Cluster-based wind turbine maintenance prioritization for a utility-scale wind farm

Abstract

At peak and off-peak periods of power generation, each wind turbine generator (WTG) on a utility-scale wind farm receives different wind speeds due to their varying geospatial locations. This variation significantly affects their long-term productivity and maintenance demand. Hence, clustering WTGs according to their resource advantages could be helpful when prioritizing maintenance activities. This study applies intelligent clustering techniques to develop WTG clusters for a utility-scale wind farm located in the Eastern Cape of South Africa using the SCADA wind speed. The wind speed variation across the 44 utility-scale wind turbines on the wind farm was analyzed. Wind resource variability on the wind farm was studied over a year across the four seasons. The average hourly wind speeds of each turbine were clustered using the k-means clustering method and the Calinski-Harabasz method was used to determine the optimal number of clusters. Based on the wind speed received by each WTG, the wind farm can be divided into seven (7) optimal clusters of WTGs. While some clusters consistently receive high wind speed, thus making the WTGs in such clusters economically viable, others receive low wind speeds, making them less economically productive. Turbines with more wind advantage will necessarily require priority maintenance owing to increased productivity and wear rate. This condition can be used to prioritize maintenance activities on the wind farm amidst other significant benefits.