Design of a deep inference framework for required power forecasting and predictive control on a hybrid electric mining truck

Abstract

Accurate required power forecasting is critical to ensure the recharge mileage and to optimize the energy utilization of heavy-duty vehicles. This paper proposed an artificial intelligence model predictive control framework for the energy management system (EMS) of the series hybrid electric vehicle with terrain and load information. It aims to achieve optimal energy distribution with increased required power prediction accuracy and optimized SoC sequence by fast analyzing high-dimensional information. Firstly, the required power is predicted by a deep inference framework (LASSO-CNN), which integrates the least absolute shrinkage selection operator (LASSO) and convolutional neural network (CNN). Secondly, the optimal SoC sequence on hilly roads is planned in advance, with the SHEV's recuperated energy being predicted based on the current driving state. Finally, MPC-based predictive energy management is achieved by combining required power prediction and SoC planning with rolling optimization and feedback correction. Simulation results show that the fuel consumption is 2.51% lower than the deterministic dynamic programming-based controller, and the computation time is decreased by 48.1%. These promising results suggest that the proposed predictive energy management strategy can play a critical role in predicting power demand, which further reduces the fuel consumption of the hybrid electric mining truck.