高速鐵路與其他軌道運輸系統在設計上最大不同處為具有嚴格的定時運行特性，亦即列車運行、號誌控制，以至於旅客或票務資訊均必須依照時刻表來運作。當高鐵系統遇到災害等緊急事故時，傳統上各調度人員必須互相溝通協調，根據目前系統狀況構思出適合的調度程序，產生對應的列車時刻表。此過程非常仰賴調度人員的溝通效率與專業經驗，耗時且易出錯，不但耽擱旅客時間與減少載客量，亦可能因調度失當造成更多人員傷亡造成；此外，在緊急事故時，高鐵系統的電力供應常成為另一不穩定來源，過去研究甚少探討此因素對於整體調度的影響。經由技術文獻回顧可知，智慧型代理人技術非常適合用於多人協調問題上，故本研究提出以本體論與智慧型代理人技術為基礎之高鐵列車緊急事故調度模式，讓智慧型代理人具備列車調度知識，能夠即刻做出最適決策。本研究首先以本體論設計知識模型來表達包含行控中心控制員、列車駕駛員、基地管理員與電力控制員等，其次，設計列車調度的推論規則來模擬各調度員之間的溝通過程。最後並以高鐵實際案例作為模型測試驗證用，驗證結果顯示此模型能產生與人工調度最佳結果一致的調度班表，且大幅減少人工調度所需時間。應用本研究模型將有助於高鐵在緊急事故時之應變效率與行車安全的提升。;There is a big difference between HSR and traditional railway system is that HSR has a characteristic of strict on-time. The operations include train scheduling, sign control and passengers ticketing information must follow the train timetable. When an incident occurred in the HSR system, each dispatcher has to communicate and coordinate with each other. They cooperate to get the suitable dispatching procedure according to the current situation and generate a corresponding train timetable. This process is dependent on the efficiency of communication and professional experience for dispatchers, it is time-consuming and error-generating. In this way, it not only delayed the passenger time and reduced the passenger capacity but also caused more casualties due to dispatching error. In addition, the power supply often be another source of instability during incident, and the few studies over the past investigated this factor that has an effect on the overall dispatching. In the technical literature review shows that intelligent agent technology is ideally suited for the issue of people coordination. Therefore, this study proposed an ontology-based multi-agent model for emergency dispatching for HSR system. The intelligent agents have the train dispatching knowledge and can instantly make optimal decisions. Firstly, the ontology knowledge model is designed to express each components of the railway system include control controller, train driver, depot manager and power controller. Then, the train dispatching inference rules are designed to simulate the communicated process between each dispatcher. Finally, the HSR real cases are used to testing in verification and validation steps. The experimental results show that proposed model can generated the dispatching procedure is the same as the best human scheduling. Application of proposed model will contribute to increase the response efficiency and train operating safety during emergency for HSR system.