Temperature Prediction of Combustion Level of Ultra-supercritical Unit Through Data Mining and Modelling

Abstract This paper presents the dynamic prediction model of different combustion levels temperature in the furnace based on the data mining and modelling. The temperature of the combustion level is the important parameter to characterize the combustion state. The infrared temperature measuring device is developed to measure the temperature of different combustion levels. Based on the operation data of 660 MW ultra-supercritical coal-fired unit, the dynamic prediction models of different combustion levels (bottom, middle and upper) temperature under different load conditions are proposed by using the autoregression integrated moving average with external input model and autoregressive distributed lag model. For the different load conditions, including high-load, middle-load and low-load, variable importance in the projection algorithm is used to obtain the key variables that affect the combustion temperature of different levels. The different prediction models are compared and analyzed. The feasibility and effectiveness of the developed dynamic model for predicting the combustion level temperature using industrial data is validated. Results demonstrate that the proposed approach provides an effective tool for predicting temperature of combustion level and the potential of further improving combustion optimization.