Optimal Design of Wireless Charging Electric Buses-Based Machine Learning: A Case Study of Nguyen-Dupuis Network

Abstract

The Wireless Charging Electric Transit Bus (WCETB) is an innovative electric transportation technology that receives power wirelessly from underground power transmitters. Since the battery in the bus can be wirelessly charged during the moving, the need to stop for the vehicle at the charging station is eliminated, and recharging time is remarkably decreased. One of the best methodologies to commercialize the WCETB technology is to economically allocate the power tracks on considered routes while optimally determining the battery size of the vehicle at the same time. In this paper, the planning of the power transmitters and the capacity of the battery size for a WCETB with a multi-route model is optimally designed by Deep Deterministic Policy Gradient (DDPG). In particular, the DDPG algorithm is adopted in this model, an appropriate reward function is defined for the multiple route problem, and the optimal problem is solved by adopting the training ability of deep neural networks (DNNs), i.e., Actor and Critic neural networks. A complex Nguyen-Dupuis (N-D) traffic network with multiple routes is considered a case study to evaluate the validity and performance of the proposed deep learning scheme from a systematic point of view. Numerical analysis along with sensitivity examination confirms the efficiency of the optimal design and solution process.