捷運列車之最小耗能行駛時間控制法

Abstract

本研究的目標是發展捷運列車的節能控制方法，讓列車在兩站之間行駛的時候，既能循遵循規劃的行駛時間也能用最小耗能抵達車站；捷運線的列車規劃系統會根據排班表和班距需求，計算出捷運線上各列車在兩站之間的行駛時間。傳統的列車控制法是盡可能按照規劃的行駛時間抵達目的地，本論文則外加最小耗能的目標條件，因此必須在列車動態特性、軌道跡線和乘客負載都可能改變的環境下，求解列車牽引馬達和煞車器的最佳控制量和作動時間。本質上，這個最佳化控制問題面對的是非線性而且受到干擾的動態系統，傳統的線性最佳控制系統求解方法並不能適用；本論文提出自評式車速跡線控制法，藉由滿足動態規劃法的最佳化原則，求解控制量和作動時間。自評式車速跡線控制法採用多層感知型類神經網路做為仿效器，用雙啟發動態規劃法建立自評式學習機制，完成學習階段的類神經網路即可計算出逼近的最佳解；本論文對捷運列車的列車動態模型和能量模型有詳細的推導，透過電腦模擬對台北捷運藍線的車站進行驗證，經由比對自評式車速跡線控制法和傳統控制法的效能，證實這個方法確實可以達成節能和控制行駛時間的目的。

In the Mass Rapid Transit (MRT) system, the train regulation system provides each train with desired run time between two adjacent stations so as to track the train schedule and designed headway. The goal of train control is traditionally to minimize the run time error and in addition in this research to minimize the energy demand subject to train dynamics, track profile, and disturbance from passenger load. Essentially, the problem dynamics is nonlinear and disturbed so that linear optimal control approach is not appropriate. This thesis proposes the adaptive critic train profile control method to solve this complex optimization problem. The optimal control quantities that may result in optimal train profile are found by satisfying the principle of optimality in dynamic programming. The dual heuristic programming method and multilayered perceptrons are employed to design the adaptive critic train profile control system. The train model and energy model of MRT train are derived in detail. The feasibility and performance of the adaptive critic train profile control are verified by simulation results from applications in stations on Blue line of Taipei MRT system. The results suggest that the method is promising in energy saving and run time control.