PNIPU Intelligent module for control of local heat supply system

Main article: Neural networks (neural networks)

2021: Announcement of intelligent module development for local heating system control

On February 26, 2021 it became known that uchenyepermsky Polytechnic University developed the intelligent module for management of the local system of heat supply. Neural networks will help accurately and quickly calculate the temperature of the coolant at the outlet of the boiler room. The technology allows you to keep it normal with consumers, avoid unreasonable overheating of the coolant and save money on heating. The developers published the results of the study in the article.

Architecture of a neuronet

As of February 20210, control units are quite widely used, which automatically maintain a given temperature at the outlet of the boiler room. The necessary values are determined by the operator, mainly focusing on the thermometer and the available inverse. communication Our development involves control using such neural networks, which use not only the current value of the ambient temperature in the calculations, but also a reasonable forecast. This allows you to estimate the carrier temperature in advance and avoid "lag." tells Vladimir Oniskiv, associate professor of computational mathematics, mechanics and biomechanics at Perm Polytechnic, Ph.D.

As explained, for the "training" of the neural network, scientists used a large amount of statistical data. It includes synchronized coolant temperatures at various points of the thermal network and ambient temperature.

Results of model operation on test set (red graph - actual temperature of coolant at boiler room outlet, blue - predicted by model, green - temperature difference module)

Scientists tested the intelligent module, integrating it into the software and hardware automated control system "Aurora. Heat balance in housing and communal services, "which was developed and used by one of the companies of the Perm Territory. As a result, the complex allows to automatically control the temperature of the coolant at the outlet of the boiler room, taking into account the forecast of changes in weather conditions.

To ensure comfortable thermal conditions in consumer homes, heat supply organizations must constantly monitor the temperature state of the network. But this service is not yet available for most heat companies, so they insure their risks by maintaining coolant temperatures. As a result, residents are often forced to overpay for utilities. explained researcher

Example of simulated tree structure of heating network (TC - heat collectors, MKD - apartment buildings)

According to scientists, the use of the neural network in the processes of managing the thermal network allows you to save fuel and prevent its overspending. With sudden changes in weather, this effect becomes especially significant. Gas savings can reach 10-15%, depending on the external air temperature and the general state of the heating networks.

Multilayer neural and deep learning networks are able to predict the required boiler temperature, taking into account the weather forecast and features of coolant movement.

In the process of creating an intellectual module, scientists analyzed various types of neural networks. The resulting architecture consists of 224 neurons ordered into three layers. The calculated temperature of the coolant at the outlet of the boiler house provides the values ​ ​ of the temperature at the entrance to the house, which are required by the standards.