Online estimation of regional water-wall temperature fields in coal-fired boilers using a CFD-calibrated data-fusion framework

During deep-peak shaving operations, coal-fired boilers encounter the issue of water-wall overheating. This study proposed a CFD-calibrated data-fusion framework for online estimation of regional water-wall temperature fields in a 660 MW ultra-supercritical boiler. First, based on computational fluid dynamics (CFD) numerical simulation results, the combustion zone was discretized into a 10×10 regular grid, from which a representative regional heat flux was extracted. By combining CFD results with distributed control system (DCS) measurements, an empirical wall temperature-heat flux ( ) relationship was established to calibrate the CFD results. Subsequently, annual DCS operational data were used to derive a statistical wall temperature-load ( ) relationship, which extended the calibrated baseline fields from representative loads to a continuous operating range. Furthermore, a data-driven dynamic deviation prediction module was introduced to estimate wall temperature deviations at the available measurement locations; the hiking optimization algorithm-based stochastic configuration network (HOA-SCN) was selected as the implementation method for this module, achieving high accuracy (MAE: 1.556–3.474°C, RMSE: 2.110–4.869°C, R 2 : 0.9401–0.9875). Finally, the prediction deviation was superimposed on the baseline temperature field to obtain dynamically corrected regional wall temperature estimates. This framework provided a computationally efficient basis for online thermal trend monitoring, relative hotspot tendency identification, and overheating risk assessment during flexible boiler operation.