Date of Publication :22nd February 2018

Abstract: In the present era, the role of solar photovoltaic (PV) in the distributed generation has become inevitable and it is also the need of the hour. Addition of distributed generations (DGs) in the utility grid demand for decent power quality, secure operation and islanding protection of the grid interconnection. In order to maintain the quality of the power, one needs to accelerate the procedure of finding the fault, reduce the downtime and bring the system back to normal condition. Many diagnostic approaches were proposed in the past to identify the PV faults but they are old school methods and sometimes become unmanageable particularly in case of multiple faults and critical PV system. In this paper recognition and categorization of all possible DC faults of a grid-connected PV system using an artificial neural network (ANN), an artificial intelligence technique is presented. The simulation of ANN was done over 100 kW solar PV system connected to an 11 kV grid. Five inputs were fed to the ANN namely, PV voltage (Vpv), current (IPV) and power (P) (PV array parameters) and irradiance (G) and module temperature (T) (environmental data). Also, there were 5 output nodes as a DC fault indicator namely, Short-Circuit; Open-Circuit, Degradation, Shading and Charging Module. The optimized neural network architecture comprised of 5 neurons in the input layer, 20 neurons in the hidden layer and 5 neurons in the output layer. The hyperbolic tangent sigmoid transfer function was used as an activation function for the hidden as well as output layers. The ANN network was trained with over 1000 samples using back propagation algorithm with the accuracy of 0.01. To achieve the set error goal of 0.01 the ANN performance converges within the 1000 epochs. The neural network was tested for additional 1040 samples which were not included in the training data. The results of the tested data were obtained with the accuracy of 99%. It is found that the proposed system has proved its best with the accuracy of 99% for the practical applications when compared with the other artificial intelligence techniques like the fuzzy system, expert knowledge, etc.

Reference :