In Russia, approximately 100 million people use centralized heating services, accounting for 70% of the total population. The operation of heating networks must ensure high-quality and energy-efficient heating while considering climate conditions and equipment status. However, manual or automatic sensor-based control systems often struggle to meet these requirements. Wear and tear on heating pipelines can cause heat losses of up to 30%, increasing the burden on residents and triggering emergency situations in buildings. Scientists at Perm Polytechnic University have developed an artificial intelligence-based intelligent management system for urban heating networks. The related research was published in the Proceedings of the 14th All-Russian Conference on Management Issues and was carried out within the framework of the "Priority-2030" strategic academic leadership program.
Centralized heating is suitable for apartment buildings connected to urban heating networks (thermal power plants or boiler houses), providing residents with heating and hot water. The heat carrier flows from the boiler house through pipelines and heating points into the building, where the temperature is regulated, and then distributed to apartments via risers. However, managing the heating system is complex. Many boiler house equipment are outdated and lack automated control, requiring operators to manually set parameters. Heating conditions are affected by multiple factors such as weather, pipelines, and heat carrier flow rate, making it crucial to obtain accurate data for each section of the network.
In recent years, artificial intelligence methods have been actively applied to solve such problems. For example, intelligent ventilation control systems in subways use neural networks to improve air quality and reduce energy consumption. However, no similar solutions have yet been developed for heating control. The urban heating network management system developed by Perm Polytechnic University is based on neural networks and takes into account weather forecasts and equipment technical conditions.
Vladimir Oniskiv, Associate Professor of the Department of Computational Mathematics, Mechanics and Biomechanics at the National Research and Policy University of the Philippines and Candidate of Technical Sciences, stated that the neural network algorithm developed for the system can analyze weather forecasts in real time based on local meteorological data and assess the status of the heating network using temperature and pressure sensors installed at the boiler house outlet and user inlets. It selects the appropriate heat carrier temperature to eliminate insufficient heating or overheating in apartments.
The scientists trained the neural network on a specially created virtual test bench, which randomly simulates heating network operation under different operating characteristics, heat losses, and predicted weather conditions. After initial training on large datasets, the algorithm is further trained under real heating conditions, allowing the development process to adapt to any heating network.
During testing, the algorithm achieved an accuracy of up to 97.9%, accurately predicting the heat carrier temperature at the boiler house outlet to ensure comfortable indoor conditions. Its advantages include rapid adaptation to weather changes, adjusting temperatures close to standard values, eliminating "overheating" phenomena, and saving approximately 10–12% in energy costs during the heating season. The algorithm can precisely and automatically adjust the heat carrier temperature according to weather forecasts—for instance, lowering the temperature in advance when a temperature rise is predicted.
The research achievements of Perm Polytechnic University scientists abandon complex physical models of heating networks in favor of flexible artificial intelligence solutions. The neural network algorithm for heating management is expected to become a powerful tool for the sustainable development of urban infrastructure, reducing costs, and providing residents with a comfortable experience.
