Battery health state estimation method, storage medium and electronic equipment

Abstract

The invention relates to the technical field of battery management, and particularly provides a battery health state estimation method, a storage medium and electronic equipment, and the method can comprise the steps: inputting a charging feature vector in a battery charging process into a mapping network, and obtaining an initial prediction value of a battery health state; determining a predicted value of the battery health state based on the initial predicted value, the charging feature vector and a dynamic network; the mapping network and the dynamic network form a fusion neural network model; solving a loss value between the predicted value and the true value of the battery health state by using the target function; optimizing model parameters of the fusion neural network model based on the loss value, and obtaining a battery health state evaluation model; the battery health state evaluation model is used for predicting the real-time value of the battery health state. According to the embodiment of the invention, the battery health state can be accurately estimated.

Description

Technical Field

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

[0002] Battery state of health (SOH) is an important indicator for evaluating battery performance. It reflects the battery's ability to store electrical energy relative to a new battery and is usually expressed as a percentage.

[0003] Currently, there are multiple methods for assessing the SOH of battery packs, including direct and indirect measurement. Direct measurement primarily uses techniques such as electrochemical impedance spectroscopy to measure parameters such as the battery's internal resistance and charge, thereby determining the battery's state of health. Indirect measurement, on the other hand, estimates the battery's SOH by analyzing data such as the battery's discharge curve and self-discharge rate. However, direct measurement is susceptible to environmental influences, which can affect the accuracy of the results. Indirect measurement relies on a full discharge curve, which is difficult to obtain in practical applications.

[0004] Therefore, how to provide a more efficient and accurate method for estimating the health status of batteries has become a technical problem that needs to be solved urgently. Summary of the Invention

[0005] The purpose of some embodiments of the present application is to provide a method, storage medium and electronic device for estimating the health status of a battery. Through the technical solutions of the embodiments of the present application, accurate estimation of the health status of the battery can be achieved with high estimation efficiency and good versatility.

[0006] In a first aspect, some embodiments of the present application provide a method for estimating a battery state of health, comprising: inputting a charging feature vector during a battery charging process into a mapping network to obtain an initial predicted value of the battery state of health; determining a predicted value of the battery state of health based on the initial predicted value, the charging feature vector and a dynamic network; wherein the charging feature vector includes multidimensional charging features and charging cycle features; wherein the mapping network and the dynamic network constitute a fusion neural network model; utilizing an objective function to solve a loss value between the predicted value and the true value of the battery state of health; wherein the objective function includes data fitting loss, physical residual loss and monotonic change trend loss of the battery state of health; optimizing model parameters of the fusion neural network model based on the loss value to obtain a battery state of health evaluation model; wherein the battery state of health evaluation model is used to predict the real-time value of the battery state of health.

[0007] Some embodiments of this application input charging feature vectors extracted during the charging process into a fusion neural network model constructed in this application to output a predicted value; then, using an objective function containing multiple losses to determine the loss between the predicted value and the true value, the fusion neural network model is optimized to obtain a battery health assessment model. This application can accurately estimate the battery health state by extracting the basic charging feature vectors during the charging process. It is applicable to various battery health estimation scenarios, has good versatility, and high estimation efficiency.

[0008] In some embodiments, before inputting the charging feature vector during the battery charging process into the mapping network, the method further includes: selecting a charging segment for feature extraction from the battery charging process; performing feature extraction on the voltage change curve and the current change curve within the charging segment to obtain a multi-dimensional original charging feature; and normalizing the multi-dimensional original charging feature and the number of charging cycles to obtain the multi-dimensional charging feature and the charging cycle feature.

[0009] Some embodiments of the present application extract features within a selected charging segment to obtain multidimensional original charging features, and then normalize them based on the number of charging cycles to obtain multidimensional charging features and charging cycle features, thereby achieving data standardization and providing a data basis for subsequent model training.

[0010] In some embodiments, determining the predicted value of the battery health state based on the initial predicted value, the charging characteristic vector and the dynamic network includes: constructing a network input vector based at least on the initial predicted value and the charging characteristic vector; and inputting the network input vector into the dynamic network to obtain the predicted value.

[0011] Some embodiments of the present application process the charging feature vector through a mapping network and a dynamic network to obtain a predicted value, thereby achieving accurate prediction of the battery health status.

[0012] In some embodiments, constructing a network input vector based at least on the initial prediction value and the charging characteristic vector includes: fitting a differential equation corresponding to the battery health state during the battery charging process; performing differential extraction on the differential equation to obtain a first differential result and a second differential result; and concatenating the charging characteristic vector, the initial prediction value, the first differential result, and the second differential result to obtain the network input vector.

[0013] Some embodiments of the present application construct a network input vector of a dynamic network through an initial prediction value and a differential result related to the battery health state, so as to achieve an accurate prediction of the battery health state.

[0014] In some embodiments, the mapping network is connected to a decay rate network; the method further includes: inputting the initial prediction value into the decay rate network to obtain a change trend of the battery health status; wherein the decay rate network is a fully connected network, a residual prediction network, or an inflection point detection network.

[0015] Some embodiments of the present application determine the change value of the battery health state through the initial prediction value and the decay rate network, thereby realizing the prediction of the change trend of the battery health state.

[0016] In some embodiments, the objective function is constructed by the following method: setting the first coefficient and the second coefficient corresponding to the physical residual loss and the monotonic change trend loss respectively; constructing the objective function based on the physical residual loss, the monotonic change trend, the first coefficient, the second coefficient and the data fitting loss.

[0017] Some embodiments of the present application construct objective functions through multiple losses, which can achieve accurate optimization of the model and improve the accuracy and robustness of the final battery health status assessment model.

[0018] In some embodiments, the data fitting loss represents the mean square error between the predicted value and the true value; the physical residual loss represents the differential residual value; and the monotonic change trend loss represents the consistency between the predicted trend of the predicted value and the true trend of the true value.

[0019] In a second aspect, some embodiments of the present application provide a method for estimating the battery health status, including: obtaining a real-time charging feature vector in the current charging process; inputting the charging feature vector into a battery health status evaluation model to obtain a real-time value of the battery health status; wherein the battery health status evaluation model is obtained by executing any method embodiment in the first aspect.

[0020] In some embodiments, the method further includes: confirming that the real-time value is less than a set threshold, triggering a battery maintenance reminder to inform relevant personnel.

[0021] In a third aspect, some embodiments of the present application provide a device for estimating a battery health state, comprising: a training module for inputting a charging feature vector during a battery charging process into a mapping network to obtain an initial predicted value of the battery health state; determining a predicted value of the battery health state based on the initial predicted value, the charging feature vector and the dynamic network; wherein the charging feature vector includes multidimensional charging features and charging cycle features; a loss calculation module for using an objective function to solve the loss value between the predicted value and the true value of the battery health state; wherein the objective function includes the data fitting loss, physical residual loss and monotonic change trend loss of the battery health state; an optimization module for optimizing the model parameters of the fusion neural network model based on the loss value to obtain a battery health state evaluation model; wherein the battery health state evaluation model is used to predict the real-time value of the battery health state.

[0022] In a fourth aspect, some embodiments of the present application provide a device for estimating the battery health status, including: an acquisition module for acquiring a real-time charging characteristic vector in the current charging process; an estimation module for inputting the charging characteristic vector into a battery health status evaluation model to obtain a real-time value of the battery health status; wherein the battery health status evaluation model is obtained by executing any method embodiment in the first aspect.

[0023] In a fifth aspect, some embodiments of the present application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can implement the method described in any embodiment of the first and second aspects.

[0024] In a sixth aspect, some embodiments of the present application provide an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the program, it may implement the method described in any embodiment of the first and second aspects.

[0025] In a seventh aspect, some embodiments of the present application provide a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, it can implement the method described in any embodiment of the first and second aspects. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of some embodiments of the present application, the following is a brief introduction to the drawings required for use in some embodiments of the present application. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0027] Figure 1 A system diagram of battery health status estimation provided for some embodiments of the present application; Figure 2 A flow chart of a method for obtaining a battery health status assessment model provided in some embodiments of the present application; Figure 3 A flow chart of a method for estimating the state of health of a battery provided in some embodiments of the present application; Figure 4 One of the block diagrams of the device for estimating the state of health of a battery provided in some embodiments of the present application; Figure 5 A block diagram of a device for estimating the state of health of a battery provided in some embodiments of the present application; Figure 6 A schematic diagram of an electronic device is provided for some embodiments of the present application. DETAILED DESCRIPTION

[0028] The technical solutions in some embodiments of the present application will be described below in conjunction with the drawings in some embodiments of the present application.

[0029] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and should not be understood as indicating or implying relative importance.

[0030] In the field of lithium-ion battery management systems (BMS), physical modeling and data-driven methods are commonly used to estimate SOH. Specifically, the physical modeling method is mainly based on equivalent circuit models (ECMs) and pseudo-2D electrochemical models (P2Ds) to achieve SOH estimation. This method requires the establishment of a large number of chemical and thermodynamic parameters, which is computationally intensive. Furthermore, parameter identification is difficult, and repeated calibration is required for different chemical compositions. Another data-driven method mainly uses regression, support vector machines, Gaussian processes, deep neural networks, and other methods to directly extract features from complete charge and discharge curves to regress SOH. However, this method has high requirements for feature versatility and requires the design of specialized features for different data sets or operating conditions, resulting in poor generalization capabilities. Furthermore, it often relies on full discharge curves, which are difficult to obtain in actual electric vehicles and energy storage systems.

[0031] From the above-mentioned related technologies, it can be seen that the existing SOH estimation methods have the problems of poor stability, high calibration cost, and difficulty in balancing versatility, estimation accuracy and computational efficiency.

[0032] In view of this, some embodiments of the present application provide a method for estimating the state of health of a battery. This method can extract a charging feature vector from a certain section of the charging process and input it into the mapping network and dynamic network designed in this application to obtain a predicted value of the battery state of health; then, an objective function containing multiple losses is used to calculate the loss value between the predicted value and the true value, and based on this, a fusion neural network model composed of the mapping network and the dynamic network is optimized to obtain a battery state of health assessment model. The battery state of health assessment model can be used to predict the battery state of health. The method for estimating the state of health provided in the embodiment of the present application uses easily extracted charging feature vectors to train the model to obtain a battery state of health assessment model. The model has good stability, can take into account versatility, estimation accuracy and estimation efficiency, and is highly practical. It can be seen that the embodiment of the present application does not require a full discharge curve and can be applied to a variety of battery types and different charging and discharging protocols; at the same time, in other scenarios, by optimizing model parameters, rapid adaptation between batteries with different chemical compositions can also be achieved.

[0033] The following is combined with Figure 1 The overall structure of the battery health status estimation system provided by some embodiments of the present application is exemplified.

[0034] like Figure 1 As shown, some embodiments of the present application provide a system diagram for estimating the state of health of a battery. The system diagram for estimating the state of health of a battery may include: a device terminal including a lithium battery pack and a data processing terminal 200. The data processing terminal 200 may obtain a charging segment of a certain cycle during the charging process of the device terminal 100; the data processing terminal 200 extracts multi-dimensional charging features from the charging segment, combines the charging cycle features to form a real-time charging feature vector, and inputs the vector into a pre-trained battery state of health assessment model that has been deployed internally to obtain a real-time prediction value of the state of health.

[0035] In order to achieve accurate estimation of the health status of lithium-ion batteries, it is first necessary to obtain a battery health status assessment model through training.

[0036] The following is combined with Figure 2 The implementation process of obtaining a battery health status assessment model performed by the data processing terminal 200 provided in some embodiments of the present application is exemplified.

[0037] Please see the attached Figure 2 , Figure 2 A flow chart of a method for obtaining a battery health status assessment model is provided for some embodiments of the present application.

[0038] In some embodiments of the present application, before executing the following method for obtaining a battery health status assessment model, it is first necessary to prepare a training data set, which contains charging feature vectors in different cycles and their corresponding real values ​​of SOH.

[0039] Specifically, in some embodiments of the present application, the charging feature vector is obtained by the following method: selecting a charging segment for feature extraction from the battery charging process; performing feature extraction on the voltage change curve and the current change curve in the charging segment respectively to obtain a multidimensional original charging feature; normalizing the multidimensional original charging feature and the number of charging cycles to obtain the multidimensional charging feature and the charging cycle feature.

[0040] For example, in a specific embodiment of the present application, the final constant current-constant voltage (CC-CV) charging stage of the battery charging process is selected (as a specific example of a charging stage); the voltage range is 4.0V to 4.2V (the latter part of the CC stage), and the current range is 0.2A to 1.0A (the initial part of the CV stage). This segment of data is relatively stable and widely present in different charging strategies, with good versatility and accessibility. The following eight statistical features are extracted from the voltage-time curve (as a specific example of a voltage change curve) and the current-time curve (as a specific example of a current change curve) of the final charging stage, for a total of 16 features (as a specific example of multi-dimensional raw charging features). The current cycle number d (as a specific example of the number of charging cycles) is added, and these and the 16 features together form a 17-dimensional input feature vector.

[0041] Specifically, the eight statistical features mainly include: mean (Mean), standard deviation (Std), skewness and kurtosis (used to measure the data distribution form), curve slope, cumulative power, entropy (Shannon entropy measures the information complexity of the curve) and charging time (that is, the time span at the end of charging).

[0042] The mean and standard deviation are calculated using conventional formulas. The slope of the curve is calculated as follows:

[0043] in, t i and x i are the known parameters in the curve respectively; For parameters t The mean of For parameters x The mean of .

[0044] Cumulative power Q: .

[0045] Where, I i is the current at the i-th moment, is the time-varying value.

[0046] Afterwards, all the above input features are scaled using the Min-Max normalization method so that all 17-dimensional features are mapped to the range of [-1, 1] to ensure the stability of model training and the balance of contribution of each feature.

[0047] The normalization formula of the Min-Max normalization method is:

[0048] in, represents the result after normalization of the i-th dimension feature (the set of all features is represented as j), Represents the i-th dimension feature.

[0049] After normalization through the above formula, multi-dimensional charging characteristics and charging cycle characteristics can be obtained.

[0050] The following is an example of the process of obtaining the battery health status assessment model. It should be understood that the battery health status assessment model is obtained by training a fusion neural network model composed of a mapping network and a dynamic network.

[0051] In some embodiments of the present application, a method for obtaining a battery health status assessment model includes: S210 , inputting a charging feature vector during the battery charging process into a mapping network to obtain an initial prediction value of the battery health state; wherein the charging feature vector includes a multi-dimensional charging feature and a charging cycle feature.

[0052] For example, the core network structure constructed in the specific embodiments of this application is a physical information neural network structure (PINN) (as a specific example of a fusion neural network model). Its core structure consists of a mapping network F and a dynamic network G, with a cascaded decay rate network connected after F to supplement the nonlinear trend change characteristics. By inputting the above-mentioned normalized 17-dimensional features into the physical information neural network structure, the prediction of SOH during the training process is achieved, and the predicted value is obtained.

[0053] Specifically, enter the variable F(d', j ; ); where d' is the current battery cycle characteristic (as a specific example of the charging cycle characteristic, that is, the result of d normalization); for example, the 1st cycle or the 200th cycle represents the time dimension of the battery aging process, reflecting the long-term trend on the SOH decline curve. jIt is a 16-dimensional statistical feature vector extracted and normalized from the end of CC-CV charging, including the mean, slope, entropy and other features of the voltage and current curves, which is highly representative and accessible. is the set of trainable parameters of F (e.g., neural network weights W and bias b), which are continuously optimized by gradient descent during training. Correspondingly, F can output the predicted value of SOH for the current cycle, which is recorded as: =F(d', j ). Specifically, the network structure of F contains multiple layers of MLP (Multilayer Perceptron), each layer uses torch.sin() as the activation function, the weight initialization uses Xavier Normal, and the Dropout probability is 0.2. In order to improve the stability of model training, the embodiment of the present application uses a periodic activation function to realize feature mapping under physical constraints, which improves the smoothness and frequency domain response capability of the SOH curve, and is more in line with the nonlinear oscillation trend in battery degradation; for small sample training, sin activation can more stably fit the SOH differential evolution trend than ReLU or GELU. This method uses a 0.2 probability Dropout, not only to prevent overfitting, but also to suppress the problem of excessive coupling between statistical features and degradation trends. When there is only one piece of battery data and the dimension is high (17 dimensions), the model is prone to overfitting of "remembering every dimension"; the introduction of Dropout enables the network to be trained in multiple suboptimal paths, thereby enhancing the generalization ability and supporting the transfer learning goal.

[0054] In some embodiments of the present application, the mapping network is connected to a decay rate network; the method further includes: inputting the initial prediction value into the decay rate network to obtain a change trend of the battery health status; wherein the decay rate network is a fully connected network, a residual prediction network or an inflection point detection network.

[0055] For example, in a specific embodiment of the present application, after the output of the F network, a small fully connected network (as a specific example of a decay rate network) is connected to learn the rate of decrease of the battery SOH (i.e., ΔSOH, as a specific example of a change trend), which is used to assist in improving the prediction sensitivity of the model at the aging inflection point. ΔSOH represents the degradation rate of a cycle, which is non-stationary, nonlinear, and has sudden changes. After the introduction of the ΔSOH branch, the model has the ability to structurally model the "change trend", especially in the stage of accelerated decay. In addition to using ΔSOH, the fully connected network of the ΔSOH branch can also be replaced with a residual prediction network. By letting the residual prediction network learn ΔSOH (ΔSOH k =SOH k -SOH k-1), which is used as the supervisory signal; thus, F has the modular trend modeling function and can learn the nonlinear mapping of the first-order derivative. In the specific implementation, a two-layer MLP can be used, and the input includes d', j and SOH k-1 In addition, a change-point-aware subnet can be introduced to train a small network to determine whether the current cycle is in the "accelerated degradation phase." This network outputs a binary value or soft score, which serves as a dynamic weighting factor for the ΔSOH output. The SOH curve typically exhibits a clear S-shaped degradation trend, and the network can determine this by the rate of change of the trend (e.g., the third-order derivative).

[0056] S220 : Determine a predicted value of the battery health state based on the initial predicted value, the charging characteristic vector, and a dynamic network.

[0057] In some embodiments of the present application, S220 may include: S221: Construct a network input vector based at least on the initial prediction value and the charging feature vector.

[0058] For example, in the specific embodiment of the present application, F output and other feature data as the network input vector of G.

[0059] In some embodiments of the present application, S221 may include: fitting a differential equation corresponding to the battery health status during the battery charging process; performing differential extraction on the differential equation to obtain a first differential result and a second differential result; and splicing the charging characteristic vector, the initial prediction value, the first differential result, and the second differential result to obtain the network input vector.

[0060] For example, in a specific embodiment of the present application, the implicit degenerate differential equation of SOH is first fitted:

[0061] Then, automatic differential extraction is performed to obtain the first differential result and the second differential result. The extraction formula of the first differential result is:

[0062] The extraction formula of the second differential result is:

[0063] Then the relevant data are spliced ​​to obtain the network input vector [d', j , , , ], a total of 35 dimensions.

[0064] S222: Input the network input vector into the dynamic network to obtain the predicted value.

[0065] For example, in a specific embodiment of the present application, the above 35-dimensional network input vector is input into G to obtain the predicted value .

[0066] To achieve precise optimization of the physical information neural network, some embodiments of the present application also construct an objective function containing multiple losses; wherein the objective function includes the data fitting loss of the battery health state, the physical residual loss, and the monotonic change trend loss of the battery health state. The data fitting loss represents the mean square error between the predicted value and the true value; the physical residual loss represents the differential residual value; and the monotonic change trend loss represents the consistency between the predicted trend of the predicted value and the true trend of the true value.

[0067] Specifically, the objective function is constructed by the following method: setting the first coefficient and the second coefficient corresponding to the physical residual loss and the monotonic change trend loss respectively; constructing the objective function based on the physical residual loss, the monotonic change trend, the first coefficient, the second coefficient and the data fitting loss.

[0068] For example, in a specific embodiment of the present application, the objective function L is obtained by adding the weighted sum of the physical residual loss, the monotonic change trend, the first coefficient, and the second coefficient to the data fitting loss. The formula of the objective function L is:

[0069] in, α is the first coefficient, β is the second coefficient, both of which are empirically adjusted parameters.

[0070] L data Fitting loss to the data to ensure that the predicted values ​​fit the true labels:

[0071] in, is the predicted value of the i-th sample predicted by the model, is the true value of SOH of the i-th sample; MSE is the mean square error.

[0072] L PDEIt is the physical residual loss, which is a differential residual constructed based on automatic differentiation and dynamic network to simulate the implicit dynamic law of the degradation process; the difference between the predicted value and the autograd value constitutes the PDE residual:

[0073] in, is the partial derivative of the model prediction value with respect to time, indicating the degradation rate of SOH with cycle time. f is the difference between the actual derivative and the model predicted derivative, f 1 and f 2 may be the residuals corresponding to data from two different samples or batches.

[0074] It should be noted that the MLP in this embodiment is not trained in isolation, but is back-propagated together with the PDE constraint term. The continuous differentiability of the sin activation function makes the autograd calculation , More stable. The network model in this application must satisfy the gradient existence and smoothness under automatic differentiation.

[0075] L mono For the monotonic change trend loss (i.e., monotonicity constraint), by introducing the monotonic physical prior, a symbolic consistency product loss function is designed to suppress unreasonable prediction rebound and strengthen the SOH degradation trend modeling. Specifically:

[0076] in, The change of SOH predicted by the physical information neural network model, The actual SOH change. If the model's predicted trend is inconsistent with the actual trend, the product is positive, the ReLU has output, and is penalized; if the trends are consistent, the product is negative, the ReLU = 0, and no penalty is imposed.

[0077] S230 , using an objective function to solve a loss value between the predicted value and the actual value of the battery health state.

[0078] For example, in a specific embodiment of the present application, the objective function L constructed above is used to solve the prediction The loss value between the value and the true SOH.

[0079] S240, optimizing the model parameters of the fusion neural network model based on the loss value to obtain a battery health status assessment model; wherein the battery health status assessment model is used to predict the real-time value of the battery health status.

[0080] For example, in a specific embodiment of the present application, the loss values ​​obtained above are used to optimize the model parameters of the physical information neural network to obtain a battery health assessment model that meets the requirements. For example, through optimization, a battery health assessment model with a set accuracy can be obtained; or after the optimization reaches the number of iterations of the training model, a battery health assessment model can be output.

[0081] During the dataset construction phase of training the physical information neural network described above, a homemade battery can be used to collect complete charge and discharge data from 200 cycles. This data can be used to construct 200 charging feature vectors and ground truth training samples, including 150 training samples, 30 validation samples, and 20 test samples. Model hyperparameters can then be set, such as batch size, total number of epochs, warm-up epochs, warm-up learning rate, initial learning rate, final learning rate, and learning rate strategy. Multiple independent experiments (e.g., 10) can be performed until the results are stable. Training can be terminated when the average MAPE is between 0.6% and 1.2% and the RMSE is less than 0.04.

[0082] After training, the model is exported to the ONNX format; a lightweight inference engine is deployed in the MCU of the embedded BMS to complete online SOH inference; this battery health status assessment model has low inference latency, small computing resource usage, and strong adaptability.

[0083] The following is combined with Figure 3 The specific process of battery health status estimation provided by some embodiments of the present application is exemplified.

[0084] Please see the attached Figure 3 , Figure 3 A flowchart of a method for estimating the battery state of health is provided for some embodiments of the present application. The method for estimating the battery state of health may include: S310, obtaining a real-time charging feature vector during the current charging process; S320, inputting the charging feature vector into a battery state of health assessment model to obtain a real-time value of the battery state of health.

[0085] For example, in a specific embodiment of the present application, by collecting signals during the battery charging process and then performing feature extraction and normalization processing, a real-time charging feature vector composed of the aforementioned 17-dimensional features can be obtained. The battery health assessment model trained in the above embodiment is loaded, and the 17-dimensional features are input into it, and the current SOH value (as a specific example of a real-time value) and the decay rate ΔSOH are output.

[0086] In some embodiments of the present application, the method for estimating the battery health status further includes: confirming that the real-time value is less than a set threshold, triggering a battery maintenance prompt to inform relevant personnel.

[0087] For example, in a specific embodiment of this application, if the current SOH value is less than 0.8 (as a specific example of a set threshold), a maintenance reminder is triggered to remind relevant personnel to perform battery maintenance. Real-time SOH prediction can support battery-related policy customization and battery life prediction.

[0088] Through some of the above-mentioned embodiments of this application, it can be seen that this application only requires the complete charge and discharge data of a single battery, without the need for additional collection of different chemical systems or working conditions, which can reduce calibration costs and realize convenient acquisition of training data sets; the fully connected network has low inference latency and is suitable for embedded real-time operation; this application does not require complex physical models, relying only on CC-CV charging statistical features and cascade networks. The feature extraction and network structure are universal and can be seamlessly migrated to other battery types or larger-scale data sets. Therefore, this application provides a lithium-ion battery SOH estimation method with high precision, high robustness and low calibration cost. This method is very suitable for fields with strict requirements on battery health management, such as electric vehicles, grid energy storage, aerospace, etc., and has high practicality.

[0089] Please refer to Figure 4 , Figure 4 A block diagram illustrating the components of an apparatus for estimating the state of health of a battery provided in some embodiments of the present application is provided. It should be understood that the apparatus for estimating the state of health of a battery corresponds to the aforementioned method embodiment and is capable of executing each step involved in the aforementioned method embodiment. The specific functions of the apparatus for estimating the state of health of a battery can be found in the description above, and a detailed description is omitted here to avoid repetition.

[0090] Figure 4 The device for estimating the battery state of health includes at least one software functional module that can be stored in a memory in the form of software or firmware or fixed in the device for estimating the battery state of health. The device for estimating the battery state of health includes: a training module 410, configured to input a charging feature vector during battery charging into a mapping network to obtain an initial predicted value of the battery state of health; determine the predicted value of the battery state of health based on the initial predicted value, the charging feature vector, and a dynamic network; wherein the charging feature vector includes multidimensional charging features and charging cycle features; wherein the mapping network and the dynamic network form a fusion neural network model; a loss calculation module 420, configured to use an objective function to solve the loss value between the predicted value and the actual value of the battery state of health; wherein the objective function includes data fitting loss of the battery state of health, physical residual loss, and monotonic change trend loss of the battery state of health; and an optimization module 430, configured to optimize model parameters of the fusion neural network model based on the loss value to obtain a battery state of health assessment model; wherein the battery state of health assessment model is used to predict the real-time value of the battery state of health.

[0091] Please refer to Figure 5 , Figure 5 A block diagram illustrating the components of an apparatus for estimating the state of health of a battery provided in some embodiments of the present application is provided. It should be understood that the apparatus for estimating the state of health of a battery corresponds to the aforementioned method embodiment and is capable of executing each step involved in the aforementioned method embodiment. The specific functions of the apparatus for estimating the state of health of a battery can be found in the description above, and a detailed description is omitted here to avoid repetition.

[0092] Figure 5 The device for estimating the battery health state includes at least one software functional module that can be stored in a memory in the form of software or firmware or solidified in the device for estimating the battery health state. The device for estimating the battery health state includes: an acquisition module 510, used to obtain a real-time charging feature vector in the current charging process; an estimation module 520, used to input the charging feature vector into a battery health state evaluation model to obtain a real-time value of the battery health state.

[0093] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working process of the device described above can refer to the corresponding process in the aforementioned method, and will not be described in detail here.

[0094] Some embodiments of the present application further provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can implement the operations corresponding to any of the above methods provided in the above embodiments.

[0095] Some embodiments of the present application further provide a computer program product, which includes a computer program, wherein when the computer program is executed by a processor, it can implement the operations corresponding to any of the above methods provided in the above embodiments.

[0096] like Figure 6 As shown, some embodiments of the present application provide an electronic device 600, which includes: a memory 610, a processor 620, and a computer program stored in the memory 610 and executable on the processor 620, wherein the processor 620 can implement a method as described in any of the above embodiments when reading the program from the memory 610 through the bus 630 and executing the program.

[0097] Processor 620 can process digital signals and can include various computing architectures, such as a complex instruction set computer architecture, a reduced instruction set computer architecture, or an architecture that implements a combination of multiple instruction sets. In some examples, processor 620 can be a microprocessor.

[0098] The memory 610 can be used to store instructions executed by the processor 620 or data related to the execution of instructions. These instructions and / or data may include code for implementing some or all functions of one or more modules described in the embodiments of this application. The processor 620 of the embodiment of the present disclosure can be used to execute the instructions in the memory 610 to implement the method shown above. The memory 610 includes dynamic random access memory, static random access memory, flash memory, optical storage, or other memory known to those skilled in the art.

[0099] The foregoing is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included within the scope of protection of the present application. It should be noted that similar reference numerals and letters represent similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined or explained in subsequent figures.

[0100] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0101] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

Claims

1. A method for estimating a battery health state, characterized in that: include: Inputting a charging feature vector during the battery charging process into a mapping network to obtain an initial prediction value of the battery health state; wherein the charging feature vector includes a multi-dimensional charging feature and a charging cycle feature; Determining a predicted value of the battery health state based on the initial predicted value, the charging characteristic vector, and a dynamic network; wherein the mapping network and the dynamic network constitute a fusion neural network model; Calculating a loss between the predicted value and the true value of the battery health state using an objective function; wherein the objective function includes a data fitting loss of the battery health state, a physical residual loss, and a monotonic change trend loss of the battery health state; The model parameters of the fusion neural network model are optimized based on the loss value to obtain a battery health status assessment model; wherein the battery health status assessment model is used to predict the real-time value of the battery health status.

2. The method according to claim 1, wherein Before inputting the charging feature vector during the battery charging process into the mapping network, the method further includes: Selecting a charging segment for feature extraction from the battery charging process; Extracting features of the voltage change curve and the current change curve in the charging section to obtain multi-dimensional original charging features; Normalization is performed on the multi-dimensional original charging characteristics and the number of charging cycles to obtain the multi-dimensional charging characteristics and the charging cycle characteristics.

3. The method according to claim 1 or 2, wherein: The determining the predicted value of the battery health state based on the initial predicted value, the charging characteristic vector, and the dynamic network includes: constructing a network input vector based at least on the initial prediction value and the charging feature vector; The network input vector is input into the dynamic network to obtain the predicted value.

4. The method according to claim 3, wherein The constructing a network input vector based at least on the initial prediction value and the charging feature vector includes: Fitting a differential equation corresponding to the battery health state during the battery charging process; Performing differential extraction on the differential equation to obtain a first differential result and a second differential result; The charging feature vector, the initial prediction value, the first differential result, and the second differential result are concatenated to obtain the network input vector.

5. The method according to claim 3, wherein The mapping network is connected to a decay rate network; the method further comprises: The initial prediction value is input into the decay rate network to obtain the change trend of the battery health state; wherein the decay rate network is a fully connected network, a residual prediction network or an inflection point detection network.

6. The method according to any one of claims 1-2, 4-5, characterized in that The objective function is constructed as follows: Setting a first coefficient and a second coefficient corresponding to the physical residual loss and the monotonic change trend loss respectively; The objective function is constructed based on the physical residual loss, the monotonic change trend, the first coefficient, the second coefficient, and the data fitting loss.

7. The method according to any one of claims 1-2, 4-5, characterized in that The data fitting loss represents the mean square error between the predicted value and the true value; the physical residual loss represents the differential residual value; The monotonic change trend loss represents the consistency between the predicted trend of the predicted value and the true trend of the true value.

8. A method for estimating a battery health state, characterized in that: include: Obtain the real-time charging feature vector during the current charging process; The charging characteristic vector is input into a battery health status assessment model to obtain a real-time value of the battery health status; wherein the battery health status assessment model is obtained by executing the method according to any one of claims 1 to 7.

9. The method according to claim 8, wherein The method further comprises: If it is confirmed that the real-time value is less than the set threshold, a battery maintenance reminder is triggered to inform relevant personnel.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein the computer program is executed by a processor to perform the method according to any one of claims 1 to 9.

11. An electronic device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the computer program executes the method according to any one of claims 1 to 9 when the processor runs the computer program.

Images