Modelling and Forecasting of 5MW Solar Photovoltaic Plant using Artificial Neural Networks

Abstract

This work illustrates Artificial neural network (ANN) based energy modelling and forecasting of a 5MW Solar Photo Voltaic Plant. The objective of the research work is to identify the impact of various atmospheric parameters on the Solar PV energy generation. The Simulation was performed using Neural network(nn) tool box in MATLAB simulation. The input parameters are total solar radiation, surface pressure, wind speed, Insolation Clearness Index, cloud amount, shortwave diffuse irradiance and total insolation index which were collected from National Renewable Energy Laboratory (NREL) database. The output parameters were collected from five-year production data of 5 MW solar PV plant in Tamilnadu, India. Six different network structures namely Cascade forward backdrop, ELMAN back prop, Feed forward back prop, Feed forward Distribute Time delay, Layer current and Regression Analysis were used. Individual models have been developed using above ANN structures and results are analysed. The model validation has been performed by comparing model predictions with power output data during the testing. Single input weather parameters have been fed and tested to find the impact of individual weather parameter on the power output of the plant. Multiple input parameters are fed and tested for improving the model. Regression analysis has shown a better performance comparing with other networks. These solar forecasting techniques help to design the electrical storage management and plan maintenance activities of smart grid system.