Using High Temporal and Spatial Resolution Forecasts to Predict Ramps for the Wind Power Industry

Wind power ramps consist in large and sudden variations in wind power production resulting from rapid changes, in the order of minutes to hours, in the available wind resource. They are difficult to predict and pose an important challenge to the wind industry, as they oblige among other things network operators to plan for other sources of generation in case the wind power production would diminish unexpectedly. What makes wind ramps even more of a complex issue is that there is no absolute definition of a ramp, as what consists in a ramp is dependent on the particular application, which changes from user to user [1].

According to the IEA Wind Task 36 (Forecasting for Wind Energy), many users like system operators and traders are concerned about wind power ramps, although a standardized way of dealing with ramp forecasts is often not known to them. As such, work is suggested on the interaction between end users and forecasters, as well as on determining how to make a forecast useful and how to interpret it [2].

The Meteorological Service of Canada (MSC) is the branch of Environment and Climate change Canada (ECCC) that generates all supercomputer weather forecasts in the country. In this project, it works with Nergica, an applied research centre in renewable energies, and the University of New Brunswick on a project that aims at improving the forecasts of wind power ramps for the industry. Commercial partners are also participating to the project via these organizations. This implies that measured wind power data are made available to the project from a number of operating wind farms, and that the interaction between the end users and forecasters is taken in consideration.

In this project, the numerical weather prediction (NWP) model GEM (Global Environmental Multiscale Model), that is used operationally by the MSC to perform numerical weather predictions, is run by the authors 4 times a day on a 2.5km grid centred around the Gaspé Peninsula and Maritime provinces in Canada. It outputs the main meteorological variables like wind velocity and direction every 3 minutes. The simulated time series of wind velocities are transformed into power using the manufacturer’s power curve. This high temporal resolution allows to catch more efficiently the beginning and end of ramp events which with a lower time resolution otherwise could be missed. An increase in the temporal resolution of the NWP output has actually been recently qualified by the IEA Wind task 36 as “the easiest low hanging fruit” research issue to harvest when it comes to wind power forecasting [2].

The three methods to identify ramps described in Bianco et al. [1] are applied on chosen time windows on the data forecasted and measured at various wind farms locations. Ramp definitions to use in these methods in terms of change in power over a certain period of time are agreed on with the commercial partners. The ability of the three methods to accurately identify the ramps is compared, as according to Bianco et al. [1], it is not clear yet which method performs the best.

Scores are calculated in order to assess the skill of the high resolution forecasts to predict ramps. This is done using the Ramp Tool and Metric developed by Bianco et al. [1], which is said to be adapted to the evaluation of ramp forecasting as it weighs the model agreement for ramp events more than during periods of near-constant power. Scores are also calculated using an in-house tool that allows to perform conditional verification as a basis for determining meteorological conditions associated with the occurrence of ramps.

The results from this work will increase the knowledge into the forecasting and identification of ramp events and their causes. Through the comparisons made with experimental data, they are also expected to contribute to the improvement of the NWP model used by, among other things, putting emphasis on improving the forecasts associated to the conditions at the heart of ramp occurrences.

References

[1] Bianco, L., Djalalova, I. V., Wilczak, J. M., Cline, J., Calvert, S., Konopleva-Akish, E., Finley, C., and Freedman, J. (2016). A Wind Energy Ramp Tool and Metric for Measuring the Skill of Numerical Weather Prediction Models, Weather Forecast., 31, 1137–1156, https://doi.org/10.1175/WAF-D-15-0144.1.

[2] Giebel, G., Cline, J., Frank, H., Shaw, W., Pinson, P., Hodge, B.-M., Kariniotakis, G., Madsen J., Möhrlen, C. (2016). Wind power forecasting: IEA Wind Task 36 & future research issues. Journal of Physics: Conference Series 753 (2016) 032042, doi:10.1088/1742-6596/753/3/03204