緊急救災物流是災害緊急管理重要的問題之一,其目的在於有效率且持續性的配送救災物資到災區,維持災民基本的生活需求。本研究針對嚴重自然災害發生後的緊急救災物流問題,發展一套即時救災物資配送決策支援系統,此系統建立在滾動平面法架構下,分為災情資訊估計與預測模組以及救災物資配送模組。災情資訊估計與預測模組利用資料融合方法與區間卡爾曼濾波器演算法,功能是預測各災區之物資需求量以及各配送路線之配送時間。為了描述災情資訊的不確定性,本研究以區間資料來表達所預測的災情資訊。預測的災情資訊將輸入至救災物資配送模組中進行最佳化求解,求解目標為最小化完成所有救災物資配送工作的總時間;最佳化模式依據決策者的風險趨避傾向分為兩種模式: 最佳最佳解模式(Best Optimal Solution, BSO),與最差最佳解模式(Worst Optimal Solution ,WSO)。本研究利用921集集大地震的資料為實驗案例,惟部分實證資料因取得不易,因而透過資料模擬的方式產生,進行數值測試。研究結果顯示本研究所發展之系統能夠有效預測救災物資需求量與配送時間,並求解出適當的配送策略,具有實務應用上的價值。
The purpose of Emergency Logistics is to distribute relief supplies efficiently to the affected areas. In this study, we present a decision support system for real-time disaster relief distribution. This system works under the rolling horizon framework and includes two modules. Disaster information prediction module contains two methods: interval data fusion and interval Kalman filter. This module can predict relief demand of each affected area and distribution time of each delivery route. In order to address the data uncertainty, we use intervals to describe the disaster information. Another module is relief supplies distribution module. This module receives the information from Disaster information prediction module and solves for optimal relief distribution flows. The goal is to minimize the total time (makespan) needed to finish all distribution works. Numerical experiments based on a real large-scale earthquake in Taiwan are conducted, and the results indicate the system can effectively predict relief demand and distribution time and solve two kinds of distribution strategy in short period of time.