Power battery health classification method, system and equipment based on deep learning algorithm and EIS technology, and medium

Abstract

The invention belongs to the technical field of new energy power battery health detection, and particularly relates to a power battery health classification method, system, equipment and medium based on a deep learning algorithm and an EIS technology, and the method comprises the steps: obtaining EIS data of a target power battery, constructing an equivalent circuit model comprising a solution resistor, an SEI layer resistor, a charge transfer impedance element and the like, and obtaining the EIS data of the target power battery; feature parameter factors are extracted, and a prediction model is constructed by using a recurrent neural network (RNN) and a long short-term memory (LSTM) network for training; according to the system, the electronic equipment and the storage medium, the practical application of the method is realized. According to the method, the defects of a traditional method in processing long-time sequence data are overcome, the accuracy and efficiency of battery health state evaluation are improved, the service life of the battery is prolonged, the operation and maintenance cost is reduced, powerful support is provided for health management and maintenance of the power battery, and the method has important practical value and application prospects.

Description

Technical Field

[0001] The present invention belongs to the technical field of new energy power battery health detection, and particularly relates to a power battery health classification method, system, device and medium based on deep learning algorithms and EIS technology. Background Art

[0002] As a core component of electric vehicles and energy storage systems, the accurate assessment of the state of health (SOH) of power batteries is crucial for ensuring system safety and extending service life. Although electric vehicle (EVs) manufacturers have developed many charge and discharge strategies and monitoring measures to extend the service life of batteries, over time, the deviation between the estimated driving range and the actual driving range is expected to become larger and larger, which is an intuitive reflection of the decline in the state of health (SOH) of the battery and capacity attenuation. This affects the driver's judgment of the vehicle condition, and even if the actual measured battery capacity loss is less than 20%, it is also a threshold for the replacement or scrapping of electric vehicle batteries.

[0003] However, the existing technologies still face many challenges in solving the above problems.

[0004] 1. The standard method for SOC estimation is through the OCV-SOC curve, which is an empirical curve and has the advantages of being fast and accurate for SOC estimation of batteries with large OCV changes. However, in SOH estimation, the method based solely on DC resistance tends to become increasingly inaccurate when changing the charge / discharge configuration, so the resistance-based SOH estimation method needs to be improved.

[0005] 2. Traditional RNNs are prone to the problems of gradient vanishing or explosion when processing long sequence data, resulting in the model being difficult to capture long-term dependencies, which limits its application in the long-term prediction of the state of health of power batteries.

[0006] 3. Although the LSTM network alleviates this problem to a certain extent, how to efficiently combine EIS test data with the LSTM network to fully exploit the time series features in the data and accurately predict the state of health of power batteries is still a hot topic and a difficult point in current research. Summary of the Invention

[0007] The object of the present invention is to provide a power battery health classification method, system, device and medium based on deep learning algorithms and EIS technology to solve the problem of accurately assessing the state of health (SOH) of power batteries, especially for the deficiencies of traditional methods in processing long time series data and the challenges of the existing technologies in efficiently predicting by combining EIS test data with the LSTM network.

[0008] The present invention achieves the above object through the following technical solutions:

[0009] In a first aspect, the present invention proposes a method for classifying the health of power batteries based on deep learning algorithms and EIS technology. The method includes:

[0010] Obtaining EIS data of a target power battery;

[0011] Constructing an equivalent circuit model based on the EIS data, and calculating model parameters according to the EIS data. The model parameters include ohmic impedance, diffusion impedance, and capacitance;

[0012] Obtaining EIS test data of the equivalent circuit model based on the model parameters, and obtaining the characteristic parameter factors of the target power battery after fitting;

[0013] Constructing a prediction model based on a deep learning algorithm of a recurrent neural network and a long short-term memory network, and training the prediction model based on the EIS test data and the characteristic parameter factors of the target power battery;

[0014] Inputting the EIS test data of the power battery to be tested into the trained prediction model, and outputting the characteristic parameter factors of the power battery to be tested;

[0015] Classifying the health of the power battery according to the characteristic parameter factors of the power battery to be tested.

[0016] Further, the obtaining of the EIS data of the target power battery includes:

[0017] Obtaining the target power battery under different temperatures and different states of charge;

[0018] Performing electrochemical impedance spectroscopy tests on the target power battery under different charge and discharge strategies to obtain EIS data.

[0019] Further, the equivalent circuit model includes a solution resistance RΩ, an SEI layer resistance Rsei, a charge transfer impedance Rct, a polarization impedance W1, and double-layer capacitors Cd1 and Cd2 at the positive and negative electrodes of the battery;

[0020] The first end of the solution resistance RΩ is connected to the positive electrode of the target battery, and the second end is connected to the first end of the SEI layer resistance Rsei; the second end of the SEI layer resistance Rsei is connected to the input end of the charge transfer impedance Rct; the output end of the charge transfer impedance Rct is connected to the polarization impedance W1; the other end of the polarization impedance W1 is connected to the negative electrode of the target battery; the double-layer capacitor Cd1 is connected in parallel across the SEI layer resistance Rsei; the double-layer capacitor Cd2 is connected in parallel between the charge transfer impedance Rct and the polarization impedance W1.

[0021] Further, obtaining the EIS test data of the equivalent circuit model based on the model parameters, and obtaining the characteristic parameter factors of the target power battery after fitting, including:

[0022] Importing data into the EIS fitting software;

[0023] Setting the equivalent circuit model based on the model parameters;

[0024] Starting the fitting process and performing fitting to obtain optimized model parameters;

[0025] Extracting characteristic parameter factors from the optimized model parameters to characterize the battery health state.

[0026] Further, the network structure of the prediction model includes:

[0027] An input layer for receiving input time series data, including EIS test data and the characteristic parameter factors;

[0028] An RNN layer for receiving time series data and performing feature extraction;

[0029] An LSTM layer for adjusting and optimizing according to the mapping relationship between the time series data and the characteristic parameter factors;

[0030] An output layer for receiving the output from the last LSTM layer;

[0031] A fully connected layer located between the input layer and the output layer for processing and transforming data.

[0032] Further, the method further includes:

[0033] Obtaining the measured value of the health state of the target power battery based on a cycle life test or a capacity test;

[0034] Updating the weights of the prediction model based on reverse adjustment of weights to obtain an optimized prediction model;

[0035] Evaluating the training result of the optimized prediction model based on the mean absolute percentage error and the root mean square error.

[0036] In a second aspect, the present invention proposes a power battery health classification system based on a deep learning algorithm and EIS technology for implementing any one of the above classification methods, and the system includes:

[0037] A data acquisition module for collecting EIS data of the target power battery under different conditions through electrochemical impedance spectroscopy technology;

[0038] An equivalent circuit model construction module for constructing an equivalent circuit model including series and parallel components according to the electrochemical characteristics of the battery;

[0039] A parameter extraction module, which is used to fit the EIS data by using EIS fitting software to extract the characteristic parameter factors of the target power battery;

[0040] A model training module, including an input layer, a convolutional layer, an LSTM layer, and a fully connected layer, which is used to build a prediction model based on a deep learning algorithm and perform model training;

[0041] A health classification module, which is used to input the EIS test data of the power battery to be tested into the trained prediction model, output the characteristic parameter factors of the power battery to be tested, and classify the health of the power battery according to the characteristic parameter factors of the power battery to be tested;

[0042] An effect evaluation module, which is used to evaluate the model training and verification results.

[0043] In a third aspect, the present invention proposes an electronic device, including:

[0044] A processor; a memory for storing instructions executable by the processor;

[0045] Wherein, the processor is configured to execute the instructions to implement the power battery health classification method as described in any one of the above.

[0046] In a fourth aspect, the present invention proposes a computer-readable storage medium, when the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device can execute the power battery health classification method as described in any one of the above.

[0047] The beneficial effects of the present invention are as follows:

[0048] The present invention uses EIS technology to obtain impedance data of the battery under different conditions, constructs an equivalent circuit model and extracts characteristic parameter factors, and then builds a prediction model based on a recurrent neural network (RNN) and a long short-term memory network (LSTM) for training. It not only overcomes the limitations of traditional methods in processing long time series data, but also improves the accuracy and efficiency of battery health state assessment. In addition, the classification system provided by the present invention further realizes the practical application of this method, provides strong support for the health management and maintenance of power batteries, helps to extend the battery life, reduce the operation and maintenance cost, and has important practical value. Description of the Drawings

[0049] Figure 1 It is a schematic flowchart of a power battery health classification method based on a deep learning algorithm and EIS technology provided by an embodiment of the present application;

[0050] Figure 2Another process schematic diagram of the power battery health classification method based on deep learning algorithm and EIS technology provided by the embodiments of the present application;

[0051] Figure 3 The circuit structure diagram of the equivalent circuit model in the power battery health classification method based on deep learning algorithm and EIS technology provided by the embodiments of the present application. Specific embodiments

[0052] The present application will be further described in detail below with reference to the accompanying drawings. It is necessary to point out here that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the protection scope of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application according to the above application content.

[0053] Embodiment 1

[0054] As Figures 1-3 shown, this embodiment proposes a power battery health classification method based on deep learning algorithm and EIS technology. The method includes the following steps:

[0055] S1. Obtain the EIS data of the target power battery; specifically, it includes preferentially obtaining the target power battery under different temperatures and different states of charge (SOC), and then performing electrochemical impedance spectroscopy (EIS) tests on the target power battery under different charge and discharge strategies to obtain the EIS data.

[0056] S2. Construct an equivalent circuit model including solution resistance, SEI layer resistance, charge transfer impedance, polarization impedance, and double-layer capacitance according to the EIS data. The equivalent circuit model includes series elements and parallel elements, and calculate the model parameters according to the EIS data. The model parameters include ohmic impedance, diffusion impedance, and capacitance.

[0057] S3. Obtain the EIS test data of the equivalent circuit model based on the model parameters, and obtain the characteristic parameter factors of the target power battery after fitting; this step is to import data into the EIS fitting software, set the equivalent circuit model based on the model parameters, start the fitting process to obtain the optimized model parameters, and then extract the characteristic parameter factors to characterize the health state of the battery.

[0058] Among them, the fitting process usually involves the least squares method to minimize the difference between the model prediction value and the actual measurement value.

[0059] S4. Construct a prediction model based on the deep learning algorithms of recurrent neural network (RNN) and long short-term memory network (LSTM), including an input layer, an RNN layer, an LSTM layer, an output layer, and a fully connected layer located between the input layer and the output layer. Train the prediction model based on the EIS test data and the characteristic parameter factors of the target power battery, so that it can learn and identify the relationship between the battery health state and the EIS data.

[0060] S5. Input the EIS test data of the power battery to be measured into the trained prediction model, and output the characteristic parameter factors of the power battery to be measured.

[0061] S6. Classify the health of the power battery according to the characteristic parameter factors of the power battery to be measured.

[0062] In this embodiment, what is obtained in step S1 is a series of EIS data (impedance data) of the target power battery under different conditions. These data are usually represented in complex form, including real part and imaginary part information, and reflect the impedance characteristics of the battery at different frequencies.

[0063] It should be noted that the purpose of obtaining the EIS test data of the equivalent circuit model in step S3 is to verify the accuracy of the equivalent circuit model and extract the characteristic parameter factors of the target power battery. After constructing the equivalent circuit model and calculating the model parameters, it is necessary to use the EIS test data to fit the model. The EIS test data in this step is obtained under the same conditions as in step 1, but more focuses on verifying the accuracy of the model. Through the fitting process, the characteristic parameter factors of the target power battery can be extracted, such as more accurate values of parameters such as ohmic impedance, diffusion impedance, and capacitance. These parameters can characterize the health state of the battery. At the same time, the accuracy of the equivalent circuit model is also verified in this step.

[0064] By setting the above solutions, the present application effectively solves the problems proposed in the background technology. First, by obtaining the EIS data under different conditions and constructing an equivalent circuit model, the electrochemical characteristics inside the battery can be more accurately reflected. Second, using the deep learning algorithm to construct a prediction model and combining the advantage of the LSTM network in processing long time series data improves the accuracy and efficiency of battery health state prediction. Finally, by extracting the characteristic parameter factors and performing health classification, an accurate assessment of the health state of the power battery is realized, providing a strong guarantee for the safe operation of electric vehicles or energy storage systems.

[0065] Combined with Figure 3, in a preferred embodiment of the present invention, the equivalent circuit model includes solution resistance RΩ, SEI layer resistance Rsei, charge transfer impedance Rct, polarization impedance W1, and double-layer capacitors Cd1 and Cd2 of the positive and negative electrodes of the battery; the first end of the solution resistance RΩ is connected to the positive electrode of the target battery, and the second end is connected to the first end of the SEI layer resistance Rsei; the second end of the SEI layer resistance Rsei is connected to the input end of the charge transfer impedance Rct; the output end of the charge transfer impedance Rct is connected to the polarization impedance W1; the other end of the polarization impedance W1 is connected to the negative electrode of the target battery; the double-layer capacitor Cd1 is connected in parallel across the two ends of the SEI layer resistance Rsei; the double-layer capacitor Cd2 is connected in parallel between the charge transfer impedance Rct and the polarization impedance W1.

[0066] Regarding the parameter description:

[0067] Solution resistance RΩ: Represents the electrolyte resistance inside the battery.

[0068] SEI layer resistance Rsei: Represents the resistance of the solid electrolyte interface (SEI) layer on the surface of the negative electrode of the battery.

[0069] Charge transfer impedance Rct: Represents the charge transfer impedance of the battery electrode reaction.

[0070] Double-layer capacitors Cd1 and Cd2 of the positive and negative electrodes of the battery: Represent the capacitance between the battery electrodes and the electrolyte.

[0071] Working process of the equivalent circuit model:

[0072] Data acquisition: First, obtain the EIS data of the target power battery through EIS testing. These data contain the impedance information of the battery at different frequencies and are the basis for constructing the equivalent circuit model.

[0073] Model construction: According to the electrochemical characteristics of the battery, construct an equivalent circuit model including series and parallel components. In the present invention, the model includes components such as solution resistance RΩ, SEI layer resistance Rsei, charge transfer impedance Rct, polarization impedance W1, and double-layer capacitors Cd1 and Cd2.

[0074] Parameter extraction: Use EIS fitting software to fit the EIS data and extract the parameters of the equivalent circuit model. These parameters include ohmic impedance, diffusion impedance, and capacitance, etc., and can characterize the health state of the battery.

[0075] Model verification: Verify the accuracy of the equivalent circuit model by comparing the actual test data and the model prediction data. If the model prediction results are consistent with the experimental results, it indicates that the model can accurately describe the electrochemical performance of the battery.

[0076] In a preferred embodiment of the present invention, EIS test data of an equivalent circuit model is obtained based on model parameters, and characteristic parameter factors of a target power battery are obtained after fitting, including: importing data into EIS fitting software; setting an equivalent circuit model based on model parameters; starting the fitting process, performing fitting to obtain optimized model parameters; extracting characteristic parameter factors from the optimized model parameters, and the characteristic parameter factors are used to characterize the battery health state, that is, the battery health state prediction value.

[0077] In a preferred embodiment of the present invention, the network structure of the prediction model includes:

[0078] An input layer for receiving input time series data, including EIS test data and characteristic parameter factors;

[0079] An RNN layer for receiving time series data and performing feature extraction;

[0080] An LSTM layer for adjusting and optimizing according to the mapping relationship between time series data and characteristic parameter factors;

[0081] An output layer for receiving the output from the last LSTM layer;

[0082] A fully connected layer located between the input layer and the output layer for processing and converting data.

[0083] In a preferred embodiment of the present invention, the method further includes reverse adjusting weights to update the weights of the neural network to minimize errors, including the following steps:

[0084] Obtaining the measured value of the health state of the target power battery based on a cycle life test or a capacity test;

[0085] Updating the weights of the prediction model based on reverse adjusting weights to obtain an optimized prediction model;

[0086] Evaluating the training result of the optimized prediction model based on the mean absolute percentage error and the root mean square error.

[0087] More specifically, reverse adjusting weights specifically includes:

[0088] 1. Forward propagation: Calculate the output of each layer until the final predicted value is obtained.

[0089] 2. Calculate the error: Use the following error calculation formula to calculate the error between the predicted value and the true value.

[0090] 3. Backward propagation: Starting from the output layer, calculate the gradient layer by layer and propagate the gradient to each layer.

[0091] 4. Update weights: Use the gradient descent method to update weights.

[0092] To evaluate the accuracy of the model, the mean absolute percentage error (MAPE) and the root mean square error (RMSE) are used to evaluate the model training and validation results. The expressions of MAPE and RMSE are as shown in the formula.

[0093]

[0094] Among them, Y is the SOH estimated value; S is the measured SOH; N is the total number of predicted samples.

[0095] In another embodiment of the present invention, based on the same inventive concept, a power battery health classification system based on a deep learning algorithm and EIS technology is proposed for implementing the classification method as described above. The system includes:

[0096] A data acquisition module for collecting EIS data of the target power battery under different conditions through electrochemical impedance spectroscopy technology;

[0097] An equivalent circuit model construction module for constructing an equivalent circuit model containing series and parallel components according to the electrochemical characteristics of the battery;

[0098] A parameter extraction module for fitting the EIS data using EIS fitting software to extract the characteristic parameter factors of the target power battery;

[0099] A model training module, including an input layer, a convolutional layer, an LSTM layer, and a fully connected layer, for constructing a prediction model based on a deep learning algorithm and training the model;

[0100] A health classification module for inputting the EIS test data of the power battery to be tested into the trained prediction model, outputting the characteristic parameter factors of the power battery to be tested, and the characteristic parameter factors are used to characterize the battery health state; classifying the health of the power battery according to the characteristic parameter factors of the power battery to be tested;

[0101] An effect evaluation module for evaluating the model training and validation results.

[0102] It should be noted here that each module in the above power battery health classification system corresponds to each step in implementing the above optimization method. The examples and application scenarios implemented by multiple modules and the corresponding steps are the same, but are not limited to the content disclosed in the above Embodiment 1.

[0103] It can be understood that the classification method and classification system proposed by the present invention can be applied to various types of power batteries, such as lithium-ion batteries, lead-acid batteries, etc.

[0104] In another embodiment provided by the present invention, an electronic device is proposed, including:

[0105] a processor; a memory for storing instructions executable by the processor;

[0106] wherein the processor is configured to execute instructions to implement the power battery health classification method as described above.

[0107] In another embodiment provided by the present invention, a computer-readable storage medium is proposed. When the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device can execute the power battery health classification method as described above.

[0108] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0109] In addition, in each embodiment of the present application, the various functional modules can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0110] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. A power battery health classification method based on deep learning algorithm and EIS technology, characterized in that: The method comprises: Obtain EIS data of the target power battery; Constructing an equivalent circuit model according to the EIS data, and calculating model parameters according to the EIS data, wherein the model parameters include ohmic impedance, diffusion impedance and capacitance; Acquiring EIS test data of the equivalent circuit model based on the model parameters, and obtaining characteristic parameter factors of the target power battery after fitting; Building a prediction model based on a deep learning algorithm of a recurrent neural network and a long short-term memory network, and training the prediction model based on the EIS test data and characteristic parameter factors of the target power battery; Inputting the EIS test data of the power battery to be tested into the trained prediction model, and outputting the characteristic parameter factors of the power battery to be tested; The health classification of the power battery is carried out according to the characteristic parameter factors of the power battery to be tested.

2. According to claim 1, a power battery health classification method based on deep learning algorithm and EIS technology is characterized by: The obtaining of EIS data of the target power battery includes: Obtain target power batteries at different temperatures and different states of charge; The target power battery is subjected to electrochemical impedance spectroscopy test under different charge and discharge strategies to obtain EIS data.

3. According to claim 1, a power battery health classification method based on deep learning algorithm and EIS technology is characterized in that: The equivalent circuit model includes solution resistance RΩ, SEI layer resistance Rsei, charge transfer impedance Rct, polarization impedance W1, and double layer capacitances Cd1 and Cd2 of the positive and negative electrodes of the battery; The first end of the solution resistor RΩ is connected to the positive electrode of the target battery, and the second end is connected to the first end of the SEI layer resistor Rsei; the second end of the SEI layer resistor Rsei is connected to the input end of the charge transfer impedance Rct; the output end of the charge transfer impedance Rct is connected to the polarization impedance W1; the other end of the polarization impedance W1 is connected to the negative electrode of the target battery; the double-layer capacitor Cd1 is connected in parallel at both ends of the SEI layer resistor Rsei; the double-layer capacitor Cd2 is connected in parallel between the charge transfer impedance Rct and the polarization impedance W1.

4. The power battery health classification method based on deep learning algorithm and EIS technology according to claim 3 is characterized by: The step of acquiring the EIS test data of the equivalent circuit model based on the model parameters and obtaining the characteristic parameter factors of the target power battery after fitting includes: Import data into EIS fitting software; Setting an equivalent circuit model based on the model parameters; Start the fitting process and execute the fitting to obtain the optimized model parameters; Characteristic parameter factors are extracted from the optimized model parameters to characterize the battery health status.

5. The power battery health classification method based on deep learning algorithm and EIS technology according to claim 1 is characterized by: The network structure of the prediction model includes: An input layer, used for receiving input time series data, including EIS test data and the characteristic parameter factors; RNN layer, used to receive time series data and perform feature extraction; LSTM layer, used to adjust and optimize according to the mapping relationship between time series data and feature parameter factors; The output layer receives the output from the last LSTM layer. The fully connected layer is located between the input layer and the output layer and is used to process and transform data.

6. The power battery health classification method based on deep learning algorithm and EIS technology according to claim 1 is characterized by: The method further comprises: Obtaining the measured value of the health status of the target power battery based on the cycle life test or capacity test; Updating the weight of the prediction model based on the reverse adjustment weight to obtain an optimized prediction model; The training results of the optimized prediction model were evaluated based on mean absolute percentage error and root mean square error.

7. A power battery health classification system based on deep learning algorithm and EIS technology, used to implement the classification method according to any one of claims 1 to 6, characterized in that: The system comprises: A data acquisition module is used to collect EIS data of the target power battery under different conditions through electrochemical impedance spectroscopy technology; An equivalent circuit model building module is used to build an equivalent circuit model containing series and parallel components according to the electrochemical characteristics of the battery; The parameter extraction module is used to fit the EIS data using the EIS fitting software to extract the characteristic parameter factors of the target power battery; Model training module, including input layer, convolution layer, LSTM layer, and fully connected layer, used to build prediction models based on deep learning algorithms and perform model training; A health classification module, used for inputting the EIS test data of the power battery to be tested into the trained prediction model, outputting characteristic parameter factors of the power battery to be tested, and performing health classification on the power battery according to the characteristic parameter factors of the power battery to be tested; Effect evaluation module, used to evaluate model training and verification results.

8. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; Wherein, the processor is configured to execute the instructions to implement the power battery health classification method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the power battery health classification method as described in any one of claims 1 to 6.

Images