Close
Perú is a challenging region for wind modelling. Winds there are strongly directional and follow a marked daily cycle, and without local data a model can miss the local wind by a wide margin.
To test how well our model performs there, we teamed up with Acciona. Acciona provided cleaned 10-minute data from three met masts, which allowed us to run a full wind data validation of Vortex TIMES in Peru.
The conclusion is clear. On its own, the model overestimated mean wind speed by 15.6%. Calibrated with on-site measurements through TIMES Remodeling, the bias fell to 1.6%, measured on data the model had never seen.
Vortex TIMES is a long-term, high-resolution wind time series at 10-minute resolution. It covers up to 30 years and heights from 10 to 440 metres above ground level, so you get wind speed, direction and variability at hub height before a single mast goes up.
TIMES Remodeling is a measure-correlate-predict (MCP) method built into TIMES. It uses your on-site measurements to correct the TIMES series in the time domain. Because the correction works step by step rather than on summary statistics, the remodelled series keeps everything TIMES offers: wind speed, direction, standard deviation and turbulence intensity, all consistent with each other.
Three met masts on the Peruvian coast, all about 25 m above sea level, provided 10-minute wind speed, direction and standard deviation.
Site | Measurement heights | Valid data | Observed mean wind speed |
|---|---|---|---|
Site 1 | 60, 80 m | 3.0 years | 8.1–8.6 m/s |
Site 2 | 20, 40, 60, 80 m | 4.0 years | 7.3–8.6 m/s |
Site 3 | 20, 40, 60 m | 4.4 years | 5.8–6.6 m/s |
The winds here are strongly directional, with a clear southeasterly prevailing sector and a marked daily cycle. TIMES reproduced these patterns, and the validation measures how much calibration adds on top.
Each dataset was split into two independent periods:
This out-of-training approach is what makes the numbers meaningful. It shows how the corrected series performs in a period that was not used for calibration, which is how an MCP-corrected series is used in practice.
Calibrating TIMES with Acciona’s measurements reduced error on every variable tested, at all sites and heights, on unseen data.
Metric (out-of-training, all sites and heights) | Original TIMES | TIMES Remodeling |
|---|---|---|
Mean wind speed absolute bias | 15.6% | 1.6% |
Wind speed RMSE, 10-min | 2.05–2.22 m/s | 1.41–1.67 m/s |
Wind speed R², daily | 0.56 | 0.71 |
Wind speed R², monthly | 0.33 | 0.71 |
Weibull A parameter, absolute error | 1.13 m/s | 0.16 m/s |
Histogram error (Earth Mover’s Distance) | 1.15 | 0.36 |
Wind direction absolute bias | 4.2° | 1.9° |
Wind rose RMSE | 1.08% | 0.59% |
Wind speed standard deviation, absolute bias | 31.5% | 1.7% |
Turbulence intensity curve MAE | 26.5% | 12.3% |
TIMES Remodeling uses Vortex TIMES as a long-term reference and applies corrections based on on-site measurements, so the series better reflects local wind conditions over the long term. Because every result in this validation comes from a period that was not used for calibration, it shows the agreement with measurements you can expect outside the measured period.
This post covers the headline numbers. The full report goes deeper, with results by site and height, calendar views of the calibration and validation periods, wind roses, daily cycle charts and turbulence intensity curves.
Download the Vortex TIMES Remodeling validation report (PDF) →
Want to see how TIMES Remodeling performs at your own site? Contact our team and bring your mast data.
Wind data validation compares modelled wind data with real measurements from met masts or lidars. It quantifies how accurate the model is for wind speed, direction, variability and turbulence, so analysts know how much they can rely on it.
Vortex TIMES is a long-term modelled wind time series at 10-minute resolution. It covers up to 30 years and heights from 10 to 440 metres above ground level, anywhere in the world.
TIMES Remodeling is a measure-correlate-predict (MCP) method built into Vortex TIMES. It uses on-site measurements to correct the TIMES series in the time domain, keeping wind speed, direction, standard deviation and turbulence intensity consistent with each other.
Each dataset was split into two independent periods. One was used to calibrate the model; the other was kept hidden and used only to score the results. Out-of-training results show how the corrected series performs in years the mast did not measure.
In this study, mean wind speed bias fell from 15.6% to 1.6%, and wind speed standard deviation bias from 31.5% to 1.7%. Monthly correlation (R²) rose from 0.33 to 0.71.
Yes. The mean error of the turbulence intensity curve fell from 26.5% to 12.3%, because turbulence intensity depends directly on the corrected wind speed and standard deviation.
Modeled wind resource data for the wind industry.
At any site around the world. Onshore and offshore.
Product news, events and webinars from Vortex. You can unsubscribe at any time.
