Investigation of the importance of criteria in potential wind farm sites via machine learning algorithms

Abstract

Wind energy has received greater attention than other energy resources due to its superior economics, low greenhouse gas emissions, and limitless wind resources. As a result, wind energy capacity has significantly increased, and the selection of the best locations for wind farms is an issue that has received extensive research. A significant step toward environmentally responsible land use planning is the site suitability assessment for the placement of wind farms. This study was conducted to determine the best locations for wind farms and to prioritize different locations and alternatives in the West of Turkey by using Maximum Entropy (MaxEnt) and Logistic Regression (LR) Methods based on Geographic Information Systems (GIS). Eight criteria were selected for creating the suitability map: air density, power density, wind speed, capacity factor, elevation, slope, aspect, and land use. Both methods were effective at choosing locations for wind farms because all the results were statistically significant in the consistency tests. MaxEnt calculated the potential wind energy fields with high accuracy and reliability with 0.915 AUC and LR multiple R square values of 0.883. Compared to the current installed power values, the MaxEnt analysis results were more consistent with the recent status. İzmir has been calculated as the province with the highest potential for wind energy area of 663 km2 by MaxEnt and 620.4 km2 by LR.