An Optimal Energy Management Strategy With Time-Varying Equivalent Factor Based on Transformer for Multi-Mode Hybrid Electric Mining Trucks

Abstract

Incorporating hybrid electric technology in mining trucks efficiently allocates energy between the diesel engine and electric motors, significantly reducing fuel consumption and emissions. This paper introduces an advanced Transformer network model aimed at an adaptive Energy Management Strategy (EMS) to improve the fuel efficiency of Hybrid Electric Mining Trucks (HEMTs). Initially, we outline models representing various HEMT subsystems, including its complex powertrain. Subsequently, leveraging the established vehicle model, we employ Dynamic Programming (DP) to ascertain the globally optimal control strategy for a standard driving cycle. Utilizing the Transformer model with ProbSparse self-attention, we determine the state of charge (SOC) associated with this globally optimal control strategy. We then implement a Proportional-Integral (PI)-based SOC tracking controller to achieve a time-varying Equivalent Factor (EF). Additionally, we compare the proposed strategy across various driving cycles, underscoring the Transformer network's superior generalization capabilities over the Long-Short Term Memory (LSTM) model. Compared to the rule-based (RB) strategy and the traditional Equivalent Consumption Minimization Strategy (ECMS), our proposed EMS with time-varying EF based on the Transformer model shows substantial performance improvements of 15% and 12%, respectively. Furthermore, our strategy exhibits a 2% fuel economy advantage over an LSTM-based EMS algorithm with time-varying EF.