Estimation of average wind speed in the city of Tianguá using Artificial Neural Network

This paper presents an Artificial Neural Network (ANN) with Time-Lagged Feedforward Network (TLFN) applied in daily wind speed forecasting. In particular, we used a multilayer perceptron (MLP) with a backpropagation algorithm and the sigmoidal activation function. Wind forecasting is relevant to aid the installation and function of wind farms and agricultural meteorology. We experimented with several MLPs topologies by varying the number of neurons and the number of layers in the ANN and collected wind speed data from a weather station at the Federal Institute of Education, Science and Technology of Ceara (IFCE), Campus Tiangua (Brazil). The data is included among the years 2014 and 2018. We evaluated each topology according to a Mean Squared Error (MSE), Mean Absolute Percentage Error (MAPE), Mean Absolute Deviation (MAD), and the Coefficient of Determination (R2). The results obtained showed the possibility of applying Artificial Neural Networks in daily wind speed forecasting.