價格遞減的多產品多期多廠多市多重運輸指派問題 及其遺傳演算求解法

Abstract

本研究提出一個具實務性的運輸規劃問題，名為「價格遞減的多產品多期多 廠多市多重運輸指派問題及其遺傳演算求解法」。完整定義此問題及其數學模型， 求解目標為極大化利潤。問題的限制條件為各期各產品從各廠的運輸量不超過工 廠的現貨量，及產品數量守恆。本文研擬遺傳演算求解法，在問題模型架構下研 擬整數型及實數型兩種染色體編碼法，分別研擬各自的交配法、突變法，並研擬 四種篩選法。最後展出最終運輸規劃結果。求解結果會顯示各期各產品從各廠使 用各交通工具運至各市場的運輸量。本研究並以C#程式語言在.NET Framework 的平台開發「遺傳演算為基的5MTP 求解系統」。本研究根據問題特性設定兩種 極端的情境及三種不同的市場價格下降率，共六種範例進行測試，並與以滿足最 大利益的市場需求的人工運輸規劃及隨機求解法進行比較，驗證本研究所提出的 求解效率及結果。

This paper presents a practical transportation planning problem for a multi-plant, multi-market, multi-product, multi-vehicle, multi-period transportation problem for price-reducing products. A mathematical model for the problem was rigorously defined and the goal to solve is to maximize the profit. Constraints on this problem include transportation amount of each period of each product of each plant doesn’t exceed its number limit and product number conservation. This paper proposed a GA based solving system for real number and integer chromosome encoding methods, and each has its own crossover and mutation methods. Each encoding method has four types of selection methods, which deploys the final transportation plan. The transportation plan reveals the numbers of each period using different vehicle from all of the plant to transport each product to each market. A prototype system, GA-based 5MTP Solver Planning System, implementing the proposed GA method was developed to test sample data. In this study, we set two extreme scenarios and three different market price decline rate based on problem characteristics, total of six samples tested. And compared our results with artificial transport planning which is aimed at the most profitable of the market demand and random solving method proposed in this study, in order to prove our method’s efficiency.