利用非對稱相異性度量求解充電車系統規劃與其他應用之最佳化問題

Abstract

在本研究中，我們考慮一個具有隨機需求位置和任意到達時間的電動車（EV）充電系統。目標為決定最佳的充電站設置位置及對應的電動車充電站途程策略（EV charging station routing policy），從而最小化充電需求的平均旅行時間或平均旅行距離（mean travel time/distance）。透過考慮基於位置的電動車充電站途程策略（location-based EV charging station routing policy）並整併 Google 地圖（Google Maps）所提供的實際交通資訊，我們得以將此問題化為一非對稱分群問題，旨在最小化資料點到所對應之群中心的差異性（dissimilarity）總和。此模型所提供之資料驅動（data-driven）方式，不但可以納入各種營運考量，更能適用於其他具有相似性質的現實應用問題。針對此問題，本研究提出了兩個創新的非對稱分群演算法，並以幾種現實情境為例展示。然而，在人造不對稱資料的穩健性衡量（robustness testing）中，儘管其中一者表現良好，另一者卻表現出其局限性。

In this research, we consider a stochastic electric vehicle (EV) charging system with random demand locations and arrival times. The objective is to determine the optimal locations for charging stations and the corresponding EV charging station routing policy to minimize the mean travel time or distance for charging demands. By considering a location-based EV charging station routing policy and utilizing real traffic information from Google Maps, we formulate this as an asymmetric clustering problem aimed at minimizing the sum of dissimilarities from data points to their respective cluster centers. This model provides a data-driven approach that not only enables the incorporation of various operational concerns but also can be applied to other similar real-world applications. Two novel asymmetric clustering algorithms are developed to address the problem, illustrated using several real-world scenarios. However, the robustness testing on synthetic asymmetric data reveals that while one algorithm demonstrates strong performance, the other exhibits limitations.