Posters

Abstract

Transition to a sustainable and renewable energy-based future still faces many challenges, O&M costs being one of them. Within the context of wind energy, both offshore and onshore wind turbine (WT) operations and maintenance (O&M) teams need reliable, comprehensive, and robust condition monitoring tools to efficiently schedule critical maintenance actions. A combination of preventive and predictive maintenance strategies has been deployed to ensure optimal operation of critical assets. Supervisory Control and Data Acquisition (SCADA) system-based power curves are commonly used to represent and monitor wind turbine operations. However, environmental, and operational variations have been shown to cause variability and data dispersion making condition monitoring difficult. Fault detection methods that do not take such variations into account are generally the least performant. Several Machine learning (ML) based techniques have shown powerful capability in fault detection and diagnosis of Wind Turbines, reducing equipment downtime, and minimizing financial losses. This work aims at improving the detection performance by proposing a comparative framework for ML based detection solutions, complete with innovative performance evaluation metric for optimal selection of fault detection strategies. The innovative strategy proposed, facilitates the selection of optimal fault detection solution countering the operational and environmental variations improving performance and reducing O&M costs.