Self-learning energy management is a promising concept, which optimizes real-world system performance by automated, on-line adaptation of control settings. In this work, the potential of self-learning capabilities related to optimization is studied for energy management in Plug-in Hybrid Electric Vehicles (PHEV). These vehicles are of great interest for the transport sector, since they combine high fuel efficiency with last mile full-electric driving. We focus on a specific use case: PHEV operation through future Zero Emission (ZE) zones of cities. As a first step towards self-learning control, we introduce a novel, adaptive supervisory controller that combines modular energy and emission management (MEEM) and deals with varying constraints and system uncertainty. This optimal control strategy is based on Pontryagin’s Minimum Principle and maximizes overall energy efficiency. The constraints are directly related to having sufficient battery energy for full electric driving and to meet real-world tailpipe NOx emissions. This control strategy is extended with a new adaptation mechanism for control parameters, including references, based on pre-view information. For a given mission, these control parameters are numerically optimized.
Simulations are done using a validated hybrid truck model with Euro-VI Diesel engine and urea-based SCR system. To demonstrate the self-learning capabilities, we study the effect of battery ageing and changing route (i.e. detour due to unexpected traffic jam). For the specified mission, the performance of the optimal MEEM is compared with a standard MEEM strategy without information on system or environment state. Cold start after the ZE zone is found to be challenging. From these results, it is concluded that vehicle operational costs can be reduced by 18% while meeting real-world emission limits if information is available on long-term battery state and route.