本文旨在實現一微控制器監控微電網於併網模式下的電壓，當微電網發生電力品質異常時，能夠即時令微電網跳脫市電，以避免電力異常造成的劇烈電壓擾動。使用離散小波轉換，分析訊號包含的暫態變化，並經由Parseval定理獲取該變化之能量值，最後將這些特徵數值輸入至決策法，輸出導通或截止的命令，以此作為本文所探討的專家系統。在專家系統中嘗試採用了模糊控制、倒傳遞類神經網路，與模糊倒傳遞類神經網路為決策法，並以容許值檢測法作為對照。且考慮小波轉換演算容易被訊號中的雜訊汙染所影響，可能導致專家系統誤判，因此，如何快速地偵測及避免雜訊影響將是本文著重的課題。 透過Matlab軟體預先模擬各類電力異常事件，並分析其特徵值變化，以建構適用的專家系統，之後輸入電壓波形訊號進行模擬測試，比較模擬的系統輸出，來找出最佳的系統架構。於模擬結果中可以發現，模糊倒傳遞類神經網路為決策法的專家系統具有較佳的表現，不但能容許雜訊波形且能有效檢測出電壓異常。 ;This thesis aims to implement a microcontroller to monitor voltage signal in microgrid, which operate in grid-connected mode. If there is abnormal voltage at the point of common coupling (PCC), it can trip the microgrid instantly to avoid voltage disturbance caused by the fault and protect the power electronics equipment. After estimating the voltage signal through discrete wavelet transform (DWT) and the Parseval’s theorem, we enter these eigenvalues into decision method to output ON/OFF command, as expert system. In this paper, the fuzzy control, back propagation neural network (BPNN), and fuzzy-back propagation neural network are used as the decision-making method, and the threshold value detection method is used as a control. Considering that DWT is easily affected by the noise pollution in the voltage signal, to overcome the misjudgment cause by noise, which makes the microgrid trip from the grid in general situation. Through Matlab software, we simulate the voltage signal of power fault event, and analysis the eigenvalues to set up the expert system. And test the proposed systems. By observing the output command, we found that the expert system of fuzzy-back propagation neural network has better performance. It not only allows noise waveforms, but also effectively detects voltage anomalies.
