Presentations

Abstract

A lack of data sharing is a widely-recognised challenge in the wind energy sector today, as demonstrated in a recent review paper. In order to improve data sharing, the WeDoWind framework has recently been developed and demonstrated in two use cases related to wind turbine fault detection. The main innovation of this framework is the way it creates tangible incentives to motivate different types of people to actually share data in practice. It helps people and organisations solve the most pressing problems collaboratively based on "challenges" and data provided by the industry. A recent assessment of participants' surveys filled out after the two case studies allowed several challenges related to data sharing to be identified, including the need to provide more clear comparison and evaluation criteria, to better integrate students into the process, to well-describe the data through standardised metadata and to prepare sufficient data for a given task. In this work, a further WeDoWind data-sharing case study is carried out, aiming to investigate and demonstrate how data sharing can be improved in practice in the wind energy sector. A "challenge" on the topic of wind turbine power curve benchmarking is created and implemented within the framework. The results allow five different data-driven power curve prediction methods to be compared. The best method reduces the model error by as much as 70% in terms of MAE and 45% in terms of RMSE compared to the standard industry method. The results of a survey filled out by the participants shows that data sharing can be improved in practice by providing clear comparison and evaluation criteria, as well as by better integrating students into the process via lectures and student projects.