Predictive energy management of hybrid electric vehicles with uncertain torque demand forecast for on-road operation
- Prädiktives Energiemanagement von Hybridfahrzeugen mit unsicherer Vorausschau des Fahrerwunschmoments für den realen Straßeneinsatz
Joševski, Martina; Abel, Dirk (Thesis advisor); Andert, Jakob Lukas (Thesis advisor)
Dissertation / PhD Thesis
Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2018
The context of this dissertation is to theoretically investigate, design and implement a real-time capable estimation and control framework for the energy management of parallel hybrid electric vehicles that is suitable for on-road operation. The pursued control objectives are to minimize fuel consumption, to improve charge-sustainability of the battery pack, to enhance driving comfort in terms of reducing gear shifts and engine on/off events, tomaximize the recuperation of kinetic energy and to supply the requested wheel torque continuously. The control framework is designed as a model predictive control (MPC) scheme. While MPC schemes for the energy management of HEV powertrains have been studied over the past 15 years, MPC schemes for on-road operation represent an active field of research. The proposed MPC scheme is particularly designed to account for uncertainties in the future driver torque demand exploited for optimization. To gain an (uncertain) guess of the future torque demands over a finite horizon, an estimator based on machine learning techniques is introduced which relies on telematics data like speed limits along the road. The control scheme is designed by means of stochastic and robust MPC to account for deviations of the estimated from the actual driver torque demand that occur in on-road operation. These deviations might cause a deterioration of control performance and constraint violations in the MPC based control scheme. The control unit is arranged as hierarchical MPC scheme, composed of a high-level and a low-level control layer. Such architecture is selected to enable optimization over different time scales, required to meet different control objectives. While high-level control focuses on fuel economy as well as gear shifting and engine on/off optimization over a long time horizon, low-level control features a continuous delivery of the requested wheel torque over a short time horizon. A comprehensive simulation study is carried out in the dSPACE Automotive Simulation Models (ASM) framework to prove the efficacy of the hierarchical distributed control approach. Finally, an energy management framework is gained that is real-time capable, expandable and suitable for on-road operation.