Abstract: Constrained by the performance of the equipment, certain issues have been existing in the main control parameters of boiler during the unit load changing process, such as large time delay and large inertia. As a result, it is not that straightforward to balance the contradiction between the needs of quick response to thermal load and stabilizing the main steam pressure. In the present work, a comprehensive method based on autoregressive moving average (ARMA) model and particle filtering is developed to perform time series forecasting on partial signals, which is designed to forecast the signals in advance such as main steam pressure such that the control delay of the main parameters could be alleviated to some extent on the boiler side. This method firstly establishes ARMA model based on historical data, then corrects the model parameters through particle filter algorithm, and at last applies the corrected model to forecast time series value. By using this method, the main steam pressure, total boiler coal quantity and main steam pressure setting value of the unit are forecasted and simulated on Matlab platform. The results show that the forecasting accuracy of this method is much better than that of ARMA model.