Towards Supervisory control for complex Propulsion subsystems

Licentiate thesis, 2018

Powertrain subsystem complexity has been on the rise with increasing legal requirements and meeting disruptive market trends. There is greater potential for cost efficient robust operation with integrated control units and software development. For systems that are interdependent, operating towards the common goal of fuel optimal operation under controlled exhaust emissions, it would be natural to integrate controls using a supervisory controller with a holistic overview of subsystem operation that utilised synergies and optimal trade-offs.

Connected cars have grown exponentially owing to consumer demand which offers rich data on vehicle operation and enables the possibility of tailoring systems to individual optimum operation. The possibility to feed external data, such as traffic information combined with the specific vehicle historic operation, enables prediction of the future vehicle trip and operating condition with greater accuracy.

A supervisory control framework for a diesel powertrain that is capable of utilising predicted look ahead information is developed. The look ahead information as a time trajectory of vehicle speed and load is considered. The supervisory controller considers a discrete control action set over the first segment of the trip ahead. The cost to optimise is defined and pre-computed off-line for a discrete set of operating conditions. A full factorial optimisation carried out off-line is stored on board the vehicle and applied in real time.

In the first approach, a set of predefined trip segments with off-line optimisation is considered. Here a library of segments is considered which would need to provide sufficient coverage of all possible trip characteristics along with a pattern matching or clustering algorithm. Another approach, to use a lumped parameter based model that can characterise the behaviour of the subsystems over the trajectory, is also examined for real-time on-line application. Simulation comparison of both controllers with the baseline controller indicates a 1% total fuel equivalent cost improvement while offering the flexibility to tailor the controller for different cost objective and improving robustness of exhaust emission control.