Abstract
The Norwegian Ministry of Petroleum and Energy’s commissioned report signifies a significant step towards the government’s goal of designating regions for 30,000 MW of offshore wind power by 2040. As per an IRENA report, operation and maintenance (O &M) costs account for a substantial portion of the overall electricity cost in G20 countries’ offshore wind farms, ranging from 16 to 25%. To address this challenge, it is crucial to explore methods that enhance operational reliability and reduce maintenance expenses in wind turbines. Predictive maintenance offers a promising solution by leveraging sensor data already present in the turbines to detect and mitigate potential issues proactively. By implementing predictive maintenance, wind farms can minimize downtime and optimize turbine performance. Wind turbines are intricate machines with multiple moving components, and failure in any part can lead to complete turbine shutdown. This can result in revenue loss for operators and increased maintenance costs if the problem is not promptly addressed. Employing a Supervisory Control and Data Acquisition (SCADA) system, which collects and analyzes data from various turbine components, enables the detection and monitoring of failures in critical parts such as the gearbox and generator using historical SCADA data. Our approach employs machine learning models, specifically extreme gradient boosting (XGBoost), and has been successfully validated in two real-world case studies involving eight different turbines. The outcomes demonstrate the effectiveness and practicality of our method in enhancing reliability and minimizing maintenance costs in wind turbines.