Decision Tree Based Methods For Voltage Stability Assessment Of Power Systems

With the development of the Voltage stability research in power system, assessing the stability of voltage rapidly and accurately has become a challenging task. For the problem about Voltage stability directly affects the security and stability of the power system.In this paper, we firstly introduce those concepts and methods about decision tree, neural network and fuzzy neural network. Then, we analyze the voltage stability respectively with those above three approaches. In the end, we design three kinds of models including decision tree (DT) model, neural network decision tree (NN-DT) model and fuzzy neural network decision tree (FNN-DT) model.This paper takes angles difference of some important nodes as candidate attributes. According to historical data, we design DT and complete predicting the voltage stability under new unknown running states. Furthermore, with some simulated examples, the analysis of voltage stability based on DT method proposed in this work shows effective results.Because of the most of misclassified cases happen on certain nodes when we analyze voltage stability with DT method, two improved methods based on NN-DT and FNN-DT algorithms are presented in order to decrease the misclassification of DT in this work.About the NN-DT method, first, this paper builds up a NN-DT model, which is more suitable to predict voltage stability of power system. Then, with a simulated case, we validate that the improved method may do effectively in assessing voltage stability. The result shows that the method has lower misclassification than traditional DT method.Another improved method of FNN-DT presents the FNN-DT prediction model, which is suitable to voltage stability of power system, and validate that it may decrease effectively misclassification by simulating.In conclusion, all these three methods, such as DT, NN-DT and FNN-DT, may assess the voltage stability of power systems rapidly and accurately. The classification accuracy of DT is better than 90%. And the NN-DT and FNN-DT may increase the classification accuracy to a different degree, over 95%.