Research On Pattern Recognition Based Power System Transient Stability Assessment

Contingency screening is a vital segment of power system transient stabilityassessment (TSA). In this dissertation, methods based on pattern recognition areused for contingency screening. The Selecting of contingency features andclassifying methods are the main contents of this paper.First, this paper summarizes a contingency feature set for transient stabilityclassification. At present, most contingency features, which are used for reflectingthe stability of system, are obtained by time-domain simulation. These features onlyfigure out the system state information at the time of simulation finished, but theycan't tell the distance between the state-point and the stability region boundary. Forthis reason, this paper chooses features that combining the transient margin with thepost-fault system state information.Back-propagation neural network (BPNN) method is used in transientcontingency screening. Results have shown that the features selected in this paperhave efficiently improved the contingency screening accuracy. Another methodbased on clustering analysis is also put forward. This method is based on Fuzzyc-means (FCM) and Vector Quantization (VQ) Method. Case studies on the10-machine New England system and IEEE 50-machine system are given to showthe validity of this method.