A real-time predictive energy management strategy of fuel cell/battery/ ultra-capacitor hybrid energy storage system in electric vehicle

Abstract

For energy management of new energy vehicles, different dynamic characteristics of different onboard power sources should be taken into consideration. In this paper, a real-time predictive energy management strategy is proposed for the fuel cell/battery/ultra-capacitor hybrid energy storage system in fuel cell electric vehicles. A LSTM neural network velocity predictor is developed to predict future velocity of the vehicle and then the future power requirement can be calculated. On this basis, the wavelet transform algorithm is adopted to protect the fuel cell and battery from fast-variation transients and peak power demand conditions; and a rule-based strategy is introduced for the control of the power sources' SOC. Simulation results show that the LSTM based predictor has an acceptable accuracy for energy management usage. The proposed energy management strategy can not only successfully reduce the power frequency of the fuel cell and battery, but also ensure the vehicle performance by keeping the SOC of the ultra-capacitor in reasonable range.