Lithium battery SOH estimation method and device based on relaxation process
Through the SOH estimation method of lithium battery based on relaxation process, using a one-dimensional convolutional neural network and Xgboost model combined with voltage and temperature data, the problem of low accuracy of SOH estimation in the prior art is solved, and a higher battery performance evaluation is achieved Improved accuracy and energy storage system performance.
Patent Information
- Application Number
- CN202411175903.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-26
- Publication Date
- 2024-12-03
AI Technical Summary
The estimation accuracy of lithium battery State Of Health (SOH) based on idealized models or fixed parameters in the prior art is low, making it difficult to reflect the actual performance under complex operating conditions.
Using the SOH estimation method of lithium battery based on the relaxation process, the lithium battery is charged and discharged, and the voltage and temperature data are collected, and the relaxation voltage value is predicted using a one-dimensional convolutional neural network model, combining the voltage and temperature statistical characteristics and two The order RC equivalent circuit model parameters are used to train the Xgboost model to improve the SOH estimation accuracy.
It improves the accuracy and estimation accuracy of lithium battery performance evaluation, can more accurately reflect the actual state of the battery under complex operating conditions, and enhances the performance of energy storage systems and the application of clean energy.
Smart Images

Figure CN119064814A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of battery energy storage, and in particular to a SOH estimation method and device for a lithium battery based on a relaxation process. Background Art
[0002] As the world pays more and more attention to environmental protection and sustainable development, the status of clean energy has become increasingly prominent, especially in the field of energy storage. Lithium-ion batteries are widely favored due to their high energy density, long cycle life and low self-discharge rate, and are widely used in many scenarios such as home energy storage, electric vehicles, photovoltaic grid energy storage, etc. In these applications, the battery management system (BMS) plays a vital role, among which the accurate estimation of the battery health state (SOH) is crucial to ensure the efficient and safe operation of the system.
[0003] In the prior art, battery SOH is often estimated based on an idealized model or fixed parameters. However, this estimation method is difficult to accurately reflect the actual performance of the battery under complex working conditions, thereby affecting the accurate evaluation of battery performance. Summary of the invention
[0004] In view of this, the present invention provides a method and device for estimating the SOH of a lithium battery based on a relaxation process, so as to solve the problem in the prior art of estimating the battery SOH based on an idealized model or fixed parameters, resulting in low estimation accuracy.
[0005] In a first aspect, the present invention provides a method for estimating the SOH of a lithium battery based on a relaxation process, the method comprising:
[0006] Carry out charge and discharge tests on lithium batteries under different conditions, and collect voltage sequence data and temperature sequence data of lithium batteries during the charge and discharge tests;
[0007] The collected voltage sequence data is input into the trained one-dimensional convolutional neural network model for prediction to obtain the predicted relaxation voltage value;
[0008] According to the collected voltage sequence data and temperature sequence data, voltage statistical characteristics and temperature statistical characteristics are calculated respectively. The voltage statistical characteristics include the standard deviation, skewness and minimum value of the relaxation voltage sequence, and the temperature statistical characteristics include the range, standard deviation and skewness of the temperature sequence.
[0009] Based on the relaxation process of lithium batteries, a second-order RC equivalent circuit model is established, and the parameters of the equivalent circuit model are calculated;
[0010] The Xgboost model is trained using relaxation voltage values, voltage statistical features, temperature statistical features, and equivalent circuit model parameters;
[0011] The voltage series data and temperature series data of the actual charging and discharging process of the lithium battery are input into the trained Xgboost model to obtain the SOH estimation result of the lithium battery.
[0012] The present invention collects voltage sequence data and temperature sequence data during the lithium battery charge and discharge test process, uses a one-dimensional convolutional neural network model to predict the relaxation voltage value, calculates voltage statistical characteristics and temperature statistical characteristics, and calculates equivalent circuit model parameters, adopts a relatively stable battery relaxation process to estimate the SOH of the lithium battery, and uses the Xgboost model to estimate the SOH of the lithium battery according to the voltage sequence data and temperature sequence data during the actual charge and discharge process of the lithium battery, thereby improving the accuracy of battery performance evaluation and effectively improving the estimation accuracy.
[0013] In an optional implementation, before inputting the collected voltage sequence data into a trained one-dimensional convolutional neural network model for prediction, the method further includes:
[0014] Dividing the collected voltage sequence data into a first training set and a first test set according to a preset ratio;
[0015] Using the first training set to train the one-dimensional convolutional neural network model;
[0016] The one-dimensional convolutional neural network model is tested using the first test set, and the estimation accuracy of the one-dimensional convolutional neural network model is measured using the mean square error value of the relaxation voltage;
[0017] When the number of iterations of the one-dimensional convolutional neural network model training reaches a preset maximum number of iterations, the training of the one-dimensional convolutional neural network model is stopped.
[0018] The present invention trains and tests a one-dimensional convolutional neural network model, and uses the mean square error value of the relaxation voltage to intuitively reflect the estimation accuracy of the one-dimensional convolutional neural network model, so as to adjust the model according to the mean square error value of the relaxation voltage, thereby improving the estimation accuracy of the one-dimensional convolutional neural network model. When a preset maximum number of iterations is reached, the training of the one-dimensional convolutional neural network model is stopped, thereby avoiding a decrease in generalization ability caused by excessive model training times.
[0019] In an optional implementation, the collected voltage sequence data is input into a trained one-dimensional convolutional neural network model for prediction to obtain a predicted relaxation voltage value, including:
[0020] Extracting voltage sequence data within a first time period from the collected voltage sequence data;
[0021] The voltage sequence data in the first time period is input into the trained one-dimensional convolutional neural network model, the voltage sequence data in the second time period is predicted, and the predicted voltage value is determined as the relaxation voltage value.
[0022] The present invention predicts the voltage sequence data in a second time period based on the voltage sequence data in a first time period through a one-dimensional convolutional neural network model, and effectively analyzes the changing trend of the voltage sequence.
[0023] In an optional implementation, the equivalent circuit model parameters include polarization resistance value, polarization capacitance value and DC internal resistance value, and the equivalent circuit model parameters are calculated according to the following formula:
[0024]
[0025] Where R1 and R2 are polarization resistors, C1 and C2 are polarization capacitors, R is the DC internal resistance, V1 and V2 are the amplitudes of the polarization voltage, I is the current, τ1 and τ2 are the time constants of the polarization process, U(0) is the charge cut-off voltage, and U ocv is the open circuit voltage.
[0026] The present invention calculates each parameter in the established second-order equivalent circuit model to intuitively associate the circuit model parameters with the actual state of the lithium battery, so as to estimate the SOH of the lithium battery.
[0027] In an optional implementation, the Xgboost model is trained using the relaxation voltage value, the voltage statistical characteristics, the temperature statistical characteristics and the equivalent circuit model parameters, including:
[0028] Dividing the relaxation voltage values, voltage statistical characteristics, temperature statistical characteristics and equivalent circuit model parameters into a second training set and a second test set according to a preset ratio;
[0029] Use the second training set to train the Xgboost model;
[0030] The second test set is used to test the Xgboost model, and the mean square error value of SOH is used to measure the estimation accuracy of the Xgboost model.
[0031] The present invention trains and tests the Xgboost model, which has the advantages of high performance and support for parallel computing. The mean square error value of SOH is used to measure the estimation accuracy of the Xgboost model, so that the model can be adjusted according to the mean square error value of SOH to improve the estimation accuracy of the Xgboost model.
[0032] In an optional embodiment, after training the Xgboost model using the relaxation voltage value, the voltage statistical characteristics, the temperature statistical characteristics and the equivalent circuit model parameters, the method further includes:
[0033] Determine the hyperparameters and hyperparameter grid corresponding to the Xgboost model;
[0034] Traverse all hyperparameter points in the hyperparameter grid and determine the hyperparameter point with the smallest mean square error of SOH;
[0035] The Xgboost model corresponding to the hyperparameter point with the smallest mean square error of SOH is determined as the trained Xgboost model.
[0036] The present invention adopts the grid search method to comprehensively traverse all hyperparameter points in the hyperparameter grid to improve the comprehensiveness of the search, avoid relying only on the local optimal solution, and improve the generalization ability of the Xgboost model.
[0037] In an optional embodiment, the voltage sequence data during the actual charging and discharging process of the lithium battery includes a relaxation voltage sequence. The voltage sequence data and the temperature sequence data during the actual charging and discharging process of the lithium battery are input into the trained Xgboost model to obtain the SOH estimation result of the lithium battery, including:
[0038] When it is detected that the lithium battery has completed charging and the static time reaches the preset time, the relaxation voltage sequence is input into the one-dimensional convolutional neural network model to obtain the voltage prediction value;
[0039] Extract statistical features and equivalent circuit model features from the voltage series data and temperature series data during the actual charging and discharging process of lithium batteries;
[0040] The voltage prediction value, statistical characteristics, and equivalent circuit model characteristics are input into the Xgboost model to output the SOH estimation result of the lithium battery.
[0041] The present invention uses the Xgboost model to comprehensively estimate the battery SOH according to the voltage prediction value output by the one-dimensional convolutional neural network model, the statistical features extracted from the voltage series data and the temperature series data in the actual charging and discharging process of the lithium battery, and the equivalent circuit model features. It takes into account the problem that the battery relaxation process may be shorter in actual operation, and combines the equivalent circuit model characteristics to effectively improve the estimation accuracy of the lithium battery SOH.
[0042] In a second aspect, the present invention provides a SOH estimation device for a lithium battery based on a relaxation process, the device comprising:
[0043] The data acquisition module is used to perform charge and discharge tests on the lithium battery under different conditions, and to collect voltage sequence data and temperature sequence data of the lithium battery during the charge and discharge test;
[0044] A prediction module is used to input the collected voltage sequence data into a trained one-dimensional convolutional neural network model for prediction to obtain a predicted relaxation voltage value;
[0045] A first calculation module is used to calculate voltage statistical characteristics and temperature statistical characteristics respectively according to the collected voltage sequence data and temperature sequence data, the voltage statistical characteristics include the standard deviation, skewness and minimum value of the relaxation voltage sequence, and the temperature statistical characteristics include the range, standard deviation and skewness of the temperature sequence;
[0046] The second calculation module is used to establish a second-order RC equivalent circuit model based on the relaxation process of the lithium battery and calculate the parameters of the equivalent circuit model;
[0047] A model training module is used to train the Xgboost model using relaxation voltage values, voltage statistical features, temperature statistical features, and equivalent circuit model parameters;
[0048] The estimation module is used to input the voltage sequence data and temperature sequence data of the actual charging and discharging process of the lithium battery into the trained Xgboost model to obtain the SOH estimation result of the lithium battery.
[0049] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, computer instructions are stored in the memory, and the processor executes the SOH estimation method for a lithium battery based on a relaxation process according to the first aspect or any corresponding embodiment thereof by executing the computer instructions.
[0050] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the SOH estimation method for a lithium battery based on a relaxation process according to the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0052] Figure 1is a flow chart of a method for estimating SOH of a lithium battery based on a relaxation process according to an embodiment of the present invention;
[0053] Figure 2 is a schematic diagram of a second-order RC equivalent circuit model according to an embodiment of the present invention;
[0054] Figure 3 is a schematic diagram of an application of a SOH estimation method for a lithium battery based on a relaxation process according to an embodiment of the present invention;
[0055] Figure 4 is a structural block diagram of a SOH estimation device for a lithium battery based on a relaxation process according to an embodiment of the present invention;
[0056] Figure 5 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION
[0057] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0058] In actual use, batteries face complex and changeable working conditions, such as temperature fluctuations, changes in charge and discharge rates, and aging, all of which will affect the capacity of the battery. Traditional capacity estimation algorithms are often based on idealized models or fixed parameters, which are difficult to accurately reflect the actual performance of batteries under these complex working conditions.
[0059] In energy storage scenarios, the battery is frequently fully charged and in a relaxed state, which provides a new idea for battery capacity prediction. Studies have shown that there is a significant correlation between the voltage of a fully charged battery after it is at rest, namely the relaxation voltage, and the battery capacity. Compared with the data of the battery under complex and changeable operating conditions, the data of the relaxation process is more stable, so it is easier to achieve high-precision battery SOH estimation. This not only improves the performance of the energy storage system, but also provides strong support for the popularization and application of clean energy. Compared with the voltage of the battery under complex and changeable operating conditions, the relaxation voltage is more stable, so it is easier to achieve high-precision battery capacity estimation based on the relaxation voltage.
[0060] However, when using the relaxation process for SOH estimation, it should be noted that although the relaxation process is relatively stable, the relaxation process collected during the actual operation of the battery may not last for a long time. In some cases, due to time constraints or operational requirements, we may not be able to obtain complete relaxation process data, which will affect our accurate evaluation of battery performance.
[0061] Therefore, in order to solve the shortcomings of the traditional SOH estimation algorithm and take into account the relaxation process, temperature influence and the problem that the relaxation process may be shorter in actual operation, an embodiment of the present invention provides a SOH estimation method for a lithium battery based on the relaxation process to comprehensively estimate the SOH of a lithium battery.
[0062] According to an embodiment of the present invention, an embodiment of a SOH estimation method for a lithium battery based on a relaxation process is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.
[0063] In this embodiment, a SOH estimation method for a lithium battery based on a relaxation process is provided. Figure 1 is a flow chart of a method for estimating SOH of a lithium battery based on a relaxation process according to an embodiment of the present invention, such as Figure 1 As shown, the process includes the following steps:
[0064] Step S101 , performing charge and discharge tests on a lithium battery under different conditions, and collecting voltage sequence data and temperature sequence data of the lithium battery during the charge and discharge tests.
[0065] In the embodiment of the present invention, the same type of lithium battery is subjected to charge and discharge tests under different conditions, and the voltage sequence data and temperature sequence data of the lithium battery are collected during each charge and discharge test. There are no specific requirements for the operating conditions of the lithium battery, which reduces the restrictions on the operating conditions of the lithium battery. Specifically, for the cyclic charge and discharge tests under different temperatures and different operating conditions, each cycle is fully charged and then left to stand for 1800 seconds before discharging. The collected battery data include: the battery voltage sequence V and temperature sequence T with a sampling interval of 10 seconds after standing for 300 seconds after each charge, the battery capacity Q released when discharged to the cut-off voltage after each charge is completed, and the current I at the last moment before standing. Q standard is the known standard capacity of lithium battery, Q / Q standard As the SOH tag value of each cycle, N is the total number of cycles.
[0066] Wherein, the voltage sequence V=[V1,V2,……V N ], temperature sequence T = [T1, T2, ... T N]; After the i-th cycle charging, the battery is left to stand for 300 seconds, and the sampling interval is 10 seconds. There are 30 voltage sampling points in total. Therefore, the voltage sequence V i It can be expressed as V i =[V i,1 , V i,2 , ... V i,30 ], the temperature sequence T after standing for 300 seconds after the i-th cycle charging i It can be expressed as T i =[T i,1 , T i,2 , ... T i,30 ]; the voltage after standing for 600 seconds after the i-th cycle charging is recorded as V i,60 , the voltage after N cycles of charging and standing for 600 seconds is combined into a sequence V 60 =[V 1,60 , V 2,60 , ... V N,60 ], the voltage after standing for 1200 seconds after the i-th cycle charging is recorded as V i,120 , the voltage after N cycles of charging and standing for 1200 seconds is combined into a sequence V 120 =[V 1,120 , V 2,120 , ... V N,120 ], the voltage after standing for 1800 seconds after the i-th cycle charging is recorded as V i,180 , the voltage after N cycles of charging and standing for 1800 seconds is combined into a sequence V 180 =[V 1,180 , V 2,180 , ... V N,180 ].
[0067] Step S102, inputting the collected voltage sequence data into a trained one-dimensional convolutional neural network model for prediction to obtain a predicted relaxation voltage value.
[0068] In an embodiment of the present invention, a one-dimensional convolutional neural network model (CNN) is established and trained, and the collected voltage sequence data is input into the trained one-dimensional convolutional neural network model to predict the relaxation voltage value.
[0069] Step S103, calculating voltage statistical features and temperature statistical features respectively according to the collected voltage sequence data and temperature sequence data.
[0070] In the embodiment of the present invention, the temperature statistical features include the range, standard deviation and skewness of the temperature series.
[0071] The standard deviation V of the relaxation voltage sequence in each cycle is calculated from the resting state to the next start-up. std , skewness V skew and minimum value V min , taking the above statistics as voltage statistical features, calculate the range T of the temperature sequence in each cycle range , standard deviation T std and skewness T skew , taking the above statistics as temperature statistical features, the feature is calculated as follows:
[0072]
[0073] T range =T max -T min
[0074] Among them, V std is the standard deviation of the relaxation voltage sequence, V skew is the skewness of the relaxation voltage sequence, V min is the minimum value of the relaxation voltage sequence, V max is the maximum value of the relaxation voltage sequence, T range is the range of the temperature series, T std is the standard deviation of the temperature series, T skew is the skewness of the temperature series, V i is the voltage sequence of the ith cycle, T i is the temperature sequence of the ith cycle.
[0075] Step S104, based on the relaxation process of the lithium battery, a second-order RC equivalent circuit model is established, and the parameters of the equivalent circuit model are calculated.
[0076] In an embodiment of the present invention, based on the relaxation process of the lithium battery, a second-order equivalent circuit model is established, and the equivalent circuit model parameters are calculated respectively. The equivalent circuit model parameters include polarization resistance value, polarization capacitance value and DC internal resistance value. The equivalent circuit model parameters and open circuit voltage value are used as features for predicting the SOH of the lithium battery.
[0077] Step S105 , training the Xgboost model using the relaxation voltage value, voltage statistical characteristics, temperature statistical characteristics and equivalent circuit model parameters.
[0078] In an embodiment of the present invention, an Xgboost model is constructed, and the relaxation voltage value, voltage statistical characteristics, temperature statistical characteristics and equivalent circuit model parameters are feature merged to form a feature data set with the SOH of each cycle to train the Xgboost model.
[0079] Step S106, inputting the voltage sequence data and temperature sequence data during the actual charging and discharging process of the lithium battery into the trained Xgboost model to obtain the SOH estimation result of the lithium battery.
[0080] In an embodiment of the present invention, voltage sequence data and temperature sequence data during the actual charging and discharging process of the lithium battery are collected, and the collected voltage sequence data and temperature sequence data during the actual charging and discharging process of the lithium battery are input into a pre-trained Xgboost model to estimate the SOH of the lithium battery.
[0081] The present embodiment provides a method for estimating the SOH of a lithium battery based on a relaxation process. By collecting voltage sequence data and temperature sequence data during a lithium battery charge and discharge test, a one-dimensional convolutional neural network model is used to predict the relaxation voltage value, voltage statistical characteristics and temperature statistical characteristics are calculated, and equivalent circuit model parameters are calculated. A relatively stable battery relaxation process is used to estimate the SOH of the lithium battery. The Xgboost model is used to estimate the SOH of the lithium battery based on the voltage sequence data and temperature sequence data during the actual charge and discharge process of the lithium battery, thereby improving the accuracy of battery performance evaluation and effectively improving the estimation accuracy.
[0082] In this embodiment, a method for estimating the SOH of a lithium battery based on a relaxation process is provided, and the process includes the following steps:
[0083] Step S201 , performing charge and discharge tests on the lithium battery under different conditions, and collecting voltage sequence data and temperature sequence data of the lithium battery during the charge and discharge tests.
[0084] For details, please see Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.
[0085] Step S202: dividing the collected voltage sequence data into a first training set and a first test set according to a preset ratio.
[0086] Step S203: training the one-dimensional convolutional neural network model using the first training set.
[0087] Step S204, using the first test set to test the one-dimensional convolutional neural network model, and using the mean square error value of the relaxation voltage to measure the estimation accuracy of the one-dimensional convolutional neural network model.
[0088] Step S205, when the number of iterations of the one-dimensional convolutional neural network model training reaches a preset maximum number of iterations, stop training the one-dimensional convolutional neural network model.
[0089] In the embodiment of the present invention, the voltage sequence V=[V1,V2,……V N ] and sequence V60 、V 120 、V 180 Merge into data set D1, and divide D1 into training set D train,1 and the test set D test,1 Two parts, the preset ratio is 4:1.
[0090] A one-dimensional convolutional neural network model CNN is established, and the number of layers, number of convolution kernels, size of convolution kernels, step size of convolution kernels, learning rate and number of iterations of the one-dimensional convolutional neural network model are determined. The model input is a cyclic voltage sequence V i =[V i,1 , Vi ,2 , ... Vi ,30 ], the output is the voltage of this cycle after 600 seconds, 1200 seconds, and 1800 seconds of rest, that is, V i,60 、V i,120 、V i,180 The purpose of this step is to establish a connection between the shorter relaxation voltage and the longer relaxation voltage, and to predict the relaxation voltage value after a longer static time through the voltage data of the first 300 seconds, so as to obtain the voltage characteristics of the complete relaxation process.
[0091] Next, we use the training set D train,1 and the test set D test,1 For CNN model training and testing, the mean square error is used to measure the estimation accuracy of the CNN model. The mean square error of the relaxation voltage is calculated as follows:
[0092]
[0093] Here, i refers to the number of cycles, and k is 1, 2, and 3, representing the voltages after standing for 600 seconds, 1200 seconds, and 1800 seconds, respectively.
[0094] When the number of iterations of the one-dimensional convolutional neural network model training reaches the preset maximum number of iterations, the training is stopped and the CNN model at this time is saved.
[0095] By training and testing the one-dimensional convolutional neural network model, the mean square error value of the relaxation voltage is used to intuitively reflect the estimation accuracy of the one-dimensional convolutional neural network model, so that the model can be adjusted according to the mean square error value of the relaxation voltage to improve the estimation accuracy of the one-dimensional convolutional neural network model. When the preset maximum number of iterations is reached, the training of the one-dimensional convolutional neural network model is stopped to avoid a decrease in generalization ability due to excessive model training times.
[0096] Step S206, inputting the collected voltage sequence data into the trained one-dimensional convolutional neural network model for prediction to obtain a predicted relaxation voltage value.
[0097] Specifically, the above step S206 includes:
[0098] Step S2061: extracting voltage sequence data within a first time period from the collected voltage sequence data.
[0099] Step S2062, inputting the voltage sequence data in the first time period into the trained one-dimensional convolutional neural network model, predicting the voltage sequence data in the second time period, and determining the predicted voltage value as the relaxation voltage value.
[0100] In the embodiment of the present invention, the voltage sequence data in the first time period is extracted from the collected voltage sequence data. Taking the first time period as the first 300 seconds and the second time period as the 600th second, or the 1200th second, or the 1800th second as an example, the voltage sequence V is input into the CNN model to obtain the predicted relaxation voltage value at the 600th second. Relaxation voltage value at 1200 seconds Relaxation voltage value at 1800 seconds The specific formula is as follows:
[0101]
[0102] The one-dimensional convolutional neural network model is used to predict the voltage sequence data in the second time period based on the voltage sequence data in the first time period, and the changing trend of the voltage sequence is effectively analyzed.
[0103] Step S207, calculating voltage statistical features and temperature statistical features respectively according to the collected voltage sequence data and temperature sequence data.
[0104] For details, please see Figure 1 Step S103 of the illustrated embodiment will not be described in detail here.
[0105] Step S208, based on the relaxation process of the lithium battery, a second-order RC equivalent circuit model is established, and the parameters of the equivalent circuit model are calculated.
[0106] In the embodiment of the present invention, Figure 2 As shown, a second-order RC equivalent circuit model is established, and the relaxation voltage V(t) can be expressed as:
[0107] V(t)=U ocv +U1(t)+U2(t)
[0108] Where, V(t) is the relaxation voltage, U ocv To represent the open circuit voltage of the battery, U1(t) and U2(t) are used to represent the polarization voltages generated by the polarization process and the concentration polarization process, respectively.
[0109] The second-order RC equivalent circuit system during the relaxation process can be regarded as the zero-state input of the system and can be expressed as:
[0110]
[0111] Then the relaxation voltage V(t) can be expressed as:
[0112]
[0113] Where V1 and V2 represent the amplitude of polarization voltage, and τ1 and τ2 represent the time constant of polarization process. ocv , τ1, τ2 as unknown parameters, time t as independent variable, relaxation voltage V as dependent variable, and use nonlinear least squares method to identify the parameters V1, V2, U in the function. ocv , τ1, τ2;
[0114] Identify V1, V2, U ocv , τ1, τ2, continue to calculate the polarization capacitance and polarization resistance, the calculation formula is as follows:
[0115]
[0116] Where R1 and R2 are polarization resistors, C1 and C2 are polarization capacitors, R is the DC internal resistance, V1 and V2 are the amplitudes of the polarization voltage, I is the current, τ1 and τ2 are the time constants of the polarization process, U(0) is the charge cut-off voltage, and U ocv is the open circuit voltage.
[0117] By calculating the parameters in the established second-order equivalent circuit model, the circuit model parameters and the actual state of the lithium battery can be intuitively associated to facilitate the estimation of the SOH of the lithium battery.
[0118] Step S209: training the Xgboost model using the relaxation voltage value, voltage statistical characteristics, temperature statistical characteristics and equivalent circuit model parameters.
[0119] Specifically, the above step S209 includes:
[0120] Step S2091 , dividing the relaxation voltage values, voltage statistical characteristics, temperature statistical characteristics and equivalent circuit model parameters into a second training set and a second test set according to a preset ratio.
[0121] Step S2092: Train the Xgboost model using the second training set.
[0122] Step S2093: Use the second test set to test the Xgboost model, and use the mean square error value of SOH to measure the estimation accuracy of the Xgboost model.
[0123] In the embodiment of the present invention, the relaxation voltage value, voltage statistical characteristics, temperature statistical characteristics and equivalent circuit model parameters are combined to form a feature data set D with the SOH of each cycle. The feature data of the i-th cycle in the feature data set D is D i The combination becomes the feature data set D, and the specific form of the feature data set D is as follows:
[0124]
[0125] Divide the data set D into training set D train , test set D test The preset ratio of the training set and the test set is 4:1.
[0126] Construct the Xgboost model and use the training set D train and the test set D test The Xgboost model is trained and tested, and the root mean square error (RMSE) of SOH is used to measure the estimation accuracy of the Xgboost model. The calculation formula of the root mean square error (RMSE) of SOH is:
[0127]
[0128] Among them SOH i is the SOH value corresponding to the i-th cycle, is the SOH value corresponding to the i-th cycle estimated by the model.
[0129] By training and testing the Xgboost model, the Xgboost model has the advantages of high performance and support for parallel computing. The mean square error value of SOH is used to measure the estimation accuracy of the Xgboost model, so that the model can be adjusted according to the mean square error value of SOH to improve the estimation accuracy of the Xgboost model.
[0130] Step S210, determining the hyperparameters and hyperparameter grid corresponding to the Xgboost model.
[0131] Step S211, traverse all hyperparameter points in the hyperparameter grid to determine the hyperparameter point with the smallest mean square error of SOH.
[0132] Step S212: The Xgboost model corresponding to the hyperparameter point with the smallest mean square error of SOH is determined as the trained Xgboost model.
[0133] In the embodiment of the present invention, in the process of training the model, the grid search method is used to optimize the hyperparameters to improve the model prediction effect. The basic idea of the grid search method is to combine the possible values of each hyperparameter into a grid, then exhaustively search all possible combinations, cross-validate each combination, and evaluate the performance of the model on the validation set. The specific method is as follows:
[0134] First, you need to determine the hyperparameters to be searched and their candidate value ranges. Then, use cross-validation to evaluate the performance of each set of hyperparameter combinations. Cross-validation divides the dataset into training and validation sets, and repeats the training and validation process multiple times to obtain a more robust performance evaluation. For each set of hyperparameter combinations in the grid, the above evaluation framework is used for training and validation, and the performance indicators of each combination are recorded. The best hyperparameter combination is selected based on the performance indicators. Usually, the combination with the best performance on the validation set is selected as the hyperparameter of the final model.
[0135] First, determine the hyperparameters to be searched, including the maximum depth of the Xgboost model, the number of decision trees, the learning rate, the minimum number of samples for leaf nodes, etc. Then determine the hyperparameter grid, that is, determine the candidate value range for each hyperparameter. The training set D is used for cross-validation. train To train, the model score is recorded as the test set D test Traverse all the hyperparameter points in the grid and find the hyperparameter point with the highest score (minimum mean square error). The Xgboost model corresponding to this hyperparameter point is used as the final trained model.
[0136] The grid search method is used to comprehensively traverse all hyperparameter points in the hyperparameter grid to improve the comprehensiveness of the search, avoid relying only on the local optimal solution, and improve the generalization ability of the Xgboost model.
[0137] Step S213, inputting the voltage sequence data and temperature sequence data during the actual charging and discharging process of the lithium battery into the trained Xgboost model to obtain the SOH estimation result of the lithium battery.
[0138] Specifically, the above step S213 includes:
[0139] Step S2131, when it is detected that the lithium battery has completed charging and the static time reaches a preset time, the relaxation voltage sequence is input into a one-dimensional convolutional neural network model to obtain a voltage prediction value.
[0140] Step S2132, extracting statistical features and equivalent circuit model features from the voltage series data and temperature series data during the actual charging and discharging process of the lithium battery.
[0141] Step S2133, input the voltage prediction value, statistical characteristics and equivalent circuit model characteristics into the Xgboost model, and output the SOH estimation result of the lithium battery.
[0142] In the embodiment of the present invention, the voltage sequence data during the actual charging and discharging process of the lithium battery includes a relaxation voltage sequence.
[0143] The trained CNN model and the trained Xgboost model are imported into the battery management system BMS. When the BMS monitors that the lithium-ion battery has completed charging and has been stationary for 300 seconds, the collected relaxation voltage series data is first input into the CNN model to obtain the predicted voltage data points, and then the statistical features and equivalent circuit model features are extracted from the voltage series data and the temperature series data. Finally, all the features are combined and input into the Xgboost model to obtain the final estimated SOH result.
[0144] The SOH estimation method for a lithium battery based on a relaxation process provided in this embodiment uses an Xgboost model to comprehensively estimate the battery SOH based on the voltage prediction value output by a one-dimensional convolutional neural network model, statistical features extracted from voltage series data and temperature series data in the actual charging and discharging process, and equivalent circuit model features. This not only takes into account the problem that the battery relaxation process may be shorter in actual operation, but also combines the equivalent circuit model features to effectively improve the SOH estimation accuracy of the lithium battery.
[0145] As a specific application example of the embodiment of the present invention, Figure 3As shown, first, the lithium battery is charged and discharged under different conditions, and the voltage data, temperature data, and battery discharge capacity during the static process after charging are collected, and the SOH label value is generated for each cycle based on the above information. A one-dimensional CNN model is constructed to predict the relaxation voltage values of the 600th, 1200th, and 1800th seconds through the voltage data of the first 300 seconds. The training set and test set are divided according to the ratio, the model is trained and tested, the root mean square error is used to evaluate the accuracy, the model is saved after iterative training to the preset number of times, and the voltage sequence V is predicted. The statistical characteristics (such as standard deviation, skewness, and minimum value) of the voltage and temperature sequences in each cycle are calculated, and a second-order RC equivalent circuit model is established at the same time. The parameters such as polarization voltage are identified by fitting through nonlinear least squares method. These parameters and features such as open circuit voltage constitute the SOH features of the predicted lithium battery. The extracted features and the corresponding SOH labels are combined into a feature data set, and divided into a training set, a validation set, and a test set. The Xgboost model is constructed, and the training and test sets are divided into the model for training and verification. During the training process, the grid search method is used to tune the model hyperparameters, and the mean square error of the test set is used as the model score. The grid of grid points is preset, and after traversing all grid points, the model corresponding to the best score is recorded, that is, the model training is completed. The trained CNN model and Xgboost model are imported into the BMS system. When the BMS monitors that the lithium battery is fully charged and has been standing for 300 seconds, the CNN model is first used to predict the voltage data points, and then the corresponding features are extracted by combining the voltage series data and the temperature series data, and then input into the Xgboost model for lithium battery SOH estimation.
[0146] In this embodiment, a SOH estimation device for a lithium battery based on a relaxation process is also provided, which is used to implement the above-mentioned embodiments and preferred embodiments, and will not be repeated hereafter. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.
[0147] This embodiment provides a SOH estimation device for a lithium battery in a relaxation process, such as Figure 4 As shown, including:
[0148] The data acquisition module 401 is used to perform charge and discharge tests on the lithium battery under different conditions, and to collect voltage sequence data and temperature sequence data of the lithium battery during the charge and discharge tests.
[0149] The prediction module 402 is used to input the collected voltage sequence data into the trained one-dimensional convolutional neural network model for prediction to obtain a predicted relaxation voltage value.
[0150] The first calculation module 403 is used to calculate voltage statistical characteristics and temperature statistical characteristics respectively according to the collected voltage sequence data and temperature sequence data. The voltage statistical characteristics include the standard deviation, skewness and minimum value of the relaxation voltage sequence, and the temperature statistical characteristics include the range, standard deviation and skewness of the temperature sequence.
[0151] The second calculation module 404 is used to establish a second-order RC equivalent circuit model based on the relaxation process of the lithium battery and calculate the parameters of the equivalent circuit model.
[0152] The model training module 405 is used to train the Xgboost model using the relaxation voltage value, voltage statistical characteristics, temperature statistical characteristics and equivalent circuit model parameters.
[0153] The estimation module 406 is used to input the voltage sequence data and temperature sequence data during the actual charging and discharging process of the lithium battery into the trained Xgboost model to obtain the SOH estimation result of the lithium battery.
[0154] In some optional embodiments, the device further comprises:
[0155] The division module is used to divide the collected voltage sequence data into a first training set and a first test set according to a preset ratio.
[0156] The training module is used to train the one-dimensional convolutional neural network model using the first training set.
[0157] The testing module is used to test the one-dimensional convolutional neural network model using the first test set, and to measure the estimation accuracy of the one-dimensional convolutional neural network model using the mean square error value of the relaxation voltage.
[0158] The stop training module is used to stop training the one-dimensional convolutional neural network model when the number of iterations of the one-dimensional convolutional neural network model training reaches a preset maximum number of iterations.
[0159] In some optional implementations, the prediction module 402 includes:
[0160] The first extraction unit is used to extract voltage sequence data within a first time period from the collected voltage sequence data.
[0161] The first prediction unit is used to input the voltage sequence data in the first time period into the trained one-dimensional convolutional neural network model, predict the voltage sequence data in the second time period, and determine the predicted voltage value as the relaxation voltage value.
[0162] In some optional implementations, the model training module 405 includes:
[0163] A division unit is used to divide the relaxation voltage value, voltage statistical characteristics, temperature statistical characteristics and equivalent circuit model parameters into a second training set and a second test set according to a preset ratio.
[0164] The training unit is used to train the Xgboost model using the second training set.
[0165] The testing unit is used to test the Xgboost model by using the second test set, and measure the estimation accuracy of the Xgboost model by using the mean square error value of SOH.
[0166] In some optional embodiments, the device further comprises:
[0167] The first determination module is used to determine the hyperparameters and hyperparameter grid corresponding to the Xgboost model.
[0168] The second determination module is used to traverse all hyperparameter points in the hyperparameter grid and determine the hyperparameter point with the smallest mean square error of SOH.
[0169] The third determination module is used to determine the Xgboost model corresponding to the hyperparameter point with the smallest mean square error of SOH as the trained Xgboost model.
[0170] In some optional implementations, the estimation module 406 includes:
[0171] The second prediction unit is used to input the relaxation voltage sequence into the one-dimensional convolutional neural network model to obtain a voltage prediction value when it is detected that the lithium battery has completed charging and the static time has reached a preset time.
[0172] The second extraction unit is used to extract statistical features and equivalent circuit model features from the voltage sequence data and temperature sequence data during the actual charging and discharging process of the lithium battery.
[0173] The output unit is used to input the voltage prediction value, statistical characteristics, and equivalent circuit model characteristics into the Xgboost model and output the SOH estimation result of the lithium battery.
[0174] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.
[0175] The SOH estimation device of the lithium battery in the relaxation process in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.
[0176] The embodiment of the present invention also provides a computer device having the above Figure 4 The relaxation process of the lithium battery SOH estimation device is shown.
[0177] See also Figure 5 , Figure 5 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 5 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 5 A processor 10 is taken as an example.
[0178] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.
[0179] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.
[0180] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0181] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.
[0182] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 5 The example of connecting through bus is taken in the following.
[0183] The input device 30 can receive input digital or character information and generate key signal input related to user settings and function control of the computer device, such as a touch screen, etc. The output device 40 can include a display device, etc.
[0184] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.
[0185] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.
Claims
1. A SOH estimation method for lithium batteries based on relaxation process, It is characterized in that The method comprises: Carry out charge and discharge tests on lithium batteries under different conditions, and collect voltage sequence data and temperature sequence data of lithium batteries during the charge and discharge tests; The collected voltage sequence data is input into the trained one-dimensional convolutional neural network model for prediction to obtain the predicted relaxation voltage value; According to the collected voltage sequence data and temperature sequence data, voltage statistical characteristics and temperature statistical characteristics are calculated respectively, wherein the voltage statistical characteristics include the standard deviation, skewness and minimum value of the relaxation voltage sequence, and the temperature statistical characteristics include the range, standard deviation and skewness of the temperature sequence; Based on the relaxation process of lithium batteries, a second-order RC equivalent circuit model is established, and the parameters of the equivalent circuit model are calculated; The Xgboost model is trained using the relaxation voltage value, voltage statistical characteristics, temperature statistical characteristics and equivalent circuit model parameters; The voltage series data and temperature series data of the actual charging and discharging process of the lithium battery are input into the trained Xgboost model to obtain the SOH estimation result of the lithium battery.
2. The method according to claim 1, It is characterized in that Before inputting the collected voltage sequence data into the trained one-dimensional convolutional neural network model for prediction, the method further includes: Dividing the collected voltage sequence data into a first training set and a first test set according to a preset ratio; Using the first training set to train a one-dimensional convolutional neural network model; Using the first test set to test the one-dimensional convolutional neural network model, and using the mean square error value of the relaxation voltage to measure the estimation accuracy of the one-dimensional convolutional neural network model; When the number of iterations of the one-dimensional convolutional neural network model training reaches a preset maximum number of iterations, the training of the one-dimensional convolutional neural network model is stopped.
3. The method according to claim 1, It is characterized in that The collected voltage sequence data is input into a trained one-dimensional convolutional neural network model for prediction to obtain a predicted relaxation voltage value, including: Extracting voltage sequence data within a first time period from the collected voltage sequence data; The voltage sequence data in the first time period is input into a trained one-dimensional convolutional neural network model, the voltage sequence data in the second time period is predicted, and the predicted voltage value is determined as the relaxation voltage value.
4. The method according to claim 1, It is characterized in that The equivalent circuit model parameters include polarization resistance value, polarization capacitance value and DC internal resistance value, and the equivalent circuit model parameters are calculated according to the following formula: Among them, R 1 and R 2 is the polarization resistance, C 1 and C 2 is the polarized capacitor, R is the DC internal resistance, V 1 and V 2 is the amplitude of polarization voltage, I is the current, τ 1 and τ 2 is the time constant of the polarization process, U(0) is the charge cut-off voltage, and U ocv is the open circuit voltage.
5. The method according to claim 1, It is characterized in that The method of training the Xgboost model by using the relaxation voltage value, voltage statistical characteristics, temperature statistical characteristics and equivalent circuit model parameters includes: Dividing the relaxation voltage values, voltage statistical characteristics, temperature statistical characteristics and equivalent circuit model parameters into a second training set and a second test set according to a preset ratio; Using the second training set to train the Xgboost model; The Xgboost model is tested using the second test set, and the mean square error value of SOH is used to measure the estimation accuracy of the Xgboost model.
6. The method according to claim 1, It is characterized in that After training the Xgboost model using the relaxation voltage value, voltage statistical characteristics, temperature statistical characteristics and equivalent circuit model parameters, the method further includes: Determine the hyperparameters and hyperparameter grid corresponding to the Xgboost model; Traversing all hyperparameter points in the hyperparameter grid, and determining the hyperparameter point with the smallest mean square error of SOH; The Xgboost model corresponding to the hyperparameter point with the smallest mean square error of SOH is determined as the trained Xgboost model.
7. The method according to claim 1, It is characterized in that The voltage sequence data during the actual charging and discharging process of the lithium battery includes a relaxation voltage sequence. The voltage sequence data and temperature sequence data during the actual charging and discharging process of the lithium battery are input into the trained Xgboost model to obtain the SOH estimation result of the lithium battery, including: When it is detected that the lithium battery has completed charging and the static time reaches a preset time, the relaxation voltage sequence is input into a one-dimensional convolutional neural network model to obtain a voltage prediction value; Extracting statistical features and equivalent circuit model features from voltage series data and temperature series data during the actual charging and discharging process of the lithium battery; The voltage prediction value, statistical characteristics and equivalent circuit model characteristics are input into the Xgboost model, and the SOH estimation result of the lithium battery is output.
8. A SOH estimation device for lithium batteries based on relaxation process, It is characterized in that The device comprises: The data acquisition module is used to perform charge and discharge tests on the lithium battery under different conditions, and to collect voltage sequence data and temperature sequence data of the lithium battery during the charge and discharge test; A prediction module is used to input the collected voltage sequence data into a trained one-dimensional convolutional neural network model for prediction to obtain a predicted relaxation voltage value; A first calculation module, used to calculate voltage statistical characteristics and temperature statistical characteristics respectively according to the collected voltage sequence data and temperature sequence data, wherein the voltage statistical characteristics include the standard deviation, skewness and minimum value of the relaxation voltage sequence, and the temperature statistical characteristics include the range, standard deviation and skewness of the temperature sequence; The second calculation module is used to establish a second-order RC equivalent circuit model based on the relaxation process of the lithium battery and calculate the parameters of the equivalent circuit model; A model training module, used to train the Xgboost model using the relaxation voltage value, voltage statistical characteristics, temperature statistical characteristics and equivalent circuit model parameters; The estimation module is used to input the voltage sequence data and temperature sequence data of the actual charging and discharging process of the lithium battery into the trained Xgboost model to obtain the SOH estimation result of the lithium battery.
9. A computer device, It is characterized in that include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the SOH estimation method for a lithium battery based on a relaxation process as described in any one of claims 1 to 7 by executing the computer instructions.
10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the SOH estimation method for a lithium battery based on a relaxation process according to any one of claims 1 to 7.
Citation Information
Cited By
Battery SOC-OCV curve determination method, device, equipment, medium and program product
CN120742126A
