Evolutionary algorithms promise more efficient plug-in hybrid EVs

Technology News | January 11, 2017

By Rich Pell

The energy management system (EMS) uses a more comlex analysis of energy use in driving that was inspired by natural phenomenon such as the way birds save energy by flying in formation and shares traffic information with other vehicles. This requires more processing power that existing systems but delivers signficant increases in efficiency and opens up the use of machine learning to further improve the efficiency over time.

PHEVs, which combine a gas or diesel engine with an electric motor and a large rechargeable battery, can be charged using mains electricity which reduces their need for fuel. However, the race to improve the efficiency of current PHEVs is limited by shortfalls in their energy management systems (EMS), which control the power split between engine and battery when they switch from all-electric mode to hybrid mode.

While not all plug-in hybrids work the same way, most models start in all-electric mode, running on electricity until their battery packs are depleted, then switch to hybrid mode. Known as binary mode control, this EMS strategy is easy to apply, but isn’t the most efficient way to combine the two power sources.

In lab tests, blended discharge strategies, in which power from the battery is used throughout the trip, have proven more efficient at minimizing fuel consumption and emissions. However, their development is complex and, until now, they have required an unrealistic amount of information upfront.

“In reality, drivers may switch routes, traffic can be unpredictable, and road conditions may change, meaning that the EMS must source that information in real-time,” said Xuewei Qi, a postdoctoral researcher at the Center for Environmental Research and Technology (CE-CERT) in UCR’s Bourns College of Engineering who led the project with Matthew Barth, CE-CERT director and a professor of electrical and computer engineering at UCR.

The developed and simulated by the team combines vehicle connectivity information (such as cellular networks and crowdsourcing platforms) and evolutionary algorithms.

“By mathematically modeling the energy saving processes that occur in nature, scientists have created algorithms that can be used to solve optimization problems in engineering,” said Qi. “We combined this approach with connected vehicle technology to achieve energy savings of more than 30 percent. We achieved this by considering the charging opportunities during the trip–something that is not possible with existing EMS.”

The current paper builds on previous work by the team showing that individual vehicles can learn how to save fuel from their own historical driving records. Together with the application of evolutionary algorithms, vehicles will not only learn and optimize their own energy efficiency, but will also share their knowledge with other vehicles in the same traffic network through connected vehicle technology.

“Even more importantly, the PHEV energy management system will no longer be a static device–it will actively evolve and improve for its entire life cycle. Our goal is to revolutionize the PHEV EMS to achieve even greater fuel savings and emission reductions,” Qi said, in a hint for the development of machine learning within the EMS.

The Office of Technology Commercialization at UCR has filed patents for the technology.

For more, see “Development and Evaluation of an Evolutionary Algorithm-Based Online Energy Management System for Plug-In Hybrid Electric Vehicles.”