基於微型相量量測器頻率資料偵測發電機跳脫事件及最低點頻率預測

Abstract

隨著台灣每年負載需求量上升，必須透過謹慎的機組調度才能度過夏天的用電高峰期，然而若此時發電機組非預期性的故障跳脫，電力系統一瞬間失去一部分的發電量，將會使電力頻率產生驟降並可能引發嚴重的停電事件。電力頻率代表著電力系統電力供需平衡，電力頻率太低會導致發電機機組連鎖效應地跳脫，致使電網崩潰全黑大停電，如同2003美加大停電一樣，但如果透過恰當的電力防衛機制，例如低頻卸載策略以及緊急啟動快速備轉機組，就能有機會避免全黑大停電的發生，2017年815大停電就是個例子，然而這些策略通常在頻率過低的情況下才開始啟動，因此本研究嘗試利用微型相量量測器實際量測台灣電網的頻率資料，自動偵測發電機跳脫事件，並預測系統最低點頻率來衡量事故的嚴重性，以利後續緊急控制與決策。 本研究提出偵測發電機跳脫事件方法，透過三個頻率資料點即時性判斷系統是否有發電機跳脫事件發生，接著，利用台電所提供的系統參數資料和本實驗室自主研製微型相量量測器所量測的頻率資料來預測事件最低點頻率。本研究方法能100%偵測出台電資料中最低點頻率低於59.8Hz的跳機事件，而在微型相量量測器資料額外偵測出的跳機事件中，有87.1%的事件可在誤差低於0.1Hz之範圍準確預測最低點頻率。

With the annual demand load increasing in Taiwan, it needs to be more cautious in dispatching generator units to get through the peak load in summer. If the generator units fail and trip unexpectedly at low spinning reserve situation, the frequency of the power system will decline quickly and cause a severe event. The frequency represents the power balance. In case the frequency is too low, the generator units will disconnect from the grid. Finally, it will lead to a power system blackout, like the North America blackout event in 2003. However, if it is treated with the proper load shedding strategy and the emergency control, it will be possible to prevent a blackout event, like the 815 Datan tripping event in Taiwan last year. The problem is that all these emergency strategies are adopted when the frequency has already been too low, so this research tries to detect the tripping event and predict its severity before the frequency drops to nadir frequency. The research hopes that the nadir frequency prediction brings benefit for the emergency control in the future. This thesis will first introduce a detection algorithm for tripping event. The algorithm uses three data points to achieve real-time event detection. Next, considering the system parameter provided by Taipower and the frequency data measured by μPMU to predict the nadir frequency. The detect algorithm can 100% detect the events which nadir frequency lower than 59.8Hz given by Taipower. If the error is lower than 0.1Hz, it represents the success of the prediction. Then 87.1% of the events that found by μPMU can predict the nadir frequency accurately.