Posters
Come meet the poster presenters to ask them questions and discuss their work
We would like to invite you to come and see the posters at our upcoming conference. The posters will showcase a diverse range of research topics, and will give delegates an opportunity to engage with the authors and learn more about their work. Whether you are a seasoned researcher or simply curious about the latest developments in your field, we believe that the posters will offer something of interest to everyone. So please join us at the conference and take advantage of this opportunity to learn and engage with your peers in industry and the academic community.
PO002: From missing data to fleet insights: extending analogs ensemble approach with wind turbine fleet data
Pierre Dagnely, Data scientist, Sirris
Abstract
Optimizing the performance of wind turbine fleets requires to assess whether each turbine is operating as expected under its current conditions and operational context. Conventional approaches often rely on trained models requiring large volumes of labeled data, making them costly to deploy and difficult to explain. To address this problem, we developed a novel method called Analog-Aware Networked Ensemble (AnAnEn). AnAnEn is a training-free, context-aware and model-agnostic methodology enabling continuous performance benchmarking, early detection of underperformance, asset drift monitoring, and production forecasting estimation before asset deployment. AnAnEn leverages the concept of analogs: past time window from the same or another asset that exhibits similar contextual conditions to the current situation, and whose observed outcomes can therefore serve as a reference for expected behavior. AnAnEn identifies the analogs that are the most relevant for the period under scrutiny, i,e., the most similar to the context, and derives an empirical distribution of expected behavior. This distribution provides point estimates and confidence intervals, ensuring transparency and explainability. We validate AnAnEn on operational datasets from two European wind farms, using missing data imputation as a controlled testbed. Across key signals (active power, generator speed, wind speed, wind direction, and ambient temperature) AnAnEn consistently outperforms baseline techniques such as forward fill and linear interpolation, reducing normalized root mean square error (NRMSE) by up to a factor of three for power-related variables. However, the method ability to generalize is limited when encountering operational regimes not represented in the historical data. The method remains computationally lightweight, capable of near real-time operation without sacrificing accuracy. Its ability to work across heterogeneous fleets without training makes it well-suited for integration into control monitoring systems, digital twins, and fleet optimization workflows. For this reason, AnAnEn can enhance operational decision-making and improve the performance and reliability of wind turbine assets.
No recording available for this poster.
