本論文針對輸電系統作最佳火力機組調派,一般火力機組調派主要是規劃各發電機組於每小時的狀態及發電量,在滿足各項限制條件下使系統總發電成本最小化。 本論文首先考慮在輸電線損失的情況下得到系統的損失係數,經由考慮包含損失的經濟調度可得各機組的發電量,然後利用蟻拓演算法作其最佳發電排程。蟻拓演算法為一啟發式的演算法,其精神主要是模擬螞蟻在尋找食物的過程中透過費洛蒙來達到知識分享以及協同合作的間接溝通,利用隨機分散搜尋方式有效地解決大型的分群問題,可避免太早收斂而侷限於區域最佳解中,對於較多螞蟻選擇的路徑,會累積較多的費洛蒙濃度,吸引更多的螞蟻前來,這個機制能夠幫助我們有效且快速的取得相當好的解答。 本論文利用提議的方法對IEEE 30-bus系統與實際台電系統做模擬,以一天24小時做機組調派,並且將調度結果和螞蟻系統與動態規劃法做比較,經由實驗可得更佳之發電成本,期望能輔助調度人員做出更經濟之調度。
This thesis aims to examine optimal thermal unit commitment of transmission system. In general, the major purpose of thermal unit commitment is to schedule the on/off status and the real power outputs of units at each hour and to minimize the system’s total production cost while satisfying each constraint. This thesis first considers transmission line loss to obtain the system’s loss coefficients, then takes into account of the economic dispatch that includes losses to obtain the output of each unit, and apply Ant Colony System (ACS) to acquire the most favorable generation schedule. ACS is a heuristic algorithm based upon the indirect communication of ants mediated by pheromone trails to achieve knowledge-sharing and collaboration during their food-seeking process. Stochastic search is adopted and performed to effectively solve problems about large-size clustering and to avoid early convergence that may provide at its best local optimal solution only. The route chosen by a substantial number of ants is very likely to attract more ants with its greater density of accumulated pheromones. This mechanism can be applied to help us obtain answers in an effective and efficient manner. This thesis simulates the proposed method for 24-hour unit commitment on an IEEE 30-bus power system and an actual Taipower system. The dispatch solutions are then compared to their counterparts obtained by Ant System (AS) and Dynamic Programming (DP) to verify the proposed method as a better approach for achieving the optimal cost of power generation. The proposed method can therefore be expected to help dispatchers perform more economical dispatch.