A sliding pressure optimization method based on RePos-DXGB heat rate prediction algorithm
The RePos-DXGB heat rate prediction algorithm, which combines a sliding window time-series reorganization and a Transformer encoder with an XGBoost model that gradually decays the learning rate, solves the problem of reduced efficiency of the thermal system under sliding pressure operation. It achieves stable prediction of heat rate and optimization of sliding pressure, thereby improving the unit's economy and regulation adaptability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- TAIYUAN UNIVERSITY OF TECHNOLOGY
- Filing Date
- 2026-04-16
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies are insufficient to effectively address the issues of decreased thermal system efficiency and increased power generation heat consumption rate caused by deviations in unit operating conditions from design conditions under sliding pressure operation. In particular, the adjustment and support capabilities are insufficient after the proportion of renewable energy power generation increases. Furthermore, existing models have shortcomings in dynamic characteristics and multi-parameter coupled interactive modeling.
The RePos-DXGB heat rate prediction algorithm is adopted. Through sliding window time-series reorganization, relative position encoding and multi-layer Transformer encoder modeling, combined with the XGBoost model with progressively decaying learning rate, the nonlinear mapping of heat rate is realized. The algorithm is optimized under the constraints of the feasible range of main steam pressure and the slope of the sliding pressure curve, and the optimal main steam pressure and sliding pressure curve are output.
It improves the stability and engineering generalization ability of heat rate prediction, enhances the economy and regulation adaptability of the unit, realizes closed-loop optimization of sliding pressure setting, and enhances the operational adaptability and economy of the unit.
Smart Images
Figure CN122413291A_ABST