Energy Management Improvement Based on Fleet Learning for Hybrid Electric Buses

This paper is focused on analysing the energetic key performance indicators of a hybrid electric bus fleet in order to improve its energy management (at local and fleet level) and profitability. The analysed fleet is composed of buses with parallel and series configurations and include energy storage systems based on batteries and ultra capacitors. The test routes have been selected from a data base of urban standardised cycles. In a first stage, a dynamic programming approach has been applied to determine the initial optimal operation performance for each bus route. Then, several disruptions (e.g. traffic jams, auxiliary consumption and passenger variations) have been added to the routes to simulate "real" road and daily operation conditions. In this paper, a fleet learning methodology is proposed to analyse, process and decide based on the collected data from "real" conditions of the whole fleet. This data is used for monitoring the energetic key performance factors by learning from the buses with the best energetic behaviour. Finally, a decision making process is applied to improve the local energy management of the less-efficient bus.