Battery degradation mitigation-oriented strategy for optimizing e-hailing electric vehicle operations

Effective management of battery degradation is crucial for electric vehicles (EVs) due to the high costs associated with replacing EV batteries. In practice, uninformed charging behaviors of EV drivers can accelerate battery wear without proper guidance. To address this challenge, this paper introduces a battery degradation mitigation-oriented charging and order-serving problem for EVs operating on the e-hailing platform. The objective is to maximize the lifespan profit for individual EVs, which encompasses order service revenue, charging expenses, and battery degradation costs. To achieve this goal, a Markov decision process model is developed to capture the dynamics of individual e-hailing EV operations, and a battery degradation cost estimation method is specifically proposed for the e-hailing scenario. Moreover, we propose a multi-agent reinforcement learning (MARL) framework with a centralized training and decentralized execution paradigm. The MARL approach integrates a reward-shaping approach and an enhanced multi-agent upper confidence bound approach to determine the optimal charging and order-serving strategy for EVs. We propose a novel order assignment method to reduce the imbalanced degradation costs across EVs during the learning process. Our simulation experiments validate that the proposed strategy can substantially prolong EV battery life while concurrently boosting driver profits. Furthermore, an explanation of the strategy is provided to ensure transparency and understanding of the decision-making process.