Optimization of minute-scale power forecasts of offshore wind farms using long-range lidar measurements and data assimilation
The ParkCast project aims to develop, optimize and evaluate new methods for short term forecasts of the performance of offshore wind farms. The power forecasts focus on the time range up to 60 minutes with high temporal resolution. The aim is to significantly improve the temporal resolution and forecasting quality of the parking performance in the above-mentioned time period and thus make a contribution to grid stability and supply security. To this end, long-range lidar measurement data are assimilated into a high-resolution, local weather model using new methods based on machine learning (ML). Physical and advanced ML-based prediction models are then used for the power prediction and validated in real time for the alpha ventus offshore wind farm as part of an online test phase.
- Measurement of the wind field in alpha ventus using long-range lidar up to a range of 10 km.
- Assimilation of lidar data into a high-resolution weather model and development of forecast methods for the wind farm performance in the time range up to 60 minutes.
- Evaluation of the different forecasts methods for different forecast horizons and demonstration of the methods in online operation.
The worldwide expansion of wind energy - especially in the offshore sector - requires the continuous improvement of forecasting methods in order to ensure grid stability and security of supply. Conventional methods for predicting the performance of wind turbines and wind farms are based on weather forecasts and - if available - current local measurement data as the feed-in. The latter allow to improve the first minutes of prediction on statistical relationships, as the spatial and temporal resolution of regional prediction models is not sufficient to capture local changes in wind speed. The aim of the ParkCast project is to close the temporal gap, which is neither captured well by local point measurements nor by weather models. We achieve this by using a long-range lidar to look a few kilometres into the inflow of the wind turbine and thus into the "future".
Since it is difficult to draw direct conclusions about future power output from the wind speed at a distance of a few kilometres, we will investigate various methods for this purpose. A very promising candidate is the use of a high-resolution, local weather model into which the lidar data is fed ("assimilated"). However, there is another shortcoming here, namely that the current assimilation methods are very (computationally) complex. We will tackle this problem by using machine learning methods, including deep neural networks. The goal is to realize and evaluate power forcasts of the wind farm up to 60 minutes using different methods.
Work packages (WP)
Coordination and partners
Contacts and partners
Stuttgart Wind Energy (SWE) at Institute of Aircraft Design
Prof. Dr. Po Wen Cheng
|+49 711 / 685-68258|
Center for Solar Energy and Hydrogen Research Baden-Württemberg (ZSW)
|+49 711 / 7870238|