Obtaining real-time and accurate temperature distribution information is of great significance for the clean and efficient operation of coal-fired boilers. Due to the difficulty in obtaining real-time temperature distribution inside the furnace

the high computational cost of computational fluid dynamics (CFD) numerical simulations

and the inability of existing data-driven methods to effectively track the changes in the boiler combustion system

this paper constructs a metric-driven digital twin model for the temperature field of coal-fired boilers based on model order reduction techniques. Taking a 600 MW front and rear wall boiler as the object

the numerical simulation is first carried out to obtain the temperature field dataset

and the metric-driven data enhancement approach is then utilized to explore the data-sparse regions and conduct targeted data augmentation. The proper orthogonal decomposition method is employed to reduce the order of the high-dimensional temperature field data through low-dimensional mode characterization. Then the relationship between the working condition parameters and the mode coefficients is fitted by the improved least-squares support vector machine method to adapt to the changes in object characteristics. Based on these

the digital twin model of the boiler temperature field is constructed. The results demonstrate that the model is capable of achieving real-time and precise mapping of the boiler temperature field. The model average absolute percentage error is 2.233%

and the root mean square error is lower than 41.066 K

indicating high accuracy and generalization. Furthermore

the computation time is reduced from approximately 45 000 s required by CFD to 9 s

representing a significant improvement in computational efficiency.