Online energy management strategy of fuel cell hybrid electric vehicles based on time series prediction

A suitable energy management strategy is essential to reduce hydrogen consumption of fuel cell hybrid electric vehicles (FCHEVs) and limiting its negative effects. Many different methods for the energy management of FCHEVs are being used. As common used optimization-based approaches, the genetic algorithm and dynamic programming (DP) are frequently used in global optimization control to improve the efficiency and performance of energy storages in FCHEVs. However, these offline strategies cannot be applied to the vehicle if the driving cycle is not known or predicted. In this paper, an online energy management strategy is proposed base d on time series prediction model nonlinear autoregressive neural network (NARANN). Then, a novel approach using the moving window method is applied, in order to 1) train the prediction model and 2) iteratively perform offline optimization-based strategies. In the proposed strategy, the prediction model can provide accurate online driving cycle. Based on these dynamically predicted driving profiles, the offline optimization-based strategies can be easily applied. The proposed strategy is simulated using actual driving cycle data from an electric Golf Cart. Simulation results show that the effectiveness of the proposed method.