廢棄物掩埋場選址及清運分區空間模式

Abstract

廢棄物掩埋場選址與清運分區促進推動民營化，是當前重要的環保業務，與一般民眾的生活更是息息相關。本研究的主要目的是發展適當的空間優選模式及啟發式解法，協助掩埋場選址分析以及清運區域的劃分。 掩埋場選址需要處理大量的空間資料，地理資訊系統雖可迅速處理資料，但並不具有優選功能，加上在選址過程中，須考量不少的複雜準則，因此，本研究針對掩埋場場址中考量不同選址因子與空間性的差異，利用混合整數規劃(mixed-integer programming, MIP)發展單區空間優選模式，協助掩埋場選址分析工作的進行。本研究以美國北卡州橘郡的選址案例，比較所發展模式與其他研究提出的兩個MIP模式之效率，結果顯示所發展模式求解時間小於其他兩個模式的1/20。同時並以半真實的苗栗縣西湖鄉案例，探討該模式輔助掩埋場選址分析問題的可行性與適用性，根據分析的結果，模式可以藉由多因子改善只考量單因子可能的誤選問題。 針對範圍較大的掩埋場選址問題時，MIP模式往往需要冗長時間求解，本研究因而發展適當的啟發式解法來改善此問題，並藉由苗栗縣西湖鄉的案例，討論該方法的優缺點，根據執行的結果，求解最久的案例約耗費10 小時，多數是在數分鐘內解得，而解得答案之目標值均有最佳解的95%以上。 將清運區域適當的劃分為幾個分區，有助於形成自由競爭市場，易於推動清運民營化。然而劃分清運區域，需要考量成本及各分區之間的差異性等問題，且前述集中性定義並不適用此清運分區問題，因而進一步定義路線集中性，然而目前並無適當清運分區模式可供使用，本研究因而利用MIP方法，發展一空間分區優選模式。在新竹市香山區的案例中，即驗證了只考量單一因子的問題，結果顯示兩個分區雖然人口數差不多，但在運輸成本的不相當。藉由本研究發展之模式，可改善此缺點，求得差異較少的結果，而在南區與頂埔區的案例中發現路線集中性高的答案，總體成本低，但會導致分區間的差異大，決策者必須在多目標間決定折衷解。 大尺度的清運分區問題亦有難以求解模式的困難，本研究因而提出啟發式的解法，以改善求解的效率。以整合法可以減少求解的土地區塊，因而縮短求解時間；而均分法則可以在每一步驟中，減少分區的數目，因而減少時間，兩個虛擬的案例說明，根據該方法，確實能在短時間內，尋得解答。 本研究建立了掩埋場選址以及清運民營化分區的空間模式及啟發式解法，藉由這些模式以及啟發式解法的協助，可以彌補地理資訊系統欠缺的場址優選能力，改善選址分析及清運分區，並且在可接受的時間內，得到解答。

Landfill siting and municipal solid waste (MSW) collection subregion districting for prompting waste collection privatization are important responsibilities of local waste management authorities. These issues have also gained increasing attention from the general public. The goals of this study are to develop appropriate spatial optimization models and heuristic methods to improve landfill siting and MSW collection subregions districting analyses. Siting a landfill typically requires processing of massive spatial data. A geographical information system (GIS), although capable of manipulating spatial data, generally lacks the ability to obtain an optimal site, especially when multiple siting rules are simultaneously considered. In this study, a spatial mixed-integer programming (MIP) siting model with consideration of varied siting factors and spatial property was developed to facilitate the landfill siting analysis. A comparison between the model and two other models was presented with a real case (Orange County, NC, USA) to demonstrate the efficiency of the proposed model. According to the comparison result, the total solving time for the proposed model is less than 1/20 of those for two other models. Application of the model for the Shihu County was implemented to explore the model applicability for a semi-real world case. With a multi-factor analysis, the model can avoid the mis-selection problem that may be encountered when only a single factor is included. Solving the MIP model for a large landfill siting problem generally require long computational time to accomplish. A heuristic method was therefore proposed to overcome this difficulty for a large siting area. A case for Shihu County was presented to demonstrate the effectiveness of the heuristic method. Results show that the longest solving time is about 10 hours and the worst objective value among all solutions still have 95% of the optimum. The heuristic method is believed to be practical. A competitive market is an essential factor for successful MSW collection privatization. Properly dividing the entire MSW collection area into several subregions is capable of prompting the competitive market. Factors such as cost, subregion variance, and region compactness should be evaluated in districting. However, the compactness definition used for landfill siting is not appropriate for this problem because MSW is collected along the road network instead of the entire region. Therefore, a road compactness index (RCI) was proposed. Currently, a model specifically developed for MSW collection subregion districting is not available. An MIP model was thus developed to resolve this districting problem. A real case for Shansan District in Hsinchu showed the bias problem for only a single factor being considered. For the two subregions of the District, populations of both subregions were almost the same, but the transportation costs were significantly different. The proposed model can avoid this bias problem by minimizing the difference among sub-regions. In the other case study for South District and Dingpu District, the result illustrated that solutions with a high RCI implied low total cost and high variance among subregions, and vice versa. A further decision-making analysis to compromise the two objectives for determining the final solution is necessary. The MIP districting model may need long time to solve for a large problem. Two heuristic methods was thus proposed to improve this difficulty. The Aggregation Method can significantly decrease the number of land parcels. The Average-Division method reduces the number of subregions in each step. Two pseudo cases were discussed to demonstrate the applicability of the methods. In summary, appropriate spatial models and heuristic methods for landfill siting and MSW collection subregion districting were developed. The models or heuristic methods can be used with a GIS to improve its optimization ability. With the assistance of the models and methods, landfill siting and MSW collection sub-region districting analyses can be significantly improved and solutions can be obtained within acceptable computational time.