Electrochemical impedance acquisition method based on large time sequence model

Abstract

The invention provides an electrochemical impedance acquisition method based on a time sequence large model. Constructing a comprehensive test matrix containing a plurality of variables of the battery material system; charging and discharging tests are carried out on a large batch of batteries under different working conditions; applying current excitation through an electrochemical workstation to obtain electrochemical impedance spectroscopy data; frequency points are extracted from the electrochemical impedance spectroscopy data to serve as candidate frequency points, electrochemical impedance values under the frequency points serve as labels, current and voltage data in the charging and discharging processes are intercepted at the same time, and a data set is divided into a training set and a testing set; on the training set, training an electrochemical impedance special large model based on a pre-trained time sequence basic model; and applying the trained electrochemical impedance special large model on the test set, obtaining an electrochemical impedance value under the candidate frequency point, and verifying the accuracy of the model. The electrochemical impedance acquisition method has the potential of being widely applied in real-time monitoring and dynamic application environments by combining a time sequence large model technology.

Description

Technical Field

[0001] The present invention relates to the technical field of Electrochemical Impedance Spectroscopy (EIS), and particularly to a method for obtaining electrochemical impedance based on a time series large model. Background Art

[0002] The internal processes of a battery involve complex electrochemical reactions, including ion migration, charge transfer, interfacial reactions, and diffusion processes, etc. These processes have important impacts on the overall performance of the battery and are directly related to key indicators such as the charge and discharge rate, cycle stability, energy density, and safety of the battery.

[0003] Electrochemical Impedance Spectroscopy (EIS) is an important tool for evaluating the performance of electrochemical systems. The EIS technique applies a small-amplitude sinusoidal voltage or current perturbation through an electrochemical workstation and measures the resulting current or voltage response to obtain impedance data at different frequencies. These impedance data can be used to analyze various kinetic processes inside the battery, including the redox reactions of electrode materials, ion transport in the electrolyte solution, and the characteristics of the electrode / electrolyte interface, etc., so as to evaluate the performance and health status of the battery.

[0004] However, there are some significant challenges in traditional methods for obtaining electrochemical impedance. First, the data processing is complex and the response speed is slow. Traditional electrochemical workstations usually take a long time to complete the measurement of impedance data at a series of frequency points during EIS measurement, which is particularly inconvenient in applications that require real-time monitoring. Second, traditional devices are large in volume and complex in operation, making it difficult to be implemented in in-vehicle systems or other mobile applications, which limits their application in dynamic battery management systems.

[0005] With the rapid growth of market demands for electric vehicles, wearable devices, and other applications with extremely high requirements for battery performance, traditional electrochemical impedance measurement methods have been difficult to meet the requirements for battery monitoring efficiency in these application scenarios. To address this challenge, there is an urgent need to develop new methods and technologies to improve the efficiency and accuracy of battery performance monitoring.

[0006] In recent years, the application of machine learning and large model technologies has provided new ideas for solving these problems. In particular, the time-frequency analysis method based on large models has shown potential in improving the efficiency of electrochemical data processing. By combining the powerful data processing capabilities of time series large models, the acquisition speed and analysis accuracy of electrochemical impedance spectra can be significantly improved.

[0007] Therefore, the present invention aims to solve the problems of long time-consuming in measuring electrochemical impedance spectra and difficulty in realizing in-vehicle measurement in the prior art, and provides a method for obtaining electrochemical impedance based on a time series large model, so as to achieve the rapid acquisition and real-time analysis of electrochemical impedance spectrum data to meet the needs of modern battery performance monitoring. Summary of the Invention

[0008] To solve the problems of long time consumption and inconvenience in the traditional electrochemical impedance spectroscopy measurement process, the present invention proposes an electrochemical impedance acquisition method based on a time series large model. The method aims to significantly shorten the acquisition time of EIS data and make it applicable to portable and in-vehicle measurement environments. To achieve the above object, the present invention provides the following technical solutions:

[0009] An electrochemical impedance acquisition method based on a time series large model, comprising:

[0010] 1. Construction of the dataset required for the battery electrochemical AC impedance proprietary large model:

[0011] Construct a test matrix including temperature, current rate, aging degree, and SOC, measure 5 minutes of constant current charging and 5 minutes of constant current discharging in all cases to obtain battery DC test information, including subsequent relaxation processes; measure EIS in all cases. Here, measuring the AC impedance is not for directly obtaining the impedance in the application, but only for use as a training label. This patent does not involve specific experimental tests, and all tests are only used to construct the training dataset for the electrochemical impedance proprietary large model. This step can be directly obtained from a public dataset and the tests may not be carried out.

[0012] Take the frequency points in the electrochemical impedance spectroscopy data obtained by the electrochemical workstation as candidate frequency points. Take the electrochemical AC impedance values at these candidate frequency points as labels, and intercept the DC current and voltage during the corresponding charging and discharging and subsequent relaxation processes as inputs to construct an electrochemical impedance dataset. The dataset should contain a sufficient number of samples to ensure the comprehensiveness and effectiveness of model training.

[0013] 2. Battery electrochemical impedance proprietary large model:

[0014] Adopt a dataset including measured electrochemical impedance, divide it into a training set and a test set, and based on a pre-trained time series basic model (such as TimesFM), train an electrochemical impedance proprietary large model by adding an electrochemical impedance-related feature acquisition module. The TimesFM model can process time series data and extract the time series features in the electrochemical impedance data, thereby improving the model's prediction ability for impedance data.

[0015] 3. Correlation analysis of the battery DC charge and discharge curves and AC electrochemical impedance characteristics:

[0016] Extract features such as incremental capacity curve, local voltage, local time, peak area, valley area, frequency domain information of Fourier transform and S transform from the DC charge and discharge process, and conduct correlation analysis on the above features to enhance the interpretability of the electrochemical impedance proprietary large model.

[0017] 4. Verification of the Effect of the Proprietary Large Model for Battery Electrochemical Impedance:

[0018] Apply the trained proprietary large model for electrochemical impedance on the test set to obtain broadband alternating current electrochemical impedance values. Plot the Nyquist and Bode diagrams of the electrochemical impedance obtained from the proprietary large model for battery electrochemical impedance and the measured electrochemical impedance to verify the accuracy of the proprietary large model for battery electrochemical impedance. Through these graphical results, evaluate the fitting effect of the proprietary large model for battery electrochemical impedance on impedance data at different frequencies.

[0019] 5. Application of the Proprietary Large Model for Battery Electrochemical Impedance:

[0020] Use the proprietary large model for electrochemical impedance to calculate the alternating current electrochemical impedance spectra at different aging states based on the charge-discharge DC signals, without the need to measure the electrochemical impedance spectra using an electrochemical workstation. Select the impedance fitting regression curve calculated by the large model at appropriate frequencies for the estimation of the battery health state, and achieve battery life prediction through extrapolation techniques. This step can be used to formulate battery maintenance plans and optimize battery usage strategies.

[0021] The present invention combines the time series large model technology to achieve the rapid acquisition and efficient analysis of electrochemical impedance spectra, significantly improving the speed and accuracy of electrochemical impedance acquisition. It does not require an electrochemical workstation and is particularly suitable for application scenarios that require real-time monitoring, such as portable devices and in-vehicle systems. This method not only improves the efficiency of electrochemical impedance data processing but also provides a powerful tool for the real-time evaluation of battery performance. Description of the Drawings

[0022] Figure 1 is a flowchart of a method for obtaining electrochemical impedance based on a time series large model according to the present invention;

[0023] Figure 2 is a schematic diagram of obtaining impedance at the required frequency points through TimesFM;

[0024] Figure 3 is the step current excitation and its response voltage model for experimental testing;

[0025] Figure 4 The output is the EIS model of the battery in the same state;

[0026] Figure 5 is the impedance fitting regression curve model. Detailed Embodiments

[0027] The present invention will be described in detail below with reference to the drawings and specific embodiments. This embodiment is based on the technical solution of the present invention and gives detailed implementation manners and specific operation procedures, but the protection scope of the present invention is not limited to the following embodiments.

[0028] This embodiment provides a method for obtaining electrochemical impedance based on a time-series large model, as Figure 1 shown, including the following steps:

[0029] S1: According to the current amplitude in the test matrix, the battery is discharged for 5 minutes and charged for 5 minutes once at each amplitude, and the input current and response voltage data of the battery are recorded respectively. This step is used to obtain the dynamic response of the battery under specific operating conditions. Electrochemical impedance spectroscopy data under current excitation is obtained through an electrochemical workstation.

[0030] S2: Perform the S1 test on the aged battery.

[0031] S3: Take the frequency points in the electrochemical impedance spectroscopy data obtained by the electrochemical workstation as candidate frequency points. Take the electrochemical impedance values at these candidate frequency points as labels, and intercept the corresponding current and voltage data during the charging and discharging processes as inputs to construct an electrochemical impedance dataset. The dataset should contain enough samples to ensure the comprehensiveness and effectiveness of model training.

[0032] S4: Divide the training set and the test set. The training set is used to train the model, while the test set is used to evaluate the performance of the model. The proportion of data division should be adjusted according to the actual data volume and model requirements to optimize the training effect and generalization ability of the model.

[0033] S8: On the training set, train based on a pre-trained time-series basic model to construct an electrochemical impedance proprietary large model with sequence input and multi-point output. Optionally, the pre-trained time-series large model is TimesFM, which has strong time-series data processing and feature extraction capabilities. By learning and modeling the electrochemical impedance data, the analysis accuracy of complex electrochemical behaviors is improved.

[0034] (1) Input data (features): As Figure 3 shown

[0035] The model input is the step current excitation and its response voltage from experimental tests

[0036] (2) EIS data (output): As Figure 4 shown

[0037] The model output is the EIS of the battery under the same state

[0038] S9: Apply the electrochemical impedance proprietary large model on the test set to obtain the electrochemical impedance values at the candidate frequency points, and compare them with the true impedance values to evaluate the fitting effect of the model on impedance data at different frequencies.

[0039] Training the electrochemical impedance proprietary large model based on the pre-trained basic large model: As Figure 5 shown

[0040] S10: Calculate the electrochemical impedance spectra at different aging states using the proprietary large model of electrochemical impedance. Select the impedance fitting regression curve at an appropriate frequency for the estimation of the battery health state, and achieve battery life prediction through extrapolation techniques. This step can be used to formulate battery maintenance plans and optimize battery usage strategies.

[0041] Generally speaking, the present invention provides a method for obtaining electrochemical impedance based on a large model of time series. Through the application of large model technology, this method significantly improves the acquisition efficiency and analysis accuracy of electrochemical impedance data, thus providing an effective tool for battery performance evaluation and optimization.

[0042] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, rather than limitations on the implementation manners of the present invention. Those skilled in the art can make other changes within the main idea of the present invention, and these obvious changes derived therefrom should all be included within the scope claimed by the present invention.

Claims

1. A method for obtaining electrochemical impedance based on a time series large model, characterized in that: The following steps are involved: Construct a comprehensive test matrix containing multiple variables of the battery material system to obtain DC response signals and electrochemical impedance spectroscopy data; train a proprietary large model of electrochemical impedance based on a pre-trained time series basic model; use the proprietary large model of electrochemical impedance to calculate the AC electrochemical impedance spectrum under different aging conditions based on the charge and discharge DC signals, fit the regression curve and perform extrapolation to achieve battery health status assessment and life prediction; the electrochemical impedance acquisition method has the potential to be widely used in real-time monitoring and dynamic application environments by combining the time series large model technology.

2. The electrochemical impedance acquisition method based on a time series large model according to claim 1 is characterized in that The specific method of constructing the dataset is as follows: (1) Construct a test matrix including temperature, current rate (excitation amplitude), aging degree, and SOC to obtain battery DC test information, including the subsequent relaxation process; EIS is measured in all cases; (2) The frequency points in the electrochemical impedance spectroscopy data obtained by the electrochemical workstation are taken as candidate frequency points, the electrochemical AC impedance values ​​at these candidate frequency points are taken as labels, and the DC current and voltage during the corresponding charging and discharging and subsequent relaxation processes are intercepted as input to construct an electrochemical impedance dataset. The dataset should contain enough samples to ensure the comprehensiveness and effectiveness of the model training.

3. The electrochemical impedance acquisition method based on a time series large model according to claim 1 is characterized in that: The specific method for constructing a large battery electrochemical impedance model is as follows: Using a dataset containing measured electrochemical impedance data, based on a pre-trained time series basic model (such as TimesFM), and by adding an electrochemical impedance related feature acquisition module, a large electrochemical impedance proprietary model is trained.

4. The electrochemical impedance acquisition method based on a time series large model according to claim 1 is characterized in that: The specific method for enhancing the interpretability of the battery electrochemical impedance large model is as follows: A correlation analysis is carried out between the battery's DC charge and discharge curves and the AC electrochemical impedance characteristics. The features extracted during the DC charge and discharge process include incremental capacity curve, local voltage, local time, peak area, valley area, frequency domain information of Fourier transform and S transform, and the AC electrochemical impedance characteristics include EIS amplitude, phase angle, real part, imaginary part, etc. at each frequency, to enhance the interpretability of the electrochemical impedance proprietary large model.

5. The electrochemical impedance acquisition method based on a time series large model according to claim 1 is characterized in that: The health assessment and life prediction based on the proprietary large model of battery electrochemical impedance are as follows: (1) Using a proprietary electrochemical impedance spectroscopy model to calculate the AC electrochemical impedance spectra under different aging conditions based on the DC charge and discharge signals; (2) The impedance fitting regression curve obtained by calculating the large model at a suitable frequency is selected to estimate the battery health status, and the battery life prediction is achieved through extrapolation technology.

Images