Disturbances Identification by Using Machine Learning Algorithms

Abstract

The quality of power and power interruptions are issues that users and power distributors are becoming more concerned about. The degradation in the quality of power comes from any disturbing phenomena that cause the mains voltage (or current) wave to depart from its nominal characteristics and are called disturbances. Identification of Power Quality Disturbances (PQD) and reliable PQD categorization are therefore particularly desirable. Additionally, identifying and categorizing PQD in distribution networks are important tasks for protecting power distribution networks. The most of disturbances are non-stationary and transient in nature, necessitating the use of advanced methods and tools for PQD analysis. The proposed method builds up to find the best model from Machine Learning (ML) classification techniques. Real three-phase voltages and frequency values are used to train ML, and according to the measured three-phase parameters, ML can identify and classify the disturbances event and find the best technique for that.