An Enhanced Frequency Analysis and Machine Learning Based Approach for Open Circuit Failures in PV Systems

Abstract

Over the last decades, the accelerated implementation of photovoltaic systems (PVS) has led to the creation of open circuit fault detection systems based on measurements made in completed facilities, growing by making the volume of data to be analyzed with each new installation, improving fault detection and location systems with various methods. In this article, an electronic adaptive device was developed that operates under a method based on the spectral analysis of signals using the Discrete Fourier Transform (DFT) and a classifier based on the k-Nearest Neighbor (k-NN) machine learning algorithm (ML) for the detection of Open Circuit Faults (OCF). The contribution of this work is that the entire photovoltaic array operated in conditions of radiance less than $10~\frac {W}{m^{2}}$ overnight with a red LED pulsed light applied on the photovoltaic array module furthest from the inverter. Under these operating conditions, the presence of an open circuit fault alters the variability in the impedances of the photovoltaic array under different fault locations in the systems compared to healthy systems without an open circuit fault, revealing that the predictability of the methodology shows values from 90% to 93% as the size of the photovoltaic system increases, concluding the effectiveness of the procedure.

Keywords

• Fault detection in photovoltaic systems
• open circuit fault detection
• discrete Fourier transform
• KNN algorithm
• dynamic impedance in photovoltaic systems