Posters

Abstract

The Importance of Database Size for Machine Learning Techniques in Power Curve Analysis Introduction Machine Learning (ML) has revolutionized various industries by enabling automation, providing predictive insights, and improving decision-making processes. In the field of wind energy, ML is used to optimize performance, predict maintenance needs, and enhance power curve analysis. This document explores the implications of database size when applying ML techniques to wind energy, focusing on methods that compare classical power curves with approaches that incorporate uncertainties and big data. The application is demonstrated on an onshore wind farm cluster with a total capacity of approximately 300 MW, including examples from different types of terrain. Classical Power Curve vs. Machine Learning Approach Traditionally, wind turbine performance has been evaluated using classical power curve tests according to IEC 61400-12, which relate wind speed to power output. This method is commonly used for power curve warranties, which link uncertainties to a commercial margin to meet power curve specifications. Power curve upgrades are often offered after the warranty period. By incorporating big data techniques, the accuracy of power curve analysis for operational wind farms can be enhanced, accounting for factors such as turbine aging, environmental conditions, and updated operational strategies. Method: The proposed approach involves using ML algorithms to analyse data from an onshore wind farm. Typically, wind farms are divided into two equal groups when assessing power curve upgrades. By comparing the uncertainties of classical power curves with those generated using ML, we can identify the challenges associated with planning power curve upgrades and ensuring their accuracy through advanced techniques. Results We will present an example involving an onshore wind farm cluster with a capacity of approximately 300 MW. The results will elaborate on important considerations for planning testing campaigns, such as dividing turbines into two equal groups, determining appropriate test durations, accounting for seasonal changes, and addressing terrain complexity. Conclusion The size and complexity of databases play a crucial role in the successful application of ML techniques in wind energy. By selecting appropriate database types and leveraging big data, we can enhance our understanding of wind farm performance, leading to more accurate predictions and optimized operations. This document provides a foundation for further exploration and development of machine learning-driven approaches in the wind energy sector. The integration of big data techniques represents a significant advancement in our ability to analyse and optimize wind farm performance, ultimately contributing to more efficient and sustainable energy production.