Posters

Abstract

Operation and Maintenance (O&M) takes up between 20-30% of the levelised cost of energy (LCOE) of wind turbines (WTs). Recent developments in artificial intelligence (AI) provide useful tools to leverage the enormous amount of data emitted by WTs and optimise O&M processes, decreasing the associated cost. ANNEA presents an intelligent solution for predictive maintenance and performance optimisation, using AI, physical modelling and normal behavior modelling to deal with unexpected WT failures. Three main issues regarding failure analysis include: non-uniform data treatment, the scarcity of available failure analyses and the lack of investigation on alternative data sources. ANNEA’s analysis of historical failure and downtime data tackles these issues. Comparison of different WT types failure behaviour shows that up to 90% of all WT failures are related to drive train components. ANNEA understands the predictability of signals using recorded measurements of a turbine at different locations. We focus on the synergistic effects of co-existing sensing technologies which construct the framework of multi-step procedure to reveal predicted responses and hidden relationships between signals. Exposure to highly variable environmental conditions can greatly impact WT reliability. ANNEA takes into account the site-specific weather conditions. With experience derived from a big database (over 5000 WTs with up to 20years of data each), we developed an approach that not only predicts future failures of components and sub-components but also identifies their root-causes. The ANNEA applied results in several wind farms show OPEX reduction by 50% and an increase of energy production by 15%.