Artificial Intelligence for Renewable Energy Systems

7Deep Feature Selection for Wind Forecasting-I

C. Ramakrishnan¹*, S. Sridhar², Kusumika Krori Dutta², R. Karthick¹ and C. Janamejaya²

¹Dept. of EEE, SNS College of Technology, Coimbatore, India

²Dept. of EEE, M S Ramaiah Institute of Technology, Bangalore, India

Abstract

The existence of abundant renewable energy and fast growing technologies in wind energy extraction creates extensive attention on the wind forecasting. The forecastings are generally very short-term forecasting (few seconds to 30 minutes ahead), short-term forecasting (30 minutes to several hour ahead), medium-term forecasting (several hours to 1 week ahead), and long-term forecasting (from 1 week to 1 year or more). The forecasting involves extraction of single or multiple features from the time series data for more accurate prediction. The different wind speed and power forecasting model include physical model, statistical model, computational model, and hybrid model. Pre-processing the raw data, feature extraction, and prediction are the steps involved in forecasting of wind speed and wind energy. The accuracy in prediction cannot be as expected due to complex computations when mathematical models have been used though it is simple. To improve the performance of prediction and reduce the computational time, computational intelligence (CI) methods like artificial neural networks, evolutionary computation, fuzzy logic, and probabilistic methods are employed for analyzing, decision-making, and optimizing the ...