The effects investigation of data-driven fitting cycle and deep deterministic policy gradient algorithm on energy management strategy of dual-motor electric bus

Abstract

Nowadays, the trend of powertrain electrification in the public transportation sector is clear. To meet the dramatic load variation and relatively high handling stability requirements for battery electric buses, the dual-motor four-wheel powertrain architecture attracts great attention in recent years. Although the bus routes are fixed, the driving speed and load vary significantly with time, season, passenger capacity, and traffic conditions, which presents a serious challenge for efficient power coupling in a dual-motor system to reduce energy consumption. This study provides a data-driven fitting cycle for the specific bus route. Then, Deep Deterministic Policy Gradient (DDPG) algorithm is introduced in Energy Management Strategy (EMS) design to improve the vehicle's economic performance with uncertain demand in the unknown cycle. The simulation results show that the proposed DDPG-EMS achieves 93.91%–97.66% of the benchmark Dynamic Programming (DP) – based EMS under various testing cycles. In addition, the comparison of DDPG-EMS agent trained by fitting cycle, standard cycle, and real driving data reached 97.2%–97.66%, 93.91%–97.0%, and 94.41%–96.0% of DP, respectively, which demonstrates the effectiveness of data-driven fitting cycle and reinforcement learning algorithm combination in EMS design for dual-motor electrified bus.