Search

Abstract The need for high‐resolution wind resource maps is increasing with the increase in the supply and development of wind power. Many physical downscaling models have been developed and applied to make these maps. However, as the existing models require extensive computations and time, statistical models with higher efficiency are being studied. Statistical models such as regression and machine learning models can quickly calculate wind resource maps, but they have a problem of low accuracy. This study proposes a machine learning model with new topography‐derived variables to interpret the physical characteristics of the wind. As the shape of topography, which was unable to be interpreted in previous studies, can be considered with new derived variables, a significant performance improvement was identified. The analysis was conducted using 1 km Weather Research and Forecasting (WRF) results and ERA5 reanalysis data from South Korea. Two Weibull distribution parameter maps were calculated and used as input and output data. Three collections of derived variables were devised and compared. Therefore, the multi‐resolution topography data showed the highest improvements with approximately 15% reduction in root mean square error (RMSE) for both the linear regression and machine learning models. In particular, the land area showed a decrease of 20%. The best proposed models showed an RMSE of 7% and 8% for two Weibull parameters. The results are expected to serve as a reference for continuing research and utilization of statistical models.

Summary A decade of research has shown that numerical weather prediction (NWP)‐modeled wind speeds can be highly sensitive to the inputs and setups within the NWP model. For wind resource characterization applications, this sensitivity is often addressed by constructing a range of setups and selecting the one that best validates against observations. However, this approach is not possible in areas that lack high‐quality hub height observations, especially offshore wind areas. In such cases, techniques to quantify and disseminate confidence in NWP‐modeled wind speeds in the absence of observations are needed. We address this need in the present study and propose best practices for quantifying the spread in NWP‐modeled wind speeds. We implement an ensemble approach in which we consider 24 different setups to the Weather Research and Forecasting (WRF) model. We construct the ensemble by considering variations in WRF version, WRF namelist, atmospheric forcing, and sea surface temperature (SST) forcing. Our analysis finds that the standard deviation produces more consistent estimates compared to the interquartile range and tends to be the more conservative estimator for ensemble variability. We further find that model spread increases closer to the surface and on shorter time scales. Finally, we explore methods to attribute total ensemble variability to the different ensemble components (e.g., atmospheric forcing and SST product) and find that contributions by components also vary depending on time scale. We anticipate that the methods and results presented in this paper will provide a reasonable foundation for further research into ensemble‐based wind resource data sets.

The need for high‐resolution wind resource maps is increasing with the increase in the supply and development of wind power. Many physical downscaling models have been developed and applied to make these maps. However, as the existing models require extensive computations and time, statistical models with higher efficiency are being studied. Statistical models such as regression and machine learning models can quickly calculate wind resource maps, but they have a problem of low accuracy. This study proposes a machine learning model with new topography‐derived variables to interpret the physical characteristics of the wind. As the shape of topography, which was unable to be interpreted in previous studies, can be considered with new derived variables, a significant performance improvement was identified. The analysis was conducted using 1 km Weather Research and Forecasting (WRF) results and ERA5 reanalysis data from South Korea. Two Weibull distribution parameter maps were calculated and used as input and output data. Three collections of derived variables were devised and compared. Therefore, the multi‐resolution topography data showed the highest improvements with approximately 15% reduction in root mean square error (RMSE) for both the linear regression and machine learning models. In particular, the land area showed a decrease of 20%. The best proposed models showed an RMSE of 7% and 8% for two Weibull parameters. The results are expected to serve as a reference for continuing research and utilization of statistical models. [ABSTRACT FROM AUTHOR]

Summary: A decade of research has shown that numerical weather prediction (NWP)‐modeled wind speeds can be highly sensitive to the inputs and setups within the NWP model. For wind resource characterization applications, this sensitivity is often addressed by constructing a range of setups and selecting the one that best validates against observations. However, this approach is not possible in areas that lack high‐quality hub height observations, especially offshore wind areas. In such cases, techniques to quantify and disseminate confidence in NWP‐modeled wind speeds in the absence of observations are needed. We address this need in the present study and propose best practices for quantifying the spread in NWP‐modeled wind speeds. We implement an ensemble approach in which we consider 24 different setups to the Weather Research and Forecasting (WRF) model. We construct the ensemble by considering variations in WRF version, WRF namelist, atmospheric forcing, and sea surface temperature (SST) forcing. Our analysis finds that the standard deviation produces more consistent estimates compared to the interquartile range and tends to be the more conservative estimator for ensemble variability. We further find that model spread increases closer to the surface and on shorter time scales. Finally, we explore methods to attribute total ensemble variability to the different ensemble components (e.g., atmospheric forcing and SST product) and find that contributions by components also vary depending on time scale. We anticipate that the methods and results presented in this paper will provide a reasonable foundation for further research into ensemble‐based wind resource data sets. [ABSTRACT FROM AUTHOR]

ABSTRACT As the penetration of wind generation increases on power systems throughout the world, the effects of wind variability on power systems are of increasing concern. This study focuses on sustained occurrences of low wind speeds over durations ranging from 1 h to 20 days. Such events have major implications for the variability of energy yields from wind farms. This, in turn, influences the accuracy of wind resource assessment. The frequency analysis techniques commonly used to study wind variability cannot represent the autocorrelation properties of wind speeds and thus provide no information on the probabilities of occurrence of such sustained, low wind events. We present two complementary methods for assessing wind variability, runs analysis and intensity-duration-frequency analysis, both with emphasis on characterising the occurrence of continuous, extended periods (up to several days) of low wind speeds. Multi-annual time series of hourly wind speeds from meteorological stations in Ireland are analysed with both techniques. Sustained 20-day periods corresponding to extremely low levels of wind generation are found to have return periods of around 10 years in coastal areas. Persistent, widespread low wind speed conditions across the entire country are found to occur only rarely. Copyright © 2012 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]