Image-based fleet-level health assessment for wind farm with power temperature curve

Abstract In the field of wind power generation, the health assessment of wind turbines (WT) is becoming increasingly important. However, most studies usually focus on a single WT and there are relatively few studies on the health assessment of the entire wind farm, thus posing challenges for their maintenance and management. Therefore, it is necessary to develop an effective method for the farm-level health assessment. To this end, this paper presents a general method for fleet-level health assessment with power temperature curve (PTC) images from Supervisory Control and Data Acquisition (SCADA) data in a wind farm. Our proposed method includes one global health assessment at the wind farm level and one local health assessment at the turbine level. The proposed method is based on the relationship between power and component temperature. The power temperature curves are first converted into grayscale images. Then the similarity between the single WT image and the reference WT image is calculated to achieve the health assessment of the temperature components of the whole wind farm. A real wind farm dataset containing 30 WTs is used to evaluate the performance of our proposed method. The results of three cases of different critical components (main bearing, gearbox, and generator rear bearing) show that our proposed method can accurately identify the health status of the components and provide easy-to-interpret results in a visualization manner, which offers a new approach for wind farm health assessment, providing vital technical support and decision-making insights for the wind power industry.