Artificial neural network-based contingency ranking and security constrained optimal dispatch.

摘要

In this dissertation work, Artificial Neural Networks have been applied to the problem of Static Security Enhancement of Power System. Static Security Enhancement essentially addresses the problem of ranking the potential contingent elements and providing preventive strategies based on security-constrained dispatch.;This new approach has been tested on the IEEE 118-bus test system. The contingency ranking obtained in the proposed method is comparable with the ranking provided by the Performance Index method. The proposed method is faster than the method based on the PI method. Once the proper training of the networks has been achieved, the new rankings are almost instantaneous while testing the trained networks.;A new method based on Hopfield Neural Network is also proposed for solving Security-Constrained Optimal Dispatch. The Hopfield network is used for the solution of the constrained optimization problem. The main idea behind solving the optimization problem is to formulate an appropriate computational energy function E(X) so that the lowest energy state would correspond to the required solution of X. The proposed method is based on transformation of the energy function minimization problem into a set of differential equations. The method is used in the active and reactive security-constrained dispatch problems with illustrations on four test systems a 6-bus system, the IEEE 14-bus test system, the IEEE 24-bus reliability test system and the IEEE 118-bus test system. Results obtained in the proposed method are comparable with the results obtained in the method based on dual-linear programming formulation.;Two three-layered perceptron networks have been used in the ranking of Line Flow contingency and Bus Voltage contingency. The training parameters for the two networks are chosen by the methods based on a regression-based correlation technique. Four new indices, two Severity Indices and two Margin Indices have been defined for ranking purposes. The Severity Indices perform better than the classical Performance Indices (PI) in the training of the networks.