Posters

Abstract

As the renewable energy sector is growing and maturing, asset owners need to manage larger and older portfolios. Therefore, they need to effectively use data to ensure the long-term, reliable, and productive operation of these portfolios. The identification of controller parameters is a fundamental aspect of optimizing the performance, efficiency, and reliability of wind turbines, which are characterized by highly nonlinear dynamics and driven by wind, rotor speed and pitch angles. Accurate models with many variables are required to capture the complex dynamics of turbine operation. These models also must balance between capturing these complexities and practical implementation. Effective parameter identification is essential for gaining a deeper insight into both current and potential performance. With more detailed and actionable insights, the owners benefit from improved Annual Energy Production, reduced operational costs and lower asset downtime. To achieve these benefits, better quality and higher frequency data should be analysed with deep domain knowledge. This study focuses on the European Academy of Wind Energy Data Science Challenge which is organized by WeDoWind. The challenge uses open-source data from Chalmers wind turbine, which includes 69 SCADA channels with time-series data sampled at 20 and 100 Hz frequency, containing information on various components such as rotor, generator, yaw motor, and pitch system. The wind turbine has a rated power of 45 kW and operates at a rated speed of 75 rpm. It is mounted on a 30-metre wooden tower with carbon fibre blades measuring 7.5 meters in length and a turbine diameter of 15.9 meters. The individual pitch system is electrically operated. Located on the island of Björkö at Skarviksvägen, 20 km west of Gothenburg, the turbine is not of commercial size but has enough similarities to represent a valuable research model for the industry (Fogelström et al., 2023). The study aims to develop a torque-rotor speed look-up table and determine the pitch controller algorithm using data-driven modelling techniques. It also seeks to analyse turbine responses to environmental and system changes to identify controller parameters across different operating modes. This study presents a data-driven approach to identify critical controller parameters using high-frequency operational data from wind turbines. By applying advanced data analysis techniques, the research demonstrates the potential of high-frequency data to enhance predictive maintenance and enable more informed decision-making. A deeper understanding of turbine conditions allows owners to make precise adjustments to system parameters, optimizing performance while reducing fatigue loads, minimizing component failure rates, and extending asset lifespan. References Sara Fogelström, Håkan Johansson, Ola Carlson, Martin Hofsäß, Oliver Bischoff, Yuriy Marykovskiy, & Imad Abdallah. (2023). Björkö Wind Turbine Version 1 (45kW) high frequency Structural Health Monitoring (SHM) data [Data set]. Zenodo. https://doi.org/10.5281/zenodo.8230330