Wind speed prediction based on simple meteorological data using artificial neural network

Abstract

Air temperature, relative humidity and vapor pressure data during 1993-2004 for city of Manjil in Iran were used for the estimation of wind speed in future time domain using artificial neural network method. The estimations of wind speed were made using three combinations of data sets namely: (i) month of the year, monthly mean daily air temperature and relative humidity as inputs and wind speed as output, (ii) month of the year, monthly mean daily air temperature, relative humidity and vapor pressure as inputs and wind speed as output and (iii) month of the year, monthly maximum daily air temperature, relative humidity and vapor pressure as inputs and wind speed as output . The measured data between 1993 and 2003 were used for training the multilayer perceptron (MLP) neural networks and the 12 months. data from 2004 as testing data. The testing data were not used for training the neural networks. The mean squared errors (MSE) for (i), (ii) and (iii) were found to be 0.003297, 0.003416 and 0.00208655 while the mean absolute percentage errors (MAPE) for testing data were 10.32%, 7.03% and 10.78%.Obtained Results show that neural networks are well capable of estimating wind speed from temperature, relative humidity and vapor pressure.