Posters

Abstract

In Post Construction Yield Assessment (PCYA), a central part involves estimating the losses experienced by individual turbines for the investigated period.In many cases, post-construction analyses rely entirely on SCADA data and nacelle-mounted anemometers. Consequently, production losses are often estimated using historical data of actual energy production and nacelle wind speed measurements. The present study focused on modelling correlated uncertainties in production losses using data from operating wind farms.However, analytical quantification of uncertainties relies on establishing representative correlation model.Here, we show how an autocorrelation model can be obtained from observed model residuals and how this model can be used in the combined uncertainty of the estimated production losses. The present study shows the effect of different assumptions in the correlation of model uncertainties on estimated production losses. The main results illustrate a large difference in uncertainties when assuming either no correlation or full correlation. Accounting for the autocorrelation of model residuals results in significantly lower total uncertainty compared to the fully correlated uncertainty. As a result, the added effort of modelling the autocorrelation becomes more attractive. This study shows the necessity of investigating autocorrelation in model residuals during time series analyses in PCYA, offering a versatile approach applicable beyond the presented models.By emphasizing the importance of the underlying correlation assumptions, we hope to contribute to refining post-construction yield assessments and offer insights that extend towards other areas in the wind energy sector.