- Stokvis reliability wins repeat work with social housing specialist
- How to engage your employees with energy saving
- Average UK businesses risking £2.8m of annual revenue through lack of energy resilience
- Rinnai - combined heating and water systems
- Are Energy Managers ready for an upsurge in on-site energy demand?
Kempten University patents advanced thermal load prediction method
For district heating networks, overal pipeline lenghts of several 100 kilometers are typical. This makes their operation a significant challenge. A more accurate prediction of future heating loads would allow operators to provide loads more efficiently and climate-friendly. A novel method developed by Kempten University of Applied Sciences in Germany is able to significantly increase the precision of thermal load predictions.
Load profiles in district heating grids depend on many parameters: If it is cold outside, higher heating loads are needed. Weekends, weekdays or holidays exhibit their own typical load patterns.
The novel method does exploit these unique characteristics: “Artificial intelligence has proven to detect patterns with very high accuracy. An example is speech recognition in mobile phones, which has become very powerful in the last years, with error rates of only a few percent. We are using similar methods for predicting future load profiles in district heating or cooling networks. Our approach does not require information of individual consumers, but only aggregated data, for example regional weather forecasts “, explains Till Faber, a PhD student and one of the main developers of the novel approach.
“Improvements in prediction accuracy can lead to significant savings in CO2 emissions and costs, since heat and power generation plants can be operated more efficiently.”, adds Prof. Matthias Finkenrath, who is leading the research project “KWK-Flex” funded by the German federal government, under which the technology has been developed. “The combination of domain knowledge from energy engineering with tools and competences from informatics, in particular by my colleague Prof. Brauer, turned out to be the key factor for success. By this we were able to develop a very effective and flexible machine learning tool – based on so-called ‘deep learning’ –, which can be used in many applications in energy and process engineering”, says Finkenrath.
A European patent application has been recently filed on the novel method for load prediction, which is currently further validated and optimized jointly with the two district heating operators, Fernwärme Ulm GmbH und ZAK Energie GmbH.
For further information visit: www.hochschule-kempten.de/kwk-flex