Wind Speed Prediction of Central Region of Chhattisgarh (India) Using Artificial Neural Network and Multiple Linear Regression Technique: A Comparative Study

There are many renewable energy sources available in the earth like the sun, wind, biomass, tides, heat of the earth etc. Wind energy is one of the important energy sources, which can be used for generating electricity, water pumping, grinding of grains with the use of wind energy conversion system. Wind is generated on the earth surface by uneven heating and it is intermittent in nature. Wind speed is a very important factor and its prediction is very useful for wind power generation and many other purposes. In this study, artificial neural network (ANN) and multiple linear regressions (MLR) techniques are utilized to predict wind speed. For this aim, one year’s data has been taken which is a total of 365 data sets. In both techniques, five parameters were taken which are relative humidity, wind direction, ambient temperature, ambient pressure and perceptible water. In the first ANN technique, an MLP model was developed using these parameters in input layer, with wind speed as output variable. The 5-20-1 neural model has been trained by LM learning algorithm and predicted minimum RMSE and MRE values as 0.4558 and 0.1589 respectively and acceptable value of coefficient of correlation (R) as 0.90162. In the second technique of MLR, same parameters have been used as independent variables and dependent variable. The regression model predicted with coefficient of correlation value as 0.77852. The comparative analysis of performance of both models shows that the MLP model is better than MLR model. Additionally, to find the most sensitive variable, sensitivity analysis has also been done. Perceptible water was found to be the most sensitive variable and the sensitivity sequence of the parameters is: Prw > Ta > WD > Pa > RH.