This paper presents a validation of atmospheric reanalysis data sets for simulating onshore wind generation time series for large-scale energy system studies. The three reanalyses are the ERA5, the New European Wind Atlas (NEWA) and DTU’s previous generation European-level atmospheric reanalysis (EIWR). An optional scaling is applied to match the microscale mean wind speeds reported in the Global Wind Atlas version 2 (GWA2). This mean wind speed scaling is used to account for the effects of terrain on the wind speed distributions. The European wind power fleet for 2015–2018 is simulated, with commissioning of new wind power plants (WPPs) considered for each year. A generic wake model is implemented to include wake losses that are layout agnostic; the wake model captures the expected wake losses as function of wind speed given the technical characteristics of the WPP. We validate both point measurement wind speeds and generation time-series aggregated at the country-level. Wind measurements from 32 tall meteorological masts are used to validate the wind speed, while power production for four years from twelve European countries is used to validate the simulated country-level power production. Various metrics are used to rank the models according to the variables of interest: descriptive statistics, distributions, daily patterns, auto-correlation and spatial-correlation. We find that NEWA outperforms ERA5 and EIWR for the simulated wind speed, but, as expected, no model is able to fully describe the auto-correlation function of the wind speed at a single point. The mean wind speed scaling is found to be necessary to match the distribution of generation on country-level, with NEWA-GWA2 and ERA5-GWA2 showing highest accuracy and precision for simulating large-scale wind generation time-series.

本文介绍了大气再分析数据集的验证，用于模拟大规模能源系统研究的陆上风力发电时间序列。这三个再分析是 ERA5、新欧洲风图集 (NEWA) 和 DTU 的上一代欧洲级大气再分析 (EIWR)。应用可选比例以匹配全球风图集第 2 版 (GWA2) 中报告的微尺度平均风速。这意味着风速缩放用于说明地形对风速分布的影响。模拟了 2015-2018 年的欧洲风力发电队，每年都会考虑新风电厂 (WPP) 的投产。实现了通用唤醒模型以包含与布局无关的唤醒损失；考虑到 WPP 的技术特性，尾流模型捕获作为风速函数的预期尾流损失。我们验证了在国家级汇总的点测量风速和发电时间序列。来自 32 个高气象桅杆的风测量用于验证风速，而来自 12 个欧洲国家的四年电力生产用于验证模拟的国家级电力生产。各种指标用于根据感兴趣的变量对模型进行排名：描述性统计、分布、日常模式、自相关和空间相关。我们发现，NEWA 在模拟风速方面的表现优于 ERA5 和 EIWR，但是，正如预期的那样，没有模型能够完全描述单点风速的自相关函数。