Analysis and Evaluation of Various Neural Network Models for Estimating wind Energy Resources Forecasting in Different Climatic Zones of Tamil Nadu India

Abstract

Precise estimation of wind velocity is essential for wind newlineenergy sectors and environmental engineering uses. In this wind speed newlineprediction research introduces three modules. The main objective of this newlineresearch is evaluate various machine learning models for estimating newlinewind speed prediction in three different locations of Tamil Nadu, India. newlineThe wind resources MERRA data for this study gathered from NASA newlineGiovanni portal. The august month data over the 13 years data (2008- newline2020) were used for this study. newlineMerg Gao et al 2018 claims through his research that India has newlinevariation in the wind speed and there is a decline in the wind speed. newlineFrom the first module, Merg Gao et al 2018 claims towards wind speed newlinevariation were tested with scientific models. newlineSecond modules elaborated hybrid wind resources forecasting newlineframe model through Neuro (Neural Network) DOE (Design of newlineExperiments) MCDM (Multi Criteria Decision Making (MCDM) for newlineforecasting and it s evaluating the forecasting performance measures. newlineHere Design of Experiments output is input for Neural Network models newlineand neural network models output is input for MCDM (Multi Criteria newlineDecision Making (MCDM) input. Forecasting experiment trails with newlineorthogonal array are conducted with collected data was used. It was also newlineformulated various neural network model related to the extrinsic factors, newlinelevels and responses. Identified best extrinsic factors levels can be used newlinefor the minimizing the forecasting error. For neural netwiv newlineaspects, the researcher considered seven factors with two levels in this newlineresearch. The secondary objective of this research module is to analysis newlinethe various neural network architecture with minimizing forecasting newlineerror. newlineThird module discussed the mathematical model for the newlinedifferent with wind resources parameters with univariate and newlinemultivariate analysis. newlineThe findings indicate that deep learning models perform better newlinethan neural network models.