Researchers at Peter the Great St. Petersburg Polytechnic University (SPbPU) have developed a hybrid architecture that combines digital twins of thermal power plants with artificial intelligence. This system optimizes plant operation modes in approximately 15 seconds, which is 16 times faster than traditional mathematical methods.
The project is part of the university's development program for 2025–2036 under the Priority-2030 initiative. The system aims to manage the equipment lifecycle by balancing reliability, fuel efficiency, and environmental impact. It is designed to replace outdated normative maintenance schedules with condition-based management.
In 2026, the team introduced a module for automatic intelligent re-parameterization of digital models, allowing for real-time calibration based on equipment degradation. Tests on a facility with five energy blocks and T-100 turbines utilized a dataset of approximately 280,000 operational states to train the surrogate models.
According to the developers, implementing these optimized modes could increase marginal income by one-third over a six-month period. The system is designed to be scalable for power plants across Russia and neighboring countries, provided that data from automated process control systems is integrated.
