Model for Wind Speed Prediction using Artificial Neural Network from Meteorological Variables: Case Study of Selected Sites of Nepal

Abstract

Wind speed data are of primary importance while designing reliable system of any kinds whose performance can be effected by the wind such as aviation planning, space vehicle launching, weather forecasting, agro-meteorology, satellite and rocket launch, control module operation of military sector and wind energy generation plant. There is no misdoubt that information of the available data of wind speed are good but in Nepal the required data of wind speed are not accessible for most of the sites due to high cost and need for regular maintenance of the measuring instruments. In this study an Artificial Neural network (ANN) was used to find the wind speed of the one place of all the seven different provinces of Nepal to identify the possible wind energy application with help of meteorological data of maximum temperature, minimum temperature, mean temperature, pressure, humidity, solar radiation, wind direction, altitude and wind speed for Tribhuvan International Airport. Data from 2008 to 2017 were used to train, test and validate the network while data of 2018 was used to find the wind speed of the seven different locations of Nepal. The ANN Fitting Tool (nftool) was used for the prediction of daily wind speed using MATLAB programming. Twelve different models with different input combinations were modeled with two layer of feed forward neural network. The result of the ANN model were compared with measured data on the basis of root mean square error (RMSE), mean bias error (MBE), mean absolute error (MAE), mean percentage error (MPE) and Coefficient of Determination (R2) in order to check the performance of developed model. The obtained result in the present work indicate that a parameter such as wind speed which is highly variable and difficult to predict can be successfully modeled using ANN and reliable prediction can be made within the certain geographical areas. Thus the optimal model designed here can be used anywhere across Nepal where the meteorological data are available. The best prediction was from Model 11 as it exhibit minimum value of RMSE (0.06763) and maximum value of R2 (99.06045%) for which input parameter were maximum temperature, minimum temperature, mean temperature, pressure, humidity, solar radiation, wind direction and altitude. The Model 8, Model 10, Model 9, Model 12, Model 3, Model 5, Model 7, Model 4, Model 6, Model 2 and Model 1 provided the respectively better prediction of wind speed.