Data-driven optimization for wind farm siting

Abstract

Both increasing power output and minimizing the variability of that output are key factors in a highly utilized wind farm. Finding a location that offers this type of wind resource requires the use of stochastic methods to explicitly model the uncertainty of wind power output. While the literature is rich with wind speed resource assessments, this work moves the field to wind farm power assessment at undeveloped locations. A wind farm power prediction is achieved with improved spatial process representation. The power prediction results are utilized directly in a wind farm location siting optimization model which endogenously incorporates uncertainty in wind power output. The overall framework for siting new wind farms, termed the Farm Finder tool, is demonstrated in a case study in the Electric Reliability Council of Texas (ERCOT). It is shown that the formulation of this model will find an optimal solution to the siting problem on the Pareto front of high power output, with high reliability. Such location decisions improve the overall social welfare in the electricity market.