Collective behavior information-based design approach to energy management strategy for large-scale population of HEVs

Abstract

Listen

Translate article

Optimization-based energy management strategies (EMS) have raised the energy-saving potential for hybrid electric vehicles (HEV). Despite this, performance of most strategies highly rely on accurate predictions on future driving demand. These predictions for long preceding horizon are inherently challenging to made due to their dependency on many uncertain factors that are complex to model. To fill the gap, this paper presents a novel EMS design method utilizing collective behavior information of large-scale population (CBLP) of HEVs. Firstly, each HEV is modeled as a particle with identical stochastic dynamics, and CBLP of such individuals is proved to be traceable based on mean-field theory, which obeys the Fokker–Planck equation. Secondly Gaussian process regression methods are developed to predict the evolution of CBLP for long preceding horizons, leveraging historical and real-time traffic data. An EMS design problem is finally formulated to minimize the energy consumption over the obtained prediction result, where working mode switching control is adopted. Following the idea, an energy management framework is proposed to offer recommended real-time strategies for vehicles. It is shown that the EMSs, designed to be optimal with respect to CBLP, will also provide near-optimal performance for each members of the population in high probability. The effectiveness of the proposed method is evaluated through numerical validations conducted on the real-world trajectory datasets.