Flame Monitoring in Power Station Boilers Using Image Processing

Combustion quality in power station boilers plays an important role in minimizing the flue gas emissions. In the present work various intelligent schemes to infer the flue gas emissions by monitoring the flame colour at the furnace of the boiler are proposed here. Flame image monitoring involves capturing the flame video over a period of time with the measurement of various parameters like Carbon dioxide (CO2), excess oxygen (O2), Nitrogen dioxide (NOx), Sulphur dioxide (SOx) and Carbon monoxide (CO) emissions plus the flame temperature at the core of the fire ball, air/fuel ratio and the combustion quality. Higher the quality of combustion less will be the flue gases at the exhaust. The flame video was captured using an infrared camera. The flame video is then split up into the frames for further analysis. The video splitter is used for progressive extraction of the flame images from the video. The images of the flame are then pre-processed to reduce noise. The conventional classification and clustering techniques include the Euclidean distance classifier (L2 norm classifier). The intelligent classifier includes the Radial Basis Function Network (RBF), Back Propagation Algorithm (BPA) and parallel architecture with RBF and BPA (PRBFBPA). The results of the validation are supported with the above mentioned performance measures whose values are in the optimal range. The values of the temperatures, combustion quality, SOx, NOx, CO, CO2 concentrations, air and fuel supplied corresponding to the images were obtained thereby indicating the necessary control action taken to increase or decrease the air supply so as to ensure complete combustion. In this work, by continuously monitoring the flame images, combustion quality was inferred (complete/partial/incomplete combustion) and the air/fuel ratio can be automatically varied. Moreover in the existing set-up, measurements like NOx, CO and CO2 are inferred from the samples that are collected periodically or by using gas analyzers (expensive). The proposed algorithm can be integrated with the distributed control system (DCS) that is used for automation of the power plant. The inferred parameters can be displayed in the centralized control room a (cost-effective solution). The major contribution of this research work is to develop an indigenous online intelligent scheme for inferring the process parameters and gas emissions in the centralized control room directly from the combustion chamber of a boiler.