Feature selection filter for classification of power system operating states

Abstract

The classification of power system operating states plays an important role in power system control and operation. Determining the state of a power system is crucial and requirements for real-time decision making in power system security assessment demand low dimensionality and low computational time. This paper investigates the benefits of using feature selection based on mutual information in power system state classification with machine learning. The AdaBoost algorithm is used for classification based on large training datasets and feature selection is applied in order to reduce their dimensionality. The selection is implemented as a filter in the pre-processing stage of AdaBoost and uses genetic algorithms to perform the search with the fitness function computed based on mutual information. The proposed method is tested on the IEEE New England 39-bus network and a comparison between the learning algorithm performances with and without feature selection is provided. Results for different genetic algorithm parameters are also presented.