- 學位論文
應用文化差值演算法訂定自備發電機組用戶之最佳契約容量
Cultured Differential Computation Algorithm for Optimal Contracted Capacity of Power Consumer with Self-Owned Generating Units
摘要
為減少電壓驟降及斷電事故造成之傷害，大型電力用戶多會自備發電機組，以提升供電品質。在考慮台電電費結構及自備發電機組發電成本下，台電電費支出及自備發電機組運轉方式與所訂定之契約容量息息相關。有鑑於此，本文研究主旨乃在如何制訂最佳契約容量，配合自備發電機組運轉，使用戶總用電成本為最低。 差值演算法透過突變、交配及選擇操作，在搜尋最佳解時具有快速及強健的收斂性。文化演算法則可擷取並儲存進化過程中從個體中所獲得之專門知識或問題特性，在差值演算法之突變操作中加入文化演算法的專門知識，可使得解之搜尋更有效率。因此，本文乃提出文化差值演算法作為求解最佳契約容量的方法，達到節省總用電成本之目的。 為驗證所提方法之可行性，本文使用某光電廠實際用電數據資料，包括台電用電度數、自備發電機組容量與燃料及運轉成本函數，以及欲制訂契約容量月份的負載需求預測。由實際系統結果證實，比較文化差值演算法及該廠現行方法，總用電成本可節省比例為14.56%，本文所提方法和其他現有最佳化演算法比較，文化差值演算法運用於自備發電機組契約容量訂定亦具有最佳化之效果。
並列摘要
To reduce the damages caused by voltage sag and interruption events, many customers have their self-owned generating units in attempt to improve the power supply quality. With the power tariff structure of the utilities and the cost functions of self-owned generating units considered at the same time, expenses due to the utility power consumed and the operation of self-owned generating units are highly related to the contracted capacity. Taking into account the corresponding operations of self-owned generating units, the thesis is thus aimed at determining the optimal contracted capacity with the utilities to obtain the lowest total power expenditure. The differential computation algorithm provides fast and robust converging characteristics in searching the optimal solution through operations of mutation, crossover, and selection. The cultural algorithm can extract and save the domain knowledge or problem properties during the evolution process. The domain knowledge in cultural algorithm can be added to the mutation operation in differential computation algorithm to make the searching more efficient. Accordingly, the thesis proposes the cultured differential computation algorithm to determine the optimal contracted capacity in order to reach the goal of saving the total power expenses. To verify feasibility of the proposed method, the thesis employs the real data obtained from an optoelectronics factory in Taiwan, data which include the amounts of power consumption from the utilities, capacities and cost functions of self-owned generating units, and load demand forecasting in the months of planning period. It is shown from the simulation results that 14.56% of electrical power expenses can be saved from the proposed cultured differential computation algorithm as compared with the method currently adopted by the factory. Also, in comparison with the other optimization methods, the proposed approach has superior results to the other existing optimization methods as revealed in the numerical results.