Intelligent Neural Learning Models for Multi-step Wind Speed Forecasting in Renewable Energy Applications

Nonlinear radial basis function neural network (RBFNN) model and a wavelet neural network (WNN) model are developed in this research study to perform multi-step wind speed forecasting of the considered wind farm target sites. Wind speed forecasting is one of the most essential predictions to be done in the power generation sector because this facilitates establishment of wind farms at locations where the wind speed level is better. Based on the prediction (forecasting) accuracy, it is decided on the establishment of wind farms at the desired locations where the forecasting was carried out. In this paper, work is carried out in developing modified variants of RBFNN and WNN. With respect to RBFNN, the learning rate parameter and momentum factor are varied during the training process and the point at which the minimized errors gets recorded is considered to be the better prediction point and the learning rate and momentum factor corresponding to that minimized error are taken as the final parametric values. In WNN, new wavelet function is employed as the activation function for evaluating the output of the network model and the network gets trained to achieve better prediction accuracy. Both the nonlinear RBFNN and WNN models are nonlinear neural network models, and these both developed novel RBFNN and WNN are tested for their effectiveness and validity on the multi-step wind speed forecasting. Simulated results attained prove the efficacy of the developed models over that of the existing models from the previous literature findings.