Performance investigation and energy optimisation in hybrid electric vehicle model using reinforcement learning and fuzzy controller. Issue 1 (16th December 2022)

• Record Type: Journal Article
• Authors: Pukkunnen, Emmanuel Babu; Joseph, Neena M.; Jos, Bos Mathew; Joy, Minu C.; Eldhose, K.A.
• Is Part Of: International journal of vehicle performance. Volume 9:Issue 1(2023)
• Page Start: 73
• Page End: 90
• Publication Date: 2022-12-16
• Subjects: HEVs -- hybrid electric vehicles -- series-parallel drivetrain -- EMSs -- energy management systems -- DRL -- deep reinforcement learning -- fuzzy control logic -- PID controllers -- speed control; Vehicles -- Design and construction -- Periodicals; 629.049
• Languages: English
• ISSNs: 1745-3194

Abstract

Hybrid electric vehicles (HEVs) are considered as one of the prominent solutions in reducing vehicular emission. Batteries and internal combustion engines (ICE) are the important components of a HEV, which acts as primary and secondary power source respectively. They simplify the refuelling process by minimising fuel consumption and by reducing virulent emissions. In this research, a series-parallel drivetrain - HEV model is proposed for investigating the performance and energy optimisation of the HEVs. The model is trained to operate at near optimum efficiency for minimising the energy loss. A deep reinforcement learning and fuzzy logic controller based energy management approach is proposed to optimise the energy consumption in HEVs. Results show that the energy management system (EMS) of the model is controlled effectively by the deep reinforcement learning (DRL) algorithm. Effective speed control is achieved by fine tuning the parameters using a fuzzy based PID controller which can be validated from the simulation results.