Study On The Multi-agent Based Distributed Fault Section Estimation System

Upon fault occur in grid, there will be a large mount of alarm message, so it is difficult for operator to make correct judgment. On the basis of analyzing the existing diagnosis methods, this thesis presents using the RBF neural network as the core algorithm. The test on a 4-bus system prove that RBF NN is very suitable to fault section estimation. For a large-scale grid, the training samples for RBF NN will mount in geometrical speed with the increasing of input and fault has local nature. In order to reduce the training difficult of RBF NN and enhance the reliability of diagnosis system, based upon the thought of "divide and conquer", we first divide the grid into desired subnets, and then develop the distributed fault section estimation system. The test on the 118-bus system proves that this system is effective in fault section estimation.