Return to search

Energy Management Strategies for Hybrid Electric Vehicles with Hybrid Powertrain Specific Engines

Energy-efficient powertrain components and advanced vehicle control strategies are two effective methods to promote the potential of hybrid electric vehicles (HEVs). Aiming at hybrid system efficiency improvement, this thesis presents a comprehensive review of energy-efficient hybrid powertrain specific engines and proposes three improved energy management strategies (EMSs), from a basic non-adaptive real-time approach to a state-of-the-art learning-based intelligent approach.
To evaluate the potential of energy-efficient powertrain components in HEV efficiency improvement, a detailed discussion of hybrid powertrain specific engines is presented. Four technological solutions, i.e., over-expansion cycle, low temperature combustion mode, alternative fuels, and waste heat recovery techniques, are reviewed thoroughly and explicitly. Benefits and challenges of each application are identified, followed by specific recommendations for future work. Opportunities to simplify hybrid-optimized engines based on cost-effective trade-offs are also investigated.
To improve the practicality of HEV EMS, a real-time equivalent consumption minimization strategy (ECMS)-based HEV control scheme is proposed by incorporating powertrain inertial dynamics. Compared to the baseline ECMS without such considerations, the proposed control strategy improves the vehicle drivability and provides a more accurate prediction of fuel economy. As an improvement of the baseline ECMS, the proposed dynamic ECMS offers a more convincing and better optimal solution for practical HEV control.
To address the online implementation difficulty faced by ECMS due to the equivalence factor (EF) tuning, a predictive adaptive ECMS (A-ECMS) with online EF calculation and instantaneous power distribution is proposed. With a real-time self-updating EF profile, control dependency on drive cycles is reduced, and the requirement for manual tuning is also eliminated. The proposed A-ECMS exhibits great charge sustaining capabilities on all studied drive cycles with only slight increases in fuel consumption compared to the basic non-adaptive ECMS, presenting great improvement in real-time applicability and adaptability.
To take advantage of machine learning techniques for HEV EMS improvement, a deep reinforcement learning (DRL)-based intelligent EMS featuring the state-of-the-art asynchronous advantage actor-critic (A3C) algorithm is proposed. After introducing the fundamentals of reinforcement learning, formulation of the A3C-based EMS is explained in detail. The proposed algorithm is trained successfully with reasonable convergence. Training results indicate the great learning ability of the proposed strategy with excellent charge sustenance and good fuel optimality. A generalization test is also conducted to test its adaptability, and results are compared with an A-ECMS. By showing better charge sustaining performance and fuel economy, the proposed A3C-based EMS proves its potential in real-time HEV control. / Thesis / Doctor of Philosophy (PhD)

Identiferoai:union.ndltd.org:mcmaster.ca/oai:macsphere.mcmaster.ca:11375/28011
Date11 1900
CreatorsWang, Yue
ContributorsEmadi, Ali, Keshavarz-Motamed, Zahra, Mechanical Engineering
Source SetsMcMaster University
LanguageEnglish
Detected LanguageEnglish
TypeThesis

Page generated in 0.0019 seconds