Pattern recognition applied to transient stability analysis of power systems with modelling including voltage and speed regulation

A pattern recognition approach to analyse transient stability using a higher order model is presented. A six order model based on two axis representation for generators including voltage and governor actions is described. Two pattern vectors based on physical considerations about the system behaviour are derived and several approximations on system trajectory are considered to deduce analytical formulas which can be computed quickly. To alleviate computational time, two discriminant analysis methods allowing reduction of pattern vectors are presented. These results are applied to an eight machine system with a three phase fault. The discriminatory power of the pattern vectors proposed is tested with the Bayes classifier and the k nearest neighbour rule. A comparison is given