Lithium ion battery health state evaluation method and system

Abstract

The invention provides a lithium ion battery health state evaluation method and system, and relates to the technical field of new energy, and the method comprises the steps: obtaining a plurality of groups of electrochemical impedance spectroscopy data of a lithium ion battery in different charging and discharging states, the electrochemical impedance spectroscopy data being associated with battery operation time and marked with an actual health state; sequencing each group of electrochemical impedance spectroscopy data according to the sequence of battery operation time to form a time sequence data set; according to the time sequence data set, taking the electrochemical impedance spectroscopy data as input and the actual health state as output, and training to obtain a battery health state evaluation model; and collecting current electrochemical impedance spectroscopy data of the lithium ion battery to be evaluated, and inputting the current electrochemical impedance spectroscopy data into the battery health state evaluation model to obtain a health state evaluation result of the lithium ion battery. The method has the beneficial effects that high-precision battery health prediction can be provided, intelligent and automatic battery management can be realized, and the safety and reliability of a battery system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy, and particularly relates to a method and system for evaluating the state of health of a lithium-ion battery. Background Art

[0002] Lithium-ion batteries are widely used in energy storage systems, electric vehicles and other fields due to their advantages such as high energy density, long cycle life, and environmental friendliness. The accurate estimation of their state of health (SOH, State of Health) or remaining service life parameters is the key to ensuring the safe and long-term stable operation of the system.

[0003] Currently, the methods for estimating the state of health (SOH) of lithium-ion batteries mainly include:

[0004] 1. Model-based estimation methods: Usually, physical models or electrochemical models are required to simulate the behavior of the battery, and the SOH is deduced according to the working state of the battery. Commonly used models include equivalent circuit models and electrochemical models.

[0005] Among them, the equivalent circuit model estimates the SOH of the battery by simulating parameters such as the voltage, current, and temperature of the battery. Commonly used circuit models include RC circuits and Thevenin models. By modeling the charge and discharge process of the battery, the change in the internal resistance of the battery can be estimated, and then the SOH can be inferred.

[0006] The electrochemical model estimates the SOH by simulating the chemical reactions and ion flow inside the battery. Common electrochemical models include the P2D (pseudo-2D) model. The electrochemical model can provide a relatively accurate SOH estimate, but the complexity of modeling and calculation is relatively high.

[0007] 2. Data-driven methods: With the progress of data acquisition technology and the rapid development of machine learning, data-driven SOH estimation methods have gradually become the mainstream. Data-driven technologies can automatically learn the complex relationship between battery performance and state of health from a large amount of historical data. Compared with traditional physical model-based methods, data-driven models do not require explicit physical process modeling, so they can more accurately capture the non-linear characteristics and complex laws during the battery degradation process, thus providing a more accurate SOH estimate.

[0008] However, the current data-driven lithium-ion battery SOH estimation methods also have certain limitations. For example, the change in the state of health of the battery is usually gradual and affected by various long-term use factors. Current and voltage data may provide sufficient information in the short term, but after the battery undergoes multiple charge and discharge cycles, extreme operations such as deep discharge, simply relying on these external data may not be able to achieve accurate SOH estimation and prediction over a long period. Summary of the Invention

[0009] In view of the problems existing in the prior art, the present invention provides a method for evaluating the health state of a lithium-ion battery, including:

[0010] Step S1, obtaining multiple groups of electrochemical impedance spectroscopy data of the lithium-ion battery under different charge and discharge states, where the electrochemical impedance spectroscopy data is associated with the battery operation time and marked with the actual health state;

[0011] Step S2, sorting each group of the electrochemical impedance spectroscopy data according to the chronological order of the battery operation time to form a time series data set;

[0012] Step S3, based on the time series data set, using the electrochemical impedance spectroscopy data as the input and the actual health state as the output, training to obtain a battery health state evaluation model;

[0013] Step S4, collecting the current electrochemical impedance spectroscopy data of the lithium-ion battery to be evaluated, and inputting the current electrochemical impedance spectroscopy data into the battery health state evaluation model to obtain the health state evaluation result of the lithium-ion battery.

[0014] Preferably, in step S1, multiple groups of the electrochemical impedance spectroscopy data have different measurement frequencies.

[0015] Preferably, the battery operation time is the charge and discharge time or the number of charge and discharge cycles.

[0016] Preferably, before performing step S2, data preprocessing is further included for the electrochemical impedance spectroscopy data, and the data preprocessing includes noise removal processing, and / or normalization processing, and / or normalization processing.

[0017] Preferably, in step S2, feature extraction is further included for multiple groups of the electrochemical impedance spectroscopy data to obtain the real part, imaginary part, modulus value, and phase angle of the impedance of the electrochemical impedance spectroscopy data, and form a multi-dimensional feature vector;

[0018] Subsequently, the corresponding multi-dimensional feature vector is used as the electrochemical impedance spectroscopy data and sorted according to the chronological order of the battery operation time to form the time series data set.

[0019] Preferably, the battery health state evaluation model is an Informer deep learning model;

[0020] The Informer deep learning model includes a multi-scale time encoder, a self-attention mechanism, and a decoder connected in sequence.

[0021] Preferably, in step S3, the mean square error is used as the loss function during the training process of the battery health state evaluation model.

[0022] The present invention also provides a lithium-ion battery state of health assessment system, which applies the above-mentioned lithium-ion battery state of health assessment method, and includes:

[0023] A data acquisition module, configured to acquire multiple groups of electrochemical impedance spectroscopy data of a lithium-ion battery under different charge and discharge states, where the electrochemical impedance spectroscopy data is associated with the battery operation time and marked with the actual state of health;

[0024] A sequence generation module, connected to the data acquisition module, configured to sort each group of the electrochemical impedance spectroscopy data according to the chronological order of the battery operation time to form a time series data set;

[0025] A model training module, connected to the sequence generation module, configured to train a battery state of health assessment model based on the time series data set, with the electrochemical impedance spectroscopy data as the input and the actual state of health as the output;

[0026] A state assessment module, connected to the model training module, configured to collect the current electrochemical impedance spectroscopy data of the lithium-ion battery to be evaluated, and input the current electrochemical impedance spectroscopy data into the battery state of health assessment model to obtain the state of health assessment result of the lithium-ion battery.

[0027] Preferably, the sequence generation module includes:

[0028] A feature extraction unit, configured to extract features from multiple groups of the electrochemical impedance spectroscopy data to obtain the real part, imaginary part, modulus value, and phase angle of the impedance of the electrochemical impedance spectroscopy data, and form a multi-dimensional feature vector;

[0029] A sorting unit, connected to the feature extraction unit, configured to use the corresponding multi-dimensional feature vector as the electrochemical impedance spectroscopy data and sort it according to the chronological order of the battery operation time to form the time series data set.

[0030] Preferably, the battery state of health assessment model is an Informer deep learning model;

[0031] The Informer deep learning model includes a multi-scale time encoder, a self-attention mechanism, and a decoder connected in sequence.

[0032] The above technical solution has the following advantages or beneficial effects:

[0033] 1) Electrochemical impedance spectroscopy (EIS) data can provide detailed electrochemical characteristic information about the internal state of the battery, capturing minute changes during the battery degradation process, ensuring high accuracy in estimating the state of health (SOH) of the battery. Meanwhile, by combining the powerful time-series prediction ability, strong adaptability, and generalization ability of the SOH assessment model, it can effectively handle changes under different battery types, usage environments, or charge-discharge strategies, and has a strong preventive maintenance ability to detect health risks in advance before the battery performance significantly degrades. In addition, it avoids the complex physical modeling process, reduces the modeling difficulty, makes battery health monitoring more flexible and scalable, and can be widely applied to multiple scenarios such as electric vehicles and energy storage systems;

[0034] 2) It avoids manual intervention, improves the automation level of battery SOH prediction, can monitor the battery health status in real time, and gives an accurate SOH estimate. Compared with traditional methods, it can not only provide high-precision battery health prediction but also achieve intelligent and automated battery management, enhancing the safety and reliability of the battery system. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 In a preferred embodiment of the present invention, it is a schematic flow diagram of a method for evaluating the state of health of a lithium-ion battery;

[0036] Figure 2 In a preferred embodiment of the present invention, it is a schematic structural diagram of an Informer deep learning model;

[0037] Figure 3 In a preferred embodiment of the present invention, it is a schematic structural diagram of a system for evaluating the state of health of a lithium-ion battery. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0038] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The present invention is not limited to this embodiment, and other embodiments can also fall within the scope of the present invention as long as they conform to the gist of the present invention.

[0039] In a preferred embodiment of the present invention, in view of the above problems existing in the prior art, a method for evaluating the state of health of a lithium-ion battery is provided, as Figure 1 shown, including:

[0040] Step S1, obtaining multiple groups of electrochemical impedance spectroscopy data of the lithium-ion battery under different charge-discharge states, where the electrochemical impedance spectroscopy data is associated with the battery operation time and labeled with the actual state of health;

[0041] Step S2, sorting each group of electrochemical impedance spectroscopy data according to the chronological order of the battery operation time to form a time-series data set;

[0042] Step S3: Based on the time series dataset, using the electrochemical impedance spectroscopy data as the input and the actual health state as the output, train to obtain a battery health state evaluation model;

[0043] Step S4: Collect the current electrochemical impedance spectroscopy data of the lithium-ion battery to be evaluated, and input the current electrochemical impedance spectroscopy data into the battery health state evaluation model to obtain the health state evaluation result of the lithium-ion battery.

[0044] Specifically, in this embodiment, for the acquisition of the electrochemical impedance spectroscopy data in step S1, the impedance spectroscopy data of the battery can be obtained by frequency scanning under different charge / discharge states. Among them, the electrochemical impedance spectroscopy data includes amplitude and phase information, presented in the form of complex numbers (complex impedance), which contains the impedance responses at different frequencies. In a preferred embodiment of the present invention, in step S1, multiple sets of electrochemical impedance spectroscopy data have different measurement frequencies. The measurement frequencies here preferably include multiple frequency points, and the frequency range covers the low-frequency region and the high-frequency region, ranging from a few Hz to dozens of kHz. Among them, the low-frequency region mainly reflects the electrochemical reaction process of the battery, and the high-frequency region reflects the internal resistance and electrode surface characteristics of the battery.

[0045] Further, after obtaining the electrochemical impedance spectroscopy data, it also includes data preprocessing of the electrochemical impedance spectroscopy data. The data preprocessing includes noise removal processing, and / or standardization processing, and / or normalization processing.

[0046] Specifically, the acquisition of the electrochemical impedance spectroscopy data is usually interfered by noise. Therefore, it is necessary to remove high-frequency noise through methods such as filtering. After removing the noise, it is necessary to standardize or normalize the data to eliminate the dimensional differences under different measurement conditions.

[0047] After the data preprocessing, it also includes feature extraction of multiple sets of electrochemical impedance spectroscopy data, that is, extracting meaningful features from the original electrochemical impedance spectroscopy data. The features here include the real part, imaginary part, modulus value, and phase angle of the impedance of the electrochemical impedance spectroscopy data, and constructing a multi-dimensional feature vector based on the real part, imaginary part, modulus value, and phase angle of the impedance;

[0048] Subsequently, the corresponding multi-dimensional feature vector is used as the electrochemical impedance spectroscopy data and sorted in the order of the battery operation time to form a time series dataset.

[0049] Specifically, the electrochemical impedance spectroscopy data is essentially time-varying. Therefore, it is necessary to associate the electrochemical impedance spectroscopy data with the operation time of the battery to form a time series dataset. Here, the time series dataset can also be understood as the lithium-ion battery aging dataset. It contains multiple multi-dimensional feature vectors, and each multi-dimensional feature vector can contain the real part, imaginary part, modulus value, and phase angle of the impedance at different measurement frequencies corresponding to the battery operation time point.

[0050] Among them, correlating the electrochemical impedance spectroscopy data with the battery operation time means that for each set of electrochemical impedance spectroscopy data, the charge-discharge time or the number of charge-discharge cycles is marked as the battery operation time. At the same time, considering that the above time series data set needs to be used to train the battery health state evaluation model, the actual health state also needs to be marked for each set of electrochemical impedance spectroscopy data. The actual health state here can be the degradation characteristics of lithium-ion batteries, such as capacity loss, internal resistance change, etc., preferably the remaining capacity of the battery.

[0051] Furthermore, step S3 uses the time series data set as the input and combines a deep learning model for SOH prediction. In terms of model selection, the present invention adopts the Informer deep learning model. The Informer deep learning model is an efficient time series prediction model based on Transformer and is suitable for modeling long time series data. Compared with traditional models such as LSTM and GRU, the Informer deep learning model has stronger long time series dependence modeling ability and is suitable for processing time series data with long-term dependence. The Informer deep learning model is based on the ideas of self-attention mechanism and time series decomposition, can process long time series data, and has high prediction accuracy. The main advantages of the Informer deep learning model include efficient calculation methods, low computational complexity, and good adaptability to large-scale time series data.

[0052] After determining the model, the previously extracted electrochemical impedance spectroscopy features (such as the real part, imaginary part, modulus value, and phase angle of the impedance at different measurement frequencies) are used as the input features of the model. Each input sequence is composed of electrochemical impedance spectroscopy data at multiple battery operation time points. The time series data set is used as the training data set for model training. The goal is to enable the Informer deep learning model to predict the future SOH state of the lithium-ion battery based on the current electrochemical impedance spectroscopy data. The mean square error (MSE) is used as the loss function during model training to measure the gap between the prediction result and the actual label. Preferably, when the mean square error (MSE) is not less than 1%, the model parameters are optimized and then training continues. When the mean square error (MSE) is less than 1%, it indicates that the model training is completed. After training is completed, the model is verified and evaluated. Finally, the trained Informer deep model is deployed to the actual battery management system (BMS) to achieve accurate estimation of the SOH of the lithium-ion battery.

[0053] Compared with the existing lithium-ion battery SOH estimation technology, the present invention combines the high-precision data acquisition ability of electrochemical impedance spectroscopy technology and the powerful time series prediction ability of the Informer deep learning model, and has significant advantages.

[0054] First of all, electrochemical impedance spectroscopy can provide detailed electrochemical characteristic information of the internal state of the battery, capture tiny changes during the battery degradation process, and ensure high accuracy in SOH estimation. The Informer deep learning model has excellent long-term time series modeling capabilities, can handle the long-term dependencies of the battery health state changing over time, and has high computational efficiency, being suitable for real-time processing of large-scale data. This method avoids manual intervention through automatic feature extraction, improves the automation level of the system, can monitor the battery health status in real time, and give accurate SOH estimation. Compared with traditional methods, the deep learning-based model has strong adaptability and generalization capabilities, can effectively cope with changes under different battery types, usage environments or charge-discharge strategies, and at the same time has a powerful preventive maintenance ability, and can detect health risks in advance before the battery performance significantly degrades.

[0055] In addition, this method also avoids the complex physical modeling process, reduces the modeling difficulty, makes the battery health monitoring more flexible and scalable, and can be widely applied to multiple scenarios such as electric vehicles and energy storage systems. Therefore, this SOH estimation method based on electrochemical impedance spectroscopy data and the Informer deep learning model can not only provide high-precision battery health prediction, but also realize intelligent and automated battery management, improving the safety and reliability of the battery system.

[0056] As Figure 2 shown, the Informer deep learning model includes a multi-scale time encoder, a self-attention mechanism, and a decoder connected in sequence.

[0057] Specifically, the input of the multi-scale time encoder is used as the input of the Informer deep learning model, and the output of the decoder is used as the output of the Informer deep learning model. Among them, through the multi-scale time encoder, feature representations at different time scales can be obtained, and then the output of the multi-scale time encoder is used as the input of the self-attention mechanism. The importance of each element is calculated through the self-attention mechanism with an adaptive length, and finally the output of the self-attention mechanism with an adaptive length is input into the decoder to obtain the final predicted health state evaluation result.

[0058] The present invention also provides a lithium-ion battery health state evaluation system, which applies the above-mentioned lithium-ion battery health state evaluation method. As Figure 3 shown, it includes:

[0059] A data acquisition module 1, which is used to acquire multiple groups of electrochemical impedance spectroscopy data of lithium-ion batteries under different charge-discharge states. The electrochemical impedance spectroscopy data is associated with the battery operation time and labeled with the actual health state;

[0060] The sequence generation module 2 is connected to the data acquisition module 1 and is used to sort each group of electrochemical impedance spectroscopy data according to the chronological order of battery operation time to form a time series data set;

[0061] The model training module 3 is connected to the sequence generation module 2 and is used to train a battery health state evaluation model with the electrochemical impedance spectroscopy data as the input and the actual health state as the output according to the time series data set;

[0062] The state evaluation module 4 is connected to the model training module 3 and is used to collect the current electrochemical impedance spectroscopy data of the lithium-ion battery to be evaluated, input the current electrochemical impedance spectroscopy data into the battery health state evaluation model to obtain the health state evaluation result of the lithium-ion battery.

[0063] In a preferred embodiment of the present invention, the sequence generation module 2 includes:

[0064] The feature extraction unit 21 is used to extract features from multiple groups of electrochemical impedance spectroscopy data to obtain the real part, imaginary part, modulus value and phase angle of the impedance of the electrochemical impedance spectroscopy data, and form a multi-dimensional feature vector;

[0065] The sorting unit 22 is connected to the feature extraction unit 21 and is used to take the corresponding multi-dimensional feature vector as the electrochemical impedance spectroscopy data and sort it according to the chronological order of battery operation time to form a time series data set.

[0066] In a preferred embodiment of the present invention, the battery health state evaluation model is an Informer deep learning model;

[0067] The Informer deep learning model includes a multi-scale time encoder, a self-attention mechanism and a decoder connected in sequence.

[0068] The above are only preferred embodiments of the present invention, and do not limit the implementation manners and protection scope of the present invention. For those skilled in the art, it should be realized that all equivalent replacements and obvious changes made by using the content of this specification and the drawings should be included in the protection scope of the present invention.

Claims

1. A method for evaluating the health status of a lithium-ion battery, characterized in that: include: Step S1, obtaining multiple sets of electrochemical impedance spectroscopy data of lithium-ion batteries under different charge and discharge states, wherein the electrochemical impedance spectroscopy data are associated with battery operation time and annotated with actual health status; Step S2, sorting each group of the electrochemical impedance spectroscopy data according to the order of the battery operation time to form a time series data set; Step S3, training a battery health status assessment model based on the time series data set, taking the electrochemical impedance spectroscopy data as input and the actual health status as output; Step S4, collecting current electrochemical impedance spectroscopy data of the lithium-ion battery to be evaluated, and inputting the current electrochemical impedance spectroscopy data into the battery health status evaluation model to obtain a health status evaluation result of the lithium-ion battery.

2. The lithium-ion battery health status assessment method according to claim 1, characterized in that: In the step S1, the multiple groups of electrochemical impedance spectroscopy data have different measurement frequencies.

3. The lithium-ion battery health status assessment method according to claim 1, characterized in that: The battery operation time is the charge and discharge time or the number of charge and discharge cycles.

4. The method for evaluating the health status of a lithium-ion battery according to claim 1, wherein: Before executing the step S2, the electrochemical impedance spectroscopy data is further preprocessed, and the data preprocessing includes noise removal processing, and / or standardization processing, and / or normalization processing.

5. The method for evaluating the health status of a lithium-ion battery according to claim 1, characterized in that: The step S2 further includes extracting features from the multiple groups of electrochemical impedance spectroscopy data to obtain the real part, imaginary part, modulus and phase angle of the impedance of the electrochemical impedance spectroscopy data, and forming a multi-dimensional feature vector; Then, the corresponding multi-dimensional feature vectors are used as the electrochemical impedance spectrum data and are sorted in chronological order according to the battery operation time to form the time series data set.

6. The lithium-ion battery health status assessment method according to claim 1, characterized in that: The battery health status assessment model is an Informer deep learning model; The Informer deep learning model includes a multi-scale temporal encoder, a self-attention mechanism, and a decoder connected in sequence.

7. The lithium-ion battery health status assessment method according to claim 1, characterized in that: In step S3, mean square error is used as a loss function during the training process of the battery health status assessment model.

8. A lithium-ion battery health status assessment system, characterized in that: The method for assessing the health status of a lithium-ion battery as claimed in any one of claims 1 to 7 comprises: A data acquisition module, used to acquire multiple sets of electrochemical impedance spectroscopy data of lithium-ion batteries under different charge and discharge states, wherein the electrochemical impedance spectroscopy data are associated with battery operation time and annotated with actual health status; A sequence generation module, connected to the data acquisition module, for sorting each group of the electrochemical impedance spectroscopy data according to the order of the battery operation time to form a time series data set; A model training module, connected to the sequence generation module, for training a battery health status assessment model based on the time series data set, with the electrochemical impedance spectroscopy data as input and the actual health status as output; The state assessment module is connected to the model training module and is used to collect current electrochemical impedance spectroscopy data of the lithium-ion battery to be assessed, and input the current electrochemical impedance spectroscopy data into the battery health state assessment model to obtain a health state assessment result of the lithium-ion battery.

9. The lithium-ion battery health status assessment system according to claim 8, characterized in that: The sequence generation module comprises: A feature extraction unit, used for performing feature extraction on a plurality of groups of electrochemical impedance spectroscopy data to obtain the real part, imaginary part, modulus and phase angle of the impedance of the electrochemical impedance spectroscopy data, and to form a multi-dimensional feature vector; A sorting unit is connected to the feature extraction unit and is used to use the corresponding multi-dimensional feature vector as the electrochemical impedance spectrum data and sort them in chronological order according to the battery operation time to form the time series data set.

10. The lithium-ion battery health status assessment system according to claim 8, characterized in that: The battery health status assessment model is an Informer deep learning model; The Informer deep learning model includes a multi-scale temporal encoder, a self-attention mechanism, and a decoder connected in sequence.

Images