A novel intelligent detection schema of series arc fault in photovoltaic (PV) system based convolutional neural network

Abstract

Series Arc Fault (SAF) can be defined as the failure that occurs between any electrical contact and any electrical circuitry. However, it considered one of the common failures that affect the operation of the PV system and causes serious problems such as fires and electrical shock hazards. Several reasons increase the possibility of this type of failures such as incorrect installation, irregular maintenance, and some environmental effects. This paper presents a new intelligent and accurate detection method of SAF in the PV system. In this method, Convolutional neural networks (CNN) which is a discriminative (supervised) deep learning algorithm used for the process of fault detection. In normal cases, the signal consists of DC component, inverter component and noise of Network. In the case of SAF, a new component will add to the signal; therefore, CNN used to discriminate against the new component to accurately detect the SAF. PSCAD is used to generate the Arc fault model; Performance evaluation and the results of the proposed method implemented using Python. The achieved accuracy of the proposed detection method is 98.9%.