串聯主動電力濾波器又稱動態電壓補償器(Dynamic Voltage Restorer, DVR)，本文硬體採用三相四線轉換器的架構，其主要功能是用於解決電力線上發生驟降(升)以及其他電壓汙染的情形。為了讓串聯主動電力濾波器可以穩定的輸出補償電壓，本文設計一具有模糊推理及線上學習能力的模糊類神經網路(Fuzzy Neural Network, FNN)，可以更好的穩定串聯主動電力濾波器在直流端的電壓，改善傳統比例積分(Proportional-Integral, PI)控制器的響應速度以及控制精度，結合所提出的雙二階廣義積分器，讓負載端可以維持良好的電力品質。 本文透過Matlab Simulink模擬，比較傳統控制法以及所提出智慧型控制法的差異，並利用Opal-RT硬體迴圈測試(Hardware-in-the-loop, HIL)平台來驗證可行性。最後證實許多電力品質問題得到改善，並將負載電壓維持在標準水平。 ;Series active power filter is also called dynamic voltage restorer (DVR). The main hardware is constructed in the three-phase four-wire inverter. Its main function is to solve the sag (swell) and other power quality problem. In order to make the output of the Series active power filter output stable compensation voltage, this paper designs a fuzzy neural network (FNN) which has fuzzy inference and online training capability, can better stabilize the voltage at the DC link of the series active power filter, improve the response speed and the control accuracy of the traditional proportional-integral (PI) control method, then combined with the proposed Dual Second Order Generalized Integrator, that makes the load side can maintain good power quality. This paper uses Matlab Simulink simulation to compare the differences between the traditional control method and the proposed intelligent control method, and implement on the Opal-RT Hardware-in-the-loop (HIL) platform to verify the feasibility. Finally proved that many voltage quality problems have been improved to maintain the load voltage at a standard level.
