Comparing Traditional Methods of Complex Networks Construction in a Wind Farm Production Analysis Problem

This work presents a comparison between two methods for complex networks construction (cross-correlation and Mutual Information based), in the assessment of wind speed prediction efficiency at different wind farms in Spain. The approach is accomplished at mesoscale, for wind speed prediction data provided by the Weather Research and Forecasting (WRF) numerical model versus the actual wind speed measurements in the wind farms. Some important differences are found in the complex networks obtained, and the corresponding global measures from them, such as the betweenness and closseness centrality. We have found out that the mutual information method better captures nonlinear relationships of the problem, obtaining complex networks with fewer spurious links than the cross-correlation based method.