本文探討在系統阻抗變化及諧波電流不確定因素下之工業用戶大容量電 力諧波濾波器之規劃，藉由使用最佳規劃與機率法來解決困難之求解問題，有 效改善諧波污染，以提升工業用戶之電力品質。由於在工業用戶中裝設大容量 被動濾波器改善諧波問題，必須考量諧波管制標準、虛功率補償及濾波器損失 限制條件，故形成非線性不可微分性質最佳化問題，以傳統嘗試法難以處理。 有鑑於此，本文先使用反應曲面法結合非線性時變粒子群演算法求解濾波器規 劃問題，反應曲面法能快速收斂並找尋出最佳解，建立濾波器系統模型，並考 量系統阻抗及諧波電流之隨機不確定因素，但爲了避免陷入局部最佳解，再結 合非線性時變粒子群演算法爲基礎的隨機捜尋法，使得求出全域最佳解之機會 大增。最後，爲驗證所提方法，本文使用工業用戶化工廠濾波器設計實例，使 用整流器的非線性負載。由模擬結果顯示，本文提出的方法能快速決定濾波器 容量，有效抑制諧波，並避免系統共振問題。

This paper provides the study of large-scale industrial harmonic filter design problems using optimization techniques and probability-based methods, considering system impedance and harmonic currents uncertainties.
The solutions of difficult optimization planning problems are achieved. It is indicated that the proposed methods not only reduce harmonic pollution but also improve power quality. In the passive harmonic filter design of the applications, this method can determine optimal filter sizes. The purposes are to minimize the total demand distortion of harmonic currents. Filters loss, reactive power compensation, constraints of individual harmonics are also considered. These are difficult to solve by conventional methods since the solution space is extremely broad. Owing to the ability dealing with optimization problems, the paper presents a combination of response surface methodology (RSM) and particle swarm optimization with nonlinear time-varying evolution (PSO_NTVE) to investigate the planning of large-scale passive harmonic filters. The basic strategy of RSM guides the search point to local optimum and the PSO_NTVE escapes from the valley of the local optimum in order to arrive at a global optimum. Finally, in order to verify the proposed methods, case studies are taken from a chemical plant where DC arc furnaces are installed. The results show that the proposed methods can quickly determine filter capacitors, reduce harmonic pollution, and avoid amplification.