Effective energy management strategy based on deep reinforcement learning for fuel cell hybrid vehicle considering multiple performance of integrated energy system

Abstract

To reduce the power burden of the fuel cell system (FCS) in fuel cell hybrid electric vehicles (FCHEVs), battery and supercapacitor are widely used as energy storage system (ESS) in FCHEVs. In this paper, an energy management strategy (EMS) based on deep reinforcement learning is proposed to solve the power allocation problem among three energy sources. Considering that three energy sources increase the complexity of the state action space, this paper proposes an adaptive fuzzy control filter to separate the load demand power to reduce the computational burden of the deep deterministic policy gradient (DDPG) algorithm. Meanwhile, a learning guidance mechanism (LGM) is proposed in this paper to guide the agent in a rational learning direction, and the reward convergence results are presented in this paper. The training results show that the equivalent hydrogen consumption reaches 99.5% of fuzzy logic controlled-based EMS (FLC-EMS), the FCS efficiency improves by 5.9%, and the stability of battery and supercapacitor is also significantly improved. Eventually, three different driving conditions are selected to verify the effectiveness of the proposed EMS, the results show that the proposed EMS in this paper outperforms (FLC-EMS) in all aspects.

Reconfiguration of low-voltage distributed power sources within electric power's distribution network based on improved particle swarm-fish swarm fusibility algorithm

“…In the research of system reliability, it is studied that the probability of power failure is the lowest among the four algorithms by reasonably configuring the line connection mode, which significantly improves the fault tolerance of the system. Hu et al 30 obtained similar results. For the economic research of the system, the energy consumption and operating cost of the system are lower than those of CNN, RF and BPNN systems.…”

A Novel Minimal-Cost Power Allocation Strategy for Fuel Cell Hybrid Buses Based on Deep Reinforcement Learning Algorithms

“…Huang et al [33] proposed an EMS based on DDPG algorithm for a range extend fuel cell hybrid vehicle to achieve optimal power allocation between fuel cell and power battery in pure electric mode and the range extend mode. DDPG introduces an actor-critical framework based on deep Q-learning compared with other deep reinforcement learning methods [34]. Zheng et al [35] proposed a DDPG-based energy management strategy, improved the efficiency of the algorithm by prioritizing experience replay techniques, and verified the effectiveness of the proposed strategy in comparison with DP.…”