Bidirectional Power Flow in DC Microgrid and its Islanding Detection Using Support Vector Machine

Abstract

With the over increasing demand on the use of renewable and dependence on electronic loads, electric vehicles and energy storage devices, DC microgrid has gained more and more importance over the past few years. In this paper, we discuss some of the aspects of a DC microgrid which includes its behavior during Islanding of the utility grid. For that, a two bus system is being implemented which contains a solar photovoltaic (PV) cell on the DC subgrid. The AC and DC bus is being connected through a bidirectional AC-DC converter which allows full controllability on voltage and bidirectional power flow in both buses. A Sinusoidal pulse width modulation (SPWM) technique is being implemented which allows independent control of active and reactive power flow between the buses. Hence the AC-DC converter posses the capability of Unity power factor operation. The DC-DC buck-boost converter is being utilized to hold a steady voltage at the DC bus. A fixed voltage at the DC bus is crucial to supply consumer load at a constant voltage. This paper presents a new machine learning based islanding detection method for a two bus DC microgrid. The proposed method uses multiple features and Support Vector Machine (SVM) classification. The method uses a set of features extracted from the voltage parameter of the DC subgrid extracted from the simulation of the proposed network under various operating conditions and power imbalance. The obtained features under islanding and non-islanding event are used to train the classifier. The results thus obtained shows a high amount of accuracy in islanding detection even under minimal power imbalance.