ERA5: The new champion of wind power modelling?

Meteorological reanalyses are crucial for wind power modeling in academia and industry, providing variables like wind speed, temperature, and pressure for applications such as generating long wind power time series and long-term correction (LTC) of wind speed measurements. Reanalyses offer the advantage of free and global availability. While MERRA-2 (NASA) has been widely used due to its hourly resolution and suitable wind speed height (50m), ERA5 (ECMWF), released in 2017-2018, offers improvements with hourly resolution, wind speed data at 100m (relevant for modern turbines), and a higher spatial resolution (around 31km) compared to its predecessor ERA-Interim and MERRA. This study directly compares the performance of MERRA-2 and ERA5 in modeling wind power generation, both aggregated for five countries (Germany, Denmark, France, Sweden, and Bonneville Power Administration in the northwest USA) and for over 1000 individual Swedish wind turbines.

Existing literature highlights the use of reanalyses in wind power modeling. Studies have shown that MERRA provides good results for country-wise wind power generation with relatively low errors compared to measurements. Previous research also explores various aspects of wind resource characterization, long-term correction techniques, and the use of reanalysis data to simulate regional wind generation variability. However, these studies often employed reanalyses with lower temporal and spatial resolution compared to ERA5, which necessitates a direct comparison to assess its potential advantages.

The study employs a relatively simple wind power model using bilinear interpolation of wind speeds from four neighboring grid points in the horizontal and a power curve representing a specific rating of 360 W/m². To account for losses, the incoming energy was reduced by a fixed amount (10% for countries, 15% for individual turbines), and the resulting energy time series were multiplied by a factor representing the measured maximum country-wise output (101% for countries, 0% for individual turbines). Mean wind speeds were determined using the measured capacity factor (CF) of each farm and a linear scaling of wind speed time series was applied, assuming constant wind shear. Downtime for individual turbines was automatically detected by identifying periods with zero actual generation but high modeled generation, removing those data points and recalculating the model. Long-term correction (LTC) was performed using a simple wind index method, randomly selecting one year of data as a baseline to predict energy production for the remaining years. Finally, a new metric, the "equivalent system radius," was introduced to quantify wind farm system size and dispersion based on the combined variance of the combined generation, assuming exponentially declining correlation with separation distance. This metric addresses limitations of existing methods such as system dimensions, area, and mean distance between farms.

ERA5 significantly outperformed MERRA-2 in modeling country-wise wind power generation. Across the five countries, ERA5 exhibited an average of 22% lower RMSE and 24% lower MAE compared to MERRA-2. The improvement was particularly pronounced in the Bonneville Power Administration (BPA) area, with errors approximately 50% lower for ERA5. Analysis of individual Swedish wind turbines revealed similar improvements. ERA5 showed a 6.6 percentage point higher average correlation and 20% lower MAE compared to MERRA-2 in hourly data. The long-term correction analysis further confirmed ERA5's superiority, demonstrating approximately 20% lower energy errors. One year of measurements long-term corrected with ERA5 provided more accurate estimates than two years of MERRA-2-corrected data. The new "equivalent system radius" metric effectively quantifies wind farm system size and dispersion, considering the geographical smoothing of wind power fluctuations.

The findings confirm ERA5's enhanced capabilities in wind power modeling compared to MERRA-2. The consistent improvements across various metrics and geographical scales highlight the benefits of ERA5's higher resolution and accuracy. The superior performance of ERA5, especially noticeable in areas with complex terrain, suggests that it is a more robust tool for wind resource assessment and long-term prediction. The new metric offers a valuable addition to wind farm characterization, enabling a more comprehensive understanding of system behavior and variability. These results have significant implications for the wind energy industry, reducing uncertainties and increasing project value.

ERA5 demonstrates superior performance in wind power modeling compared to MERRA-2, offering higher accuracy, lower errors, and improved long-term prediction capabilities. The new "equivalent system radius" metric provides a valuable tool for wind farm characterization. Future research could explore more sophisticated wind power models, incorporating additional parameters to further enhance accuracy and address potential biases. Investigating the application of ERA5 in future renewable energy system scenarios, considering higher capacity factors and increased offshore wind power, is also warranted.

The study employed a relatively simple wind power model, potentially limiting the extent of improvement achievable through model optimization. The specific parameters used might not be universally applicable across all geographical locations and wind regimes. The automatic downtime detection method, while effective, may still contain some level of error in classification. The study primarily focuses on data from specific regions and time periods, limiting generalizability to other areas and climates.