Session: 01-04: AI for Energy Sustainability IV

Paper Number: 131342

131342 - Decentralized Condition Monitoring for Distributed Wind Systems: A Federated Learning-Based Approach to Enhance SCADA Data Privacy 

Abstract: 

Background: The development of distributed wind systems (DWSs) is remarkably showing promising growth, with an increasing number of small-scale wind turbines being deployed to harness clean and renewable energy, which can reduce inefficiencies and vulnerabilities embedded in long-distance power delivery, particularly in rural areas.

Motivation: While the resource potential for distributed wind suggests very favorable conditions, the distributed wind industry suffers significant challenges in optimizing the performance and mitigating condition monitoring associated with DWSs. Traditional centralized data processing approaches are often inadequate due to the sensitive nature of data shared between different companies operating these turbines.

Method: The paper presents a novel approach to condition monitoring in DWSs, integrating federated learning and supervisory control and data acquisition (SCADA) data. In this paradigm, multiple wind turbines collaboratively contribute to the training of machine learning models without compromising data privacy. Federated learning emerges as a revolutionary approach to offer a paradigm where a global model can be trained across multiple decentralized SCADA datasets without actual data exchange, thus ensuring data privacy and security. The proposed condition monitoring framework aims to enhance the accuracy and efficiency of condition monitoring for DWSs by leveraging the vast amount of operational data generated by SCADA systems. Distributed wind turbines, often located in diverse geographical areas, contribute to a federated learning ecosystem, creating a collaborative model that reflects the collective knowledge of the entire system. This collaborative learning approach enables the identification of subtle patterns and anomalies indicative of potential faults or performance issues across the distributed wind infrastructure. The integration of SCADA data is instrumental in providing real-time insights into the health and performance of individual turbines. Key operational parameters, such as vibration metrics, temperature readings, wind speeds, and turbine outputs, are continuously monitored, allowing for the early detection of deviations from normal operating conditions. The federated learning models, trained on this dynamic SCADA dataset, evolve over time to adapt to changing environmental and operational factors, ensuring robust and adaptive condition monitoring, while maintaining data confidentiality. Privacy preservation is a paramount consideration in the proposed framework. Federated learning allows wind system operators to collectively improve the monitoring models without exposing sensitive turbine-specific information. This ensures that the benefits of collaborative learning can be harnessed without compromising proprietary or confidential data. Furthermore, the incorporation of federated learning models enhances the security and transparency of the condition-monitoring process. 

Conclusion: The synergy of federated learning and SCADA data in condition monitoring for DWSs represents a pioneering paradigm shift. This innovative framework promises to significantly improve the accuracy, efficiency, and privacy of monitoring processes while fostering a collaborative approach to knowledge sharing within the distributed wind infrastructure. The proposed solution holds great potential for advancing the reliability, performance, and sustainability of wind energy systems in the face of evolving operational challenges.

Presenting Author: Gang Li Mississippi State University

Presenting Author Biography: Dr. Gang Li is an Assistant professor in the Department of Mechanical Engineering at Mississippi State University.
From 2021 to 2023, he was an Assistant Research Professor in the Department of Mechanical Engineering at the University of Maryland, Baltimore County (UMBC). From 2017 to 2021, he was a Postdoctoral Research Associate in the Department of Mechanical Engineering at UMBC. From 2016 to 2017, he was a visiting Ph.D. student in the Department of Mechanical Engineering at UMBC. He received the B.S., M.S., and Ph.D. degrees from the University of Shanghai for Science and Technology, Shanghai, China, in 2010, 2013, and 2017, respectively, all in mechanical engineering.
His research interests include the fields of renewable energy technologies, dynamics and vibration, control theory, condition monitoring algorithms, structural health monitoring, and life cycle assessment, and involve analytical development, numerical simulation, experimental validation, and industrial application.

Authors:

Gang Li Mississippi State University
Yusen Wu University of Miami
Yelena Yesha University of Miami