Machine Learning Based Islanding Detection During Blackout

Abstract

In modern era, demand of electricity is increasing rapidly due to fast development in industrial sector, communication technologies, commercial and domestic usage. The need of the hour is to protect the power system from blackouts during abnormal conditions, like voltage collapse and frequency disturbance etc between the generation and consumption. The sudden change in weather conditions, loading of power system, and over ageing of transmission and distribution lines becomes the cause of blackouts. Islanding detection and protection is one of the best possible techniques to improve the power quality, reliability, system security, transmission congestion relief and efficient energy utilization. Distribution generators (DGs), if properly placed have many positive impacts on the operation of power system. The islanding can improve the reliability by counting the operation of electrical power to keep the balance between consumption and generation. The main problem is sustained islanding with fast timing. Plant should sustain its island mode condition when fault occurs in transmission lines. Most islanding detection algorithms suffer from large NDZ and Non detection zone must be minimized to protect system properly and avoid blackouts. This study presents a technique to develop a reliable system to prevent complete blackouts, while maintaining the stability of the power system with high variability and uncertainty under the abnormal conditions. It is also needed to present the period all the DGs need to connect and disconnect immediately when the island occurs. For this purpose, each DG must be equipped with islanding detection devices. In this research work the presented system is a two-step process, based on the collection of some of the key characteristics in the current and voltage signals to analyze second harmonic by the Fourier transform (FT).