Optimal sizing of a series PHEV : comparison between convex optimization and particle Swarm optimization

M. Pourabdollah, E. Silvas, N. Murgovski, M. Steinbuch, B. Egardt

Research output: Contribution to journalConference articleAcademicpeer-review

39 Citations (Scopus)
7 Downloads (Pure)


Building a plug-in hybrid electric vehicle that has a low fuel consumption at low hybridization cost requires detailed design optimization studies. This paper investigates optimization of a PHEV with a series powertrain configuration, where plant and control parameters are found concurrently. In this work two of ten used methods are implemented to find optimal energy management with component sizes. In the first method, the optimal energy management is found simultaneously with the optimal design of the vehicle by using convex optimization to minimize the sum of operational and component costs over a given driving cycle. To find the integer variable, i.e., engine on-off, dynamic programming and heuristics are used. In the second method, particle swarm optimization is used to find the optimal component sizing, together with dynamic programming to find the optimal energy management. The results show that both methods converge to the same optimal design, achieving a 10.4% fuel reduction from the initial powertrain design. Additionally, it is highlighted that the usage of each of the method poses challenges, such as computational time (where convex optimization outperforms particle swarm optimization by a factor of 20) and the tuning effort for the particle swarm optimization and the ability to handle integer variables of convex optimization.
Original languageEnglish
Pages (from-to)16-22
Issue number15
Publication statusPublished - 28 Oct 2015
Event4th IFAC Workshop on Engine and Powertrain Control, Simulation and Modeling (E-COSM '15), August 23-26, 2015, Columbus, OH, USA - Ohio State University, Columbus, United States
Duration: 23 Aug 201526 Aug 2015


Dive into the research topics of 'Optimal sizing of a series PHEV : comparison between convex optimization and particle Swarm optimization'. Together they form a unique fingerprint.

Cite this