Application of pattern recognition technique to power system security analysis and optimization

The pattern recognition method is applied to the related power system problems of security analysis and security constrained optimum power dispatch. Various techniques, such as box data creation and wafer structure, the method of search, short-term load trend prediction, feature reduction, and on-line training set creation, have been developed to achieve high evaluation accuracies as well as computational savings. The formulation of the method used is discussed, and several study cases are presented to demonstrate the feasibility of its possible implementation in real systems. These study cases include an application of the method to evaluate post-contingency voltage prediction in the Hong Kong transmission network and the optimum power dispatch in typical IEEE networks