Support Vector Data Description (SVDD) based Inverter Fault Diagnostic Method

Abstract

The grid connected Photovoltaic (PV) systems have attracted significant attention in the recent years, thus, the reliability of grid-connected inverters has become a major concern. Any kind of failure at the inverter side may severely affect the output power leading to instability at the grid end. To improve the system availability and reliability, an inverter fault diagnostic mechanism becomes imperative. This paper proposes novel Support Vector Data Description (SVDD)-based fault detection technique for grid connected single-phase PV Inverters. It possesses the ability to address the problem of anomaly detection. It seeks to find the hypersphere with minimum volume containing most of the relevant fault-free data objects. With the occurrence of fault, the faulty data points will be the outside of the hypersphere. The fault detection and classification algorithm is developed in MATLAB/Simulink environment. The simulation results verify the effectiveness of the proposed technique.