摘要 面對石化燃料價格不斷上漲的趨勢,以及國際溫室氣體排放限制等問 題,促使人們正視電力結構必須調整的問題,逐漸由高汙染的發電方式改 變為採取低污染或無污染的方式進行。然而,能源結構與產業政策的調整 需要十到十五年的時間,期間充滿許多不確定因素的影響,故本研究應用 依時性隨機規劃模式探討在二氧化碳排放量限制下之電力系統擴增計畫問 題,一併考慮電力需求、石化燃料價格、水力及風力發電效用等不確定因 素於電力系統擴建計畫中,配合情境分析方法,規劃各電廠、各種發電型 態(包括火力發電、水力發電、核能發電、風力發電等)未來之裝置容量。因 為本問題為NP-hard,本研究發展出一以和聲搜尋機制為基礎之啟發式演算 法進行求解,並透過數值例測試證明此演算法之正確性。經由參數分析進 行此和聲搜尋法之參數設定,以提升其求解品質與運算速度。此外,藉由 敏感度分析了解採用太陽能發電系統、生質能發電系統及高污染發電機組 退休等因素對電力裝載計畫之影響。最後,則根據本研究之測試結果提出 一些結論與建議。
Abstract The raising price of fossil fuels, the uncertainty of power demand and supply in future, and the international limitation of greenhouse gas emission by Kyoto Protocol prompt people to face the problem of transferring the structure of power system. Non-polluted or Low-polluted power generating should replace high-polluted one in the long run. However, it will take ten to fifteen years to adjust the structure of energy systems and industrial strategies because of much uncertainty. This thesis proposes a time-dependent stochastic programming model to formulate the power expansion problem under limitation of greenhouse gas emission. The uncertainty of producing cost, green power supply, and regional electric loads in future can be considered simultaneously. By using scenario analysis to present environment parameters, we can determine the appropriate capacity allocation among different methods of power generation, such as thermal power generating, hydraulic power generating, nuclear power generating, and wind power generating. Because this problem is NP-hard, this study develops a meta-heuristic algorithm based on harmony search to solve it. Several numerical examples are utilized to demonstrate the validness of the developed algorithm. The parameter analysis of this harmony search-based algorithm is done to increase the quality and the speed of searching processes. Sensitivity analysis is performed to understand the influence of increasing the renewable power generation, and decreasing the capacity of high-polluted electric generating sets. Finally, some conclusions and suggestions are provided according to the testing results.