Adjustable ε smooth support vector regression for combustion state analysis

For the purpose of reducing calculation complexity, the smooth support vector regression (SSVR) is put forward to improve the support vector regression algorithm. In this paper, SSVR is used to make furnace combustion state time series analysis model. In the analysis of the furnace combustion state time series based on SSVR, the influence of different kernel functions and parameter epsiv on the model is researched and compared; at the same time, the phenomenon of false alarm generated by the interference with camera from coal ash and coal slag is discovered in the time series fitting, and the method of adjustable epsiv smooth support vector regression (AepsivSSVR) is presented to mask the phenomenon of false alarm. About 200 furnace flame images as the sample series are used to simulate experiment. The experiment results show that the method of AepsivSSVR can obtain the good effect of analysis.