A Bayesian model for wind farm capacity factors

Abstract

Capacity factors are an important performance metric for offshore wind energy projects as they indicate how efficiently a given project generates electricity. Given the intermittent nature of the wind resource, there is substantial variability between observed capacity factors seasonally and between years. However, little work has focused on extracting trends in the variable wind farm energy generation data. This paper proposes applying hierarchical Bayesian techniques to historical capacity factors to enable the prediction of capacity factor distributions. The proposed model relies on data from UK offshore wind farms, the most developed market for offshore wind energy, but is equally applicable to other countries. The resulting capacity factor distributions highlight a substantial variability in capacity factors both when modelled as yearly and monthly (which accounts for seasonality). The model shows that newer wind farms have higher capacity factors than older farms, with an improvement of approximately 20%. It is also demonstrates that capacity factor variability has a smaller impact on levelized cost of energy estimates. The results from this study can be used both to predict capacity factors for individual wind farms or to make predictions for a generic wind farm.