結合傳統呼叫系統與目的地呼叫系統之電梯調度最佳化問題

Abstract

本研究探討如何解決並最佳化電梯調度問題，旨在平衡能源效率與乘客服務品質。文獻上大多把最佳化電梯調度問題被歸類為 NP-hard 問題，因為其涉及從眾多可行的電梯運行組合中識別最佳調度方案。

我們研究了一種混合控制系統的效率，該系統可將樓層設定為傳統呼叫或目的地呼叫模式。其目標是最大限度地降低電梯運行能耗成本，並減少長時間等待乘客的有效等待時間。

我們透過求解一系列靜態問題來解決動態電梯調度問題，並利用窮舉演算法和基因演算法進行求解。實驗結果表明，相較於傳統呼叫系統，目的地呼叫系統和混合呼叫系統在大多數情境下，於能耗和有效等待時間方面均表現出更佳的性能。此外，基因演算法在計算成本方面提供了顯著優勢，使其成為實際電梯系統更可行的方案。

我們開發了一個與電梯模擬系統串接的電梯調度系統，以模擬現實世界的運作，證實了我們所提出的方法在實務上的可行性。在利用模擬真實乘客需求進行的動態情境案例研究中，結果顯示我們所提出的基於基因演算法的派遣策略，相較於原本基於規則派遣的方法，在能源消耗和有效等待時間方面均顯著降低。

In this study, we explores optimization strategies for the Elevator Dispatching Problem (EDP), aiming to balance energy efficiency with passenger service quality. The EDP is classified as an NP-hard problem, involving the identification of optimal scheduling solutions from numerous feasible elevator operation combinations.

We investigate the efficiency of a hybrid control system, which sets floors for either conventional or destination call system. The objective is to minimize the total energy consumption cost of the elevators and the effective waiting time of passengers experiencing prolonged waiting periods.

Our approach addresses the dynamic EDP by solving a sequence of static problems, utilizing both exhaustive and genetic algorithms. Experimental results indicate that, compared to the All-Conventional call system, both All-Destination and Hybrid call systems tend to achieve better performance in energy consumption and effective waiting time across most scenarios. Moreover, the genetic algorithm offers substantial advantages in computational cost, rendering it a more feasible approach for practical elevator systems.

We developed an elevator dispatch system that integrates with an elevator simulation system to replicate real-world operations. Case studies in dynamic scenarios, utilizing simulated real passenger demand, revealed that the proposed GA-based strategy significantly reduces energy consumption and effective waiting time compared to the rule-based method.