本研究提出一種應用於孤島型微電網(Microgrid， MG)之智慧型動態適應性虛擬阻抗控制策略，用以在負載切換條件下快速改善功率分配並穩定負載電壓。在微電網中，分散式電源(Distributed Generators， DGs)之間的精確功率分配，對於維持系統穩定性並提供負載所需功率而言相當重要。然而，在低壓微電網中，由於饋線阻抗不匹配，往往會造成各分散式電源之間產生不同程度的電壓降，進而導致功率分配不均與電力品質惡化。因此，本文提出一種適應性虛擬阻抗控制器(Adaptive Virtual Impedance， AVI)，以達成精確的功率分配。所提出之控制器可根據功率分配誤差動態調整虛擬阻抗，使其能夠適應不同的饋線阻抗條件。相較於傳統固定式虛擬阻抗方法，本文所提出之控制策略能夠在多台分散式電源並聯運轉的情況下有效運作。 由於傳統比例積分(Proportional-Integral， PI)控制器難以有效處理外部擾動與參數變化，本文進一步發展一種新型派翠機率赫米特模糊類神經網路(Petri Probabilistic Hermite Fuzzy Neural Network， PPHFNN)控制器，用於適應性虛擬阻抗之估測。此智慧型網路取代傳統 PI 控制器，以進一步提升功率分配性能。本文亦詳細介紹所提出 PPHFNN 之網路架構。此外，所發展之方法不僅能提升系統性能並改善功率分配的精確度，同時亦可實現實功率與虛功率之解耦。最後，本文透過包含兩組分散式電源單元之模擬與實驗驗證所提出之 PPHFNN 控制器的有效性，並在負載切換條件下，將其性能與未加入虛擬阻抗之傳統控制方法進行比較。;An intelligent dynamic and adaptive virtual impedance control scheme for an islanded microgrid (MG) is proposed in this study to rapidly improve power sharing and stabilize the load voltage under load switching conditions. Accurate power sharing among distributed generators (DGs) in MGs is essential to preserve system stability and deliver the required load power. However， mismatched feeder impedances in low-voltage MGs often result in disparate voltage drops， leading to unequal power sharing and deteriorated power quality. Consequently， an adaptive virtual impedance (AVI) controller is presented to achieve accurate power sharing. The proposed controller dynamically adjusts the virtual impedance based on power sharing errors， allowing it to adapt to various feeder impedances. Unlike conventional fixed virtual impedance methods， the proposed control scheme can operate effectively with multiple connected DGs. Since traditional proportional-integral (PI) controllers cannot effectively handle external disturbances and parameter variations， a novel Petri probabilistic Hermite fuzzy neural network (PPHFNN) controller is developed for the estimation of virtual impedance. This intelligent network replaces the traditional PI controller to further enhance power sharing performance. The network architecture of the proposed PPHFNN is introduced in detail. Moreover， the developed approach not only enhances system performance and improves the accuracy of power sharing， but also achieves the decoupling of active and reactive power. Finally， the effectiveness of the proposed PPHFNN-based controller is validated through simulations and experiments involving two DG units. Its performance is also compared with that of conventional control methods without virtual impedance under load switching conditions.
