SSGC-GAT: Synergistic Similarity Graph Construction Strategy Combined With GAT Network for Wind Turbine Anomaly Identification Using SCADA Data

The supervisory control and data acquisition (SCADA) system is the standard installation on large wind turbine (WT) to monitor all major WT subcomponents. By analyzing SCADA data, the anomaly of the WT can be timely identified. However, the complex coupling relationship between different sensors poses a great challenge to the high accuracy of WT anomaly identification. In this article, a novel synergistic similarity graph construction (SSGC)-graph attention network (GAT) method that integrates the SSGC strategy into GAT is proposed to realize high-accuracy anomaly identification of WT. The GAT has a strong graph data modeling capability to accurately capture important relationships between nodes. Furthermore, the proposed SSGC strategy constructs similar graph data by fusing the adjacency matrices computed by four different methods. The SSGC strategy can adaptively learn the complex relationships among multiple parameters to improve the accuracy of anomaly identification. A large number of experiments are conducted to verify the effectiveness and superiority of the proposed SSGC-GAT. The experimental results show that, compared with other several benchmark methods, the proposed SSGC-GAT has the best identification performance. In addition, the ablation experiment results demonstrate that the proposed SSGC strategy can effectively improve the accuracy of WT anomaly identification.