This Aeolius Wind Systems Wind Forecasting Lessons Learnt Report outlines the key learnings Aeolius Wind Systems has learnt in the period from October 2019 – April 2020.
Report extract
The skill of four machine leaning models were evaluated using data sourced from the Macarthur windfarm. The forecast models were successful in predicting the broad trend in the incoming wind field in the 5 minutes ahead time frame, however failed to capture the turning points with the required level of accuracy. Further model development is required to improve forecast skill to permit the use of these tools in operational settings.
Similarly, a short-range (2 to 5 km) scanning Doppler lidar was trialled to evaluate its suitability for use in the power forecast applications. The lidars measurement and 5 minutes ahead forecast capability was found to be suboptimal under the environmental conditions at the test site. However, the investigation confirmed the accuracy of the wind retrievals in the far field using data from meteorological masts and provided insights into the required scanning strategies for the planned dual Doppler lidar demonstration at Macarthur windfarm in late 2020.