ESTIMATION OF ELECTRIC VEHICLE MULTIPLE PACKAGE LITHIUM-ION BATTERY STATE OF CHARGE USING AN EXTENDED MACHINE LEARNING

The estimation of battery state-of-charge (SOC) in electric or hybrid vehicle has vital importance in the designing process of battery management systems. The state-of-charge estimation is implemented using different modelling approaches, model-based estimators such as Kalman filtering and Luenberger observer and data-driven based modelling techniques like artificial neural network and machine learning methods. This study aimed to develop a battery state-of-charge estimation method and proposed a novel architecture for multiple battery back SOC estimation using an extended learning machine (ELM). The ELM approach is applied considering battery operating conditions using global vehicle driving profiles, New European Driving Cycle and Worldwide harmonized Light vehicles Test Procedure. The performance of the proposed SOC estimation method is evaluated by using statistical criteria (RMSE, R2, MAPE). Consequently, the obtained results show that a data-driven ELM approach with a less complex structure can obtain better performance compared with other advanced estimator methods under different operating conditions.