Scientists develop AI to save heat in urban networks

Researchers from Perm National Research Polytechnic University (PNRPU) have created a neural network–based control system for centralized heat supply that can reduce heat losses by 10–12% and cut expenses during the heating season. The AI-driven development improves the energy efficiency of urban heating networks and can be adapted for any region in Russia, according to the PNRPU press service.

The Heating Supply Problem in Russia

Centralized heating serves about 100 million Russians — roughly 70% of the country’s population. However, worn-out pipelines and inefficient temperature regulation lead to serious issues:

• Overheating of buildings: Indoor temperatures often exceed standards, causing discomfort and increasing costs.
• Heat losses: Up to 30% of heat is lost due to outdated infrastructure.
• Breakdowns: Poor network management increases the risk of failures.

These factors raise heating expenses and reduce the resilience of urban infrastructure. Notably, the problem is also relevant in other countries, including Armenia.

How the AI System Works

The PNRPU neural network uses weather forecasts and temperature and pressure data from boilers and consumption points. Key features include:

• Real-time operation: The algorithm analyzes incoming data and instantly determines the optimal temperature of the heat carrier, preventing overheating.
• Adaptability: The system responds to changing weather, for example lowering the heat carrier temperature during warmer conditions to match outdoor air.
• Accuracy: In testing, temperature prediction accuracy reached 97.9%.

The system was initially trained on a virtual stand simulating various heating networks, and later refined with real operational data. This ensured its adaptation to real-life conditions.

Advantages and Prospects

The development by PNRPU scientists offers several major benefits:

• Resource savings: Reducing heat loss by 10–12% lowers fuel and maintenance costs.
• Ease of implementation: The system does not require complex physical models and can be integrated into existing heating networks.
• Versatility: Suitable for any urban heat supply system, regardless of scale or region.

According to the authors, the system can become an important tool for improving energy efficiency and promoting sustainable urban development. It not only saves resources but also reduces environmental impact by cutting boiler emissions.

Context and Significance

The PNRPU development is especially relevant during modernization of public utilities in Russia and other countries, where heating networks in some regions are up to 70% worn out. Using AI for heat supply management could become part of a national energy efficiency strategy, supported by programs such as Energy Saving and Energy Efficiency Improvement through 2030.

Previously, scientists have applied AI in the energy sector for tasks like predicting energy consumption and optimizing power plant operations. The new PNRPU system expands these applications by focusing on heat supply.

In Brief…

Scientists from PNRPU have developed a neural network system that reduces heat loss in urban networks by 10–12% through precise control of heat carrier temperature. Using weather forecasts and sensor data, it achieved 97.9% prediction accuracy and can be adapted to any heating network. Announced in 2025, this innovation promises to boost energy efficiency, lower costs, and improve comfort for millions of Russians, highlighting AI’s role in modernizing urban infrastructure.