On-line identification of biomass fuels based on flame radical imaging and application of radical basis function neural network techniques

In biomass fired power plants a range of biomass fuels are used to generate electric power. It is desirable to identify the type of biomass fuels on-line continuously in order to achieve an improved combustion efficiency, and reduced pollutant emissions. This paper presents the recent investigations into the on-line identification of biomass fuels based on the combination of flame radical imaging and radical basis function (RBF) neural network (NN) techniques. The characteristic values of flame radicals (OH*, CN*, CH* and C₂*), including the intensity ratio, intensity contour, mean intensity, area and eccentricity, are computed to reconstruct two types of RBF NN, that is, accurate and probabilistic RBF networks. Experimental results obtained for three types of biomass fuels (flour, willow sawdust and palm kernel shell) firing on a laboratory-scale combustion test rig are presented to demonstrate the effectiveness of the proposed method.