A Bayesian Hierarchical Model to Understand the Effect of Terrain on Wind Turbine Power Curves

Abhinav Prakash, Se Yoon Lee, Bani Mallick, Yu Ding

Abstract:

This paper is concerned with explicitly modeling the
effect of terrain on wind power curves. Terrain characteristics are
spatially-varying but temporally constant, whereas other power
curve-affiliating variables such wind speed, temperature, and wind
power vary both spatially and temporally. In order to effectively
model such two modes of variation in the data, we employ a
Bayesian hierarchical model (BHM) that connects the terrain characteristics with the parameters in a power curve. BHM jointly
models the data from all turbines on a wind farm for attaining the
turbine-specific, terrain-incorporating power curves. Our analysis
shows that, out of the three terrain variables available in our data,
ruggedness has the strongest effect on the power curve. We also
evaluate the applicability of using the resulting power curve model
for turbines on a different terrain and find that incorporating
terrain information explicitly is beneficial. The specific BHM mechanism of using terrain information leads to 7–10% improvement
over the group averaging approach.