Very short term prediction of solar radiation for residential load scheduling in smartgrid

Abstract

An ever growing electric demand and diminishing fossil fuel reserve lead the power sector to give importance to renewable power generation. Among various renewable sources, solar and wind are the most preferred renewable power generation in India. In addition to the large-scale solar PV power generation, residential consumers are also encouraged by the state and central governments to install small-scale roof top solar PV power generation to meet out their own demand partially or completely. Very short term prediction of solar radiation and temperature is essential for a smart prosumer with PV installation to enjoy uninterruptible power supply, make profit, reduce grid dependency and support the grid operation during peak demand periods. In this paper, an ANN model is developed in MATLAB to predict the solar power in very short duration while considering the different factors representing present climatic changes and residential consumer's constraints like computational complexity and data management. The developed model is tested with three different approaches based on the nature of input data set. The results of three approaches are compared with the actual weather data recorded from weather monitoring system installed at National Institute of Technology Tiruchirappalli, India. As a result of the comparative study a suitable approach for better prediction is suggested in this paper.