Fourth Workshop on
Best Practice in the Use of
Short-term Forecasting of Wind Power

The idea of this workshop series is to bring together utility and TSO personnel using wind power forecasts daily, and to share some tips and tricks for making the best use of them. This is flanked with the latest news from research and notes from operational forecasters. For previous workshops, see here:  the first, second and third workshop.

This workshop is again co-organised with the 9th International Workshop on Large-Scale Integration of Wind Power into Power Systems as well as on Transmission Networks for Offshore Wind Power Plants on 18-19 October 2010. From 13-15 October, the regular UWIG Fall Technical Meeting was held as well.
News: the next Workshop will be in Århus, Denmark, on 24 October 2011, again in conjunction with the Offshore Workshop. Please stand by for details.

Program

0900: Morning session: Wind integration with hydro power

Gregor Giebel, Risø DTU: The new State-of-the-Art report and the European Experience in SafeWind and other projects
In 2003, I authored a 35 page report on the State-of-the-Art in Short-term Prediction of Wind Power for the Anemos project. In the last two years, I updated the report to contain 70 pages and about 300 references (to be published just after this event). A general overview is given, plus the differences between the old version and the new version is worked out. Some emphasis is given to the current two large European development projects which sponsored this exercise, the SafeWind project related to better extreme event forecasts, and the Anemos.plus project giving better decision tools based on forecasting to end users.

Wei Yu, Environment Canada: Wind Energy Forecasting R&D and Demonstration Project in Canada
W. Yu, A. Plante, S. Dyck, L. Chardon, S. Beauregard, N. Gauthier, R. Benoit, A. Glazer, Luu-Dung Tran, F. Petrucci (Environment Canada), A. Forcione, G. Roberge, P. Dionne, J. Choisnard, S. Antic (Hydro-Quebec)  
Environment Canada (EC) and Hydro-Quebec (HQ) established a joint RDD (Research and Development, and Demonstration) project on wind energy forecasting in 2006. The objective of the project is to provide the industry a state-of-the-art forecasting tool for better daily management of wind power and grid integration. This presentation gives an overview on this RDD project   The Canadian Limited Area model, GEM-LAM 2.5km, was used for prediction of meteorological variables, such as wind speed, wind direction, temperature, etc. The model was calibrated with 24 historical events in 2005 over Gaspe regions of Quebec Province. This experimental forecasting system has undergone tests in real-time for wind plants in Quebec, Canada since May 2007. The meteorologists of EC and HQ evaluate the daily forecasts using real-time observations from both EC meteorological stations and masts installed in the wind plants. The hourly forecasts are archived over the last 3 years and provide a valuable database for analysis of the model’s performance and of other meteorological phenomena. The forecasting system and validation results will be presented. Research needs for further improvement of this forecasting system will also be discussed.

Chris Otton, Transpower, NZ: What do we need from Wind Generation Forecasting?
A number of the forecasting systems available approach the generation forecasting challenge from the perspective of a centralised forecaster. Transpower, as New Zealand's system operator, finds itself in the position of requiring a forecast for it's security studies whilst maintaining a separation of the market forecasting for generation scheduling that is required of Wind Farm operators. This talk discusses Transpowers current requirements for a wind generation forecast with a view to maintaining security of supply.

Coffee Break

Late morning session: Users Area

Bill Henson, ISO New England, US: The New England Wind Integration Study - or how to learn from the early adopters of forecasting
Many recent studies as well as power system operators' experiences have demonstrated the value of windpower forecasting.  Windpower and windpower forecasting are fields where tremendous technological progress is taking place with advances being made in such areas as integrated remote sensing, windplant grid support capability, ramp forecasting, etc.  Is there an optimal way forward for the power system integration of weather driven generation (e.g. wind, solar) and the forecasting that facilitates these resources?  Is it possible for the systems being developed and implemented to "leave room" for the incorporation of new capabilities-both in the forecasting itself and in the data acquisition used to drive the forecasts?  What lessons can be learned from the pioneering integration experiences and early adopters of windpower forecasts?  ISO New England is finalizing the results of the New England Wind Integration Study and has been considering these questions as the region moves forward to implement the recommendations developed during the study.

László Varga, E.On Hungary: Probabilistic forecasts for balance circles included wind generation
In market driven electricity trading the balance circle is a group of producers, marketers, and consumers where a balancing mechanism is used to match supply and demand in all time periods (e.g. every quarter hour). The difference between the actual consumption and supply is balanced by external sources or the transmission system operator, which results in the purchase costs and imbalance charges. In this presentation an illustrative model of a balance circle is presented to plan the cost of balancing energy. The circle includes a wind turbine and an electricity consumer. The air temperature and the wind speed influence the electric loads and the wind power output, consequently, the accuracy of meteorological forecasts effects the planned electricity purchase.   Mathematical models are developed for the wind generator and the consumer loads. The application of these models allows us to predict the balancing energy requirements using deterministic (best guess) and probabilistic (ensemble) weather forecasts. Since the imbalance price is much more expensive than the spot market price a good schedule of energy purchase can reduce the balancing charges. Based on the ensemble and ensemble mean weather forecasts for the wind speed and air temperature we formed two schedules for the spot energy purchases. Employing the probabilistic (ensemble) weather prediction decreasing in operational costs could be achieved.

1230-1330: Lunch

After lunch session: Operational Forecasters Area

Craig Collier, GL GarradHassan US: Extreme Volatility of the Wind:  Forecaster Challenges and Perspectives from Forecast Users
This presentation focuses on the causes of large fluctuations in wind for western U.S. sites in complex terrain.  The challenges faced by forecasters for capturing “ramps” both deterministically and probabilistically are described, and perspectives from forecast users are presented.  Cost and profitability impacts are defined, and metrics for measuring forecast model fidelity for ramp detection and simulation are provided.

Amanda Fong, WindLogics, US: WindLogics Operational Wind Power Forecasting: Need, Methodology, and Challenges
The downturn of the United States economy has emphasized the need for an efficient, timely plan for incorporating renewable energy into the grid system.  Feeling the financial pinch from power consumers, the independent system operators need to accurately balance generation and load on the power grid in order to mitigate overall cost.  To do so, the system operators and utilities work together to schedule and coordinate power generation, transmission and load demand.  The utilities need an accurate wind power forecast to provide the system operator with the best future wind generation prediction.  WindLogics intelligently combines public weather data with wind plant information to create wind speed and power forecasts.  In order to provide the best forecast possible for our customers, a mathematical computational learning system (CLS) is used to find trends comparing past power forecasts and actual generation.  Then going forward, the forecast uses scaled data learned from the numerical relationships identified by the CLS, thus improving overall forecast accuracy.  Challenges with wind power forecasting include impediments with forecasting small scale weather features, as well as quality and availability of meteorological and operational data.

Ulrich Focken, energy&meteo systems, DE: Experiences with Extreme Event Warning and Ramp Forecasting for US Wind Farms
A reliable ramp forecast and warnings that indicate extreme events, like shut-downs due to high wind speeds or icing are of high value for a system operator. We present recent advances and practical experiences with ramp event forecasting and warnings for wind farms in the United States (in particular, the BPA ramp event project in the Pacific Northwest) and from south-eastern Australia.

Coffee Break

Closing session: News from research

Audun Botterud, Argonne National Laboratory, US: Advances in probabilistic forecasting with application to wind power trading
A. Botterud, J. Wang (Argonne National Laboratory), R. J. Bessa, H.Keko, J. Sumaili, V. Miranda (INESC Porto)
The first part of the talk will give a brief overview of different approaches to probabilistic wind power forecasting, with focus on two new kernel density forecast methods. We will compare initial results from the two methods to traditional approaches based on quantile regression. The second part will present an application of probabilistic forecasts, i.e. a model for optimal trading of wind power in electricity markets under uncertainty in wind power and prices. We will show examples from an illustrative case study of a hypothetical wind farm in Illinois. 

Jian Ma, Yuri Makarov, Pacific Northwest National Laboratory, US: A New Wind Forecast Error Model and Integrating Wind Generation Forecast Uncertainty into Dispatch and Unit Commitment  
This talk covers two projects regarding wind forecast error simulation and integrating wind generation forecast uncertainty into dispatch and unit commitment. 1) A new method using multiple cross-correlated random processes to simulate wind forecast errors is developed. The method is based on a transition probability matrix computed from an empirical error distribution function. The derivation of the method and some experimental results obtained by generating new forecast errors together with their statistics are presented. 2) An approach to evaluate the uncertainties of the ramping capacity and requirements caused by intermittent generation is developed. A probabilistic algorithm based on histogram analysis is used. A framework and a demonstration tool for integrating the proposed method with an EMS system are developed. 

Joel Bedard, École de technologie supérieure, CA: Geophysic Model Output Statistics for Short-Term Numerical Wind Predictions over Complex Sites
Joel Bedard, Wei Yu, Yves Gagnon, Christian Masson: A Geophysic Model Output Statistic (GMOS) module is developed and applied for optimal use of the NWP for a complex, coastal site in Atlantic Canada.  GMOS differs from other MOS that are widely used by meteorological centers in the following aspects: 1) it takes into account the surrounding geophysical parameters such as surface roughness, terrain height, etc., along with the wind direction; 2) the proposed GMOS can be directly applied for model output correction without any training, although a training of the GMOS will further improve the results. The GMOS was applied to improve the Environment Canada GEM-LAM 2.5km forecasts at North Cape (Prince Edward Island, Canada).  It was found that the GMOS improves the predictions RMSE by 25-30% for all time horizons and almost all meteorological conditions.  Also, the topographic signature of the forecast error due to insufficient grid refinement is eliminated and the NWP combined with the GMOS now outperforms the persistence model from a 2h horizon, instead of 4h without the GMOS.  Finally, the GMOS developed for North Cape was applied and validated at another site located in Bouctouche (New Brunswick, Canada).  Similar improvements on the GEM-LAM 2.5km forecasts were observed, thus showing the general applicability of the GMOS.

1700 (actually, was 1800): Close of workshop

 

Date and Venue: LOEWS Hotel Le Concord, Québec City, Canada, Room Jean Paul Lemieux. 16 October 2010, 09:00 to 17:00.

Registration: Please go here (you will need to set up an account, it's quick and afterwards you can change your data yourself, if need arises). There is a workshop fee of 199 EUR.