The Research Of Static Security Assessment Of Power System Based On Binary Tree SVM

Power system static security assessment is one of the most important problems which relatepower system secure-stable performance. The development of modern intelligent method hasprovided favorable conditions to it. Static security can be rapidly assessed using the modernintelligence technology. Support vector machine (SVM) is such a kind of machine learningmethod. SVM is based on statistical learning theory, which effectively solves the problems ofsmall sample size, high dimension, nonlinear, local optimal solution and so on. On the basic of theexisting intelligence methods of SSA, this paper does some in-depth research on the SVMalgorithm and its application to SSA.The main innovations in this paper are as following:1. A new Decision Tree (DT) based feature selection algorithm which applies to the expectedaccident analysis of power system is proposed. The features in the nodes of DT appear indescending order of their importance. When analyze the features we can select the relativeimportance ones according to the decision tree to solve the problem of choosing features of powersystem.2. A Approximate Complete Binary Tree Support Vector Machine (ACBT-SVM) whichbetween the partial binary tree and complete binary tree is proposed. The basic types of samplesin the feature space are distributed unbalanced and each cluster center are not equidistant betweeneach other. Then the ACBT-SVM algorithm which has a better geometric significance solves thisproblem.3. Due to the large number of power system, this paper proposes an improved simplifyingBT-SVM method. It combines the concept of equivalent distance, decreases the number oftraining samples and reduces the sample size. This method can simplify the classification scale ofSVM and improve the classification speed and accuracy.