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 the academic community. We look forward to seeing you there!
PO312: Explainable Artificial Intelligence Techniques for optimizing wind farm Operation & Maintenance
Ludovico Terzi, Technical Performance and Analysis Manager, Engie
Abstract
A robust comprehension of wind farm operation is a prerequisite for optimizing the Operation & Maintenance, but this is a complex task due to the fact the power extracted by wind turbines has a multivariate dependence on environmental conditions and working parameters. It can be stated that highlighting an under-performance is relatively easy by using state of the art methods, while individuating the root cause is far from trivial. Thanks to the widespread use of SCADA systems, wind energy technology is definitely projected into the era of data. This guarantees vast opportunities for improving O&M practices, but the downsides should not be overlooked. In particular, an excessive reliance on blackbox Machine Learning (ML) methods poses potential risks like over-parameterization and limited generality. Therefore, it does not come as a surprise that there is a growing interest to the concepts of models' interpretability and explainability, which deal in general with the transparency in the relation between input variables, models' parameters and output. Based on this, the present work illustrates how innovative eXplainable Artificial Intelligence (XAI) techniques have been incorporated in the ENGIE Italia fleet monitoring thanks to a partnership between several the company and academia (University of Perugia). The workflow is based on a space-time comparison at the wind farm level, which is made quantitative and explainable through the computation of the Shapley coefficients for the various input variables of a multivariate normal-behavior model for the power of the machines. By using the above framework, several anomalies have been individuated and a statistical analysis of the Shapley coefficients has led to circumscribe the involved input variables, which in turn means the subcomponent of the machine. There is a high potentiality of generalizing and automating the workflow, which would be beneficial for improving the O&M.
