Battery Aging State Estimation Method and System Based on Improved Gaussian Process Regression

Abstract

The present invention relates to a method and system for estimating the aging state of a battery based on an improved Gaussian process regression, and relates to the technical field of batteries. The method includes: obtaining an original data set of a lithium battery at different SOC intervals and different discharge current rates through experiments; analyzing the original data set to determine a model data set; dividing the model data set into a training set and a test set; establishing an improved Gaussian process regression model; training and testing the improved Gaussian process regression model using the training set and the test set to generate a trained improved Gaussian process regression model for predicting the aging state of the lithium battery. By coupling the input feature values and adding the model estimated value of the previous moment as an input feature value, the present invention reduces the model dimension, lowers the training difficulty, significantly improves the accuracy of the battery aging state estimation, and greatly reduces the uncertainty of the prediction result.

Description

Technical Field

[0001] The present invention relates to the technical field of batteries, and in particular to a method and system for estimating the aging state of a battery based on improved Gaussian process regression. Background Art

[0002] Since modern times, with the extensive use of fossil fuels such as petroleum, the ecological environment has been severely damaged. For the sustainable development of the economy and the friendly harmony between humans and nature, there is an urgent need to develop new environmentally friendly energy sources. As an energy storage component, lithium batteries have been widely used in many fields due to their high energy density, high power density, small specific gravity, rechargeable and dischargeable, and low environmental impact. As the number of charge and discharge cycles of lithium batteries increases, the aging state of lithium batteries will continue to deepen, and their available capacity will decrease significantly. Incorrect use of lithium batteries may cause serious safety accidents. Therefore, accurate estimation of the aging state of lithium batteries has important practical significance.

[0003] Currently, the research on the aging state estimation of lithium batteries mainly falls into three categories: mechanism models, equivalent models, and data models. Among them, the method based on the data model generally uses the voltage, current, temperature, SOC (State of Charge), and other intervals during the charge and discharge process of the battery as the input values of the model to estimate the aging state of the battery. Gaussian process regression has strong non-linear fitting ability and is suitable for regression problems with high dimensions, non-linearity, and small samples. It has been widely used in system identification, control system design, system prediction, etc. Since battery aging is a complex non-linear process, Gaussian process regression is also widely used in the estimation of lithium battery aging. However, due to the fact that the median value of the input feature value SOC, the depth of discharge, and the discharge rate are constant throughout the aging cycle, and due to the characteristics of the Matern kernel function, the more similar the feature values are, the larger the output value of the kernel function will be. Therefore, only the equivalent cycle number of the feature values changes during the entire aging cycle. The constant feature values make the output of the kernel function large and basically consistent, which will have a negative impact on training, resulting in poor hyperparameter optimization effect and ultimately poor prediction accuracy. In addition, the 95% confidence interval is relatively wide, and during the aging prediction process, only the equivalent decay rate under other aging stresses is relied on for prediction, and the time attribute between the capacity decay rates cannot be related, resulting in large uncertainty in the prediction results. Summary of the Invention

[0004] The object of the present invention is to provide a method and system for estimating the aging state of a battery based on improved Gaussian process regression, so as to greatly reduce the uncertainty of the battery aging state prediction result and improve the prediction accuracy.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A method for estimating the aging state of a battery based on an improved Gaussian process regression, comprising:

[0007] Obtaining an original data set of a lithium battery at different SOC intervals and different discharge current rates through experiments;

[0008] Analyzing the original data set to determine a model data set; the model data set includes the median of the SOC interval, the depth of discharge, the discharge current rate, and the equivalent number of cycles as input feature values, and includes the lithium battery aging state index as the output value;

[0009] Dividing the model data set into a training set and a test set;

[0010] Establishing an improved Gaussian process regression model; the input of the improved Gaussian process regression model is the coupled processed feature values and the battery aging state at the previous moment, and the output is the battery aging state at the current moment;

[0011] Training and testing the improved Gaussian process regression model using the training set and the test set to generate a trained improved Gaussian process regression model;

[0012] Predicting the aging state of the lithium battery using the trained improved Gaussian process regression model.

[0013] Optionally, the obtaining an original data set of a lithium battery at different SOC intervals and different discharge current rates through experiments specifically includes:

[0014] Setting multiple SOC intervals and multiple discharge current rates of the lithium battery;

[0015] Performing multiple cycle charge and discharge experiments on multiple lithium batteries of the same model at different SOC intervals and different discharge current rates, and recording the SOC interval and the discharge current rate corresponding to each lithium battery during the multiple cycle charge and discharge experiments;

[0016] After the multiple cycle charge and discharge experiments, performing another charge and discharge experiment on the lithium battery at a preset discharge current rate, and recording the discharge capacity during this charge and discharge experiment as the current available capacity of the lithium battery after experiencing multiple charge and discharge cycles;

[0017] Generating the original data set according to the SOC interval and the discharge current rate corresponding to each lithium battery during the multiple cycle charge and discharge experiments and the current available capacity of the lithium battery after experiencing multiple charge and discharge cycles.

[0018] Optionally, the analyzing the original data set to determine a model data set specifically includes:

[0019] Calculate the median of the corresponding SOC interval according to the different SOC intervals;

[0020] Calculate the depth of discharge according to the different SOC intervals;

[0021] Calculate the aging state index of the lithium battery after the nth charge and discharge cycle according to the current available capacity of the lithium battery after experiencing multiple charge and discharge cycles;

[0022] Use the median of the SOC interval, the depth of discharge, the corresponding discharge current rate, and the equivalent number of cycles as input feature values, and use the corresponding aging state index as the output value to form the model data set.

[0023] Optionally, the establishment of the improved Gaussian process regression model specifically includes:

[0024] Construct Gaussian process regression models using different kernel functions;

[0025] Initialize the hyperparameters to zero, train the Gaussian process regression models using different kernel functions, optimize the hyperparameters by the conjugate gradient method, and use RMSE to evaluate the error of the Gaussian process regression models using different kernel functions;

[0026] Determine the Gaussian process regression model using the Matern+LIN kernel function with the minimum error as the improved Gaussian process regression model.

[0027] Optionally, the training and testing of the improved Gaussian process regression model using the training set and the testing set to generate a trained improved Gaussian process regression model specifically includes:

[0028] Multiply the equivalent number of cycles by the corresponding median of the SOC interval, the depth of discharge, and the discharge current rate respectively to obtain the coupled feature values;

[0029] Use the coupled feature values and the battery aging state at the previous moment as the input values of the improved Gaussian process regression model, and use the battery aging state at the current moment as the output value of the improved Gaussian process regression model to train the improved Gaussian process regression model;

[0030] During the training process, use the testing set to test the improved Gaussian process regression model, and use the goodness of fit and the root mean square error as the model evaluation indicators;

[0031] Determine the improved Gaussian process regression model that meets the preset model evaluation indicators as the trained improved Gaussian process regression model.

[0032] A battery aging state estimation system based on an improved Gaussian process regression, comprising:

[0033] An original data acquisition module, configured to obtain an original data set of a lithium battery at different SOC intervals and different discharge current rates through experiments;

[0034] An original data analysis module, configured to analyze the original data set to determine a model data set; the model data set includes the median of the SOC interval, the depth of discharge, the discharge current rate, and the equivalent cycle number as input feature values, and includes the lithium battery aging state index as an output value;

[0035] A model data set division module, configured to divide the model data set into a training set and a test set;

[0036] A model establishment module, configured to establish an improved Gaussian process regression model; the input of the improved Gaussian process regression model is the coupled processed feature values and the battery aging state at the previous moment, and the output is the battery aging state at the current moment;

[0037] A model training module, configured to train and test the improved Gaussian process regression model by using the training set and the test set to generate a trained improved Gaussian process regression model;

[0038] A battery aging state prediction module, configured to predict the lithium battery aging state by using the trained improved Gaussian process regression model.

[0039] Optionally, the original data acquisition module specifically includes:

[0040] An SOC interval and rate setting unit, configured to set multiple SOC intervals and multiple discharge current rates of the lithium battery;

[0041] A cyclic charge and discharge experiment unit, configured to perform multiple cyclic charge and discharge experiments on multiple lithium batteries of the same model at different SOC intervals and different discharge current rates, and record the corresponding SOC interval and discharge current rate of each lithium battery during the multiple cyclic charge and discharge experiments;

[0042] A current available capacity experiment unit, configured to perform one more charge and discharge experiment on the lithium battery at a preset discharge current rate after the multiple cyclic charge and discharge experiments, and record the discharge capacity during this charge and discharge experiment as the current available capacity of the lithium battery after experiencing multiple charge and discharge cycles;

[0043] An original data set generation unit, configured to generate the original data set according to the corresponding SOC interval and discharge current rate of each lithium battery during the multiple cyclic charge and discharge experiments and the current available capacity of the lithium battery after experiencing multiple charge and discharge cycles.

[0044] Optionally, the original data analysis module specifically includes:

[0045] SOC interval median calculation unit, configured to calculate the corresponding SOC interval median according to the different SOC intervals;

[0046] Discharge depth calculation unit, configured to calculate the discharge depth according to the different SOC intervals;

[0047] Aging state index calculation unit, configured to calculate the aging state index of the lithium battery after the nth charge and discharge cycle according to the current available capacity of the lithium battery after experiencing multiple charge and discharge cycles;

[0048] Model data set generation unit, configured to use the SOC interval median, the discharge depth, the corresponding discharge current rate, and the equivalent cycle number as input feature values, and use the corresponding aging state index as the output value to constitute the model data set.

[0049] Optionally, the model establishment module specifically includes:

[0050] Gaussian process regression model construction unit, configured to construct Gaussian process regression models using different kernel functions;

[0051] Gaussian process regression model evaluation unit, configured to initialize the hyperparameters to zero, train the Gaussian process regression models using different kernel functions, optimize the hyperparameters by the conjugate gradient method, and use RMSE to evaluate the errors of the Gaussian process regression models using different kernel functions;

[0052] Improved Gaussian process regression model construction unit, configured to determine the Gaussian process regression model using the Matern+LIN kernel function with the smallest error as the improved Gaussian process regression model.

[0053] Optionally, the model training module specifically includes:

[0054] Feature value coupling unit, configured to multiply the equivalent cycle number by the corresponding SOC interval median, discharge depth, and discharge current rate respectively to obtain the coupled feature values;

[0055] Model training unit, configured to use the coupled feature values and the battery aging state at the previous moment as the input values of the improved Gaussian process regression model, and use the battery aging state at the current moment as the output value of the improved Gaussian process regression model to train the improved Gaussian process regression model;

[0056] A model testing unit, which is used to test the improved Gaussian process regression model with the test set during the training process, and uses the goodness of fit and the root mean square error as model evaluation indicators;

[0057] A model training completion unit, which is used to determine the improved Gaussian process regression model that meets the preset model evaluation indicators as the trained improved Gaussian process regression model.

[0058] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0059] The present invention provides a method and system for estimating the battery aging state based on an improved Gaussian process regression. The method includes: obtaining an original data set of a lithium battery under different SOC intervals and different discharge current rates through experiments; analyzing the original data set to determine a model data set; the model data set includes the median value of the SOC interval, the depth of discharge, the discharge current rate, and the equivalent cycle number as input feature values, and includes the lithium battery aging state index as an output value; dividing the model data set into a training set and a test set; establishing an improved Gaussian process regression model; the input of the improved Gaussian process regression model is the coupled processed feature values and the battery aging state at the previous moment, and the output is the battery aging state at the current moment; using the training set and the test set to train and test the improved Gaussian process regression model to generate a trained improved Gaussian process regression model; using the trained improved Gaussian process regression model to predict the lithium battery aging state. By studying the cyclic discharge of the lithium battery under different SOC intervals and different discharge current rates, the present invention qualitatively and quantitatively analyzes the influence of the SOC interval and the discharge current rate on the lithium battery aging rate; by coupling the input feature values and adding the model estimated value at the previous moment as an input feature value, the model dimension is reduced, the training difficulty is lowered, and at the same time, the accuracy of the battery aging state estimation is significantly improved, making the correlation coefficient R 2 significantly increase, the RMSE significantly decrease, the confidence interval width decrease from 1.25% to 0.4%, and the uncertainty of the prediction result is greatly reduced. Description of the Drawings

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0061] Figure 1 It is a flowchart of a method for estimating the battery aging state based on an improved Gaussian process regression of the present invention;

[0062] Figure 2 Schematic diagram of the principle of a battery aging state estimation method based on improved Gaussian process regression according to the present invention;

[0063] Figure 3 Schematic diagram of the improved Gaussian process regression model provided by an embodiment of the present invention. Specific implementation manners

[0064] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0065] The object of the present invention is to provide a battery aging state estimation method and system based on improved Gaussian process regression. By coupling the characteristic input values and introducing the model estimation value of the previous moment, the correlation coefficient R 2 significantly increases, the RMSE significantly decreases, the confidence interval width drops from 1.25% to 0.4%, greatly reducing the uncertainty of the battery aging state prediction result and improving the prediction accuracy.

[0066] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific implementation manners.

[0067] Figure 1 Flowchart of a battery aging state estimation method based on improved Gaussian process regression according to the present invention, Figure 2 Schematic diagram of the principle of a battery aging state estimation method based on improved Gaussian process regression according to the present invention. Refer to Figure 1 and Figure 2 , a battery aging state estimation method based on improved Gaussian process regression according to the present invention includes:

[0068] Step 101: Obtain the original data set of the lithium battery at different SOC intervals and different discharge current rates through experiments.

[0069] The present invention obtains the discharge data of different SOC cycling intervals and different discharge rates through experiments to form the original data set.

[0070] The step 101 of obtaining the original data set of the lithium battery at different SOC intervals and different discharge current rates through experiments specifically includes:

[0071] Step 1.1: Set multiple SOC intervals and multiple discharge current rates of the lithium battery;

[0072] In the present invention, M lithium batteries of the same model are selected for N times of cyclic charge and discharge experiments. The SOC operating range (abbreviated as SOC range) of the lithium battery is expressed as SOC l ~SOC h , where SOC l represents the lower limit of the SOC range, and SOC h represents the upper limit of the SOC range. In the embodiments of the present invention, multiple SOC operating ranges of the lithium battery are set as SOC li ~SOC hi , where i = 4, and SOC li represents the lower limit of the i-th SOC range, and SOC hi represents the upper limit of the i-th SOC range. The 4 different SOC operating ranges are specifically:

[0073] SOC l1 ~SOC h1 = 15% - 40%

[0074] SOC l2 ~SOC h2 = 40% - 65%

[0075] SOC l3 ~SOC h3 = 65% - 90%

[0076] SOC l4 ~SOC h4 = 15% - 90%

[0077] There are m1 lithium batteries in each of the 4 working ranges, that is, M = 4 * m1.

[0078] In the embodiments of the present invention, the set charging current rate (abbreviated as charging rate) is 1C, and multiple discharge current rates (abbreviated as discharge rates) include three rates of 2C, 4C, and 10C.

[0079] Step 1.2: Perform multiple cyclic charge and discharge experiments on multiple lithium batteries of the same model under different SOC ranges and different discharge current rates, and record the SOC range and discharge current rate corresponding to each lithium battery during the multiple cyclic charge and discharge experiments.

[0080] Select M lithium batteries of the same model for N times of cyclic charge and discharge experiments. In the examples of the present invention, M is equal to 36 and N is equal to 4500. The lithium batteries in each working range are cyclically charged and discharged within the corresponding range, the charging current is 1C rate, and the constant current charging is up to SOC h ; the discharge currents are 2C, 4C, and 10C rates respectively, and the constant current discharge is up to SOC l。There are m2 lithium batteries for each of the three discharge rates in the present invention, i.e., m1 = 3 * m2. During the experiment, the discharge SOC range and discharge rate of the corresponding lithium batteries are recorded.

[0081] Step 1.3: After the above-mentioned multiple cycle charge-discharge experiments, perform one more charge-discharge experiment on the lithium battery at a preset discharge current rate, and record the discharge capacity during this charge-discharge experiment as the current available capacity of the lithium battery after experiencing multiple charge-discharge cycles.

[0082] In the embodiment of the present invention, N = 50. After every 50 repeated cycle charge-discharge experiments, charge all the lithium batteries at a constant current of 1C to the charge cut-off voltage, and then use constant voltage charging until the charging current is less than the preset threshold to stop charging; perform constant current discharge on all the lithium batteries at a constant rate of 1C until the voltage of the lithium battery drops to the discharge cut-off voltage; record the discharge capacity of the whole process as the current available capacity Q of the lithium battery after experiencing 50 charge-discharge cycles. now 。The definition of the lithium battery aging state index adopted by the present invention is as follows:

[0083]

[0084] where is defined as the aging state index of the lithium battery after the nth repetition of 50 cycle charge-discharge experiments, representing the aging state after the nth cycle of 50 times; Q0 is defined as the initial rated capacity of the lithium battery, is defined as the current available capacity of the lithium battery after the nth cycle of 50 times.

[0085] Step 1.4: Generate the original data set according to the SOC range and discharge current rate corresponding to each lithium battery during the multiple cycle charge-discharge experiments and the current available capacity of the lithium battery after experiencing multiple charge-discharge cycles.

[0086] Step 102: Analyze the original data set to determine the model data set.

[0087] Analyze the original data set to determine that the characteristic values are the median of the SOC range, discharge depth, discharge current rate, and number of cycles. The model data set includes the median of the SOC range, discharge depth, discharge current rate, and equivalent number of cycles as input characteristic values, and includes the lithium battery aging state index as the output value.

[0088] The above-mentioned Step 102 analyzes the original data set to determine the model data set, which specifically includes:

[0089] Step 2.1: Calculate the corresponding median of the SOC range according to the different SOC ranges.

[0090] Calculate the median SOC of the SOC range based on the charge-discharge SOC range m ; The median SOC of the i-th SOC range m The calculation formula is as follows:

[0091]

[0092] where i = 1 to 4.

[0093] Step 2.2: Calculate the depth of discharge according to the different SOC ranges.

[0094] Calculate the depth of discharge DOD based on the charge-discharge SOC range; the calculation formula for the DOD of the i-th depth of discharge is as follows:

[0095] DOD i =(SOC hi -SOC li )

[0096] where i = 1 to 4.

[0097] Step 2.3: Calculate the aging state index of the lithium battery after the n-th cycle of multiple charge-discharge cycles according to the current available capacity of the lithium battery after experiencing multiple charge-discharge cycles.

[0098] Calculate the aging state after the 50th cycle of the n-th cycle as Aging state The calculation formula is as follows:

[0099]

[0100] Step 2.4: Use the median SOC, the depth of discharge, the corresponding discharge current rate, and the equivalent cycle number as input feature values, and use the corresponding aging state index as the output value to form the model data set.

[0101] During the experiment, record the battery discharge rate corresponding to the i-th SOC range as C disi , where i = 1 to 4.

[0102] In the present invention, to unify the calculation of capacity release, one cycle at 75% depth of discharge is used as the standard cycle (Standard cycle, C s ), so 3 cycles in the 25% interval are equal to 1 standard cycle. Then the calculation formula for the equivalent cycle number (Equivalent cycle times, C e ) is as follows:

[0103] C s = 3C e1 = 3C e2 = 3Ce3 = C e4

[0104] where C ei represents the equivalent cycle number corresponding to the i-th SOC interval, where i = 1 to 4.

[0105] Taking the median SOC of the calculated SOC interval mi , the depth of discharge DOD i , the corresponding discharge current rate C disi and the equivalent cycle number C e ; as input feature values, and taking the corresponding aging state index as the output value to form the model data set.

[0106] After the lithium battery has experienced multiple charge and discharge cycles for the nth time, the collected model data set includes:

[0107] Input values:

[0108]

[0109]

[0110]

[0111]

[0112] Output values:

[0113]

[0114] where DODn, and respectively represent the median of the SOC interval, the depth of discharge, the discharge current rate, the equivalent cycle number, and the aging state index of the lithium battery after repeating 50 charge and discharge cycles for the nth time.

[0115] Step 103: Divide the model data set into a training set and a test set.

[0116] Divide the model data set composed of input feature data into a training set and a test set, and the relationship between the two sets of data is shown in the following table:

[0117]

[0118] Among them, the training data set is marked with "√", and the test data set is marked with "#".

[0119] Step 104: Establish an improved Gaussian process regression model.

[0120] First, Gaussian process regression models with different kernel functions are constructed. The hyperparameters are initialized to zero, and the Gaussian process regression models with different kernel functions are trained. Hyperparameter optimization is performed by the conjugate gradient method, and the RMSE (Root Mean Squard Error) is used to evaluate the model errors of the Gaussian process regression models with different kernel functions. Finally, it is determined to select the Matern + linear kernel function (Linear, LIN) as the kernel function of the improved Gaussian process regression model of the present invention. Figure 3 It is a schematic diagram of the improved Gaussian process regression model provided by the embodiment of the present invention. Refer to Figure 3 , the input of the improved Gaussian process regression model is the eigenvalue after coupling processing and the battery aging state at the previous moment, and the output is the battery aging state at the current moment.

[0121] Therefore, step 104 of establishing the improved Gaussian process regression model specifically includes:

[0122] Step 4.1: Construct Gaussian process regression models with different kernel functions;

[0123] Step 4.2: Initialize the hyperparameters to zero, train the Gaussian process regression models with different kernel functions, perform hyperparameter optimization by the conjugate gradient method, and use RMSE to evaluate the errors of the Gaussian process regression models with different kernel functions;

[0124] Step 4.3: Determine the Gaussian process regression model with the Matern + LIN kernel function having the minimum error as the improved Gaussian process regression model.

[0125] Step 105: Use the training set and the test set to train and test the improved Gaussian process regression model to generate a trained improved Gaussian process regression model.

[0126] Use the training set to train the improved Gaussian process regression model, test the model performance through the test set, and use the goodness of fit R 2 and the root mean square error RMSE as the model evaluation indicators.

[0127] Step 105 of using the training set and the test set to train and test the improved Gaussian process regression model to generate a trained improved Gaussian process regression model specifically includes:

[0128] Step 5.1: Multiply the equivalent cycle number by the median value of the corresponding SOC interval, the discharge depth, and the discharge current multiple respectively to obtain the eigenvalue after coupling processing.

[0129] Multiply the equivalent number of cycles by three fixed input eigenvalues (the median value of the SOC interval, depth of discharge, and discharge current rate) respectively as the input values of the improved Gaussian process regression model, so that it has cycle number information while retaining the original eigenvalue information. Therefore, the eigenvalue after coupling processing obtained in the present invention includes and

[0130] Step 5.2: Use the eigenvalue after the coupling processing and the battery aging state at the previous moment as the input values of the improved Gaussian process regression model, and use the battery aging state at the current moment as the output value of the improved Gaussian process regression model to train the improved Gaussian process regression model.

[0131] Also add the model estimated value at the previous moment into the Gaussian process regression model to jointly estimate the aging state at the current moment. Specifically, import the input features of the training set and the model estimated value at the previous moment into the improved Gaussian process regression model for training; import the input features of the test set and the model estimated value at the previous moment into the Gaussian process regression model for testing to obtain the output feature as shown in Figure 3 .

[0132] During the training process, use the test set to test the improved Gaussian process regression model, and use the goodness of fit and root mean square error as the model evaluation indicators.

[0133] The present invention uses the goodness of fit R 2 and the root mean square error RMSE as the evaluation indicators of the algorithm. Among them, the calculation formula of R 2 is as follows:

[0134]

[0135] The calculation formula of RMSE is as follows:

[0136]

[0137] Among them, y i is the true value of the aging state of the i-th; is the model estimated value corresponding to the aging state, which is the value of the battery aging state index predicted by the improved Gaussian process regression model at the current moment in the present invention; is the arithmetic mean of the aging states at all moments, and m is the number of true values or estimated values of the aging state.

[0138] Step 5.4: Determine the improved Gaussian process regression model that meets the preset model evaluation index as the trained improved Gaussian process regression model.

[0139] Step 106: Use the trained improved Gaussian process regression model to predict the aging state of the lithium battery.

[0140] When predicting the aging state of the lithium battery, the input features of the lithium battery to be predicted and the model estimated value at the previous moment are input into the trained improved Gaussian process regression model, and the output features can be obtained The output features characterize the aging state of the lithium battery.

[0141] In the present invention, by studying the cyclic discharge of the lithium battery in different SOC working ranges and different discharge rates, the influence of the SOC working range and the discharge rate on the aging rate of the lithium battery is qualitatively and quantitatively analyzed. By performing coupling processing on the input feature values and adding the model estimated value at the previous moment as the input feature value, the model dimension is reduced, the training difficulty is lowered, and at the same time, the accuracy of estimating the battery aging state is significantly improved, making the correlation coefficient R 2 significantly increase, the RMSE significantly decrease, the width of the confidence interval decrease from 1.25% to 0.4%, and the uncertainty of the prediction result is greatly reduced.

[0142] Based on the method provided by the present invention, the present invention also provides a battery aging state estimation system based on an improved Gaussian process regression. The system includes:

[0143] An original data acquisition module, configured to obtain the original data set of the lithium battery at different SOC intervals and different discharge current rates through experiments;

[0144] An original data analysis module, configured to analyze the original data set to determine the model data set; the model data set includes the median value of the SOC interval, the depth of discharge, the discharge current rate, and the equivalent cycle number as input feature values, and includes the lithium battery aging state index as the output value;

[0145] A model data set division module, configured to divide the model data set into a training set and a test set;

[0146] A model establishment module, configured to establish an improved Gaussian process regression model; the input of the improved Gaussian process regression model is the coupled feature value and the battery aging state at the previous moment, and the output is the battery aging state at the current moment;

[0147] A model training module, which is used to train and test the improved Gaussian process regression model by using the training set and the test set, and generate a trained improved Gaussian process regression model;

[0148] A battery aging state prediction module, which is used to predict the aging state of a lithium battery by using the trained improved Gaussian process regression model.

[0149] Among them, the original data acquisition module specifically includes:

[0150] An SOC interval and rate setting unit, which is used to set multiple SOC intervals and multiple discharge current rates of the lithium battery;

[0151] A cyclic charge and discharge experiment unit, which is used to perform multiple cyclic charge and discharge experiments on multiple lithium batteries of the same model under different SOC intervals and different discharge current rates, and record the SOC interval and discharge current rate corresponding to each lithium battery during the multiple cyclic charge and discharge experiments;

[0152] A current available capacity experiment unit, which is used to perform one more charge and discharge experiment on the lithium battery at a preset discharge current rate after the multiple cyclic charge and discharge experiments, and record the discharge capacity during this charge and discharge experiment as the current available capacity of the lithium battery after experiencing multiple charge and discharge cycles;

[0153] An original data set generation unit, which is used to generate the original data set according to the SOC interval and discharge current rate corresponding to each lithium battery during the multiple cyclic charge and discharge experiments and the current available capacity of the lithium battery after experiencing multiple charge and discharge cycles.

[0154] The original data analysis module specifically includes:

[0155] An SOC interval median calculation unit, which is used to calculate the corresponding SOC interval median according to the different SOC intervals;

[0156] A discharge depth calculation unit, which is used to calculate the discharge depth according to the different SOC intervals;

[0157] An aging state index calculation unit, which is used to calculate the aging state index of the lithium battery after the nth time of experiencing multiple charge and discharge cycles according to the current available capacity of the lithium battery after experiencing multiple charge and discharge cycles;

[0158] A model data set generation unit, which is used to form the model data set by using the SOC interval median, the discharge depth, the corresponding discharge current rate and the equivalent cycle number as input feature values and the corresponding aging state index as the output value.

[0159] The model establishment module specifically includes:

[0160] A Gaussian process regression model construction unit for constructing Gaussian process regression models using different kernel functions;

[0161] A Gaussian process regression model evaluation unit for initializing hyperparameters to zero, training the Gaussian process regression models using different kernel functions, optimizing hyperparameters through the conjugate gradient method, and using RMSE to evaluate the error of the Gaussian process regression models using different kernel functions;

[0162] An improved Gaussian process regression model construction unit for determining the Gaussian process regression model using the Matern+LIN kernel function with the minimum error as the improved Gaussian process regression model.

[0163] The model training module specifically includes:

[0164] An eigenvalue coupling unit for multiplying the equivalent cycle number with the median value of the corresponding SOC interval, the depth of discharge, and the discharge current rate respectively to obtain the coupled eigenvalues;

[0165] A model training unit for using the coupled eigenvalues and the battery aging state at the previous moment as the input values of the improved Gaussian process regression model, and using the battery aging state at the current moment as the output value of the improved Gaussian process regression model to train the improved Gaussian process regression model;

[0166] A model testing unit for testing the improved Gaussian process regression model using the test set during the training process, and using the goodness of fit and the root mean square error as model evaluation indicators;

[0167] A model training completion unit for determining the improved Gaussian process regression model that meets the preset model evaluation indicators as the trained improved Gaussian process regression model.

[0168] The present invention discloses a method and system for estimating the battery aging state using an improved Gaussian process regression based on the SOC cycle interval and the discharge rate, including: experimentally obtaining discharge data in different SOC cycle intervals and different discharge rates, analyzing the data set, determining the model input values, constructing an improved Gaussian process regression model, and evaluating the improved Gaussian process regression model. The present invention can qualitatively and quantitatively analyze the influence of different SOC working intervals and different discharge rates on the aging rate of lithium batteries. By performing coupling processing on the eigenvalues and adding the model estimation value at the previous moment as the input value, this technical means reduces the model dimension, lowers the training difficulty, and significantly improves the accuracy of battery aging state estimation.

[0169] In the present specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple. For the relevant parts, reference can be made to the description in the method section.

[0170] Specific examples are used in this article to elaborate on the principles and implementation manners of the present invention. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present invention. At the same time, for those of ordinary skill in the art, based on the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation on the present invention.

Claims

1. A method for estimating the battery aging state based on improved Gaussian process regression, characterized in that Including: Obtaining the original data set of the lithium battery at different SOC intervals and different discharge current rates through experiments; Analyzing the original data set to determine the model data set; the model data set includes the median value of the SOC interval, the depth of discharge, the discharge current rate, and the equivalent cycle number as input feature values, and includes the lithium battery aging state index as the output value; Dividing the model data set into a training set and a test set; Establishing an improved Gaussian process regression model; the input of the improved Gaussian process regression model is the coupled processed feature values and the battery aging state at the previous moment, and the output is the battery aging state at the current moment; Training and testing the improved Gaussian process regression model using the training set and the test set to generate a trained improved Gaussian process regression model; The training and testing the improved Gaussian process regression model using the training set and the test set to generate a trained improved Gaussian process regression model specifically includes: Multiplying the equivalent cycle number by the corresponding median value of the SOC interval, the depth of discharge, and the discharge current rate respectively to obtain the coupled processed feature values; Using the coupled processed feature values and the battery aging state at the previous moment as the input values of the improved Gaussian process regression model, and using the battery aging state at the current moment as the output value of the improved Gaussian process regression model to train the improved Gaussian process regression model; During the training process, using the test set to test the improved Gaussian process regression model, and using the goodness of fit and the root mean square error as the model evaluation indexes; Determining the improved Gaussian process regression model that meets the preset model evaluation indexes as the trained improved Gaussian process regression model; Using the trained improved Gaussian process regression model to predict the aging state of the lithium battery.

2. The method according to claim 1, wherein The obtaining the original data set of the lithium battery at different SOC intervals and different discharge current rates through experiments specifically includes: Setting multiple SOC intervals and multiple discharge current rates of the lithium battery; Performing multiple cycle charge and discharge experiments on multiple lithium batteries of the same model at different SOC intervals and different discharge current rates, and recording the corresponding SOC interval and discharge current rate of each lithium battery during the multiple cycle charge and discharge experiments; After the multiple cycle charge and discharge experiments, performing one more charge and discharge experiment on the lithium battery at the preset discharge current rate, and recording the discharge capacity during this charge and discharge experiment as the current available capacity of the lithium battery after experiencing multiple charge and discharge cycles; Generating the original data set according to the corresponding SOC interval and discharge current rate of each lithium battery during the multiple cycle charge and discharge experiments and the current available capacity of the lithium battery after experiencing multiple charge and discharge cycles.

3. The method according to claim 2, wherein The analyzing the original data set to determine the model data set specifically includes: Calculating the corresponding median value of the SOC interval according to the different SOC intervals; Calculating the depth of discharge according to the different SOC intervals; Calculate the aging state index of the lithium battery after the n-th multiple charge and discharge cycles according to the current available capacity of the lithium battery after experiencing multiple charge and discharge cycles; Use the median of the SOC interval, the depth of discharge, the corresponding discharge current rate, and the equivalent cycle number as input feature values, and the corresponding aging state index as the output value to form the model data set.

4. The method according to claim 3, wherein The establishment of the improved Gaussian process regression model specifically includes: Construct Gaussian process regression models using different kernel functions; Initialize the hyperparameters to zero, train the Gaussian process regression models using different kernel functions, optimize the hyperparameters by the conjugate gradient method, and use RMSE to evaluate the error of the Gaussian process regression models using different kernel functions; Determine the Gaussian process regression model using the Matern+LIN kernel function with the minimum error as the improved Gaussian process regression model.

5. A battery aging state estimation system based on improved Gaussian process regression, characterized in that, Include: An original data acquisition module for obtaining the original data set of the lithium battery at different SOC intervals and different discharge current rates through experiments; An original data analysis module for analyzing the original data set to determine the model data set; the model data set includes the median of the SOC interval, the depth of discharge, the discharge current rate, and the equivalent cycle number as input feature values, and includes the aging state index of the lithium battery as the output value; A model data set division module for dividing the model data set into a training set and a test set; A model establishment module for establishing an improved Gaussian process regression model; the input of the improved Gaussian process regression model is the coupled processed feature values and the battery aging state at the previous moment, and the output is the battery aging state at the current moment; A model training module for training and testing the improved Gaussian process regression model using the training set and the test set to generate a trained improved Gaussian process regression model; The model training module specifically includes: A feature value coupling unit for multiplying the equivalent cycle number with the corresponding median of the SOC interval, the depth of discharge, and the discharge current rate respectively to obtain the coupled processed feature values; A model training unit for using the coupled processed feature values and the battery aging state at the previous moment as the input values of the improved Gaussian process regression model, and the battery aging state at the current moment as the output value of the improved Gaussian process regression model to train the improved Gaussian process regression model; A model testing unit for testing the improved Gaussian process regression model using the test set during the training process, and using the goodness of fit and the root mean square error as the model evaluation indicators; A model training completion unit for determining the improved Gaussian process regression model that meets the preset model evaluation indicators as the trained improved Gaussian process regression model; A battery aging state prediction module for predicting the aging state of the lithium battery using the trained improved Gaussian process regression model.

6. The system according to claim 5, wherein The original data acquisition module specifically includes: An SOC interval and rate setting unit for setting multiple SOC intervals and multiple discharge current rates of the lithium battery; The cyclic charge and discharge experiment unit is used to perform multiple cyclic charge and discharge experiments on multiple lithium batteries of the same model in different SOC intervals and at different discharge current rates, and record the SOC interval and discharge current rate corresponding to each lithium battery during the multiple cyclic charge and discharge experiments; The current available capacity experiment unit is used to perform one more charge and discharge experiment on the lithium battery at a preset discharge current rate after the multiple cyclic charge and discharge experiments, and record the discharge capacity during this charge and discharge experiment as the current available capacity of the lithium battery after experiencing multiple charge and discharge cycles; The original data set generation unit is used to generate the original data set according to the SOC interval and discharge current rate corresponding to each lithium battery during the multiple cyclic charge and discharge experiments and the current available capacity of the lithium battery after experiencing multiple charge and discharge cycles.

7. The system according to claim 6, wherein The original data analysis module specifically includes: The SOC interval median calculation unit is used to calculate the corresponding SOC interval median according to the different SOC intervals; The discharge depth calculation unit is used to calculate the discharge depth according to the different SOC intervals; The aging state index calculation unit is used to calculate the aging state index of the lithium battery after the nth time of experiencing multiple charge and discharge cycles according to the current available capacity of the lithium battery after experiencing multiple charge and discharge cycles; The model data set generation unit is used to form the model data set by taking the SOC interval median, the discharge depth, the corresponding discharge current rate, and the equivalent cycle number as input feature values and the corresponding aging state index as the output value.

8. The system according to claim 7, wherein The model establishment module specifically includes: The Gaussian process regression model construction unit is used to construct Gaussian process regression models with different kernel functions; The Gaussian process regression model evaluation unit is used to initialize the hyperparameters to zero, train the Gaussian process regression models with different kernel functions, optimize the hyperparameters by the conjugate gradient method, and use RMSE to evaluate the error of the Gaussian process regression models with different kernel functions; The improved Gaussian process regression model construction unit is used to determine the Gaussian process regression model with the Matern+LIN kernel function with the smallest error as the improved Gaussian process regression model.

Images