Calibrating Modelled High-Resolution Time Series with Measurements: A Case Study in New Zealand
In wind energy projects, obtaining precise and reliable wind resource data is crucial, yet the available data often remains limited in both time and space. To extend this data into the long term, integrating real-world measurements with model results becomes essential. This study demonstrates the process of improving turbulence calibration in modeled high-resolution time series, using data from a single site in New Zealand (specific location and quantitative results are kept confidential). The figures presented correspond to the out-of-training period, which represents a timeframe without measured data, illustrating how calibration influences the model’s performance.Wind distribution
The first figure showcases the wind speed distribution and highlights how turbulence calibration refines the model’s ability to capture wind variability. Calibration adjusts the modeled wind speeds to align more accurately with observed values, improving the representation of both the central tendencies and extreme values. As a result, the recalibrated model’s Weibull fitting parameters are now more consistent with those derived from the observed data, indicating an improved capacity for predicting the frequency and intensity of different wind speeds.
More specifically, the metrics show a significant improvement in all the metrics after remodeling the TIMES.The most notable improvement is seen in turbulence, where the bias decreased dramatically from 7.08% to just 0.04%.
| Metrics | Wind Speed | Wind Direction | Standard Deviation Speed |
|---|---|---|---|
| TIMES Bias (%) | 15.31 | 8.88 | 7.08 |
| TIMES Remodeled Bias (%) | 1.22 | 2.44 | 0.04 |
| TIMES RMSE | 2.59 m/s | 29.95º | 0.46 m/s |
| TIMES Remodeled RMSE | 1.98 m/s | 28.16º | 0.38 m/s |
Turbulence as a function of wind speed
The second figure examines turbulence intensity as a function of wind speed, offering a clearer view of the calibration’s impact. The introduction of real-world measurements results in a significant alignment between the modeled and observed turbulence trends across varying wind speeds. This adjustment enhances the model’s accuracy, ensuring that it better represents how turbulence fluctuates under different wind conditions. This step is particularly crucial for optimizing wind farm design, where an accurate understanding of turbulence intensity is vital for both performance and durability.
Turbulence as a function of wind direction
The third figure, a polar plot, explores turbulence intensity relative to wind direction. Before calibration, notable discrepancies existed between the model and observed data, especially in specific wind sectors. Post-calibration, these discrepancies are substantially reduced, resulting in a model that not only captures wind speed and turbulence more reliably but also the directional variability of turbulence. This improvement is essential for regions with complex wind patterns, as directional turbulence can significantly influence turbine wear and energy output.
References:
- Cavero, Gerard (2024):Extrapolating turbulence measurements to the long term. Wind Europe Technology Workshop 2024, 10-11 June, Dublin Speaker Session Programme
- Cavero, Gerard (2024): Extending Turbulence Data Over Time. ACP Resource and Technology Conference 2024, Sep 30 - Oct 2, Phoenix, AZ American Clean Power