Forecasting ECMS for Hybrid Electric Vehicles

Abstract

This paper aims to propose a real-time suitable method to tackle the problem of energy and pollutant management of Hybrid Electric Vehicles. Methods proposed in the literature often limit the underlying optimal control problem to single-instant optimizations (Paganelli, 2002) due to the difficulty of taking future into account and to onboard limited computational resources. The point of the present paper is to propose an online oriented method based on a long-term vehicle speed prediction, using cartographic information such as speed limitation, road curvature, traffic and road signs. Pontryagin Maximum Principle applied on this speed prediction signal allows to convert the optimal control problem into a root-finding problem. This problem is solved using a Pegasus algorithm initialized by a black-box method trained offline, allowing high computational efficiency. The results are near-optimal and significantly better than classical methods: in the real-driving trip used in this paper, forecasting-ECMS showed a consumption 1.1% better and NOx emissions 4.4% better than a SOC-feedback adaptive-ECMS.