Reinforcement Learning-Based Adaptive Mobile Charging Station Placements in Mobile-Fixed Charging Stations Collaboration Network

Today, electric vehicles (EVs) have gained significant recognition in the global market as an innovative mode of transport. However, the development of EVs is based on the interaction with the energy Internet. The increasing number of EVs has presented significant challenges to the existing charging infrastructure. Traditional fixed-location charging stations are increasingly inadequate to meet fluctuating charging demand, leading to inefficiencies such as long waiting times and uneven distribution of charging load. To address the problem, this paper proposes a Mobile-Fixed Charging Stations Collaboration Network (MFCSCN), which integrates fixed charging stations with mobile charging stations to dynamically adapt to real-time charging demands. Firstly, LightGBM is utilized to predict the charging load at fixed charging stations considering historical data and real-time EV mobility patterns. Secondly, the reinforcement learning algorithm is employed to optimize the placement of mobile charging stations based on predicted demand. Third, within the MFCSCN framework, LPPO (LightGBM and Proximal Policy Optimization) is proposed, which combines predictive modeling and reinforcement learning to optimize the dynamic placement of mobile charging stations. Through extensive simulations, we demonstrate that the MFCSCN significantly improves the responsiveness and scalability of the EV charging infrastructure, offering a robust solution to the evolving needs of urban mobility.