.png){kind=icon}
.png){kind=icon}
Return
State of Health Estimation Method for Pure Electric Vehicle Power Batteries Based on Grid Search Cross-Validation-Extreme Gradient Boosting
.png){kind=icon}
OA
AI
S
Y
D
G
X
J
Z
DOI：10.1002/ese3.70334.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Accurately estimating the state of health (SOH) of power batteries is beneficial for their maintenance, delaying aging, ensuring safety, and providing a basis for their secondary use to enhance resource utilization efficiency. However, existing data-driven methods rely heavily on laboratory data and lack adequate adaptability to real-world vehicle conditions. Moreover, traditional gradient boosting algorithms such as gradient boosting decision trees (GBDT) and LogitBoost encounter precision and generalization issues when faced with the complex operating conditions of real vehicles, thereby limiting their practical applications. To address these challenges, this paper proposes a method for estimating the SOH of power batteries in pure electric vehicles using an extreme gradient boosting (XGBoost) model optimized by the grid search cross-validation (GSCV) method, based on data from a vehicle manufacturer's monitoring platform. First, data are divided according to a “discharge + charge” pattern, and 16 capacity degradation feature factors from six categories are extracted from the discharge-charge segments as input variables for the XGBoost model, while partial charged capacity is extracted from the charge segments as the output label for the model. Subsequently, to overcome the XGBoost model's sensitivity to hyperparameters and its susceptibility to overfitting, the GSCV method is employed for parameter optimization of the XGBoost model, and the GSCV-XGBoost model is used to estimate partial charged capacity. Finally, an SOH correction method is applied to the output of the GSCV-XGBoost model to obtain the corrected SOH. Experimental results demonstrate that the SOH estimated by the GSCV-XGBoost model combined with the SOH correction method exhibits smaller errors and remains consistently below 2% compared to SOH corrected based on the Ampere-hour integral method. In estimating partial charged capacity, the GSCV-XGBoost model significantly outperforms the XGBoost model. Compared to the CBDT and linear regression (LR) models, the GSCV-XGBoost model achieves the highest goodness of fit (R²), with the smallest mean absolute error (MAE) and root mean squared error (RMSE). The research findings presented in this paper are expected to provide effective solutions for real-world vehicle power battery SOH monitoring.
Keywords:
power batteries
pure electric vehicles
SOH|GSCV-XGBoost
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
3.4
Papers:
2.3K
Citations:
5.7K
Organization
Cited Papers