本論文旨在針對三相三線式UPQC之直流鏈電壓控制策略進行研究與優化。透過智 慧型控制演算法，解決傳統控制方法在面對複雜電力環境時的性能侷限。 隨著電力電子設備與各類非線性負載在現代電網中廣泛應用，系統極易出現電流諧 波污染及功率因數下降等電力品質問題；同時，市電側常見的電壓驟升與驟降亦會嚴重 影響終端設備的穩定性。UPQC作為一種結合串聯型與並聯型補償器的綜合電力品質解 決方案，能同時兼顧電壓與電流端的補償效能，而其直流鏈電容電壓的穩定度則是決定 補償性能的關鍵核心。 在直流鏈電壓控制方面，傳統的比例積分控制雖結構簡單，但在負載劇烈變動或電 網電壓擾動時，往往存在動態響應較慢、過衝量較大且穩定時間較長等缺點。為此，本 論文引入了徑向基函數神經網路與脈衝徑向基底混合型神經網路兩種智慧型控制策略， 並詳細推導其控制架構與學習法則。 本論文利用 MATLAB & Simulink R2021b 建立三相三線式UPQC系統模型，模擬 在非線性負載環境下，以及市電發生驟升與驟降之極端情境，針對比例積分、徑向基函 數神經網路與脈衝徑向基底混合型神經網路三種控制、演算法進行模擬分析與綜合評比。 實驗結果顯示，在改善電網電壓波動與抑制負載諧波的過程中，本論文所提出的脈衝徑 向基底智慧型控制法能較傳統比例積分控制更有效地穩定直流鏈電壓，具備更優異的動 態跟蹤能力與穩態精度，進而證實其在提升UPQC整體電力品質補償效能上的優越性。;This thesis aims to conduct an in-depth research and optimization on the DC-link voltage control strategy for a three-phase three-wire Unified Power Quality Conditioner (UPQC). By introducing an intelligent control algorithm， this study addresses the performance limitations of conventional control methods under complex power grid environments. With the widespread application of power electronic devices and various non-linear loads in modern power grids， systems are highly susceptible to power quality issues such as current harmonic pollution and power factor degradation. Meanwhile， grid-side disturbances such as voltage swells and voltage sags severely undermine the stability of terminal equipment. As a comprehensive power quality solution combining series and shunt compensators， the UPQC can simultaneously manage compensation performance at both the voltage and current sides， where the stability of its DC-link capacitor voltage serves as the critical core determining the overall compensation efficiency. Regarding DC-link voltage control， although conventional Proportional-Integral (PI) control possesses a simple structure， it often suffers from drawbacks such as slow dynamic response， large overshoot， and prolonged settling time under severe load fluctuations or grid voltage disturbances. To overcome these limitations， this thesis introduces two intelligent control strategies: the Radial Basis Function Neural Network (RBFNN) and a hybrid spiking radial basis function neural network(SNN-RBF)， accompanied by detailed mathematical derivations of their control architectures and learning rules.In this thesis， a three-phase three-wire UPQC system model is established using MATLAB & Simulink R2021b. Simulation analyses and comprehensive performance evaluations are conducted for three distinct control algorithms—namely PI control， RBFNN， and the hybrid spiking radial basis function neural network—under non-linear load conditions as well as extreme scenarios involving grid-side voltage swells and sags. Simulation results demonstrate that during the process of mitigating grid voltage fluctuations and suppressing load harmonics， the proposed spiking radial basis function intelligent control strategy stabilizes the DC-link voltage more effectively than the conventional PI control. It exhibits superior dynamic tracking capability and steady-state precision， thereby validating its superiority in enhancing the overall power quality compensation performance of the UPQC.
