Short-term Wind Speed Prediction using ANN

Abstract

With the advent of 21st century, all countries of the world are striving to meet their needs from renewable energy and leave as low carbon footprint as possible; depletion of fossil fuels and climate change being the root reasons. India has the intent of achieving half of its energy needs by renewables by the year 2030 and as of 31st March, 2021, the wind capacity of India was found to be thirty-nine GW. Producing energy from wind is one of the cleanest and environment friendly ways of producing electricity as it is omnipresent. This paper focuses on estimating the unpredictable wind speeds at one of the windiest sites in India - Mahabaleshwar taking eight meteorological parameters as input for a period of twenty-seven months (from IMD) with the help of neural network tool in MATLAB using Levenberg-Marquardt method under Nonlinear Autoregressive with External Input consisting of more than two thousand datapoints. The model predicts the wind speed with agreeable regression and mean square error values. Accurate prediction of wind speed helps in locating wind farm sites, predicting power output from wind farms, scheduling maintenance of wind turbines and preparation against catastrophic wind speeds.