Battery modeling using grey box and black box data driven techniques

The ability to predict battery lifespan accurately is very crucial in many applications to enhance performance and reliability. This paper models the battery using grey and black box approaches to predict Lithium-ion battery charging and discharging scenarios. This study presents a comparative battery modeling framework for terminal voltage prediction using Grey Wolf Optimization-based grey-box modeling and Random Forest-based black-box modeling. Paper also investigates correlation analysis study on battery parameters for improved prediction accuracy. The implementation of the proposed work uses data from NASA Prognostics Center of Excellence Data Set Repository. The proposed approaches are evaluated using experimental battery data to assess their capability in accurately representing battery voltage behavior under different operating conditions. This study assesses the developed battery models against manufacturer data and compares with Mean Absolute Percentage Error (MAPE) metric. The black box model records lower MAPE for maximum number of cycles under any charging/discharging rate. Reliable battery model increases accurate estimation of battery behavior. Therefore, in this paper, correlation analysis study attempts to identify the appropriate features that have a significant impact on reliable prediction of charging characteristics and discharging characteristics of the battery. It reveals that temperature of the battery needs to be considered for accurate battery model prediction. Hence, this paper also provides battery modeling (BM) considering the temperature effect with black box technique.