Lithium ion battery health state prediction method and system based on Bayesian optimization

Abstract

The invention relates to a Bayesian optimization-based lithium ion battery health state prediction method and system, and the method comprises the steps: 1, obtaining the charging and discharging data of a battery, carrying out the preprocessing of the charging and discharging data of the battery, and extracting a key health factor from the preprocessed data; the key health factors comprise constant-current charging time, voltage fluctuation and the change rate of discharge capacity along with time; 2, constructing an initial BiSLTM model based on the key health factors, setting an initial hyper-parameter combination, adjusting hyper-parameters by using Bayesian optimization to obtain an optimal hyper-parameter combination, and constructing a BiLSTM optimal model by using the obtained optimal hyper-parameter combination; and step 3, predicting the health state of the battery by using the BiLSTM optimal model. According to the method, the hyper-parameters of the BiLSTM model are optimized through Bayesian optimization, the problem of BiLSTM hyper-parameter tuning is solved, the globally optimal solution can be efficiently found under the condition that the number of evaluation times is limited, the search duration is greatly shortened, and the generalization efficiency of the model is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of lithium battery health state prediction, and in particular relates to a lithium ion battery health state prediction method and system based on Bayesian optimization. Background Art

[0002] There are two main methods for predicting the health status of lithium batteries: model-based prediction methods and data-driven prediction methods. Model-based methods are mainly based on electrochemical models and equivalent physical models. Electrochemical models provide detailed battery behavior information by simulating the chemical reaction process and physical process inside the battery. The equivalent circuit model is usually a circuit model, which mainly reflects the external current and voltage of the battery and the internal equivalent resistance information, simplifying the complex reaction process of the battery. The data-driven evaluation method uses the battery's historical data and machine learning methods. For example, by discarding the crossover probability and mutation probability fixed by the traditional GA, the method of optimizing the initial weight threshold of the BP neural network with an adaptive probability calculation strategy; a lithium battery SOH (state of health, SOH, health status) prediction method based on the self-attention mechanism and gated recurrent unit (GRU); a lithium battery SOH and RUL prediction model based on a bidirectional long short-term memory network (BiLSTM), etc.

[0003] Among the existing SOH prediction methods for lithium-ion batteries, the prediction methods based on electrochemical models generally have the problems of complex models, large parameter requirements, high computational costs, and difficulty in accurately obtaining model parameters; the prediction methods based on equivalent physical models generally have the problem that the models are difficult to accurately reflect the real changes in the battery and have low accuracy. As for the prediction methods based on data-driven, in the existing technical methods, the method of optimizing the initial weight threshold of the BP neural network by abandoning the crossover probability and mutation probability fixed by the traditional GA and using an adaptive probability calculation strategy effectively improves the accuracy of the prediction, but the addition of the adaptive probability calculation optimization algorithm inevitably increases the complexity of the algorithm; the SOH prediction method for lithium batteries based on the self-attention mechanism and gated recurrent unit (GRU) has limitations in capturing long time series features, and it is difficult to solve the problem of gradient disappearance when facing long sequences, so this method has great requirements for the form of data; the SOH and RUL prediction model for lithium batteries based on the bidirectional long short-term memory network (BiLSTM) has difficulties in hyperparameter tuning, and requires very fine parameter tuning to achieve good performance.

[0004] BiLSTM can effectively process long sequence data and learn long-term dependencies by introducing the gate mechanism of LSTM (input gate, forget gate and output gate). In addition, unlike unidirectional LSTM, BiLSTM can process both forward and reverse information of the sequence at the same time, thereby capturing the features in the sequence more comprehensively. However, since the BiLSTM model contains multiple hyperparameters, such as learning rate, number of hidden layer neurons, number of iterations (epoch), batch size, dropout rate, etc. These hyperparameters influence each other and jointly determine the performance and training effect of the model. Therefore, multiple factors need to be considered comprehensively during the tuning process, which makes BiLSTM very difficult to tune hyperparameters. Summary of the invention

[0005] The purpose of the present invention is to provide a lithium-ion battery health status prediction method and system based on Bayesian optimization, provide a SOH prediction model based on Bayesian optimization BiLSTM (BO-BiLSTM), and use Bayesian optimization to optimize the hyperparameters of the BiLSTM model, thereby solving the problem of BiLSTM hyperparameter tuning. Moreover, the Bayesian optimization algorithm, as a global hyperparameter optimization algorithm, can efficiently find the global optimal solution with a limited number of evaluations compared to other traditional optimization algorithms, thereby greatly reducing the search time and improving the generalization efficiency of the model.

[0006] The present invention provides a method for predicting the health status of a lithium-ion battery based on Bayesian optimization, comprising the following steps:

[0007] Step 1: Obtain battery charge and discharge data, preprocess the battery charge and discharge data, and extract key health factors from the preprocessed data; wherein the battery charge and discharge data include voltage, current, and temperature; the key health factors include constant current charging time, voltage fluctuation, and the rate of change of discharge capacity over time;

[0008] Step 2: construct an initial BiLSTM model based on key health factors, set an initial hyperparameter combination, use Bayesian optimization to adjust the hyperparameters to obtain the optimal hyperparameter combination, and use the obtained optimal hyperparameter combination to construct a BiLSTM optimal model;

[0009] Step 3: Use the BiLSTM optimal model to predict the health status of the battery.

[0010] Furthermore, the step 2 comprises:

[0011] The hyperparameters are modeled through Gaussian process, and the probability improvement acquisition function is used to select the next evaluation point. This process is iterated continuously until the set target value is met or the maximum number of iterations is reached. The optimal hyperparameter combination of the BiLSTM model is found, and the optimal BiLSTM model is constructed using the optimal hyperparameter combination.

[0012] Furthermore, the step 3 comprises:

[0013] The optimal hyperparameter combination obtained by Bayesian optimization is used to train the BiLSTM model and conduct evaluation tests. The SOH is predicted by inputting test data into the BiLSTM optimal model and the predicted results are compared with the actual SOH values ​​to evaluate the accuracy and robustness of the BiLSTM optimal model and verify the effectiveness of the model in practical applications.

[0014] Furthermore, the step 1 comprises the following steps:

[0015] Step 11: Obtain battery charge and discharge data from a battery cycle charge and discharge experiment under laboratory conditions;

[0016] Step 12: Serialize the battery data, sort the battery data in order of time, and calculate the SOH value of each cycle;

[0017] Step 13: Delete the cycle charge and discharge data of the battery data whose charge and discharge capacity exceeds the threshold, and normalize the data;

[0018] Step 14: Extract the time required for the battery voltage to reach 4.2V under 0.5C constant current rate charging from the battery charging and discharging data as the battery health factor HI1;

[0019] HI1=t 4.2V -t 0.5C

[0020] Where, t 4.2V Indicates the time when the battery voltage reaches 4.2V; t 0.5C Indicates the time point when the battery starts charging at a constant current rate of 0.5C;

[0021] Step 15: Extract the time required for the current to gradually drop to the cut-off current of 0.05A from the battery charge and discharge data after the battery reaches the 4.2V constant voltage stage as the battery health factor HI2;

[0022] HI2=t 0.05A -t 4.2V

[0023] Where, t 0.05A Indicates the time point when the current drops to the cut-off current of 0.05A;

[0024] Step 16: Extract the voltage fluctuation data of the battery during constant current discharge from the battery charge and discharge data as the battery health factor HI3;

[0025]

[0026] Where N is the number of sampling points during the constant voltage discharge process in one cycle, V i is the voltage value of each sampling point, is the average value of the voltage during the discharge process;

[0027] Step 17: Extract the discharge rate data of the battery in one cycle from the battery charge and discharge data as the battery health factor HI4;

[0028]

[0029] Where ΔQ dis Indicates the change in discharge capacity within Δt time;

[0030] Step 18: Extract the peak voltage and current data of the battery during constant current discharge from the battery charge and discharge data as battery health factors HI5 and HI6 respectively;

[0031] Step 19: Use the Pearson correlation coefficient as a measure to measure the degree of association between each health factor and SOH;

[0032]

[0033] In the formula, r represents the Pearson correlation coefficient; n is the number of cycles; H i and SOH i Health factors and SOH values ​​for each cycle; and is the average value of health factor and SOH during the entire cycle;

[0034] Step 110: At the input and output ends of the model, the health factor with a correlation coefficient greater than 0.9 is used as the input of the model, the time series step length of the input data is 16, and the output is the predicted SOH for one time step in the future.

[0035] The present invention also provides a lithium-ion battery health status prediction system based on Bayesian optimization, comprising a prediction module, wherein the prediction module executes the lithium-ion battery health status prediction method based on Bayesian optimization.

[0036] The present invention also provides a non-transitory computer-readable storage medium, which stores computer instructions. When the computer instructions are executed by a processor, the method for predicting the health status of a lithium-ion battery based on Bayesian optimization is implemented.

[0037] The present invention also provides an electronic device, comprising:

[0038] 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 method for predicting the health status of a lithium-ion battery based on Bayesian optimization by executing the computer instructions.

[0039] Through the above scheme, through the lithium-ion battery health status prediction method and system based on Bayesian optimization, the hyperparameters of the BiLSTM model are optimized using Bayesian optimization, which solves the problem of BiLSTM hyperparameter tuning. Moreover, the Bayesian optimization algorithm is a global hyperparameter optimization algorithm. Compared with other traditional optimization algorithms, it can efficiently find the global optimal solution with a limited number of evaluations, greatly reducing the search time and improving the generalization efficiency of the model.

[0040] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention and implement it according to the contents of the specification, the following is a detailed description of the preferred embodiments of the present invention in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flow chart of a lithium-ion battery health status prediction method based on Bayesian optimization of the present invention;

[0042] Figure 2 is a capacity decay curve diagram of each battery after data processing of the present invention;

[0043] Figure 3 is the prediction result of the CS2_35 battery in one embodiment of the present invention;

[0044] Figure 4 is the CS2_36 picture prediction result in one embodiment of the present invention;

[0045] Figure 5 is the prediction result of the CS2_37 picture in one embodiment of the present invention;

[0046] Figure 6 is the prediction result of the CS2_38 picture in one embodiment of the present invention;

[0047] Figure 7 It is a schematic diagram of the structure of the electronic equipment of the present invention. DETAILED DESCRIPTION

[0048] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0049] Ginseng Figure 1 As shown, this embodiment provides a lithium-ion battery health status prediction method based on a Bayesian optimized bidirectional long short-term memory network (BiLSTM). The model test data uses the experimental data set of the Center for Advanced Lifecycle Engineering at the University of Maryland. The data set contains four types of battery aging data: CS2_35, CS2_36, CS2_37, and CS2_38. All CS2 batteries have a rated capacity of 1.1Ah and have undergone the same charging curve, i.e., a standard constant current / constant voltage protocol with a constant current rate of 0.5C until the voltage reaches 4.2V, and then continues at 4.2V until the charging current drops below 0.05A. The battery maintains a discharge rate of 1C during the discharge phase until the voltage reaches 2.7V. The capacity decay curves of each battery after data processing are shown in Figure 2. Figure 2 As shown, for this data set, the lifetime failure threshold is reached when the capacity decays to 80% of the rated capacity, which is 0.88Ah.

[0050] The specific steps are as follows:

[0051] Step 1: Obtain battery charging and discharging data, preprocess the battery charging and discharging data, and extract key health factors from the preprocessed data.

[0052] Collect battery charge and discharge data from the experiment, including parameters such as voltage, current, and temperature. Clean and normalize the collected data to remove outliers and missing values ​​to ensure data quality and consistency. Then, extract key health factors from the preprocessed data, such as constant current charging time, voltage fluctuation, and the rate of change of discharge capacity over time. Finally, divide the data set into training and test sets to facilitate model training and evaluation.

[0053] Step 2: Build the initial BiLSTM model based on the key health factors, set the initial hyperparameter combination, use Bayesian optimization to adjust the hyperparameters to obtain the optimal hyperparameter combination, and use the obtained optimal hyperparameter combination to build the BiLSTM optimal model. Bayesian optimization models the hyperparameters through the Gaussian process, and uses the probability improvement (PI) acquisition function to select the next evaluation point, and iterates this process until the set target value is met or the maximum number of iterations is reached, thereby finding the optimal hyperparameter combination of the BiLSTM model and using the optimal hyperparameters to build the BiLSTM optimal model.

[0054] Step 3: Use the optimal hyperparameter combination obtained by Bayesian optimization to train the BiLSTM model and conduct evaluation tests. By inputting the test data into the optimized BiLSTM model, SOH prediction is performed, and the prediction results are compared with the actual SOH value to evaluate the accuracy and robustness of the model and verify the effectiveness of the model in practical applications. Finally, the optimized BiLSTM model is used to accurately predict the SOH of the battery.

[0055] The step 1 comprises the following steps:

[0056] Step 11: Obtain battery charge and discharge data from the lithium battery cycle charge and discharge experiment under laboratory conditions, mainly parameters such as voltage, current, and temperature.

[0057] Step 12: Serialize the battery data, sort the battery data in chronological order, and calculate the SOH value of each cycle.

[0058] Step 13: Delete the cycle charge and discharge data in the battery data whose charge and discharge capacity exceeds the threshold, and normalize the data.

[0059] Step 14: Extract the time required for the battery voltage to reach 4.2V under 0.5C constant current rate charging from the battery charge and discharge data as the battery health factor HI1.

[0060] HI1=t 4.2V -t 0.5C

[0061] Where, t 4.2V Indicates the time when the battery voltage reaches 4.2V; t 0.5C Indicates the time point when the battery starts charging at a constant current rate of 0.5C.

[0062] Step 15: Extract from the battery charge and discharge data the time required for the current to gradually drop to a cut-off current of 0.05A after the battery reaches the 4.2V constant voltage stage as the battery health factor HI2.

[0063] HI2=t 0.05A -t 4.2V

[0064] Where, t 0.05A Indicates the time point when the current drops to the cut-off current of 0.05A.

[0065] Step 16: Extract the voltage fluctuation data of the battery during constant current discharge from the battery charge and discharge data as the battery health factor HI3.

[0066]

[0067] Where N is the number of sampling points during the constant voltage discharge process in one cycle, V i is the voltage value of each sampling point, is the average voltage during the discharge process.

[0068] Step 17: Extract the discharge rate data of the battery in one cycle from the battery charge and discharge data as the battery health factor HI4.

[0069]

[0070] Where ΔQ dis It represents the change of discharge capacity within Δt time.

[0071] Step 18: Extract the peak voltage and current data of the battery during constant current discharge from the battery charge and discharge data as battery health factors HI5 and HI6 respectively.

[0072] Step 19: Use the Pearson correlation coefficient as a measure to measure the degree of association between each health factor and SOH.

[0073]

[0074] In the formula, r represents the Pearson correlation coefficient; n is the number of cycles; H i and SOH i Health factors and SOH values ​​for each cycle; and is the average value of the health factor and SOH during the entire cycle. Taking the battery data from the University of Maryland as an example, the calculation results of the Pearson correlation coefficient are shown in Table 1:

[0075] Table 1 Pearson correlation coefficient calculation results

[0076]

[0077] The calculation results of each health factor on each battery data set are shown in the table. The value range of each calculation result is between -1 and 1. The closer the result is to 1, the higher the positive correlation is. The closer it is to -1, the higher the negative correlation is. The closer it is to 0, the lower the correlation is. The health factor with an average coefficient greater than 0.9 is selected as the model input.

[0078] Step 110: At the input and output of the model, the health factor with a correlation coefficient greater than 0.9 is used as the input of the model, the time series step of the input data is 16, and the output predicts the SOH of the next time step. The first 35% of the cycle data of each data set is used as the training set, and the remaining cycle data is used as the test set.

[0079] On each battery aging data, SOH prediction is performed using four models: BO-BiLSTM, BiLSTM, BO-LSTM, and BO-GRU. The parameter settings of each model are shown in Table 2. The prediction comparison results are shown in Figures 3 to 6 shown.

[0080] Table 2 Parameter settings

[0081]

[0082] The results show that the BO-BiLSTM model has excellent prediction capabilities on the four battery aging data sets, while the other models have obvious deficiencies in prediction capabilities on CS2_35 and CS2_36 batteries.

[0083] Table 3 Error indicators of each model

[0084]

[0085] Table 3 shows the error index results of each model for the SOH prediction of four batteries. The BO-BiLSTM model performs well in the SOH prediction of all batteries, with the lowest RMSE and MAE values ​​and the highest R2 value, indicating that it is superior to other models in terms of overall error, absolute error and data variation interpretation ability. The BiLSTM model also performs well in some cases, but is slightly inferior to BO-BiLSTM overall. In contrast, the performance of the BO-LSTM and BO-GRU models is relatively weak, especially on the CS2_36 and CS2_37 datasets, indicating that their prediction capabilities when processing these data are not as good as those of BO-BiLSTM and BiLSTM.

[0086] The SOH prediction model based on Bayesian optimization bidirectional long short-term memory network (BiLSTM) first preprocesses the battery charging and discharging data to extract health factors (HI) such as constant current charging time, voltage fluctuation and discharge capacity change rate; then, by calculating the Pearson correlation coefficient, the HI with high correlation with SOH is extracted from each HI as the model input; finally, the hyperparameters of the BiLSTM model are optimized using Bayesian optimization, and the SOH prediction model based on Bayesian optimization BiLSTM is established. This prediction model uses Bayesian optimization to optimize the hyperparameters of the BiLSTM model, solving the problem of BiLSTM hyperparameter tuning. Moreover, as a global hyperparameter optimization algorithm, the Bayesian optimization algorithm can efficiently find the global optimal solution with a limited number of evaluations compared to other traditional optimization algorithms, greatly reducing the search time and improving the generalization efficiency of the model.

[0087] This embodiment also provides a lithium-ion battery health state prediction system based on Bayesian optimization, including a prediction module, and the prediction module executes the lithium-ion battery health state prediction method based on Bayesian optimization.

[0088] This embodiment also provides a non-transitory computer-readable storage medium, which stores computer instructions. When the computer instructions are executed by a processor, the method for predicting the health status of a lithium-ion battery based on Bayesian optimization is implemented.

[0089] Ginseng Figure 7 As shown, this embodiment also provides an electronic device, including:

[0090] A memory 201 and a processor 202, wherein the memory 201 and the processor 202 are communicatively connected to each other, the memory 201 stores computer instructions, and the processor 202 executes the method for predicting the health status of a lithium-ion battery based on Bayesian optimization by executing the computer instructions.

[0091] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the technical principles of the present invention, and these improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for predicting the health status of lithium-ion batteries based on Bayesian optimization, characterized in that: The steps include: Step 1: Obtain battery charge and discharge data, preprocess the battery charge and discharge data, and extract key health factors from the preprocessed data; wherein the battery charge and discharge data include voltage, current, and temperature; the key health factors include constant current charging time, voltage fluctuation, and the rate of change of discharge capacity over time; Step 2: construct an initial BiLSTM model based on key health factors, set an initial hyperparameter combination, use Bayesian optimization to adjust the hyperparameters to obtain the optimal hyperparameter combination, and use the obtained optimal hyperparameter combination to construct a BiLSTM optimal model; Step 3: Use the BiLSTM optimal model to predict the health status of the battery.

2. The method for predicting the health status of a lithium-ion battery based on Bayesian optimization according to claim 1, characterized in that: The step 2 comprises: The hyperparameters are modeled through Gaussian process, and the probability improvement acquisition function is used to select the next evaluation point. This process is iterated continuously until the set target value is met or the maximum number of iterations is reached. The optimal hyperparameter combination of the BiLSTM model is found, and the optimal BiLSTM model is constructed using the optimal hyperparameter combination.

3. The method for predicting the health status of a lithium-ion battery based on Bayesian optimization according to claim 2, characterized in that: The step 3 comprises: The optimal hyperparameter combination obtained by Bayesian optimization is used to train the BiLSTM model and conduct evaluation tests. The SOH is predicted by inputting test data into the BiLSTM optimal model and the predicted results are compared with the actual SOH values ​​to evaluate the accuracy and robustness of the BiLSTM optimal model and verify the effectiveness of the model in practical applications.

4. The method for predicting the health status of a lithium-ion battery based on Bayesian optimization according to claim 1, characterized in that: The step 1 comprises the following steps: Step 11: Obtain battery charge and discharge data from a battery cycle charge and discharge experiment under laboratory conditions; Step 12: Serialize the battery data, sort the battery data in order of time, and calculate the SOH value of each cycle; Step 13: Delete the cycle charge and discharge data whose charge and discharge capacity exceeds the threshold in the battery data, and normalize the data; Step 14: Extract the time required for the battery voltage to reach 4.2V under 0.5C constant current rate charging from the battery charging and discharging data as the battery health factor HI1; HI1=t 4.2V -t 0.5C Where, t 4.2V Indicates the time when the battery voltage reaches 4.2V; t 0.5C Indicates the time point when the battery starts charging at a constant current rate of 0.5C; Step 15: Extract the time required for the current to gradually drop to the cut-off current of 0.05A from the battery charge and discharge data after the battery reaches the 4.2V constant voltage stage as the battery health factor HI2; HI2=t 0.05A -t 4.2V Where, t 0.05A Indicates the time point when the current drops to the cut-off current of 0.05A; Step 16: Extract the voltage fluctuation data of the battery during constant current discharge from the battery charge and discharge data as the battery health factor HI3; Where N is the number of sampling points during the constant voltage discharge process in one cycle, V i is the voltage value of each sampling point, is the average voltage during the discharge process; Step 17: Extract the discharge rate data of the battery in one cycle from the battery charge and discharge data as the battery health factor HI4; Where ΔQ dis Indicates the change in discharge capacity within Δt time; Step 18: Extract the peak voltage and current data of the battery during constant current discharge from the battery charge and discharge data as battery health factors HI5 and HI6 respectively; Step 19: Use the Pearson correlation coefficient as a measure to measure the degree of association between each health factor and SOH; In the formula, r represents the Pearson correlation coefficient; n is the number of cycles; H i and SOH i Health factors and SOH values ​​for each cycle; and is the average value of health factor and SOH during the entire cycle; Step 110: At the input and output ends of the model, the health factor with a correlation coefficient greater than 0.9 is used as the input of the model, the time series step length of the input data is 16, and the output is the predicted SOH for one time step in the future.

5. A lithium-ion battery health status prediction system based on Bayesian optimization, characterized in that: It includes a prediction module, which executes the lithium-ion battery health status prediction method based on Bayesian optimization as described in any one of claims 1 to 4.

6. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, a lithium-ion battery health status prediction method based on Bayesian optimization as described in any one of claims 1 to 4 is implemented.

7. An electronic device, 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 method for predicting the health status of a lithium-ion battery based on Bayesian optimization as described in any one of claims 1 to 4 by executing the computer instructions.

Images