- 學位論文
負載預測與契約用電最佳化之研究
Load Prediction and Contract Capacity Optimization Research
摘要
近年來綠色環保議題廣泛受到重視，有效率地使用能源成為永續經營之關鍵，資訊科技迅速進步，生活品質要求度高，伴隨著電能消耗量大幅提升，透過負載管理能夠改善當前用電供需端之現況，有鑑於此，電力公司提出簽訂用電契約來進行輸電端之負載管理，企業等大型用戶與電力公司簽訂用電契約，電力公司可以根據契約量預先進行電力調度及排程，減少電能之浪費及提升電力服務品質，用戶可以藉由簽訂契約，了解自身所需的電量並達到有效用電。訂定適當的契約量能夠使用戶有效地節省用電成本，過低的契約量造成罰款問題，導致用電成本提升；簽訂過高的契約量，雖可避免罰款但也因此造成不必要的電能浪費，亦造成用電成本增加。合理地評估負載量及契約量可以解決上述問題，因此本研究提出負載預測及契約訂定最佳化，其目的在於幫助用戶評估適當的契約量，維持用戶穩定用電品質下，並節省用電成本。由於粒子群演算法具備解空間資訊之多樣性，以及不受限於資料量限制等特性，我們透過粒子群演算法搭配時間序列模型進行負載預測，並透過隨機模擬機制進行調控，預測下一年可能的負載量，在最佳化模組部分，藉由詢問式學習概念提升粒子群演算法效能[1] 並計算全年契約量。本研究以實際案例進行實驗，其結果指出，本研究方法在負載預測中達約90%之準確度，在各產業領域中，在公營單位、商業、製造業及服務業準確度為90%、92%、90%、85%；根據實驗結果指出可幫助用戶在公營單位、商業、製造業及服務業，分別節省$$195,374、$4,031、$30,978、$39,905之用電成本。
並列摘要
In Taiwan, most industrial and commercial enterprises sign power contracts with Taiwan Power Company. Problems occur when deciding the capacity in contracts: the high power capacity leads to increase of total electronic consuming cost. However, if the power capacity is set low, consumers run the risk of high penalty when the actual consumption exceed. The aim of this thesis is to optimize power demand for Taiwanese industries through the model of forecast and optimization. This thesis presents a new combination method by using particle swarm optimization (PSO) to forecast the load capacity, and control uncertainty of forecasting with stochastic simulation. Then optimize the capacity of contract with the improved particle swarm optimization by query based learning (QBLPSO) [1] algorithm. There are two main purposes in this thesis. First, the proposed method will be compared with the other methods, and we analyze separately the forecast and optimization. Second, we make decision analysis framework for determining the optimal power contract capacity and an empirical study in real cases, which included the industry, the commerce. The load forecast has about 90% of accuracy in government units, 92% of accuracy in commerce, 90% of accuracy in manufacturing industry, and 85% of accuracy in service industry. And optimization model help user to save about $195,374 in government units, $4,031 in commerce, $30,978 in manufacturing industry, and $39,905 in service industry. Therefore, the result of experiment explain that this proposed method can help user efficiently to make appropriate contract capacity.