterized in that: Before synchronously training each of the life prediction models to be trained in the group drive model according to the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data to obtain the trained target group drive model, the method further includes: Obtaining a cycle life of the lithium-ion battery according to the maximum available capacity of the lithium-ion battery; Obtaining battery operation data and battery material data of the lithium-ion battery under any charge and discharge cycle as the sample data; the sample data includes voltage data and current data of the lithium-ion battery; the battery material data includes solid phase particle radius; According to the difference between the cycle life and the number of cycles corresponding to the sample data, a remaining service life label corresponding to the sample data is obtained.

5. A method for predicting the remaining service life of a lithium-ion battery, characterized in that: The method comprises: Inputting the charging time, current data and solid phase particle radius of the lithium-ion battery into each trained life prediction model in the target group driving model, respectively, to obtain the predicted remaining service life of the lithium-ion battery output by each trained life prediction model; the target group driving model is obtained according to the steps of the method described in any one of claims 1 to 4; According to the predicted remaining service life, a target life prediction model that meets a preset model performance condition is selected from the plurality of trained life prediction models included in the target group driving model; The target remaining service life of the lithium-ion battery is obtained according to the predicted remaining service life output by the target service life prediction model.

6. The method according to claim 5, characterized in that The predicted remaining useful life is obtained by: Using the inner network in the trained life prediction model, based on the charging time, the current data and the solid phase particle radius, the battery aging state of the lithium-ion battery is evaluated to obtain battery aging information of the lithium-ion battery; The voltage data, the current data and the battery aging information of the lithium-ion battery are input into the outer network of the trained life prediction model to obtain the predicted remaining service life of the lithium-ion battery.

7. The method according to claim 5, characterized in that The method of selecting a target life prediction model that meets a preset model performance condition from a plurality of trained life prediction models according to the predicted remaining service life includes: Performing average processing and standard deviation processing on all the output predicted remaining useful lives to obtain a mean value and a standard deviation of the predicted remaining useful lives; According to the mean and the standard deviation, a target life prediction model with a value less than a quantile threshold is screened out from the plurality of trained life prediction models; the quantile threshold is set according to the mean and the standard deviation.

8. A device for constructing a lithium-ion battery life prediction model, characterized in that: The device comprises: A model acquisition module, used to acquire an original network model; the original network model includes an original inner network and an outer network; the outer network is used to predict the remaining service life of the lithium-ion battery; A model optimization module, used for fusing the mechanism information of the lithium-ion battery with the original inner network to obtain an inner network and a life prediction model to be trained composed of the inner network and the outer network; the inner network is used for evaluating the battery aging of the lithium-ion battery; A model building module, used to build a group driving model to be trained according to the plurality of life prediction models to be trained; The model training module is used to synchronously train each life prediction model to be trained in the group drive model according to the sample data of the lithium-ion battery and the remaining service life label corresponding to the sample data, so as to obtain a trained target group drive model.

9. A device for predicting the remaining useful life of a lithium-ion battery, characterized in that: The device comprises: A life prediction module, used for inputting the charging time, current data and solid phase particle radius of the lithium-ion battery into each trained life prediction model in the target group driving model, respectively, to obtain the predicted remaining service life of the lithium-ion battery output by each trained life prediction model; the target group driving model is obtained according to the steps of any one of the methods of claims 1 to 4; A performance screening module, configured to screen out a target life prediction model that meets a preset model performance condition from a plurality of the trained life prediction models included in the target group driving model according to the predicted remaining service life; The life determination module is used to obtain the target remaining service life of the lithium-ion battery according to the predicted remaining service life output by the target life prediction model.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4, or 5 to 7 are implemented.

Images