Theses

Predictive energy management has gained interest in automotive research, primarily aimed at optimising power delivery between energy sources, since the introduction of hybrid electric powertrains. As battery electric vehicles and fuel cell hybrid electric vehicles emerge as leading alternatives to internal combustion engines both introduce distinct thermal challenges that have spurred interest in predictive thermal management. However, few studies have explored integrated strategies that jointly optimise propulsion and thermal energy within a unified framework. This thesis investigates integrated predictive thermal and energy management through two case studies focused on fuel cell hybrid electric vehicle and battery electric vehicle applications. A Dynamic Programming framework is used to determine globally optimal control strategies in offline simulations, then adapted to shorter prediction horizons to assess suitability for online application.

The first case study extends the classic hybrid power split problem by applying predictive control to a hydrogen fuel cell bus model in MATLAB, incorporating battery state of charge and fuel cell temperature as system states. The Dynamic Programming controller aims to minimise hydrogen consumption while avoiding thermal violations of +80°C through a multi-objective cost function, achieving an average reduction of 0.25kg/100km in fuel use and 3.57% increase in efficiency over the rules-based benchmark. The second case study applies the framework to an electric vehicle holistic thermal management system model provided by AVL GmbH. Thermal energy is distributed across the cabin, battery, and motor using multi-mode control. The Dynamic Programming controller selects optimal thermal modes and heater power across varying conditions, achieving an average 2.6% efficiency gain and a 15.2km increase in range. However, cost balancing proved challenging, leading to a 1.04°C average reduction in cabin temperature. Applying Dynamic Programming with shorter prediction horizons produced mixed results for both studies, sometimes under-performing compared to the rules-based benchmarks, while introducing irregularities as it attempts to optimise with less foresight.

This research explores the feasibility and limitations of applying Dynamic Programming for integrated predictive thermal and energy management where prior studies have only addressed propulsion and thermal energy separately. The results highlight the practical challenges of implementing thermal models for predictive control and the challenge of balancing trade-offs between efficiency and thermal performance. None the less, the learning gained supports a foundation for future work toward real-time, holistic vehicle control..