The Study Of Automatic Contingency Selection Based On Fuzzy Inference And Neural Network

Power system static security analysis is of great importance to the power system operation and planning, the main objective of which is to predict whether there are potential safety hazard by contingency analysis. AC power flow calculation is a traditional method which result is accurate, but it has a large calculation considered the each contingency in the fault set, so the method can not meet the requirements of online. Therefore, Automatic Contingency Selection (ACS) should become the first priority.ACS can select the contingencies which would endanger the safe operation of the system including exceeding the limit of active power or bus voltage. Also it can give a ranking according to the performance index. Only the contingency having a top rank should be taken into account so it can save a lot of time and increase the speed of security analysis.This paper applies the FIS and RBF neural network to the study of ACS, The main work is as follows:Firstly, a new active power performance index is defined which can reflect the severity degree of contingency, and a fuzzy compensation factor is included taking into account the masking problem when the index is not high.Secondly, combining fuzzy theory with the practical problems of contingency selection, A Mamdani fuzzy inference system is created to calculate the fuzzy compensation factor which is part of performance index. The main works of modeling include determining the input and output, selecting membership functions, setting fuzzy rules, selecting fuzzy inference algorithm and so on.Thirdly, a three-layer RBF neural network is constructed, which treats dynamo output, load power output and network topology as its inputs, while the active performance index is its output. The sample sets are obtained by off-line load flow calculation. In the end, suitable algorithms are chosen to train and test the neural network model.In the end, this algorithm is applied to IEEE-30 bus system and IEEE-118 bus system, the performance index of active power is calculated and ranked. Then this paper also compared and analyzed the results of the several methods, finally effectiveness of the algorithm is assessed.The results of the case show that the proposed algorithm has a high capture rate, and the calculation accuracy and speed are satisfactory, so it can meet the requirements of on-line.