基于改进Smote策略的动力电池故障检测OA

Fault detection in automotive power batteries represents a critical technology for preventing vehicle safety incidents.However,current research faces two major challenges:imbalanced distribution of fault data and sample scarcity,which leads to insufficient prediction accuracy and limited model generalization capability.To tackle these issues,a Stacking-integrated diagnostic framework that merges data augmentation with feature reconstruction techniques is introduced.Firstly,the fault level classification system is reconstructed based on Fault Tree Analysis(FTA),an improved Border-Smote algorithm is employed to achieve sample augmentation.Secondly,time,working condition and driving-related features are extracted through feature engineering to construct a multi-dimensional feature space.Finally,based on Bayesian hyperparameter optimization,aStacking ensemble model comprising LightGBM,XGBoost and random forest(RF)as base-level classifiers with logistic regression(LR)as the meta-classifier is constructed.The experimental validation shows that when combined with the enhanced Smote data augmentation and feature reconstruction techniques,the proposed ensemble approach attains a remarkable detection accuracy of over 97％.