An innovative methodology for the determination of wind farms installation location characteristics using GIS and Delaunay Triangulation.

Renewable energy development and more specifically Wind Farm (WF) installation has been increased during the last years by most countries. A discipline that has been studied thoroughly is the impact of wind turbines on installation locations, during energy production or at the end of their life. Citizens' attitudes, social and environmental factors are among the criteria affecting the selection of WFs installation locations. Various machine learning tools have been used in order to analyze the allocation of the Renewable Energy Systems (RES). In this paper a methodology based on computational geometry and Geographical Information Systems (GIS) is proposed in order to cluster the locations of Wind Turbines (WT) and define the boundaries of the areas that are affected by them. Delaunay Triangulation is selected as the main algorithm, since it is appropriate for handling geospatial data and it uses a geometric approach. Furthermore, algorithms based on graph theory are proposed in order to improve the efficiency of the methodology. The proposed methodology was implemented in locations of WTs in Greece and was compared to other known clustering algorithms. It was found that the locations of WTs in Greece are separated into 158 clusters. Some characteristics of the areas defined by the clusters, such as the distance from the protected areas, the distance from the closest airport and the number of residential areas that are within them, were computed and it was shown that these areas can be separated into 4 groups. The application aims to help policy makers to improve the effectiveness of their decisions by understanding in a better way which areas are more impaired or beneficial by the installation of WFs. • This study deals with the Wind Farm installation problem for an entire country. • Wind Farm locations in Greece are clustered into 158 clusters. • The areas that are more beneficial or impaired by wind turbines are determined. • The mean number of residential areas per cluster is equal to 6. • 79 out of 158 clusters do not contain any residential area. [ABSTRACT FROM AUTHOR]