Wind field reproduction using neural networks and conditional simulation

A procedure based on the use of artificial neural networks (ANNs) and conditional simulation (CS) is presented as a tool to simulate wind fields. The aim of this work is to show a complete methodology to reproduce partially correlated wind fields for a two-dimensional space, starting from the knowledge of the local mean velocity and the level of roughness on the soil. The use of a multilayer ANN to predict wind time series in four strategic points in the two-dimensional space, together with the implementation of image recognition techniques, are explained first; afterwards the CS algorithm is applied to obtain intermediate wind series in order to complete the simulation. The effectiveness of this procedure is evaluated by comparison of its results to those reported by theory, for a specific example.