Long, Medium, and Short-Term Wind Speed Forecasting using Artificial Neural Networks

Abstract

In this research, two solution methods including an Artificial Neural Network (ANN) as well as an ANN combined with Genetic Algorithm (GA) are developed for forecasting wind speeds in the city of Binalood in Iran considering long, medium, and short-term periods. The input data included site atmospheric pressure, mean sea level atmospheric pressure, temperature at a height of 100 meters, soil temperature at a depth of 0-10cm, absolute humidity, relative humidity, total accumulated precipitation, total cloud cover, plus speed and direction of wind at a height of 100 meters, which were collected using windPRO software. The models were run in MATLAB with all mentioned parameters considered as the inputs, and wind speed at of 100 m height considered as the output. Hourly meteorological data pertaining to a prolonged period of time from 2008 to 2019 were used for training the ANN models for the forecasting. Mean Squared Error (MSE) and correlation coefficient between outputs and targets were used as the measure of accuracy. After a sensitivity analysis, the ANN-GA hybrid, with an MSE of 0.29192 and a correlation coefficient of 0.99750, was found to be the very accurate method for wind speed forecasting in the study area.