Posters

Abstract

Least Squares Regression Method of MCP: Performance Evaluation of Long-Term Correction Methods of Scaling and Data-Filling in Different Time Domains   Author: Barış Gündoğan Contributors: Murat Hazer Uygunol, Ufuk Yaman   Department of Data Analytics, USENS Energy Solutions, January 24th, 2025   The MEASNET publication Evaluation of Site-Specific Wind Conditions v3 specifies two methods for long-term extrapolation—scaling and data filling—but does not clarify how or when to choose one method over the other. The purpose of this study is therefore threefold: (1) to report long-term validation results for several state-of-the-art data sources used in the long-term correction of meteorological wind measurements, (2) to present the performance of the Least Squares Regression MCP method when applied using scaling at hourly, daily, weekly, and monthly resolutions, as well as data filling at an hourly resolution, and (3) to define a set of conditions for selecting between scaling and data-filling approaches. This investigation involves 10 meteorological measurement stations spread across Türkiye, each with approximately five years of continuous high-availability data. For the long-term datasets, EMD-WRF Europe+, ERA-5, and MERRA-2 were employed concurrently with the station data. One-year and six-month subsets of measurements were extracted from each station to quantify error margins resulting from long-term corrections, projected against the full five-year dataset. Wind speed and energy output error margins were calculated using 33 power curves from established turbine variants at standard air density. Two main criteria guided the analysis. First, each method was evaluated to determine if it could achieve an error margin below 1% in wind speed and 2% in energy output. Second, the long-term-corrected short-term data were validated against actual on-site measurements to confirm they remained within the calculated long-term uncertainty range. These resulting error margins were then compared with MCP performance parameters, specifically the correlation coefficient (R) and MCP uncertainty. Overall, EMD-WRF Europe+ provided the lowest error margins for both the one-year and six-month correction periods, with the data-filling approach yielding the highest R values. For scaling, ERA-5 performed best in terms of meeting the most criteria and recording the lowest wind speed error, although the highest correlation coefficient was still observed with EMD-WRF Europe+. Notably, weekly and monthly scaling produced the largest wind speed errors—sometimes exceeding those found when using only six months of measurement data—likely due to fewer time steps and lower resolution. Furthermore, the data-filling approach is only suitable when using a dataset with higher horizontal resolution.