- 學位論文
應用基因演算法於巨大廢棄物處理廠之隨機區位選擇問題
Applying genetic algorithms to the stochastic facility location problem for bulky waste recycling
摘要
巨大廢棄物包括廢棄的沙發、床鋪、桌椅、櫥櫃、腳踏車及修剪庭院的樹枝等，不同種類的巨大廢棄物所需處理的方式有所不同，而巨大廢棄物回收再利用以拆解破碎和修繕為主，本研究將巨大廢棄物處理廠內部分成多種處理流程，來處理不同類型的巨大廢棄物。為了增加巨大廢棄物處理廠的設備使用率，地方政府在設置巨大廢棄物處理廠時，應需考慮巨大廢棄物產生的種類與特性。而實際上，巨大廢棄物隨地區的不同，產生的數量與類型也不同。本研究在模式建構中，隨機產生巨大廢棄物的產生量情境，探討隨機產生量的區位選擇問題，並決定巨大廢棄物處理場內部處理流程的組合，另外，區位選擇問題為NP-Hard的問題，本研究使用基因演算進行求解，利用基因演算法的編碼特性，針對問題的類型，設計適合本研究的二元編碼，使用多種的交配機制與突變機制來增加染色體多樣性，避免陷入局部最佳解。並加入基因修補模式，來產生合適的處理流程數，使得染色體能符合容量限制要求，以改善求解品質。探討隨機產生巨大廢棄物產生量情境下的區位問題，並透過本研究設計的數值去證實演算法的有效性。在參數分析中，尋找合適的參數，以改善基因演算法的求解速度與求解品質。敏感度分析中，測試不同的參數對於區位選擇問題影響。根據測試結果，提出結論和建議。
並列摘要
Bulky waste embraces disused furniture, old bicycles, and lopping, etc. It depends on the kind of bulky waste to dispose them. The main disposition of bulky waste includes disintegrating, crumbling and repairing. A variety of procedures will be adopted in a recycling processing plant. In order to increase the capacity utilization rate of a recycling processing plant, the distributions of different kinds of bulky waste should be considered when the authorities determine the location of recycling processing plant. In reality, quantities of different kinds of bulky waste are very hard to forecast precisely. In this study, the uncertainties within the distributions of different kinds of bulky waste are regarded as scenario variables in the proposed stochastic location model of bulky waste. It decides whether a recycling processing plant should be opened and which disposition procedure should be introduced. In addition, a heuristic algorithm integrating genetic algorithm with scenario generation method is developed in this thesis because the stochastic location problem is an NP-hard problem. A binary coding is designed to represent the decision variables of the proposed model. In order to escape the local optima, genetic operators are conceived to increase the variety of solution combinations. Furthermore, a repairing strategy is addressed to improve the solution quality. A numerical example is utilized to demonstrate the validness of the developed algorithm. The corresponding parameter analysis of genetic algorithm is done to increase the quality and the speed of searching processes. Sensitivity analysis is performed to understand the influence of different parameters on the stochastic location problem. According to the testing results, some conclusions and suggestions are provided at last.