Clustering of Temperature Measurement Points for Key Components of Offshore Wind Turbines Considering Spatiotemporal Multi-Modality Characteristics

The offshore environment is constantly changing and complex, causing changes in environmental temperature and operating conditions with seasonal variations, resulting in a spatiotemporal multimodal correlation between temperature measurement points of key components within wind turbines. A clustering method for temperature measurement points of key components of wind turbines based on Spearman correlation analysis and VGAE algorithm is proposed in this paper, which provides a reasonable modeling strategy for temperature characteristic perception modeling of offshore wind turbines. Analyze the raw data of wind turbines based on seasonal changes, generate spatiotemporal multimodal characteristic data, and use Spearman correlation coefficient to preliminarily analyze the nonlinear correlation of characteristic variables, defining the undirected graph relationship nodes for feature variable; Then, the potential feature dependencies and temperature variation characteristics of the undirected graph relationship nodes are excavated, and the Variational graph auto encoders algorithm is used. The encoder composed of GCN and variational inference method are used to analyze the potential representations and reconstruct the undirected graph relationship nodes; Finally, the K-means clustering method is used to cluster the temperature measurement points of key components of offshore wind turbines. The results indicate the effectiveness of the method.