Wind Speed Prediction using Artificial Neural Networks Based on Grey Wolf Optimizer

doi: 10.14456/mijet.2022.6

Abstract

This paper presented the optimization of Multi-Layer Perceptron (MLP) Artificial Neural Networks (ANNs) using the Grey Wolf Optimizer (GWO) algorithm. The objective was to develop a prediction model for wind signal using artificial neural networks by using the principle of numerical statistical prediction and time-series data from air pressure, temperature, and wind speed. For accuracy and efficiency of the developed prediction model, the model consisted of the time series data divided into two sets as 70% for the learning data set   and 30% for the test data set of the total data for the model to learn from the actual data set. The results obtained were Model 3-24-12-1 using input data with 3 nodes, 1 hidden layer with 24 nodes using tansig activation function, 2 hidden layers with 12 nodes using the tansig activation function and the output layer had 1 node using purelin activation function. The mean squared error in the prediction was obtained at 0.0054. It can be concluded that the prediction model could be used to forecast wind signals very well. In future work, the training algorithms by the optimizer may be used to achieve the least erroneous results.