- 學位論文
面板廠電力契約容量最佳化
Optimization of Electric Contract Capacities For A TFT-LCD Factory
摘要
目前全球產業已經面臨到微利時代的來臨，加上臺灣地窄人稠，四面環海，資源匱缺，使得能源開採或電廠的設置越來越不容易，造成電價不斷攀升，迫使企業不得不採取更積極的措施來因應，其中預估電力契約需量的決策就是企業降低成本的一個重要方向。 一般製造業者在向電力公司申請用電之初必須訂定契約需量，企業必須選擇最佳契約容量，避免廠內超約用電受罰或契約容量訂定過高而增加了基本電費支出，如果契約需量訂得太高，每月所繳交之基本電費形同浪費，反之，若訂定過低則將面臨2~3倍的罰款，兩者無形中皆會增加企業成本及未來的營運費用。 本文提出一個以倒傳遞類神經網路，建構一個電力需量之預測模型，以提供預測值供決策者作決策時參考，並以基因演算法進行選定最佳電力契約容量。本文以某一面板製造廠為個案研究對象，配合其所提供的數據作為實例討論與驗證其可行性。先以NeuroSolutions建構一系統進行推論運算，再以MATLAB GATool工具實際進行最佳解運算，研究結果發現藉由NeuroSolutions所得到的電力契約需量十分接近實際結果，以該預測值利用基因演算法進行最佳契約容量訂定之依據，可節省該廠不必要之電費支出。
並列摘要
At present the global industries are faceing a meager profit time. Taiwan is a small island and its resources are limited. The power plant construction becomes more and more difficult. That makes the electricity cost increase constantly. One of methods that an enterprise may reduce its cost is making optimal contract capacities. A manufacturing industry must stipulate the contract capacities before applying for the electricity. An enterprise must choice the optimal contract capacities to avoid the forfeit and increasing the cost. If the contract capacities is too high, the electricity costs becomes waste. On the contrary if the contract capacities is too low, the business must to pay twice or triple forfeit. It will be increase manufacturing industry cost and operation expenses in the future. In this thesis, a neural network for demand forecasting is proposed. Policy maker can use the result to make decision and use Genetic Algorithm to choice the optimal contract capacities. The thesis employs the real data from a TFT-LCD factory. The purpose of this study is usage of NeuroSolutions system to forecast the demand in future and use MATLAB GATool to determine optimal contract capacities. It is found that the simulation results of NeuroSolutions is very close to the actual result. A manufacturing industry can save the electrical bill cost if the proposed method is used.