Fuzzy logic based power-split hybrid propulsion control system using digital twin assisted parallel learning

This paper explores an innovative concept of cyber-physical real-time optimisation, in which a digital twin-assisted parallel learning (DTPL) mechanism is proposed to improve the performance of a fuzzy logic (FL-) based power-split hybrid propulsion control system in terms of stability and energy consumption. This mechanism enables parallel learning between the actual supervisor and its digital twin in real driving situations. If the virtual controller dominates the driving process, the new parameter functions are synchronised to the real controller at the same time. Based on an analysis of the conformation of the hybrid propulsion model and its FL-based control system, a chaos-enhanced accelerated particle swarm optimisation algorithm is applied to the parallel learning of the membership functions. By hardware-in-the-loop testing, the result shows that the DTPL-driven control system leads to better fuel economy. Fuel consumption can be reduced by up to 15% compared to a system using charge sustaining and charge depleting strategy, and by up to 12% compared to a system using FL control strategy over an in-house driving cycle collected from the driving simulator.