Blade Icing Fault Identification Based on Domain-Adaptive Spatiotemporal Attention Model

As one of the key components of wind turbines, the health of wind turbine blades directly affects the utilization rate of wind energy. Blade icing fault identification based on SCADA data is low-cost and effective. However, SCADA data has strong spatiotemporal correlation, and there are strong distribution differences in the spatiotemporal characteristics of different wind turbines, which makes the identification of cross-turbine blade icing faults extremely poor. And the actual state labels are precious and difficult to obtain. To address the above problems, this paper proposes a domain-adaptive spatiotemporal attention model for cross-turbine blade icing fault identification. By introducing a feature extraction network based on spatiotemporal attention and using the maximum mean difference algorithm to quantify and minimize the feature differences between source domain data and target domain data. It reduces the dependence on a large amount of labeled data and improves the generalization ability of the model. Experimental results show that this method significantly improves the identification rate of cross-turbine blade icing faults, verifying the effectiveness and practicality of this method.