Reinforcement learning based Islanding detection technique in distributed generation

Abstract

As the scientific, technical, and economic developments of the world continue to advance, there will be an increased requirement for Distribution Generation(DG) technology. There is a widespread disruption in the primary power grid, the procedure known as islanding involves the construction of a power island that operates in a manner analogous to a section of the utility system. The Islanding Detection (ID), in which the development of islanding, actions carried out during islanding, and the approaches utilized to recognize islanding are detailed, is entirely necessary for this study. In this paper the author used reinforcement learning method which is a technique for introducing machines to learn by rewarding appropriate conduct and/or penalizing inappropriate behavior. The Remote technique, the Local approach, and the Hybrid approach were all observable in the Islanding detection methodologies. Passive islanding detection algorithms for inverter-oriented distributed generation systems based on Variational Mode Decomposition (VMD) and the microgrid approach are implemented throughout this research attempt. According to the comparison results of the study, the proposed system is more reliable and better than the compared technique. The comparison results show the detection time of suggested model which is 0.06 s. In​ contrast to active techniques, it is capable of functioning normally and without disrupting the usual operation of the system. As a result, it can be used effectively for real-time applications.

Keywords

• ID (Islanding Detection)
• DG technology
• Point of Connection (PoC)
• VMD (Variable Mode Decomposition)
• PLL (Phase Locked Loop) and microgrid