Predictive energy management strategy on hybrid electric vehicle utilizing available future information

Abstract

Hybrid electric vehicles (HEV) have been introduced to the market as a solution to meet the strict fuel and emissions regulations. As they propel the vehicle using two energy sources, a proper energy management strategy is needed. Among existing energy management optimization methods for HEVs, the horizon optimization methods (e.g. Dynamic Programming) guarantee the global optimality when minimizing fuel consumption. However, implementing such the optimization method for real-world operation is impossible as horizon optimization methods require the prior knowledge of the full driving cycle including the future driving cycle. While prior studies proposed various approaches to predict the future driving cycle based on the past and present driving information, their prediction accuracy was low. Therefore, this study proposes the energy management optimization method of a parallel HEV through using future driving conditions that are predicted based on not only past and present information, but also available future information. In order to minimize the fuel consumption at each time, the instantaneous optimization method is used because the method does not need an unavailable full driving cycle. The future equivalent factor, which is the converting parameter from electric power to fuel consumption used in the instantaneous optimization, is predicted using not only the past and present information, but also available future traffic and road slope information by a Long Short-Term Memory network. The results revealed that in various driving environments, the proposed predictive energy management strategy improved fuel economy and charge-sustenance performances on the extreme driving environment than other energy managements.