An Efficient Li-ion Battery Management System with Lossless Charge Balancer for RUL and SoH Prediction

Electric vehicles (EV) employ batteries to generate their mechanical power for transportation, but the main challenge is to improve the battery management system (BMS) and increase the lifespan of the EV battery. In the existing battery management system, energy loss during charge balancing operation and prediction errors happens in remaining useful life (RUL) and state of health (SoH). Hence a novel Efficient Li-ion Battery Management System with Lossless Charge Balancer for RUL and SoH Prediction is proposed to improve the Battery Management System (BMS) and lifespan of the EV battery. The existing battery management systems have various cell-balancing approaches, but the energy losses in the form of heat create unavoidable instant charge imbalance. Thus, a novel Optimized Multi Input Multi Output-Bi Directional Long Short-Term Memory (MIMO-Bi-LSTM) has been proposed, in which the MIMO-Bi-LSTM Unit is providing better SoC estimation of each cell, and the FFOA (Fruit Fly Optimization Algorithm) is utilized in this state of charge (SoC) estimation of battery and improved accuracy. Moreover, an Adaptive Matrix Gate Switch Balancer is introduced in which the Adaptive Matrix Switch Algorithm is used to avoid charge imbalance and the DGTO (Duplex Gate Turn-Off Thyristors) switches reduce the energy loss during switching and improving the cell life cycle. Furthermore, the existing technique did not consider the variation of the EV motor's efficiency that changes throughout the operation and the motor terminal resistance which also affects the cycle life of the battery. So, the novel Optimized UK-ANFI Network is introduced in which a UK (Unscented Kalman) Filter eliminate the non-linearity in the measured values of parameters and the ANFI (Adaptive Neuro-Fuzzy Inference) Network receives the linearized data and predicts the RUL and SoH of the battery pack. Then a GWO (Grey Wolf Optimizer) minimize prediction errors and provide better life cycle prediction. The result obtained by the proposed model have low RMSE in RUL and SoH prediction, high accuracy and low prediction time.