EWE AG operates and markets two CHP units to supply the University of Oldenburg with heat. The CHP units operate heat driven and the power generated is fed into the public grid. It is necessary to market the produced power via direct marketing to operate the CHP units revenue optimised. So for marketing success a power production forecast which is as precise as possible is essential.

Since the CHP units operate heat driven, an accurate forecast of the heat demand on the consumer side is needed initially with which to optimise operation and forecast power production. To achieve this perpendo developed, tested and evaluated a method based on artificial neural networks (ANN).

Some of the main areas of focus were the essential boundary conditions, the weather influence, the data structure, the training period and the required resolution.

In a first step perpendo developed a method for the annual heat demand forecast, whereby the hourly heat demand for an average year can be considered as the basis for the assessment of operational optimisation as well as for expansions and retrofits of the CHP plant. The method thus serves to support the evaluation of long-term investment decisions. In a subsequent step the method was extended to create short-term heat demand forecasts on the basis of current weather forecasts and recorded consumption data. This short-term heat demand forecast can then be used with the optimal operating strategy as a basis for the electrical power generation forecast.

It could be shown that short-term heat demand forecasts can be created for day-ahead trading on the basis of weather forecasts and recorded consumption data. The application example shows that a reliable heat demand forecast is already possible merely with an hourly weather forecast of temperature and relative humidity.

Heat Demand Forecasts (ANN)

 

 
zurück to top