Lithium ion battery health state evaluation method and equipment based on deconstruction physical information, medium and product

Abstract

The invention discloses a lithium ion battery health state evaluation method and device based on deconstruction physical information, a medium and a product, and relates to the technical field of lithium ion battery health state evaluation, and the method comprises the steps: extracting a multi-dimensional health factor, inputting the multi-dimensional health factor into a deep physical information neural network fusing a self-attention mechanism module and a Koopman neural operator module, and obtaining a deep physical information neural network; and outputting the SOH estimation value. By extracting the universal health factors suitable for multiple working conditions, the problem that the universality of the health factors is insufficient is solved; a self-attention mechanism is utilized to enhance features and reduce redundancy, and the perception ability of the model to key information is enhanced; physical information is deconstructed by means of a Koopman neural operator, a physical mechanism of battery degradation is fused into the model, and the physical interpretability of the model is enhanced; the deep physical information neural network is fused with multi-dimensional information for estimation, individual differences and complex working conditions of different batteries are effectively dealt with, and therefore high-precision and high-robustness lithium ion battery SOH estimation is achieved.

Description

Technical Field

[0001] The present application relates to the technical field of lithium-ion battery health status assessment, and in particular to a lithium-ion battery health status assessment method, device, medium and product based on deconstruction of physical information. Background Art

[0002] Lithium-ion batteries have been increasingly widely used in new energy vehicles and grid-side energy storage due to their many advantages, including high energy density, low self-discharge rate, long cycle life, and lack of memory effect. However, during long-term operation, lithium batteries often face complex and changing operating conditions and external environmental conditions, which causes irreversible chemical and physical degradation within the battery, manifested as a continuous increase in battery impedance and continuous capacity decay. In extreme cases, it may even cause internal short circuits, leading to thermal runaway accidents, resulting in huge economic losses and serious safety accidents. Therefore, achieving high-precision and robust battery state of health (SOH) estimation is crucial for building a reliable battery management system (BMS), which can effectively ensure the safe and stable operation of the entire system.

[0003] However, the current lithium-ion battery health status estimation method cannot guarantee the high accuracy and strong robustness of lithium-ion battery SOH evaluation performance. Summary of the Invention

[0004] The purpose of this application is to provide a lithium-ion battery health status assessment method, equipment, medium and product based on deconstructed physical information, which can achieve high-precision and strong robustness of lithium-ion battery SOH estimation.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a method for assessing the health status of a lithium-ion battery based on deconstructed physical information, comprising:

[0007] Real-time collection of operating data of the target lithium-ion battery charging stage at each cycle number;

[0008] Performing feature extraction on the operating data under each set of cycle numbers to obtain a multi-dimensional universal health factor set;

[0009] The multidimensional universal health factor set is used as input, and a trained lithium-ion battery state of health assessment model is used to output an estimated SOH value for each cycle number, wherein the lithium-ion battery state of health assessment model is a deep physical information neural network that integrates a self-attention mechanism module and a Koopman neural operator module. The self-attention mechanism module is used to perform feature enhancement on the multidimensional universal health factor set and output an enhanced feature set. The Koopman neural operator module is used to perform physical information deconstruction on the enhanced feature set and output a physical decay rate. The deep physical information neural network is used to output the estimated SOH value based on the enhanced feature set and the physical decay rate.

[0010] In a second aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the lithium-ion battery health status assessment method based on deconstructed physical information as described in the first aspect above.

[0011] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the lithium-ion battery health status assessment method based on deconstruction of physical information as described in the first aspect above.

[0012] In a fourth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the lithium-ion battery health status assessment method based on deconstruction of physical information as described in the first aspect above.

[0013] According to the specific embodiments provided in this application, this application has the following technical effects:

[0014] The present application provides a lithium-ion battery health status assessment method, device, medium and product based on deconstruction of physical information. The method collects the operating data of the charging stage of the target lithium-ion battery at each cycle in real time, extracts a multidimensional universal health factor set from it, and then inputs it into a deep physical information neural network (a trained lithium-ion battery health status assessment model) that integrates a self-attention mechanism module and a Koopman neural operator module: the self-attention mechanism module performs feature enhancement on the multidimensional universal health factor set to output an enhanced feature set, and the Koopman neural operator module performs physical information deconstruction on the enhanced feature set to output a physical decay rate. Finally, the deep physical information neural network combines the enhanced feature set and the physical decay rate to output the SOH estimation value at each cycle. This solution solves the problem of insufficient universality of health factors by extracting universal health factors applicable to multiple working conditions; uses the self-attention mechanism to enhance features and reduce redundancy, thereby enhancing the model's perception of key information; uses the Koopman neural operator module to deconstruct physical information and integrate the physical mechanism of battery degradation into the model, enhancing the physical interpretability of the model; the deep physical information neural network fuses multi-dimensional information for estimation, effectively addressing the individual differences and complex working conditions of different batteries, thereby achieving high-precision and robust SOH estimation of lithium-ion batteries. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 This is a flow chart of a method for assessing the health status of a lithium-ion battery based on deconstructing physical information in Example 1 of the present application;

[0017] Figure 2 Schematic diagram of the overall flow of the KNO-SA-DPINN estimation model in Example 1 of the present application;

[0018] Figure 3 This is the current-voltage curve of the CCCV stage under the aging cycle in Example 1 of the present application;

[0019] Figure 4 This is the feature correlation matrix diagram in Example 1 of this application, where: Figure 4 (a) is the feature correlation matrix diagram corresponding to HI1-HI4, Figure 4 (b) is the characteristic correlation matrix diagram corresponding to HI5-HI8. Figure 4(c) is the feature correlation matrix corresponding to HI9-HI12. Figure 4 (d) is the feature correlation matrix corresponding to HI13-HI15;

[0020] Figure 5 This is a schematic diagram of the self-attention mechanism structure in Example 1 of this application;

[0021] Figure 6 This is a schematic diagram of the KNO structure in Example 1 of the present application;

[0022] Figure 7 This is a schematic diagram of the capacity attenuation trajectory in Example 1 of the present application, wherein: Figure 7 (a) is a schematic diagram of the capacity decay trajectory corresponding to the NEEPU dataset. Figure 7 (b) is a schematic diagram of the capacity decay trajectory corresponding to the XJTU dataset;

[0023] Figure 8 Schematic diagram of the SOH estimation results of the proposed method under simulated vehicle operating conditions in Example 1 of the present application;

[0024] Figure 9 Schematic diagram of the SOH estimation results of the proposed method under conventional discharge conditions in Example 1 of the present application;

[0025] Figure 10 Schematic diagram of the SOH estimation results of the proposed method under random discharge conditions in Example 1 of the present application. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0027] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0028] Example 1

[0029] Research has found that current data-driven SOH estimation methods still have shortcomings: most mathematically driven models are "black box" models, which do not require in-depth research on the characteristics of aging mechanisms, resulting in a lack of physical interpretability; the extracted health indicators (HIs) are often limited to specific operating conditions and cannot be applied to multiple operating conditions, lacking a certain degree of universality. Extracting multiple features often leads to feature redundancy, affecting the model's comprehensive estimation performance; in actual applications, there are significant inconsistencies between different batteries and the operating conditions are complex and changeable. This complex characteristic is difficult to effectively couple through the model, greatly affecting the model's estimation performance, making it impossible to ensure high accuracy and strong robustness in the SOH evaluation performance of lithium-ion batteries.

[0030] In order to solve the problems of poor cross-battery estimation accuracy and weak robustness caused by the poor universality of HI and poor physical interpretability of the model when facing complex working conditions, the current lithium-ion battery SOH estimation model is proposed. Figure 1 As shown, this embodiment provides a lithium-ion battery health status assessment method based on deconstruction of physical information, including:

[0031] SA: Real-time collection of operating data of the target lithium-ion battery charging stage at each cycle number;

[0032] SB: performing feature extraction on the operating data under each set of cycle numbers to obtain a multidimensional universal health factor set;

[0033] SC: Taking the multidimensional universal health factor set as input, the trained lithium-ion battery health status assessment model is used to output the SOH estimation value at each cycle number, wherein the lithium-ion battery health status assessment model is a deep physical information neural network that integrates a self-attention mechanism module and a Koopman neural operator module. The self-attention mechanism module is used to perform feature enhancement on the multidimensional universal health factor set and output an enhanced feature set. The Koopman neural operator module is used to perform physical information deconstruction on the enhanced feature set and output a physical decay rate. The deep physical information neural network is used to output the SOH estimation value based on the enhanced feature set and the physical decay rate.

[0034] By extracting universal health factors applicable to multiple working conditions, the problem of insufficient universality of health factors is solved; the self-attention mechanism is used to enhance features and reduce redundancy, thereby enhancing the model's perception of key information; with the help of the Koopman neural operator to deconstruct physical information, the physical mechanism of battery degradation is integrated into the model, enhancing the physical interpretability of the model; the deep physical information neural network fuses multi-dimensional information for estimation, effectively coping with individual differences and complex working conditions of different batteries, thereby achieving high-precision and robust SOH estimation of lithium-ion batteries.

[0035] The following describes in detail the lithium-ion battery health status assessment method based on deconstructed physical information of this embodiment from the perspective of training a lithium-ion battery health status assessment model.

[0036] This embodiment proposes a KNO-SA-DPINN model (i.e., a lithium-ion battery health status assessment model) for lithium-ion battery SOH estimation. First, 15 common features highly correlated with battery capacity are extracted based on battery aging data, and a self-attention mechanism is introduced for secondary screening. This process not only effectively overcomes the problem of feature redundancy, but also significantly improves the universality of HIs, providing high-quality feature input for subsequent modeling; secondly, a generalized dynamic degradation equation for batteries is constructed to characterize the complex physical degradation characteristics of batteries under the coupling of multiple factors, and more accurately reflect the aging law of batteries. The generalized dynamic degradation equation for batteries is efficiently solved through the Koopman Neural Operator (KNO) module, the physical information of the battery is deeply mined, and the estimation accuracy and interpretability of the model are enhanced; finally, the self-attention mechanism module and the KNO module are combined to build the KNO-SA-DPINN model. By deeply integrating domain knowledge with neural network training, the shortcomings of the traditional physical information neural network (PINN) in physical constraint capabilities are effectively overcome, and high-precision and robust SOH estimation is achieved. The flow chart of SOH estimation using KNO-SA-DPINN model is as follows Figure 2 The specific steps are as follows:

[0037] S1: Obtain the battery's full life cycle operation data and pre-process the aging data

[0038] Online data collection is performed on the full battery cycle of the vehicle or grid-side energy storage system. The data is then cleaned, including Gaussian noise filtering and filling in missing values ​​and outliers based on the 3σ principle. The cleaned data is grouped by cycle number and labeled with the SOH value according to formula (1). This constructs time series data, which helps the model effectively capture the time-dependent characteristics of battery degradation. Cycle number refers to the cumulative count of a lithium-ion battery completing a full charge and discharge cycle and is a key time variable for quantifying the degree of battery aging. Finally, the data is divided into training, validation, and test sets for subsequent model performance verification.

[0039]

[0040] Among them, C i is the maximum available capacity value under the i-th cycle, and C0 is the initial capacity value.

[0041] S2: Extract common features based on charging data and perform correlation analysis

[0042] Considering that the data obtained in actual applications are generally non-constant charge and discharge data, the discharge stage is random, while the charge stage is relatively fixed. Therefore, the general features that are strongly correlated with the capacity are extracted from the specific intervals of the constant current (CC) charging stage and the constant voltage (CV) charging stage to characterize the battery degradation process. The voltage-current curve of the CCCV stage under the long-term aging cycle is as follows: Figure 3 As shown, the voltage curve with a voltage window of 0.4V (i.e. [3.8V, 4.2V]) is selected to extract the common features (voltage mean: V-mean, voltage standard deviation: V-std, voltage kurtosis: V-kurt, voltage skew: V-skew) based on statistics, i.e. HI1-HI4. Figure 3 It can be seen that the voltage curve of a specific voltage window is linearly fitted, and then the angle α (V-LFA, i.e., the voltage linear fitting angle) formed with the horizontal line becomes smaller as the number of cycles increases. The change in incremental capacity (IC) on continuous voltage steps also contains many characteristics of battery aging, and these characteristic changes further map the aging mechanism of the battery. Therefore, the voltage of a specific voltage window is subjected to IC analysis, and the position of the IC peak (P-place), the height difference between the two ends of the peak (AD), the area corresponding to the peak (P-area), and V-LFA are used as HI5-HI8.

[0043] The current curve with a current window of 1A (i.e., [0A, 1A]) was selected for HI extraction. As in the CC stage, statistical features (current mean: I-mean, current standard deviation: I-std, current kurtosis: I-kurt, and current skewness: I-skew) and current linear fitting angle β (I-LFA) were extracted. In addition, in order to more comprehensively characterize the characteristics of the CV stage, the area (I-area) enclosed by the current curve and the x-axis and the sampling time (T-CV) corresponding to [0A, 1A] were extracted, i.e., HI9-HI15. HI was normalized by Min-Max to eliminate the potential impact of different feature dimension differences on subsequent analysis. The correlation matrix between HI and capacity was calculated based on the Pearson correlation coefficient (PCCs) to quantitatively characterize the linear correlation between the variables. The characteristic correlation matrix is ​​shown in the figure below. Figure 4 As shown, the average absolute value of the correlation coefficient of HI1-HI15 is 0.93, and all show a strong linear correlation trend with SOH.

[0044] S3: Feature enhancement of the extracted multidimensional universal health factor set based on the self-attention mechanism with residual

[0045] Among the 15 common features extracted by S2, Figure 4 It can be seen that the PCCs between some HIs are as high as 1, indicating that there is significant redundancy in the feature set, which can easily lead to model overfitting and weaken the generalization ability. Therefore, a self-attention mechanism that integrates residuals and layer normalization is constructed. By dynamically allocating weights and integrating multi-dimensional general feature information, the expressive power of features is further enhanced, the problem of feature redundancy is effectively avoided, and the inherent law of SOH attenuation is more accurately mapped. The introduction of residual connections helps alleviate the gradient vanishing problem in deep neural networks, and adding normalization operations can ensure the scale consistency of input data, thereby maintaining the stability of the model training process. The specific structure is as follows: Figure 5 The specific implementation steps are as follows:

[0046] S301: Assume that the input time series is X={x1,x2,…,x T}, where x i is a d-dimensional vector (d=15), and T is the sequence length. Perform a linear transformation on the input sequence to generate query, key, and value respectively:

[0047] Q=XW Q ; K = XW K ; V = XW V (2)

[0048] Among them, W Q 、W K 、W V The dimensions are T×d q 、Τ×d k 、Τ×d v , d q =d k =d v .

[0049] S302: Calculate the attention score, introduce a scaling factor to perform scaling operations and perform normalization processing to obtain the attention weight matrix A:

[0050]

[0051] Among them, d k is the dimension of the key vector.

[0052] S303: Perform a weighted sum of A and V to generate a weighted feature set Z. A residual connection is introduced to Z and normalized to obtain the final HI set Z′. This feature set Z′ serves as input data for subsequent neural operator solutions, providing a solid data foundation for efficient model training and accurate estimation.

[0053] S4: Solving the generalized dynamic equations of batteries based on the Koopman neural operator to achieve deep deconstruction of physical information

[0054] In order to better characterize the complex degradation process of the battery, a generalized dynamic equation is constructed. The specific steps are as follows:

[0055] S401: Construct a multivariate capacity characterization function that accurately describes the complex degradation characteristics of lithium batteries. The specific definition is as follows:

[0056] u=f(t,X)(4)

[0057] Where t represents time (in this embodiment, the number of cycles), and X represents a feature set consisting of various influencing factors such as state of charge, depth of discharge, temperature, charge and discharge rate, and health indicators. In this embodiment, X is the feature set Z′.

[0058] Without loss of generality, in order to better characterize the dynamic degradation mechanism of the battery, the decay rate of SOH is defined as follows:

[0059]

[0060] S402: Equation (5) is an explicit partial differential equation (PDE) parameterized by the set θ, where h(θ) represents a nonlinear function of t, X, and u. By changing h(θ), various forms of degradation models can be expressed. However, the specific parameters of the explicit PDE are difficult to obtain, and the form lacks fixedness, making it difficult to directly apply it to the actual scenario of lithium battery degradation modeling. In view of the above limitations, this paper defines a more generalized dynamic degradation equation with parameter Θ to extract the degradation mechanism of specific HI and aging time data, which is defined as follows:

[0061] u t =H(t,Z′,u,u t ,u Z′ ,u Z′Z′ ,…;Θ) (6)

[0062] Among them, u t is the decay rate of SOH, is the first-order partial derivative of u with respect to the set Z′, H() is the explicit characterization function of the generalized dynamic degradation equation of the battery, u is the capacity, and Θ is the parameter set of the generalized equation.

[0063] As a new type of neural operator solver, KNO maps nonlinear dynamics to linear space and efficiently learns the solutions of nonlinear partial differential equations through frequency division. Under the premise of ensuring prediction accuracy, KNO can accurately solve the generalized dynamic degradation equation of batteries and effectively address the challenge of predicting the long-term complex dynamic behavior of batteries. Therefore, as a solver of the generalized dynamic equation of batteries, KNO efficiently mines the degradation physical information hidden in the data. In order to improve the solution efficiency and reduce the computational complexity, equation (6) can be simplified to:

[0064] u t =H(t,Z′,u,u t ,u z′ ;Θ) (7)

[0065] S403: Assume that the input of the PDE in equation (7) is where D is (t, Z′, u, u t ,u Z′ ). Encoder is the observation function g(·), Mapping to space Transformed into And perform a fast Fourier transform:

[0066]

[0067] The Fourier series is truncated at the maximum number f of the frequency model, and the system is divided into a low-frequency part and a high-frequency part. The Koopman operator is learned in the low-frequency part, and the high-frequency fluctuation information is extracted through a one-dimensional convolutional layer in the high-frequency part.

[0068] S404: For the low frequency part, define Hankel matrix where t∈ε + , and train the o-dimensional linear layer to learn the Koopman operator Used to predict the latest observable Future state: Mapping to

[0069] S405: For the high-frequency part, the high-frequency part is independently predicted forward through a one-dimensional convolutional layer:

[0070]

[0071] S406: Two future states and Weighted summation, training Decoder to represent the observation function g -1 (·)Analytical space The target state of the solution of the equation in

[0072]

[0073] Among them, λ∈[0,1] represents the relative weight of controlling low-frequency and high-frequency information.

[0074] S5: Build the KNO-SA-DPINN model and use the trained model for online SOH estimation of lithium batteries

[0075] Based on the Deep Physics-Informed Neural Network (DPINN), combined with the residual self-attention mechanism and the KNO module, a comprehensive estimation model is constructed: KNO-SA-DPINN. The overall framework of the model is as follows Figure 6 As shown in Figure 2, the model construction steps are as follows:

[0076] S501: In order to truly integrate physical information into the neural network training process rather than just serving as a constraint, the KAPINN used to fit the generalized dynamic degradation equation of the battery (Formula 7) is defined as follows:

[0077]

[0078] S502: Fit u using a deep feedforward neural network Γ(·) t , and the final KAPINN equation is obtained:

[0079]

[0080] Wherein, t represents time (in this embodiment, the number of cycles), X represents a multi-dimensional universal feature set, and ψ represents a parameter set of a neural network.

[0081] S503: KAPINN combines the feature-enhanced KNO network to integrate the mined physical information into model training, effectively overcoming the limitations of traditional PINNs in lithium battery SOH estimation, such as poor interpretability and insufficient constraint capabilities. KAPINN's parameters ψ and Θ are trained by minimizing the mean squared error loss. The specific loss function is as follows:

[0082] ζ PINN =ζ data +αζ momo +βζ PDE (14)

[0083] Among them, data is the data loss, ζ momo is the monotonicity loss, ζ PDESolve the term loss for KNO. α and β are loss weights. For the data term loss, the data term loss function is determined by calculating the mean square error between the estimated value and the target value:

[0084]

[0085] in, is the estimated capacity of Γ(·), u i is the actual value of capacity, and i is the number of cycles.

[0086] For monotonic loss, we mainly focus on the comprehensive analysis of the overall decline trend of battery capacity and capacity regeneration phenomenon. The predicted difference is judged by the ReLU function. and the true difference (u i -u i+1 ), thereby determining the monotonicity loss function:

[0087]

[0088] For the KNO solution loss, it mainly includes prediction loss and data reconstruction loss:

[0089]

[0090] Among them, λ p ,λ r are the weights for the prediction and reconstruction processes in the control loss function, respectively.

[0091] S504: According to the principle of minimizing the loss function value, the constructed estimation model is optimized using the Adam optimizer to optimize the model hyperparameters. The optimal model is then used for experimental verification, and the estimation performance of the model is verified through the test set. Four evaluation indicators are selected to scientifically describe the experimental results, namely, Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Root Mean Squared Error (RMSE), and Fit Coefficient (R 2 ), the specific calculation formula of the indicator is as follows:

[0092]

[0093] Where N is the number of samples, u i 、 are the true capacity value and estimated capacity value of the i-th cycle respectively.

[0094] This embodiment proposes a deep physical information neural network framework that integrates KNO and self-attention mechanisms, which effectively solves the limitations of existing methods in cross-battery SOH estimation tasks. First, by constructing a multi-dimensional universal feature set, the model can capture the nonlinear dynamic characteristics of the battery degradation process, significantly improving the universality under complex working conditions. Secondly, the self-attention mechanism is introduced to realize the dynamic adjustment of feature weights, effectively overcoming the inconsistency problem caused by individual differences in batteries. Then, the nonlinear dynamic system is linearized by KNO, and the hidden physical information is deeply mined, which significantly enhances the interpretability of the model. Finally, the accuracy and robustness of the proposed model are verified by simulation examples.

[0095] To verify that the proposed lithium-ion battery state of health estimation method based on the KNO-SA-DPINN model can solve the problems of low cross-battery estimation accuracy and weak robustness caused by the weak universality of health factors and poor physical interpretability of the model in complex dynamic conditions, a self-built vehicle-based simulated operating condition lithium-ion battery aging dataset (NEEPU) and a Xi'an Jiaotong University lithium-ion battery aging dataset (XJTU) were selected for simulation examples. The specific steps for implementation are as follows:

[0096] (I): Analyze the original battery data in the battery aging dataset. The data includes the voltage and current curves during the charging phase, as well as the capacity degradation curve. This time, we used the first batch of NEEPU lithium-ion battery cycle aging datasets and the first, third, and fourth batches of XJTU lithium battery aging datasets, namely N11-4, B11-4, B31-4, and B41-4, for a total of 16 batteries for simulation verification. The battery capacity decay curve is as follows: Figure 7 The specific experimental details of the NEEPU dataset and the XJTU dataset are shown in Table 1.

[0097] Table 1 Specific experimental details of each batch of the dataset

[0098]

[0099] (II): Based on the original voltage, current and capacity data obtained, a multi-dimensional universal health factor is extracted. The extracted HI1-HI15 are labeled with SOH labels in sequence according to the number of cycles to facilitate the subsequent division of training and test sets. The labeled group data is subjected to Min-Max normalization processing. The specific formula is as follows:

[0100]

[0101] Among them, x i is the original group data, y iis the normalized group data, max(x) and min(x) are the corresponding maximum and minimum values ​​in the original group data, respectively.

[0102] (III): Divide the processed feature data into training set, validation set and test set in a ratio of 2:1:1, i.e., 2 batteries in the training set, 1 battery in the validation set and 1 battery in the test set. T} is input into the KNO-SA-DPINN model. The specific steps are as follows:

[0103] (1) Perform a linear transformation on the input sequence to obtain the query, key, and value. Then calculate the attention score, introduce a scaling factor for scaling and normalization, perform a weighted sum of the weight matrix A and V, perform a residual connection on the result, and perform layer normalization to obtain the enhanced feature combination Z′. Z′ serves as one of the input features for the subsequent KNO module solution.

[0104] (2) When the input sequence X is input into the SA module, it is input into the Deep Multilayer Perceptron (DMLP) for estimation, and the feedforward estimation is performed. Serves as one of the input features to the KNO module for solving the generalized dynamic equations of a battery.

[0105] (3) Integrate the output of the SA module and the DMLP network and use it in the KNO module to solve the battery degradation equation u t =H(t,Z′,u,u t ,u z′ ; Θ), KNO performs frequency division on the input data, mines the degradation physical information inside the battery from different dimensions, and performs dimensionality reduction on the data after fast Fourier transform in order to calculate the loss function.

[0106] (4) The entire deep physical information neural network is trained according to the principle of minimizing the loss function value until the optimal model is obtained. The optimal model is then used for test set testing, the final test results are output, and case analysis is performed.

[0107] (IV) The divided test set is used to verify the performance of the KNO-SA-DPINN model. Through cross-experiments on four batteries under different operating conditions, the comprehensive estimation performance of the proposed model is verified to test whether the proposed model meets the expected requirements. The specific steps are as follows:

[0108] (1) Set the performance evaluation index of the model. Select MAE, MAPE, RMSE and R 2 As the evaluation criteria. The smaller the MAE, MAPE and RMSE, the higher the estimation accuracy of the model; R2 The closer it is to 1, the better the model fit is and the better the estimation effect is.

[0109] (2) For the simulated vehicle operating conditions (N-Batch 1), the SOH estimation results of the proposed method are as follows: Figure 8 As shown. Figure 8 It can be seen that: in the four batteries, the SOH values ​​estimated by the KNO-SA-DPINN model show a significant linear correlation with the true value, and the data points are concentrated, with the average absolute error less than 0.0058, which proves that the proposed method has high accuracy and certain universality in cross-battery SOH estimation. Among them, the MAE, MAPE and RMSE of KNO-SA-DPINN are 0.57%, 0.64% and 0.71% respectively, and R 2 The average value reached 97.28%,

[0110] (3) For the conventional discharge condition (X-Batch 1), the SOH estimation results of the proposed method are as follows: Figure 9 As shown. Figure 9 As can be seen from the data, within each battery, the estimated results of the proposed method gradually converged towards the diagonal line, exhibiting a more tightly clustered pattern, with the average absolute error below 0.0057. This fully demonstrates that the KNO-SA-DPINN model can fully utilize its algorithmic advantages to mine the deep value of data when handling different operating conditions, significantly demonstrating the method's excellent estimation accuracy and robustness.

[0111] (4) For random discharge conditions (X-Batch 3 and X-Batch 4), the SOH estimation results of the proposed method are as follows: Figure 10 As shown. Figure 10 As can be seen from the results, KNO-SA-DPINN demonstrates excellent estimation results for different random discharge strategies. The vast majority of data points are distributed near the diagonal red line, and the average absolute error is less than 0.0061, demonstrating the accuracy and robustness of the model's cross-battery SOH estimation under different operating conditions. Among the four battery estimation results, the proposed method achieves average MAE, MAPE, and RMSE of 0.61%, 0.66%, and 0.82%, respectively, demonstrating excellent estimation accuracy.

[0112] This embodiment proposes a method for estimating the health status of lithium-ion batteries based on the KNO-SA-DPINN model. First, general features are constructed. General features applicable to a variety of complex working conditions are extracted from battery aging data to provide high-quality input features for subsequent modeling; secondly, multi-dimensional features are enhanced. The extracted multi-dimensional features are enhanced to avoid the redundancy problem caused by too many features, and the self-attention mechanism with residuals is used to reconstruct the feature set to further amplify the implicit physical information and improve the universality of HI; finally, the physical information is deeply deconstructed. The neural operator is used to solve the nonlinear partial differential equations that characterize battery degradation, deeply explore the domain knowledge hidden in the data, and truly integrate physical information into the model training process. The evolution law of the battery state is accurately portrayed, effectively overcoming the shortcomings of the traditional PINN in physical constraint capabilities, thereby achieving high-precision and robust SOH estimation.

[0113] The key to this implementation lies in the introduction of KNO to solve the generalized dynamic degradation equation for batteries, achieving a deep fusion of physical information and estimation models. This fusion provides effective physical mechanism guidance for the model, overcoming the problem of physical information imbalance during the cross-battery estimation process, thereby ensuring the model's generalization across different battery samples and operating conditions. Furthermore, the self-attention mechanism integrated with KNO-SA-DPINN further amplifies the universality of multidimensional universal features, providing more flexible solutions for the SOH estimation task of unknown lithium-ion batteries. By combining physical information with a data-driven integrated estimation framework, the proposed method provides an advanced solution for lithium-ion battery SOH estimation. The excellent performance demonstrated by this technique in simulation examples fully demonstrates its advantages in handling complex battery degradation problems, providing new research ideas and application solutions for the field of lithium-ion battery state estimation and an efficient and intelligent decision-making support tool for battery management systems. The application of this technology not only accurately assesses the health status of batteries and reduces operation and maintenance costs, but also prevents safety incidents caused by battery failures, thus possessing significant social and economic value.

[0114] Example 2

[0115] This embodiment provides a computer device, which can be a server or a terminal. The computer device includes a processor, a memory, an input / output interface (I / O for short) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. 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 for data involved in a lithium-ion battery health status assessment method based on deconstruction of physical information. 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 via a network connection. When the computer program is executed by the processor, it implements a lithium-ion battery health status assessment method based on deconstruction of physical information in Example 1.

[0116] Example 3

[0117] This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0118] Example 4

[0119] This embodiment provides a computer program product, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0120] 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.

[0121] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented 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 memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0122] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0123] The technical features of the above embodiments can 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.

[0124] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A lithium-ion battery health status assessment method based on deconstruction of physical information, characterized in that: The lithium-ion battery health status assessment method based on deconstruction of physical information includes: Real-time collection of operating data of the target lithium-ion battery charging stage at each cycle number; Performing feature extraction on the operating data under each set of cycle numbers to obtain a multi-dimensional universal health factor set; The multidimensional universal health factor set is used as input, and a trained lithium-ion battery state of health assessment model is used to output an estimated SOH value for each cycle number, wherein the lithium-ion battery state of health assessment model is a deep physical information neural network that integrates a self-attention mechanism module and a Koopman neural operator module. The self-attention mechanism module is used to perform feature enhancement on the multidimensional universal health factor set and output an enhanced feature set. The Koopman neural operator module is used to perform physical information deconstruction on the enhanced feature set and output a physical decay rate. The deep physical information neural network is used to output the estimated SOH value based on the enhanced feature set and the physical decay rate.

2. The lithium-ion battery health status assessment method based on deconstruction of physical information according to claim 1, characterized in that: The operation data includes current data and voltage data in the constant current stage and the constant voltage stage; Feature extraction is performed on the operating data under each set of cycle times to obtain a multi-dimensional universal health factor set, specifically including: Performing feature extraction on the specific voltage window data in the constant current stage to obtain the voltage average, voltage standard deviation, voltage kurtosis, voltage skewness, voltage linear fitting angle, incremental capacity peak position, incremental capacity peak height difference and incremental capacity peak area; Performing feature extraction on the specific current window data in the constant voltage stage to obtain the current average value, current standard deviation, current kurtosis, current skewness, current linear fitting angle, current curve area and sampling time; Normalizing the voltage average, the voltage standard deviation, the voltage kurtosis, the voltage skewness, the voltage linear fitting angle, the incremental capacity peak position, the height difference between the two ends of the incremental capacity peak, the incremental capacity peak area, the current average, the current standard deviation, the current kurtosis, the current skewness, the current linear fitting angle, the current curve area, and the sampling time to obtain a normalized factor; The normalized factors constitute the multidimensional universal health factor set.

3. The lithium-ion battery health status assessment method based on deconstruction of physical information according to claim 1, characterized in that: The training process of the lithium-ion battery health status assessment model includes: Obtaining full life cycle operation data of lithium-ion batteries; Preprocessing the full life cycle operation data of the lithium-ion battery to obtain time series data with SOH labels indexed by the number of cycles; Performing feature extraction on charging data in the time series data to obtain a sample multi-dimensional universal health factor set, wherein the charging data includes constant current stage charging data and constant voltage stage charging data; Performing feature enhancement on the sample multidimensional universal health factor set based on a self-attention mechanism with residual connections to obtain a sample enhanced feature set; Performing physical information deconstruction on the sample enhancement feature set based on the Koopman neural operator, and outputting the sample physical attenuation rate; The sample enhancement feature set and the sample physical decay rate are used as input, and a deep physical information neural network is trained by minimizing the loss function to output the SOH estimation value of the lithium-ion battery at each cycle number.

4. The lithium-ion battery health status assessment method based on deconstruction of physical information according to claim 3, characterized in that: Based on the self-attention mechanism with residual connection, the features of the multi-dimensional universal health factor set of the sample are enhanced to obtain the sample enhanced feature set, which specifically includes: Performing a linear transformation on the sample multidimensional universal health factor set to generate a query matrix, a key matrix, and a value matrix; Calculating an attention weight matrix based on the query matrix and the key matrix; Performing weighted summation on the attention weight matrix and the value matrix to obtain a sample feature set; A residual connection is introduced into the sample feature set and layer normalization is performed to obtain the sample enhanced feature set.

5. The lithium-ion battery health status assessment method based on deconstruction of physical information according to claim 3, characterized in that: The physical information of the sample enhancement feature set is deconstructed based on the Koopman neural operator, and the sample physical attenuation rate is output, specifically including: Constructing a generalized dynamic degradation equation for a battery based on the sample enhanced feature set and the number of cycles; Solving the generalized dynamic degradation equation of the battery based on the Koopman neural operator to obtain the sample physical attenuation rate; The Koopman neural operator is used to solve the generalized dynamic degradation equation of the battery to obtain the sample physical attenuation rate, which specifically includes: Mapping the sample enhancement feature set and the number of cycles to an observation space through an observation function, and performing Fourier transform to obtain frequency domain data; dividing the frequency domain data into a low-frequency part and a high-frequency part; constructing a Hankel matrix based on the low-frequency portion; Learning the Koopman operator from the Hankel matrix by training the linear layer, and using the Koopman operator to perform a linear transformation on the latest fragment of the matrix to predict the low-frequency prediction results under the future number of cycles; The one-dimensional convolution layer extracts the fluctuation information of the high-frequency part and obtains the high-frequency prediction results under the future cycle number; The low-frequency prediction result and the high-frequency prediction result are integrated to obtain the sample physical attenuation rate.

6. The lithium-ion battery health status assessment method based on deconstruction of physical information according to claim 5, characterized in that: The expression of the generalized dynamic degradation equation of the battery is: he t =H(t,Z′,u,u t ,he z′ ΛΘ); Among them, u t is the physical attenuation rate of the sample; H() is the explicit characterization function of the generalized dynamic degradation equation of the battery; u is the capacity; t is the number of cycles; u z' is the first-order partial derivative of u with respect to the sample enhanced feature set Z′; Θ is the parameter set of the generalized equation.

7. The lithium-ion battery health status assessment method based on deconstruction of physical information according to claim 3, characterized in that: The loss function includes data term loss, monotonicity loss and KNO solution term loss.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the lithium-ion battery health status assessment method based on deconstructed physical information as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for evaluating the health status of a lithium-ion battery based on deconstructed physical information according to any one of claims 1 to 7 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for evaluating the health status of a lithium-ion battery based on deconstructed physical information according to any one of claims 1 to 7 is implemented.

Images