A hybrid BiLSTM-MPC predictive control strategy for energy and thermal management of electric vehicles with real-time validation

Energy management efficiency in battery electric vehicles (xEVs) is vital for enhancing their driving performance, battery safety, and energy management efficiency. Traditional controllers, such as PID, Fuzzy logic, ANN, and traditional Model Predictive Control (MPC), have provided an acceptable performance level. However, these methods do not provide enough capacity for tackling drive cycle dynamics, future uncertainty, and heat effects, resulting in higher tracking errors and lower SoC preservation. For solving this problem, a predictive intelligent control scheme combining BiLSTM-based future speed prediction along with constrained MPC optimization of control input was proposed for energy management in xEVs. Using the BiLSTM model, future speed trajectory over the prediction horizon is predicted, whereas the MPC method minimizes tracking error, energy expenditure, deviation of the battery temperature from its optimal level, and the control effort. The proposed method was evaluated under various drive cycles, such as UDDS, WLTP, HWFET, and MIDC, with simulations performed on MATLAB/Simulink, as well as on OPAL-RT OP4512 hardware-in-the-loop validation platform. As a result, the proposed controller provided the lowest RMSE value of 1.48 km/h, highest final SoC of 85.9%, regenerative energy recovery of 210 Wh, and maximum battery temperature of 42.1 °C.