Development of suitable Data Mining Algorithm for Power System Fault Detection using Machine Learning Approach

Abstract

A Modern power system (PS) is a complex interconnected network which includes four different parts such as generation, transmission, distribution, and loads. Since many advanced measurements and protection instruments are equipped with modern power system, a huge amount of data has been collected with larger dimensionality and quantity. To utilize the long previous data efficiently for post fault analysis, large capacity memory devices are required at PDC level. Accurate prediction. Past literarature reports indicates that the traditional analytical methods which are presently used in practice are not accurate and also not providing enough speed. Therefore, Machine Learning algorithms are used to create emulation of power system.   These algorithms are used to make fast and accurate decisions. Here an attempt is made to implement various Machine Learning algorithms such as K-Nearest Neighbour (KNN), Decision Tree (DT), Support Vector Machine (SVM) on power system fault data collected by PMU from Indian power system for protection. Here an attempt is made to check the performance of different Machine learning algorithms using different normalization techniques such as standard scaler and Minmax scaler which helps to extract the useful information for improving situational awareness in power system. .