A MPC-Based Robust HDP Online Energy Management Strategy for Series Hybrid Loaders With Input Disturbances

For further reducing the fuel consumption of the series hybrid loaders (SHLs) with input disturbances, this paper develops a robust heuristic dynamic programming (HDP) online strategy based on model predictive control (MPC). The dynamic model of SHLs with disturbances is built by employing recurrent neural network (RNN), to efficiently represent its actual operating process. Then, the MPC-based robust HDP (MPC-RHDP) local optimal control algorithm is presented, and the convergence of proposed algorithm and the stability of control system are proved. Furthermore, a MPC-RHDP online strategy is proposed to execute the energy management controller. At last, the RNN dynamic model and designed MPC-RHDP strategy are verified and compared on the hardware in loop platform. After analyzing the experiment results, in comparison with analytic model, the RNN model in higher precision is capable of more effectively describing the real process for SHLs with disturbances. In addition, under sand and stone scenarios, the proposed MPC-RHDP strategy can separately achieve average fuel-saving rate by 24.07% and 10.27% in contrast to rule-based strategy and equivalent consumption minimization strategy, providing a novel method for real-time energy optimization of the SHLs with disturbances.