In the course of the energy turnaround, more and more electricity from decentralized photovoltaic and wind power plants is flowing through the power grids. In order for transmission system operators to be able to control plants when necessary, distribution system operators will soon have to create more transparency in their grids. This is provided for in the new regulation of grid management, Redispatch 2.0 for short, which will come into force on 1 October 2021. However, it will be difficult for many distribution system operators to implement the requirements on their own due to a lack of know-how and too little time. In order to support companies with the required feed-in forecasts, the Centre for Solar Energy and Hydrogen Research Baden-Württemberg (ZSW) has now developed the “GridSage” tool. It uses artificial intelligence to accurately forecast the generation of green power plants in the distribution grid for the next 36 hours. It is also possible to predict the load. Stadtwerke Schwäbisch Hall is already using GridSage successfully.
Adjusting the feed-in of green electricity power plants in the event of an overload
The legal basis is the Grid Expansion Acceleration Act. The previously applicable regulations on feed-in management for renewable energy and combined heat and power plants will be supplemented by the new redispatch system. As of the fourth quarter, green electricity and CHP plants with an installed generation capacity of 100 kilowatts or more must be included in the redispatch. For this purpose, distribution system operators must, among other things, prepare feed-in forecasts and identify redispatch potentials.
GridSage: Forecasts for Redispatch 2.0
The forecast of generation output is of central importance in the Redispatch 2.0 scenario. Good forecasts ensure that redispatch measures can be carried out cheaply and efficiently. However, this poses major challenges for small distribution system operators in particular. The ZSW forecasting tool GridSage supports them in this. The use of GridSage makes distribution networks transparent, equips them for the future and helps to avoid network bottlenecks. “GridSage forecasts the power generation in the distribution grid for the next 36 hours with a resolution of 15 minutes,” explains Dr. Jann Binder from ZSW. “We update the forecasts for the individual EEG plants and grid nodes every hour and make them available to the grid operator in an automated way.” Other time intervals and forecast horizons or flexible forecast delivery are also possible on request.
The ZSW researchers create the forecasts with the help of artificial intelligence methods: neural networks learn from past data which generation plant will produce how much power under which weather conditions. GridSage uses this information to automatically generate high-resolution forecasts. At many points, the ZSW subjects the measured data to a plausibility check, for example by comparing the nominal plant output with the annual generation or by comparing it with the generation of measured neighbouring plants, in order to exclude erroneous data.
GridSage can also do more than is required by Redispatch 2.0: “In addition to generation, the tool forecasts the load in the distribution grid,” says Binder. “In the future, this will become increasingly important. The more electromobility becomes established and heat pumps are installed in homes, the greater the importance of precise forecasts of consumption in distribution grids. In this way, the grids can continue to be operated efficiently in the future and the need for grid expansion can be reduced.”