Minimum-fuel Energy Management of a Hybrid Electric Vehicle via Iterative Linear Programming

Nicolò Robuschi, Mauro Salazar (Corresponding author), Nicola Viscera, Francesco Braghin, Christopher H. Onder

Research output: Contribution to journalArticleAcademicpeer-review

2 Citations (Scopus)
161 Downloads (Pure)

Abstract

This paper presents models and optimization algorithms to compute the fuel-optimal energy management strategies for a parallel hybrid electric powertrain on a given driving cycle. Specifically, we first identify a mixed-integer model of the system, including the engine on/off signal and the gear-shift commands. Thereafter, by carefully relaxing the fuel-optimal control problem to a linear program, we devise an iterative algorithm to rapidly compute the minimum-fuel energy management strategies including the optimal gear-shift trajectory. We validate our approach by comparing its solution with the globally optimal one obtained solving the mixed-integer linear program and with the one resulting from the implementation of the optimal strategies in a high-fidelity nonlinear simulator.
We showcase the effectiveness of the presented algorithm by assessing the impact of different powertrain configurations and electric motor size on the achievable fuel consumption.
Our numerical results show that the proposed algorithm can assess fuel-optimal control strategies with low computational burden, and that powertrain design choices significantly affect the achievable fuel consumption of the vehicle.
Original languageEnglish
Article number9220815
Pages (from-to)14575-14587
Number of pages13
JournalIEEE Transactions on Vehicular Technology
Volume69
Issue number12
Early online date26 Oct 2020
DOIs
Publication statusPublished - Dec 2020

Keywords

  • Hybrid electric vehicles
  • convex optimization
  • energy management
  • linear programming
  • mixed-integer linear programming
  • supervisory control

Fingerprint

Dive into the research topics of 'Minimum-fuel Energy Management of a Hybrid Electric Vehicle via Iterative Linear Programming'. Together they form a unique fingerprint.

Cite this