Real time energy management of electrically turbocharged engines based on model learning

Engine downsizing is a promising trend to decarbonise vehicles but it also poses a challenge on vehicle driveability. Electric turbochargers can solve the dilemma between engine downsizing and vehicle driveability. Using the electric turbocharger, the transient response at low engine speeds can be recovered by air boosting assistance. Meanwhile, the introduction of electric machine makes the engine control more complicated. One emerging issue is to harness the augmented engine air system in a systematical way. Therefore, the boosting requirement can be achieved fast without violating exhaust emission standards. Another raised issue is to design an real time energy management strategy. This is of critical to minimise the required battery capacity. Moreover, using the on-board battery in a high efficient way is essential to avoid over-frequent switching of the electric machine. This requests the electric machine to work as a generator to recharge the battery. The capability of generating power strongly depends on the engine operating point. One big challenge is that the calibration of generating power capability is time-consuming in experiments. This paper proposes a neuro-fuzzy approach to model the engine. Based on the virtual engine model, the capability of generating power at arbitrary engine operating point can be obtained fast and accurately, which is applicable to implement in real time