Is it always windy somewhere? Occurrence of low-wind-power events over large areas

TL;DR

This study shows that aggregating wind power from multiple geographically dispersed sites significantly reduces the occurrence of low-wind events, enhancing grid reliability and economics.

Contribution

It introduces a theoretical framework combining Large Deviations theory and normal distribution to predict wind power aggregation effects without adjustable parameters.

Findings

Low-wind hours decrease exponentially with more sites aggregated.

Hours below 5% capacity drop from 2140 to 36 annually when combining nine sites.

The tail behavior of generation levels is well described by Large Deviations theory.

Abstract

The incidence of widespread low-wind conditions is important to the reliability and economics of electric grids with large amounts of wind power. In order to investigate a future in which wind plants are geographically widespread but interconnected, we examine how frequently low generation levels occur for wind power aggregated from distant, weakly-correlated wind generators. We simulate the wind power using anemometer data from nine tall-tower sites spanning the contiguous United States. We find that the number of low-power hours per year declines exponentially with the number of sites being aggregated. Hours with power levels below 5% of total capacity, for example, drop by a factor of about 60, from 2140 h/y for the median single site to 36 h/y for the generation aggregated from all nine sites; the standard deviations drops by a factor of 3. The systematic dependence of generation-level probability distribution "tails" on both number and power threshold is well described by the theory of Large Deviations. Combining this theory for tail behavior with the normal distribution for behavior near the mean allows us to estimate, without the use of any adjustable parameters, the entire generation duration curve as a function of the number of essentially independent sites in the array.