Forecast of control energy for energy trading

Using historical data and forecasts from the energy sector, the balance of physical control energy calls is to be
in the German grid control network (NRV) are forecast.

L

Identification of the sign of the balance with a quality of 83% for positive balances

!

Significant improvement of the absolute deviation in comparison for simple updating



Walkforward model that is constantly improving with new data

Challenge

An energy service provider would like to Machine Learning model for Forecast of the positive or negative balance of control energy in the Grid Regional Network (NRV). The aim is to be able to operate on the volatile to be able to trade more efficiently in the electricity market.

Solution

Numerous data sources are cleaned up and integrated into a unified Time series format merged. A visual and data-based Exploration of correlations and possible influencing factors will be is carried out. Subsequently, several regression and Classification models for the prediction of the balance or its Sign and application of a two-stage ensemble model created.

Result

Es there is a prototype of a forecasting model that is geared towards this, with currently available data, the balance of control energy in 15 minutes to predict. This allows the electricity producer to position itself in the market. better position.