Aiming at the problems of boiler heating surface from cleaning to ash deposition and slagging, the heat transfer efficiency of boiler is reduced. With the cleaning factor as the monitoring index, a real-time soot blowing prediction method based on unscented Kalman filter algorithm is proposed. The clean factor degradation data was analyzed by double exponential function fitting, and the model parameters were updated by the unscented Kalman filter algorithm, and the future trend of the cleaning factor was predicted. At the same time, a soot-blowing optimization model with the largest heat transfer per unit time is proposed to further optimize the sootblowing time. Taking the cleaning factor data of a economizer as an example, by comparing with the extended Kalman filter algorithm, it is found that the proposed method can predict the soot blowing time more accurately and perform the optimization of soot blowing optimization to verify the optimization proposed in this paper. The feasibility of the model.