Condition Monitoring of Photovoltaic Systems Using Machine Leaming Techniques

Abstract

Condition monitoring of any system is essential to maintain its healthy operation as it results in getting maximum revenue with minimum maintenance and operation costs. The main objective of this paper is to develop a fault detection algorithm capable of classifying different faults that can be occur in a Photovoltaic (PV) systems. Output characteristics of the PV system are used as valuable information to observe various types of faults and their locations. Wavelet transforms and neural network systems were adapted to filter the non-significant anomalies and make it easier to detect faults that are to be taken care of in a timely manner. The neural network (NN) classification adapts Multilayer perceptron (MLP) to identify the type and location of occurring faults. Wavelet transform (WT) based signal processing technique is utilized in the feature extraction process to provide inputs to the NN. The developed detection algorithm is adapted for 24/7 automated surveillance. The developed algorithm achieved 98.2% accuracy when tested on a predetermined fault data set.