Data-driven Energy Management Strategy for Plug-in Hybrid Electric Vehicles with Real-World Trip Information

TL;DR

This paper introduces a data-driven, cloud-assisted energy management system for plug-in hybrid electric vehicles that learns from real-world trip data to improve fuel efficiency, demonstrating significant gains over baseline methods.

Contribution

It proposes a novel two-layer EMS leveraging Vehicle-to-Cloud connectivity and real-world trip data to optimize energy management in PHEVs.

Findings

Achieved up to 7.3% improvement in MPGe on real-world routes.

Demonstrated the effectiveness of cloud-based learning for energy management.

Validated on over 3000 miles of real-world driving data.

Abstract

This paper presents a data-driven supervisory energy management strategy (EMS) for plug-in hybrid electric vehicles which leverages Vehicle-to-Cloud connectivity to increase energy efficiency by learning control policies from completed trips. The proposed EMS consists of two layers, a cloud layer and an on-board layer. The cloud layer has two main tasks: the first task is to learn EMS policy parameters from historical trip data, and the second task is to provide the policy parameters along a certain route requested from the vehicle. The on-board layer receives the learned policy parameters from the cloud layer and computes a real-time solution to the powertrain energy management problem, using a model predictive control scheme. The proposed EMS is evaluated on more than 3000 miles (48 independent driving cycles) of real-world trip data, collected along three commuting routes in California. For the routes, the proposed algorithm shows 3.3%, 7.3%, and 6.5% improvement in average MPGe when compared to a baseline EMS.