Grant: Research Tools and Instruments

Wind power generation has seen a remarkable increase globally due to the renewable and sustainable nature of wind energy. Canada is set for massive growth in wind power generation, expected to increase ninefold by 2050 to meet its carbon neutrality target. However, significant operation and maintenance (O&M) costs for wind turbines (WTs), primarily due to drivetrain system failures, are still limiting wind energy's potential. Additionally, accurate energy prediction remains challenging due to the diverse degradation modes of WTs. To reduce O&M costs, minimize system downtime, and improve wind energy prediction, accurate fault diagnosis and prognosis (FDP) of WT drivetrain systems is essential. Data-driven approaches, particularly those leveraging the rapid growth of machine learning, have shown promise for FDP. However, challenges such as data availability, model interpretability, and prediction reliability still limit the effectiveness of these methods.     A wind turbine simulator with a sensing system is requested in this RTI proposal to develop digital twin and transfer learning-based FDP and energy prediction methodologies for WTs. With this system, the applicants aim to address these issues by creating a novel methodology that integrates both data-driven and physics-based models. A digital twin (DT) model of the WT drivetrain system will be developed, which will be adaptively updated and optimized by comparing the system responses of the real WT simulator with the DT model. The optimized DT model will generate accurate simulated condition data, including fault conditions and system degradation. Advanced transfer learning-based models will then be investigated to achieve accurate FDP and energy prediction for the real WT simulator using both simulated data from the DT and collected monitoring data from the WT.     The research, enabled by the requested equipment, will also provide extensive training opportunities for a broad range of HQPs, who will receive comprehensive training in dynamic system modeling, sensing and data acquisition, edge computing, model optimization, and advanced machine learning. The proposed research will have transformative impacts on Canada's wind energy sector, enhancing competitiveness, productivity, and safety while reducing overall costs, energy consumption, and carbon emissions.     Furthermore, this research will address critical industry challenges (e.g., lack of data, low reliability, and limited transferability) in achieving effective condition monitoring of WTs. The knowledge gained from this research could also be applied to a wide range of industrial assets, including jet engines (airplanes) and other manufacturing machinery. Effective condition monitoring, combined with advanced machine learning and cyber-physical systems, could also benefit transportation, healthcare, and civil engineering sectors.