DECENTRALIZED CONDITION MONITORING FOR DISTRIBUTED WIND SYSTEMS: A FEDERATED LEARNING-BASED APPROACH TO ENHANCE SCADA DATA PRIVACY

The paper presents a new condition monitoring method for distributed wind systems (DWSs) by combining federated learning with supervisory control and data acquisition (SCADA) data. Federated learning facilitates training of a global model across decentralized SCADA datasets without actual data exchange. This method allows multiple DWSs to collectively contribute to training machine learning models without compromising data privacy. The framework aims to enhance the accuracy and efficiency of condition monitoring by leveraging operational data from SCADA systems. It enables collaborative learning across diverse geographical areas, identifying subtle patterns and anomalies indicative of potential faults or performance issues. The integration of SCADA data provides real-time insights into individual turbine health and performance, allowing for early detection of deviations from normal operating conditions. The federated learning model evolves to adapt to changing environmental and operational factors while maintaining data confidentiality. This condition monitoring method allows for collective improvement of monitoring models without exposing sensitive turbine-specific information, enhancing security and transparency. The synergy of federated learning and SCADA data represents a pioneering paradigm shift, promising to improve monitoring processes and knowledge sharing within the wind infrastructure, advancing reliability, performance, and sustainability in the face of operational challenges.