Communication Dans Un Congrès Année : 2023

CONWEY: an Energy Yield Assessment based on Operational Data for Repowering Projects

Résumé

Repowering consists of replacing a plant reaching its end of life with a new one, benefiting from local acceptance, existing permits, and infrastructure. It should represent up to 4 GW of yearly wind power capacity installation by 2030 in Europe [1]. Errors in energy yield assessment (EYA) negatively impact profitability [2]. Consequently, decreasing errors is at the core of many research initiatives in wind energy [3]. State-of-the-art EYA tools are typically adapted for greenfield projects and involve physics-driven models [4,5]. This study presents an alternative EYA process that benefits from systematic physics-driven models errors. The process applies only to repowering projects for which existing plant operational data permits to evaluate physics-driven models error at a site. We set the ground for applying Gaussian Process methods in the first part. We demonstrate physics-driven models errors are jointly Normal distributed. We give a theoretical explanation relying on existing physics-driven models error models [6,7]. Then, we cross-compare annual energy prediction (AEP) from physics-driven models and operational data for 28 nearby plants. We demonstrate the Normality using Mardia's test. We also identify a systematic effect primarily related to the site, i.e. the correlation between observed errors is non null. In the second part, we demonstrate the physics-driven models error covariance function equals a weighted sum of covariances of wind speed, site-specific power curve, and losses estimation errors. We comment on the expected values of all covariances. We link the wind speed error covariance to the Clerc model described in [8]. In the third part, we use the Normality of physics-driven model errors and the covariance function to define an unbiased AEP predictor using Gaussian Process. The predictor, named CONWEY, corresponds to AEP and uncertainty prediction equations. Finally, we compare CONWEY to state-of-the-art EYA on a real-life project: AEP predicted by CONWEY has a 1.3% lower uncertainty.

Fichier non déposé

Dates et versions

hal-04637798 , version 1 (07-07-2024)

Identifiants

  • HAL Id : hal-04637798 , version 1

Citer

Paul Mazoyer, Thomas Duc, Andreas Bechmann, Georges Kariniotakis. CONWEY: an Energy Yield Assessment based on Operational Data for Repowering Projects. Wind Energy Science Conference, WESC 2023, European Academy of Wind Energy, May 2023, Glasgow, United Kingdom. ⟨hal-04637798⟩
165 Consultations
0 Téléchargements

Partager

  • More