本計畫建立一套配電系統快速復電之計算機輔助決策系統，當配電系統遭受意外事故時，用以輔助系統調度員進行最適之復電規劃。計畫中首先利用模糊因果網路模式建立配電系統在不同分歧負載及饋線容量情況下之多目標復電規劃，並據以建立網路訓練用資料集，再藉本計畫所提出之徑向基函數神經網路進行學習訓練。徑向基函數神經網路具有架構簡單及非線性輸入/輸出映射的特性，故非常適合於求解配電系統之復電問題。該網路中隱藏層中心節點神經元的數目及隱藏層至輸出層間的權值為決定網路學習效果的重要參數。本計劃中首先利用垂直最小平方演算法決定網路中隱藏層中心節點神經元的數目，以精減網路架構，再藉混合粒子群最佳化方法調整網路中隱藏層至輸出層間的連結權值，以快速、有效地求解配電系統之復電問題。當模式訓練完成，操作者能非常快速獲得復電規劃。本計畫以一典型台電配電系統為例進行研究，所提出的方法亦與類神經網路方法進行比較，以驗證其在實際應用上的有效性。The project aims at establishing a computer-based assistance system to help the dispatchers achieve an adequate restoration plan when the system suffers contingent events. To set up the training data, the fuzzy cause-effect network (FCEN) is first used to build up the multi-objective restoration plans based on diverse lateral loads and feeder capacity. A radial basis function neural network (RBFNN) is then proposed to train the obtained training data. Since the RBFNN has the characteristics of simple structure and nonlinear input-output mapping, it is very adequate to solve the problem of service restoration. The parameters of the node number in the hidden layer and weights between hidden layer and output layer are important for constructing the networks. To quickly solve the restoration plans reasonably and effectively, this project adopts the orthogonal least squares (OLS) algorithm to determine the number of center in hidden layer, which can simplify the network structure. Then the hybrid particle swarm optimization (HPSO) approach is used to tune the weights between hidden layer and output layer. Once the RBFNN is trained properly, the dispatcher can quickly acquire the restoration plans as soon as the inputs are given. The proposed approach has been tested on a typical Taipower distribution system. To verify the effectiveness of the proposed networks, comparison has been made to the existing artificial neural networks (ANN) method.