Using Logistic Regression-Cellular Automata to Project Future Sites for Commercial Wind Energy Development

Keywords: Wind Energy; Environmental Modeling; Geographic Information Systems; Logistic Regression; Cellular Automaton
Abstract Type: Poster Abstract

Abstract

Expansion of the United States' wind energy sector places onus on managing competition for different forms of land use, minimizing the environmental drawbacks of commercial wind farms, and ensuring wind energy's continued socio-economic appeal. Existing means of modeling Wind Farm Site Suitability (WiFSS) rely on construction of composite suitability surfaces in a Geographic Information System, though such models are not temporally explicit and seldom account for qualitative predictors (e.g., demographics, legislation in effect, lobbyism) of wind farm siting decisions. A logistic regression approach allows an equation-based relationship between the likelihood of wind farm occurrence and a comprehensive set of predictors of WiFSS to be established. By applying a logistic regression model to grid cells that cover a single state, or the Continental United States, the importance of these predictors (e.g., wind speed, distance to existing infrastructure, public opinion), and the likelihood of grid cells currently containing a commercial wind farm or not, can be determined. Preliminary tests of the model show that between 65 and 80% of currently existing wind farm locations across the United States can be correctly identified. Such high predictive power means that the model's combination with a cellular automaton could allow future commercial wind farm locations to be similarly projected. Using the United States Department of Energy's 2015 Wind Vision report to inform scenario construction, the ultimate objective is to use the decision rules inherent to cellular automata to project suitable wind farm sites given set capacity targets, presented as a modeling tool for end users.