Wind farm flow reconstruction and prediction from high frequency SCADA Data

Abstract

The flow inside a typical large wind from propagates through the array from turbine to turbine. We present an algorithm that intuitively processes measured values from as much as possible turbines in real-time and uses them to determine a flow reconstruction and minute-scale forecast for the downstream turbines. For the validation we used full-field measurements from the offshore wind farm Global Tech I. The flow reconstruction is compared to the measurements of one turbine, which was excluded from the algorithm and achieved a root mean square error of 0.55 ms−1 for the wind speed estimation. The flow forecasting is tested for three prediction horizons 30s, 60s and 120s. Together with automated error correction to account for calibration errors and wake effects, the flow prediction achieves a root mean square error of 0.52 ms−1 for the 120s-forecast of the wind speed, which beats the persistence forecasting method. The reconstruction allows to analyse the flow in the wind farm, to detect abnormal turbine behaviour and to estimate fatigue loads, and the minute-scale forecasting is a useful tool for predictive wind farm control and estimating the available power of a wind farm, which becomes more and more necessary for grid stability.