Advanced systems of artificial neural networks allow to predict the behavior of heat consumption. Deep learning systems and other forms of artificial intelligence are now widely used in various fields of human activity. ORTEP has been involved in machine learning research for modeling heat consumption for more than 2 decades.

Consumption model based on neural networks also addresses long-term adaptation to consumption behavior changes. This model is an integral part of DYMOS product and can also be optionally used with the MOP tool. Within DYMOS, this consumption model is complemented by a very efficient system of short-term consumption adaptation.

To achieve satisfactory results, the neural network training system must be complemented by an effective training data preparation system. ORTEP has developed a suitable methodology and framework of software tools for such data preparation during a research project supported by a grant from the Ministry of Industry and Trade of the Czech Republic.

As a part of this project, ORTEP has also developed a software for comprehensive testing of created neural models of heat consumption and a stabilization system that ensures that the neural models of consumption produces reasonable values even in unexpected situations.

The used neural model of heat consumption also includes the accumulation effect of buildings during the transition from warmer to colder periods and other relevant effects. Unlike consumption models based on the thermal model of buildings, this neural model is universal and also includes the consumption of hot tap water.

In addition to MOP and DYMOS products that include the resulting trained neural networks for modeling and predicting heat consumption, ORTEP also offers separatelly the framework for data preparation and learning and testing of neural heat consumption models, including both methodology and licenses for all software tools of the framework.