隨著全球電力系統朝向低碳化與高電氣化發展，再生能源占比與終端用電需求同步成長，電動車與充電基礎設施已逐漸成為能源轉型的重要組成之一。相較於一般私人乘用車，電動巴士具有固定路線、固定時間、高功率充電與高電量需求等特性，因此其充電系統除需具備足夠功率供應能力外，亦須同時滿足穩定通訊、充電排程與場站整體調度等要求。在此背景下，ISO 15118 已成為電動車與充電樁之間進行高階通訊的重要標準，可支援充電參數協商、智慧充電、Plug and Charge 與未來雙向充電功能。然而，既有研究指出，ISO 15118 在實際部署中仍可能面臨通訊降級、身分綁定不足、TLS 支援不足與通訊訊號不佳或受限等問題，進而影響充電通訊之穩定性與充電可靠度。 有鑑於此，本研究以高密度電動巴士充電場域為對象，提出結合充電恢復機制與機器學習風險預測之自適應控制架構。本文先以電池電量判斷中斷事件是否仍值得恢復，再依據失敗次數與風險分數決定下一次嘗試恢復充電的時機，並搭配模型層自適應機制與控制層策略切換，使系統可在功率調降、自動重啟與混合控制之間選擇較合適的處理方式。模擬結果顯示，所提出之方法可降低人工介入需求，並提升充電穩定性；進一步導入整體自適應策略後，亦能在中斷率、目標電量達成率與完成延遲之間取得較均衡之營運表現。整體而言，本文方法可作為高密度電動巴士充電場域之設計參考，並提供後續實場驗證與控制策略擴充之基礎。;Reliable charging communication is essential for high-density electric bus charging stations， where vehicles require high-power charging， fixed schedules， and stable station-level coordination. Although ISO 15118 supports charging negotiation， smart charging， Plug and Charge， and future bidirectional charging， practical deployment may still suffer from communication degradation， authentication issues， TLS limitations， and unstable signal quality. This thesis proposes an adaptive recovery control method for electric bus charging communications. The proposed framework integrates SoC-Driven Session Resume， Adaptive Back-off Resume (ABR)， machine-learning-based risk prediction， and adaptive control strategies. When an interruption occurs， the system determines whether the session should be recovered according to the battery state of charge and recovery conditions. It then adjusts the resume timing based on retry status， risk score， and communication quality. Model-level and control-level adaptation are further applied to select suitable prediction models and control actions. Simulation results show that the proposed method can reduce manual intervention， improve charging stability， and achieve a better balance among interruption rate， target SoC achievement rate， recovery time， and operational delay. Therefore， the proposed method can provide a practical reference for improving reliability in high-density electric bus charging stations.
