隨著電動車的普及，充電基礎設施的需求日益增長。本研究的目標是開發一個基於深度強化學習的最佳充電站選擇演算法，以提升電動公車的充電效率。本文提出利用深度強化學習（Deep Reinforcement Learning, DRL）的方法，使用Deep Q Network（DQN）來作為最佳充電站的選擇。 本研究針對電動公車的充電需求建立模組，並引入DQN來改善充電站選擇。通過使用歷史充電資料和公車軌跡資料，演算法能夠學習到在不同情況下的最佳充電站選擇策略。DQN使用Q值函數來預測每個狀態下的期望獎勵，並基於這些預測來進行決策。利用訓練過的模型，我們可以使用當前資料來選擇充電站。在實驗結果部分，本研究比較不同獎勵函數對選擇充電站的影響。 本研究主要是分析出最佳充電站選擇策略，以減少人工選擇充電站的成本，對於推動電動車充電基礎設施的發展具有重要意義。;With the widespread adoption of electric vehicles, the demand for charging infrastructure is increasingly growing. This study aims to develop an optimal charging station selection algorithm based on deep reinforcement learning to enhance the charging efficiency of electric buses. This paper proposes using Deep Reinforcement Learning (DRL) methods, utilizing the Deep Q Network (DQN) for optimal charging station selection. This study establishes a model for the charging needs of electric buses and introduces DQN to improve the selection of charging stations. By using historical charging data and bus trajectory data, the algorithm can learn the optimal charging station selection strategy under different circumstances. The DQN uses the Q-value function to predict the expected reward for each state and makes decisions based on these predictions. With the trained model, we can use current data to select charging stations. In the experimental results section, this study compares the impact of different reward functions on selecting charging stations. The main focus of this research is to analyze the optimal charging station selection strategy to reduce the cost of manual selection of charging stations, which is of significant importance for promoting the development of electric vehicle charging infrastructure.
