Battery health state monitoring method and device of energy storage battery

Abstract

The invention discloses a battery health state monitoring method for an energy storage battery, and the method comprises the steps: obtaining historical operation data of the energy storage battery, carrying out the labeling of the historical operation data through the health state of the battery, and enabling the historical operation data and a label to form a data set; constructing an initial model, wherein the initial model comprises an input module, a multi-branch feature extraction module, a dynamic gating fusion module, a time sequence dependence enhancement module and an output module; training the initial model by using a data set so as to output a monitoring model for predicting the health state of the battery; and inputting the operation data of the energy storage battery into the monitoring model to obtain the battery health state of the current energy storage battery. The invention also provides a battery health state monitoring device. The method provided by the invention can provide a reliable basis for state maintenance, service life management and system scheduling of the battery.

Description

Technical Field

[0001] This invention belongs to the field of battery management system technology, and particularly relates to a method and device for monitoring the health status of energy storage batteries. Background Technology

[0002] With the large-scale integration of renewable energy and the development of smart grids, energy storage systems are playing an increasingly important role in applications such as power peak shaving, microgrids, and smoothing the output of renewable energy. Among these, lithium-ion batteries and other energy storage batteries are widely used due to their high energy density and long cycle life. However, energy storage batteries are affected by various factors during long-term operation (such as temperature, charge / discharge rate, and cycle count), leading to gradual performance degradation and potentially even safety hazards. Therefore, accurate and efficient monitoring of the State of Health (SOH) of energy storage batteries has become a crucial issue in ensuring the reliability and safety of energy storage systems.

[0003] Currently, common battery health status monitoring methods mainly include model-driven methods, data-driven methods, and hybrid methods. Model-driven methods rely on physical modeling and are limited by the complexity of modeling and the difficulty of obtaining parameters. Data-driven methods, such as machine learning and deep learning models, have strong nonlinear fitting and feature learning capabilities and have achieved initial success in the field of health prediction. However, traditional data-driven methods often rely on single-modal signals, such as using only voltage, current, or capacity sequences, making it difficult to comprehensively capture the multidimensional features of the battery degradation process. In addition, most methods ignore the time dependence and dynamic changes of intermodal information during battery degradation, resulting in insufficient accuracy and robustness of health assessment results. Therefore, there is an urgent need for a health status monitoring method that integrates multimodal features, has time-series modeling capabilities, and can adaptively adjust feature weights to improve the accuracy of SOH prediction and system adaptability, meeting the needs of energy storage systems for high-reliability and highly intelligent battery management.

[0004] Patent document CN120294587A discloses a method and system for monitoring and predicting the health status of a battery by fusing multimodal features. It obtains multimodal data by periodically collecting static data and real-time dynamic data of the battery, and then realizes the prediction through steps such as data monitoring, shared low-level feature extraction, SOH prediction, RUL prediction, and multi-objective joint optimization.

[0005] Patent document CN119807906A discloses a deep learning method and apparatus for predicting cell failure in a battery management system (BMS). The method includes: collecting multimodal operating data of the battery cell and dividing it into time-series samples; extracting time-domain features, frequency-domain features, and difference features of the time-series samples, using Pearson correlation coefficient and mutual information for feature selection, and constructing a dataset through principal component analysis for dimensionality reduction; constructing a cell failure prediction model based on a gated recurrent unit model, and optimizing the model performance by combining an attention mechanism; predicting battery health status and failure time based on cell failure prediction, introducing an adaptive failure threshold mechanism to dynamically adjust the failure judgment criteria; and constructing an incremental learning framework based on sample selection to dynamically adjust the prediction model through new operating condition samples. This scheme relies on traditional indicators such as Pearson coefficient and mutual information for feature selection, and the overall feature modeling is biased towards statistical and non-deep representation. Summary of the Invention

[0006] The purpose of this invention is to provide a method and device for monitoring the health status of energy storage batteries. This method can provide a reliable basis for battery status maintenance, life management, and system scheduling.

[0007] To achieve the first objective of this invention, the following technical solution is provided: a method for monitoring the health status of an energy storage battery, comprising the following steps: Acquire historical operating data of energy storage batteries, and label the historical operating data according to the battery health status. Combine the historical operating data and labels to form a dataset. An initial model is constructed, which includes an input module, a multi-branch feature extraction module, a dynamic gating fusion module, a temporal dependency enhancement module, and an output module. The input module is used to input the current operating data; The multi-branch feature extraction module includes a time-domain branch, a frequency-domain branch, and a statistical branch set in parallel. The time-domain branch is used to extract time-domain abrupt change features in the running data, the frequency-domain branch is used to extract frequency-domain trend features in the running data, and the statistical branch is used to extract physical statistical features in the running data. The dynamic gating fusion module is used to weight and fuse time-domain abrupt change features, frequency-domain trend features, and physical statistical features to generate fused features. The temporal dependency enhancement module performs a self-attention mechanism operation on the fused features based on the time series to output enhanced features; The output module makes predictions based on the enhancement features and outputs the prediction results. The initial model is trained using the dataset to output a monitoring model for predicting battery health status. The operating data of the energy storage battery is input into the monitoring model to obtain the current battery health status.

[0008] This invention constructs a multi-branch feature extraction architecture consisting of time-domain, frequency-domain, and statistical branches to capture local mutations, frequency-domain energy changes, and statistical indicators during battery degradation from multiple perspectives. It also introduces a learnable dynamic gating fusion mechanism to automatically adjust the weights of each modal feature according to the battery degradation stage, thereby achieving adaptive fusion of information from different modalities and effectively improving the model's generalization ability and robustness under various operating conditions.

[0009] Specifically, the historical operating data includes voltage, current, temperature sequence, battery capacity, and battery internal resistance.

[0010] Specifically, the time-domain branch uses three layers of dilated convolution to extract features from the voltage and current time-domain signals in the historical operating data, so as to output the corresponding time-domain abrupt change features; The frequency domain branch generates a time-frequency graph by performing a short-time Fourier transform on all time-series signals in the historical running data, and extracts the long-distance frequency domain dependence and the energy attenuation mode of the focused internal resistance frequency band from the time-frequency graph through the Transformer encoder to output frequency domain trend features. The statistical branch calculates the peak position of the incremental curve of battery capacity and the internal resistance temperature coefficient, and maps them to a high-dimensional feature vector through a fully connected layer to output the corresponding physical statistical features.

[0011] Specifically, the dynamic gating fusion module adjusts the feature weights of each mode according to the different stages of battery aging.

[0012] Specifically, the expression for the dynamic gating fusion module is as follows:

[0013] In the formula, This represents the gate weight matrix; the closer an element's value is to 1, the more important the corresponding feature. W 1 , W 2 , W 3 For learnable parameters, the dimensions are optimized through backpropagation as follows: , This is the bias vector.

[0014] Specifically, the time-series dependency enhancement module uses bidirectional LSTM to process the fused feature sequence in order to capture the dependency relationship between aging accumulation and future trends, and outputs the hidden state sequence. The self-attention mechanism calculates the weights for each time step based on the output hidden state sequence, and then generates a weighted feature vector as the enhanced feature output based on the time step weights.

[0015] Specifically, the expression for the timing dependency enhancement module is as follows:

[0016] In the formula, For the first t The attention weights for each loop sum to 1. For Bi-LSTM t The bidirectional hidden state at each time step, D For the hidden layer dimension; Let H be the attention weight matrix, where H = [h1, h2, ..., h T ] represents the hidden state sequence.

[0017] To achieve the second objective of this invention, the following technical solution is provided: a battery health status monitoring device for performing the steps of the above-described method for monitoring the battery health status of an energy storage battery.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: A multi-branch feature extraction architecture consisting of time-domain branch, frequency-domain branch, and statistical branch was constructed, which can capture local mutations, frequency-domain energy changes, and statistical indicators in the battery degradation process from multiple perspectives. This breaks through the limitations of traditional single-modality feature extraction and improves the comprehensiveness and accuracy of feature representation. By introducing a learnable dynamic gating fusion mechanism, the weights of each modal feature are automatically adjusted according to the battery degradation stage, realizing adaptive fusion of information from different modalities and effectively improving the model's generalization ability and robustness under various operating conditions. Attached Figure Description

[0019] Figure 1 This is a flowchart of a battery health status monitoring method for an energy storage battery provided in this embodiment. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0021] like Figure 1 As shown, this embodiment provides a method for monitoring the battery health status of an energy storage battery, and its steps are as follows: Acquire historical operating data of energy storage batteries, and label the historical operating data according to the battery health status. Combine the historical operating data and labels to form a dataset. An initial model is constructed, which includes an input module, a multi-branch feature extraction module, a dynamic gating fusion module, a temporal dependency enhancement module, and an output module. The input module is used to input the current operating data; The multi-branch feature extraction module includes a time-domain branch, a frequency-domain branch, and a statistical branch set in parallel. The time-domain branch is used to extract time-domain abrupt change features in the running data, the frequency-domain branch is used to extract frequency-domain trend features in the running data, and the statistical branch is used to extract physical statistical features in the running data. The dynamic gating fusion module is used to weight and fuse time-domain abrupt change features, frequency-domain trend features, and physical statistical features to generate fused features. The temporal dependency enhancement module performs a self-attention mechanism operation on the fused features based on the time series to output enhanced features; The output module makes predictions based on the enhancement features and outputs the prediction results. The initial model is trained using the dataset to output a monitoring model for predicting battery health status. The operating data of the energy storage battery is input into the monitoring model to obtain the current battery health status.

[0022] More specifically, the overall structure of the model consists of the following five layers: input layer, multi-branch feature extraction layer, dynamic gating fusion layer, temporal dependency enhancement layer, and output layer.

[0023] For the original signal of the input layer, the time-domain abrupt change, frequency-domain trend and physical statistical features of the battery signal are extracted in parallel through time-domain branch, frequency-domain branch and statistical branch. The time-domain abrupt change, frequency-domain trend and physical prior features of the battery degradation signal are deconstructed, forming a parallel cross-modal extraction architecture of time-domain-frequency-statistical features, breaking through the representation limitations of single mode.

[0024] Based on the feature data generated by each branch, the contribution of multimodal features is dynamically adjusted using learnable gating weights in the dynamic gating fusion layer to generate fused features. The weights of each modal feature are automatically adjusted according to the different battery aging stages to achieve dynamic feature selection.

[0025] The time-domain branch employs a 3-layer dilated CNN to process voltage / current time-series signals, with hole ratios of 1, 2, and 4 respectively. Each layer is followed by BatchNorm and ReLU activation to capture local abrupt changes at different time scales. The time-series signal is then subjected to a Short-Time Fourier Transform (STFT) to generate a time-frequency map, converting the time-domain signal into frequency-domain energy distribution features. A 2-layer Transformer encoder is used to extract long-range frequency-domain dependencies, focusing on the energy decay mode in the internal resistance-related frequency band (10-100Hz). The peak position of the capacity increment curve (dQ / dV) and the internal resistance temperature coefficient are calculated and mapped to a high-dimensional feature vector through a fully connected layer.

[0026] The output dimensions of each branch layer are as follows: ; In the formula, F CNN Indicates the output of the time-domain branch. C 1 For the number of CNN channels, T For sequence length, F Transformer For frequency domain branch output, C 2 Hide dimensions for Transformer F stat For statistical branch output, C 3 For statistical dimensions.

[0027] The expressions for calculating the gating weights and the feature fusion of the adaptive gating network are as follows:

[0028] In the formula, This represents the gate weight matrix; the closer an element's value is to 1, the more important the corresponding feature. W 1 , W2 , W 3 For learnable parameters, the dimensions are optimized through backpropagation as follows: , This is the bias vector.

[0029] Bi-LSTM (Bi-LSTM) processing of fused feature sequences F fusion Capture the dependency between aging accumulation and future trends, and output the hidden state sequence H=[h1,h2,…,h T The self-attention mechanism calculates the weights at each time step based on the output hidden state sequence, and then generates a weighted feature vector based on the time step weights. The overall expression is as follows:

[0030] In the formula, For the first t The attention weights for each loop sum to 1. For Bi-LSTM t The bidirectional hidden state at each time step, D For the hidden layer dimension; This is the attention weight matrix.

[0031] This embodiment also provides a battery health status monitoring device for performing the steps of the battery health status monitoring method for energy storage batteries provided in the above embodiments.

[0032] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0033] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 Devices that specify the functions in one or more boxes. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0034] This invention does not rely on expert experience, greatly saves manpower, helps improve the operational stability of DC distribution networks, and has significant engineering application value and promotion prospects.

[0035] Furthermore, the terms "upper," "lower," "inner," "outer," "front," and "rear" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Unless otherwise specifically stated, the relative steps, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention.

[0036] Of course, the above description is only a specific embodiment of the present invention and is not intended to limit the scope of the present invention. All equivalent changes or modifications made to the structure, features and principles described in the claims of the present invention should be included in the scope of the claims of the present invention.

[0037] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method of monitoring the state of health of an energy storage battery, characterized by, Includes the following steps: Acquire historical operating data of energy storage batteries, and label the historical operating data according to the battery health status. Combine the historical operating data and labels to form a dataset. An initial model is constructed, which includes an input module, a multi-branch feature extraction module, a dynamic gating fusion module, a temporal dependency enhancement module, and an output module. The input module is used to input the current operating data; The multi-branch feature extraction module includes a time-domain branch, a frequency-domain branch, and a statistical branch set in parallel. The time-domain branch is used to extract time-domain abrupt change features in the running data, the frequency-domain branch is used to extract frequency-domain trend features in the running data, and the statistical branch is used to extract physical statistical features in the running data. The dynamic gating fusion module is used to weight and fuse time-domain abrupt change features, frequency-domain trend features, and physical statistical features to generate fused features. The temporal dependency enhancement module performs a self-attention mechanism operation on the fused features based on the time series to output enhanced features; The output module makes predictions based on the enhancement features and outputs the prediction results. The initial model is trained using the dataset to output a monitoring model for predicting battery health status. The operating data of the energy storage battery is input into the monitoring model to obtain the current battery health status.

2. The method of state of health monitoring of an energy storage cell according to claim 1, characterized in that, The historical operating data includes voltage, current, temperature sequence, battery capacity, and battery internal resistance.

3. The method of claim 1, wherein, The time-domain branch uses three layers of dilated convolution to extract features from the voltage and current time-series signals in the historical operating data, so as to output the corresponding time-domain abrupt change features. The frequency domain branch generates a time-frequency graph by performing a short-time Fourier transform on all time-series signals in the historical running data, and extracts the long-distance frequency domain dependence and the energy attenuation mode of the focused internal resistance frequency band from the time-frequency graph through the Transformer encoder to output frequency domain trend features. The statistical branch calculates the peak position of the incremental curve of battery capacity and the internal resistance temperature coefficient, and maps them to a high-dimensional feature vector through a fully connected layer to output the corresponding physical statistical features.

4. The method of state of health monitoring of an energy storage battery of claim 1, wherein, The dynamic gating fusion module adjusts the feature weights of each mode according to the different stages of battery aging.

5. The method of state of health monitoring of an energy storage cell according to claim 1 or 4, characterized in that The expression for the dynamic gating fusion module is as follows: ； wherein, denotes the gating weight matrix, the closer the element value is to 1, the more important the corresponding feature is, W 1 , W 2 , W 3 are learnable parameters, optimized by backpropagation, with dimensions , is a bias vector.

6. The method of state of health monitoring of an energy storage battery of claim 1, wherein, The temporal dependency enhancement module uses bidirectional LSTM to process the fused feature sequence to capture the dependency relationship between aging accumulation and future trends, and outputs the hidden state sequence. The self-attention mechanism calculates the weights for each time step based on the output hidden state sequence, and then generates a weighted feature vector as the enhanced feature output based on the time step weights.

7. The method of claim 1 or 6, wherein, The expression for the timing dependency enhancement module is as follows: ； wherein, is the attention weight for the t th cycle, and the sum is 1, is the bidirectional hidden state of the t th time step of the Bi-LSTM, D is the hidden layer dimension; is the attention weight matrix, H = [h1, h2, …, h T ] represents the hidden state sequence.

8. A battery state of health monitoring apparatus characterized by comprising: The steps are for performing the battery health status monitoring method for energy storage batteries as described in any one of claims 1 to 7.

Images