A lithium ion battery state of health estimation method based on bayesian optimization ensemble model

Abstract

The application discloses a lithium ion battery state of health estimation method based on a Bayesian optimization integrated model, and is applied to the technical field of battery state of health estimation. The method comprises the following steps: constructing an integrated SOH prediction model, taking XGBoost and LightGBM as base learners, and selecting a weighted average as an integrated strategy; initializing Bayesian optimization parameters, optimizing hyperparameters of XGBoost and LightGBM by using a Bayesian optimization algorithm, searching for optimal weighting coefficients based on a Bayesian optimization strategy; fusing according to optimal weight proportions to obtain an integrated SOH prediction model; and collecting charging and discharging data of a lithium ion battery to input the integrated SOH prediction model and obtain an SOH prediction value. The application optimizes hyperparameters and integrated weights by using a Bayesian algorithm, and excellent prediction performance is achieved.

Description

Technical Field

[0001] This invention relates to the field of battery health state estimation technology, and more specifically to a method for estimating the health state of lithium-ion batteries based on a Bayesian optimization ensemble model. Background Technology

[0002] Accurate assessment of the State of Health (SOH) of lithium-ion batteries is crucial for the safe and reliable operation of energy storage systems and electric vehicles. With the continuous development of intelligent algorithms, data-driven SOH estimation methods have attracted widespread attention. Data-driven methods predict SOH by mining health feature information from historical battery operating data. Existing technologies mainly involve manually extracting key features such as charge-discharge curves and capacity as model inputs. Although manual feature extraction methods have achieved some success in existing research, the process is highly subjective, potentially leading to information loss and difficulty in comprehensively capturing useful information from the data. Therefore, providing a lithium-ion battery state of health estimation method based on a Bayesian optimization ensemble model is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0003] In view of this, the present invention provides a method for estimating the state of health (SOH) of lithium-ion batteries based on a Bayesian optimized ensemble model. The method optimizes the hyperparameters of the model based on Bayesian optimization and enhances robustness and generalization ability by utilizing the diversity among multiple models, thereby improving the overall SOH estimation performance of the battery.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A method for estimating the state of health of lithium-ion batteries based on a Bayesian optimization ensemble model includes the following steps: S1. Construct an ensemble SOH prediction model, using XGBoost and LightGBM as base learners and weighted average as the ensemble strategy. S2. Initialize Bayesian optimization parameters, define the range and dimensions of the hyperparameter search space, and set the number of iterations and the upper and lower bounds of the hyperparameters; S3. Use the Bayesian optimization algorithm to perform a global search optimization of the hyperparameters of XGBoost and LightGBM, record the optimal parameter combination and complete the training; S4. Search for the optimal weighting coefficients of XGBoost and LightGBM in the integrated SOH prediction model based on the Bayesian optimization strategy; S5. The optimized XGBoost and LightGBM are fused together according to the optimal weight ratio to form a Bayesian optimization-driven integrated SOH prediction model. S6. Collect the charging and discharging data of the lithium-ion battery and preprocess it. Input the preprocessed data into the trained integrated SOH prediction model to obtain the SOH prediction value.

[0005] Optionally, the objective function for Bayesian optimization is: ; In the formula, The target is the optimal hyperparameter. For hyperparameter vectors, For the search space of hyperparameters; Using Gaussian process modeling, calculate the prior and posterior distributions: ; ; ; In the formula, Describe the objective function. Indicates Gaussian process modeling, and These are the prior and posterior mean functions, respectively. For observation models, To observe the noise, For the current observation dataset, and These are the prior and posterior covariance functions, respectively; Based on the posterior probability distribution, a sampling function is used to determine the next sampling point, exploring the optimal solution while minimizing the number of experiments: ; ; In the formula, This is the smallest target value observed so far. This represents the desired improvement value.

[0006] Optionally, five-fold cross-validation is used as the performance evaluation method for Bayesian optimization. The training set is randomly divided into five mutually exclusive subsets. Each subset is used as the validation set, and the remaining subsets are used as the training set. The model is trained, and the mean squared error (MSE) on the validation set is calculated. The average of the five-fold cross-validation is calculated and the negative value is taken to obtain the result. The calculation formula is as follows: ; ; In the formula, For the hyperparameter score, For the first k The validation set of the fold. For the first k The mean square error of the fold. For the firstk Zhe Di i The true value of each sample; For the first k Zhe Di i The predicted value for each sample.

[0007] Optionally, the prediction result of the integrated SOH prediction model is a weighted combination of the predictions from the two learners, calculated using the following formula: ; In the formula, The prediction results of the XGBoost model. The prediction results of the LightGBM model, For XGBoost weights, For the weights of LightGBM, , The SOH prediction results are for the integrated SOH prediction model.

[0008] Optionally, after collecting the charge and discharge data of lithium-ion batteries in S6, battery aging characteristics are extracted from the data, and the battery aging characteristics with the strongest correlation to SOH are selected by Pearson coefficient as input to the SOH prediction model.

[0009] Optionally, S6 and later also include: S7. Compare the predicted SOH values ​​with the actual measured values, and evaluate the performance of the SOH prediction model using mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination.

[0010] As can be seen from the above technical solution, compared with the prior art, the present invention provides a method for estimating the health status of lithium-ion batteries based on a Bayesian optimization ensemble model, which has the following beneficial effects: 1. This invention optimizes key parameters such as learning rate, tree depth, and number of trees of XGBoost and LightGBM using the Bayesian algorithm, and further optimizes the ensemble weights of the two. The resulting weighted ensemble model achieves excellent prediction performance. Compared with the ensemble method of single learner or traditional grid search optimization, the joint tuning strategy using Bayesian optimization performs better in terms of accuracy and generalization ability, and is more efficient under the condition of limited samples. 2. This invention extracts health features from nonlinear lithium-ion battery charge and discharge data. The constructed SOH estimation model has high accuracy and good robustness in health state estimation. Bayesian optimization improves the efficiency of hyperparameter search and avoids local optima. The automatically optimized ensemble weights reasonably combine the complementary advantages of the two types of models, thereby improving the generalization and robustness of the model. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0012] Figure 1 This is a flowchart of the lithium-ion battery health state estimation method based on a Bayesian optimization ensemble model according to the present invention. Figure 2 This is a schematic diagram of the XGBoost tree structure growth process of the present invention; Figure 3 This is a flowchart of the LightGBM histogram construction process of the present invention; Figure 4 This is a schematic diagram of the tree structure of LightGBM based on leaf growth according to the present invention; Figure 5 This is a flowchart illustrating the application of the lithium-ion battery health status estimation method of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] This invention discloses a method for estimating the state of health of lithium-ion batteries based on a Bayesian optimization ensemble model, such as... Figure 1 As shown, it includes the following steps: S1. Construct an ensemble SOH prediction model, using XGBoost and LightGBM as base learners and weighted average as the ensemble strategy. S2. Initialize Bayesian optimization parameters, define the range and dimensions of the hyperparameter search space (such as learning rate, maximum depth, number of trees, etc.), and set the number of iterations and the upper and lower bounds of the hyperparameters. S3. Use the Bayesian optimization algorithm to perform a global search optimization of the hyperparameters of XGBoost and LightGBM, record the optimal parameter combination and complete the training; S4. Search for the optimal weighting coefficients of XGBoost and LightGBM in the integrated SOH prediction model based on the Bayesian optimization strategy; S5. The optimized XGBoost and LightGBM are fused together according to the optimal weight ratio to form a Bayesian optimization-driven integrated SOH prediction model. S6. Collect the charging and discharging data of the lithium-ion battery and preprocess it. Input the preprocessed data into the trained integrated SOH prediction model to obtain the SOH prediction value.

[0015] In this embodiment, the tree structure growth process of XGBoost is as follows: Figure 2 As shown, this algorithm employs a layer-by-layer growth strategy, which allows for the parallel partitioning of multiple leaf nodes within the same depth layer, thereby improving training efficiency and supporting multi-threaded optimization. Simultaneously, this layer-by-layer expansion approach can, to some extent, control model complexity and reduce the risk of overfitting.

[0016] The LightGBM build process is as follows: Figure 3 As shown, the growth process is as follows Figure 4 As shown, by introducing a histogram algorithm for feature discretization and statistical optimization, the core idea is to divide continuous feature values ​​(float numbers) into a finite number of discrete intervals (bins) and construct histograms on these intervals to statistically analyze feature distribution. Based on histogram aggregation statistics, the model can quickly determine the optimal split point with low computational cost, thus significantly improving training efficiency. Furthermore, LightGBM employs a leaf-by-leaf growth strategy, which, compared to traditional layer-by-layer growth, accelerates the decision tree construction process while maintaining model accuracy.

[0017] Furthermore, the objective function for Bayesian optimization is: ; In the formula, The target is the optimal hyperparameter. For hyperparameter vectors, The search space for hyperparameters; in this embodiment, the learning rate... Maximum tree depth Number of trees ; Since the objective function is unknown, a Gaussian process (GP) is used for modeling to calculate the prior and posterior distributions: ; ; ; In the formula, Describe the objective function. Indicates Gaussian process modeling, and These are the prior and posterior mean functions, respectively. For observation models, To observe the noise, For the current observation dataset, and These are the prior and posterior covariance functions, respectively; Based on the posterior probability distribution, a sampling function is used to determine the next sampling point, exploring the optimal solution while minimizing the number of experiments: ; ; In the formula, This is the smallest target value observed so far. This represents the desired improvement value.

[0018] Furthermore, the core of Bayesian optimization is to define an objective function that needs to be minimized. To evaluate the merits of different hyperparameter combinations, five-fold cross-validation is used as a performance evaluation method for Bayesian optimization. The training set is randomly divided into five mutually exclusive subsets, each of which is used as the validation set, and the remaining subsets are used as the training set. The model is trained, and the mean squared error (MSE) on the validation set is calculated. The average of the five-fold cross-validation is calculated and the negative value is taken to obtain the result. The calculation formula is as follows: ; ; In the formula, For the hyperparameter score, For the first k The validation set of the fold. For the first k The mean square error of the fold. For the first k Zhe Di i The true value of each sample; For the first k Zhe Di i The predicted value for each sample.

[0019] Furthermore, after optimizing the hyperparameters of the two base learners, the optimal weighting coefficients in the ensemble model are searched using a Bayesian optimization strategy to minimize the overall prediction error. The prediction result of the ensemble SOH prediction model is a weighted combination of the predictions from the two learners, calculated using the following formula: ; In the formula, The prediction results are from the XGBoost model. The prediction results of the LightGBM model, For XGBoost weights, For the weights of LightGBM, , The SOH prediction results are for the integrated SOH prediction model.

[0020] Furthermore, such as Figure 5 As shown, after collecting the charge and discharge data of lithium-ion batteries in S6, battery aging characteristics are extracted from the data. The battery aging characteristics with the strongest correlation to SOH are selected by Pearson coefficient and used as input for the SOH prediction model.

[0021] In this embodiment of the invention, multi-dimensional dynamic data such as voltage, current, and capacity are collected during the charging and discharging process of lithium-ion batteries. The raw data is preprocessed, including removing outliers and noise, and the data scale is standardized. Based on this, eight battery aging characteristics, such as mean voltage, standard deviation, and peak voltage, are extracted from the battery data. The mean voltage of the individual battery cells with the strongest correlation to State of Health (SOH) is selected as the model input using the Pearson coefficient. Then, the hyperparameters of the XGBoost and LightGBM models are automatically optimized using Bayesian optimization to improve the estimation performance of each model. Furthermore, Bayesian optimization is used to determine the optimal weight coefficients for the two models during the integration process. Finally, based on this integrated model, an accurate estimate of the battery's State of Health (SOH) is achieved.

[0022] Furthermore, after S6, it also includes: S7. Compare the predicted SOH values ​​with the actual measured values, and evaluate the performance of the SOH prediction model using mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination.

[0023] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0024] Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for estimating the state of health of lithium-ion batteries based on a Bayesian optimization ensemble model, characterized in that, Includes the following steps: S1. Construct an ensemble SOH prediction model, using XGBoost and LightGBM as base learners and weighted average as the ensemble strategy. S2. Initialize Bayesian optimization parameters, define the range and dimensions of the hyperparameter search space, and set the number of iterations and the upper and lower bounds of the hyperparameters; S3. Use the Bayesian optimization algorithm to perform a global search optimization of the hyperparameters of XGBoost and LightGBM, record the optimal parameter combination and complete the training; S4. Search for the optimal weighting coefficients of XGBoost and LightGBM in the integrated SOH prediction model based on the Bayesian optimization strategy; S5. The optimized XGBoost and LightGBM are fused together according to the optimal weight ratio to form a Bayesian optimization-driven integrated SOH prediction model. S6. Collect the charging and discharging data of the lithium-ion battery and preprocess it. Input the preprocessed data into the trained integrated SOH prediction model to obtain the SOH prediction value. The objective function of Bayesian optimization is: ； In the formula, The target is the optimal hyperparameter. For hyperparameter vectors, For the search space of hyperparameters; Using Gaussian process modeling, calculate the prior and posterior distributions: ； ； ； In the formula, Describe the objective function. Indicates Gaussian process modeling, and These are the prior and posterior mean functions, respectively. For observation models, To observe the noise, For the current observation dataset, and These are the prior and posterior covariance functions, respectively; Based on the posterior probability distribution, a sampling function is used to determine the next sampling point, exploring the optimal solution while minimizing the number of experiments: ； ； In the formula, This is the smallest target value observed so far. This represents the desired improvement value.

2. The method for estimating the health status of a lithium-ion battery based on a Bayesian optimization ensemble model according to claim 1, characterized in that, Five-fold cross-validation is used as the performance evaluation method for Bayesian optimization. The training set is randomly divided into five mutually exclusive subsets. Each subset is used as the validation set, and the remaining subsets are used as the training set. The model is trained, and the mean squared error (MSE) on the validation set is calculated. The average of the five-fold cross-validation is calculated and the negative value is taken to obtain the result. The calculation formula is as follows: ； ； In the formula, The score for hyperparameters, For the first k The validation set of the fold. For the first k The mean square error of the fold. For the first k Zhe Di i The true value of each sample; For the first k Zhe Di i The predicted value for each sample.

3. The method for estimating the health status of a lithium-ion battery based on a Bayesian optimization ensemble model according to claim 2, characterized in that, The prediction result of the integrated SOH prediction model is a weighted combination of the predictions from the two learners, calculated using the following formula: ； In the formula, The prediction results are from the XGBoost model. The prediction results of the LightGBM model, For XGBoost weights, For the weights of LightGBM, , The SOH prediction results are for the integrated SOH prediction model.

4. The method for estimating the health status of a lithium-ion battery based on a Bayesian optimization ensemble model according to claim 3, characterized in that, After collecting charge and discharge data of lithium-ion batteries in S6, battery aging characteristics are extracted from the data. The battery aging characteristics with the strongest correlation to SOH are selected by Pearson coefficient and used as input for the SOH prediction model.

5. The method for estimating the health status of a lithium-ion battery based on a Bayesian optimization ensemble model according to claim 4, characterized in that, S6 and later include: S7. Compare the predicted SOH values ​​with the actual measured values, and evaluate the performance of the SOH prediction model using mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination.

Images