A Multi-Stage Anomalies Recognition and Revision Method for Operation Data of Solid-State Batteries

Accurate measurement of operation parameters such as voltage of solid-state batteries (SSBs) has great potential, since they have shown lots of promise on reliable operation for electric vehicles (EVs). However, abnormalities pervade the operation data due to the faults in sensors or transmission errors, which are difficult to be recognized and revised due to the short-term fluctuation of operation data and the uncertainty of abnormalities. To meet these challenges, an anomalies recognition and revision method for improving the quality of SSBs data for EVs is proposed. First, an improved two-stage ensemble local outlier factor (TELOF) anomalies recognition method is proposed to alleviate the problem of the misrecognition for normal operational data during data cleaning process owing to the large short-term fluctuations. Next, XGBoost-assisted multinomial timeseries (MTXGBoost) is developed to improve the precision and generalization ability of revised results. The real experimental data from two types of SSBs under different conditions is utilized to validate the effectiveness of the proposed method. Comparative results show that the proposed method can accurately recognize anomalies and stably revise results. Both f1score and R² are above 0.95. Thus, the proposed approach can effectively pre-process data to improve decision making and the management ability for SSBs.