Combustion Diagnosis And Emission Prediction By Means Of Flame Visualization

The process of combustion in a large-scale power plant boiler involves complex physical and chemical reactions occurred in large space, having the characteristics of frequent fluctuations and conspicuous three dimensions. The combustion visualization is a new subject with flourishing development in the latest forty years. Its successful applies in the combustion diagnose of boiler in power plant have aroused many people's interests.This paper presents a study on the calibration method for a flame/temperature measurement system based on an optical sensing system. This system was calibrated using a blackbody furnace as a standard temperature source. The relationship between flame temperatures and the grey levels of the images was established through image processing and function correlation. The results could be used for temperature measurement and flame monitoring.On-line measurements were conducted on a 50MW and a 300MW coal-fired boilers by the use of the flame image processing-based detection system. This paper also presents the outcome of the measurement of temperature distribution inside a pulverized coal boiler in a power station by using a novel flame/combustion visualization system. In addition, an analysis has been performed on the relationship between the flame image parameters and characteristics of the coal and the actual output of the boiler respectively. The results of the analysis indicate that the energy signals of the image and the furnace temperature field can reflect the combustion condition in the boiler correctly and quantitatively.Experimental results produce the flame images and temperature distributions that are further translated into digitized data. The characteristic parameters are defined as the average and the maximum grey levels, the variance, the entropy, the abundance and the energy of the flame images acquired at different optical wave lengths, and theirvalues are extracted from the experimental data. Then a feature space are formed consisting of the above six characteristic parameters for the LS-SVM (Least Square Support Vector Machines) analysis, based on which the NOx emission is predicted to find out the relationship between emission and the parameters of the flame images. The results show that the predicted values and the measured values are in good agreement.