Correlation Analysis of Wind Farms through Short-term Probablitic Analysis

Abstract

The intermittency of renewable energy resources such as wind farms is causing a significant challenge on power grids due to possible short-term steep variations in generation. In order to properly quantify the probabilistic nature of these short-term changes, in this paper, several correlation coefficient methods are applied on the power generation data of Australian wind farms to define the correlation between their steep generation variations. The study has compared Pearson Correlation Coefficient (PCC), Sørensen-Dice coefficient (DC), Overlaps similarity (OS), and K-means clustering methods. The results show that the PCC provides better performance on long-term generation datasets, while, DC and OS show better results for short-term conditions. The Standard Deviation (SD) computed with K-means can present the interrelationship between wind farms, and this method could also split the aggregated SD into different groups to show the impact from steep variation in generation with probabilistic analysis. This study presented the feasible correlation method for different wind farms with short-term generation data and proved the “extreme” SD can cause contingency in power systems with high penetration wind farms.