Federated Real-Time Monitoring Method for Industrial IoT Devices and Its Application in Wind Farm Cluster

Abstract

Driven by industrial Internet of Things (IIoT) technologies, wind farms are evolving into highly digitalized and interconnected energy systems. With the accelerating expansion of renewable energy and the large-scale deployment of wind farm clusters, the demand for collaborative real-time monitoring across multiple farms has become increasingly urgent. However, online monitoring methods focus on individual wind farms and face limitations such as data privacy and cross-farm information collaboration in wind farm clusters with multiple operators. Meanwhile, existing multifarm fault detection/diagnosis methods based on federated learning (FL) are mostly designed for offline scenarios and rely on a large number of fault labels. Motivated by the above challenges, this article develops a federated analysis and real-time monitoring (FARM) strategy for wind farm clusters, which enables online collaborative monitoring for multiple farms while preserving data privacy. Specifically, a farm-level federated sparse principal component analysis (SPCA) method is proposed to enable cross-farm information collaboration, in which the central server exchanges encrypted principal component (PC) projections rather than raw data. Furthermore, a multivariate statistical process control scheme is designed at the central server based on the multifarm information, which enhances the sensitivity and reliability of anomaly detection. Finally, the superiority of the proposed method is demonstrated through both numerical experiments and real wind farm datasets.