A Machine-learning based approach for Hybrid Electric Vehicle Redesign and Processor-in-Loop Based Power Management Validation

Abstract This article presents a design planning method for a power-split hybrid electric vehicle, which optimizes the powertrain components and the power management strategy (PMS) for better fuel efficiency and lower emission, following the principles of ecodesign. The road gradient is taken into account in the optimization process, which uses an offline constrained method. The powertrain components are tuned for real-time driving data using a Surrogate assisted evolutionary algorithm, which generates multiple design alternatives and selects the best one using a Modified Technique for Order of Preference by Similarity to Ideal Solution with vehicle weight reduction of 4%. The PMS is based on a model predictive control equipped with two stage optimization approach, which determines the optimal values of engine torque, engine speed, motor torque, motor speed and brake signal to ensure the desired power supply, with minimum fossil fuel consumption and emission. The proposed method achieves more than 5% improvement in fuel efficiency and 10% improvement in emission reduction compared to the existing methods that use Dynamic programming or Fuzzy logic.