Abstract
Abstract In this article, a multi-objective optimization-oriented energy management strategy is investigated for fuel cell hybrid vehicles on the basis of rule learning. The degradation of fuel cells and lithium-ion batteries are considered as the objective function and translated into the equivalent hydrogen consumption. The optimal fuel cell power sequence and state of charge trajectory, considered as the energy management input, are solved offline via the Pontryagin’s minimum principle. The K-means algorithm is employed to hierarchically cluster the optimal data set for preparation of rules extraction, and then the rules are excavated by the improved repeated incremental pruning to production error reduction algorithm and fitted by the quasi-Newton method. The simulation results highlight that the proposed rule learning-based energy management strategy can effectively save hydrogen consumption and prolong fuel cell life with real-time application potential.
Highlights
As a promising segment of transportation electrification, fuel cell hybrid vehicles (FCHVs) have been the research hotspot in automotive industry, owing to their zero emission, high efficiency and low noise [1]
Proper power allocation between different energy sources is spurred to optimize the operation performance of FCHVs, and it is often tackled by the socalled energy management strategies (EMSs), which have been intensively investigated by industry and academia [2]
The main contributions of this paper are attributed to the following two aspects: 1) A novel multi-objective EMS for FCHV is proposed based on the rule learning, and 2) the parameters of the fitted formulas are solved by the BFGS algorithm
Summary
As a promising segment of transportation electrification, fuel cell hybrid vehicles (FCHVs) have been the research hotspot in automotive industry, owing to their zero emission, high efficiency and low noise [1]. [15] applies the PMP to optimize the battery life while reducing the battery energy loss, fuel consumption and power system cost Another kinds of global optimization methods are learning-based algorithms, such as reinforcement learning (RL) and its extensions, including Q-learning and deep RL [16]. Their controlling performance is difficult to guarantee all the time due to complicated time-varying driving conditions In this context, instantaneous optimization-based strategies emerge, including ECMS and model predictive control (MPC) algorithms. The main contributions of this paper are attributed to the following two aspects: 1) A novel multi-objective EMS for FCHV is proposed based on the rule learning, and 2) the parameters of the fitted formulas are solved by the BFGS algorithm. The motor driving system is supposed to be able to effectively cope with voltage variations, and the specific working process is not considered
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
