Application Research Of Information Fusion Clustering Method In Boiler Combustion System

The issues of how to run the boiler in a safe and economical manner and reduce thegeneration cost in an efficient way is the top urgency in thermal power plants. The controllevel of boiler combustion system has a great effect on the safety and economy of plantoperation, thus research on boiler control is of highly theoretical and realistic significance.This paper focuses on how to apply the information fusion clustering method into boilercombustion system control for the sake of a safe and economical running and lowergeneration cost.A deep and thorough study has been carried on into the problems of uncontrollabilityof the combustion system, large number of variables and unsatisfactory clustering effect inthe combustion process. It is conducted as follows.First of all, based on the comprehensive study of multi-sensor information fusion,this research describes and designs the fusion structure and function structure inmulti-sensor information fusion system, then the applied ART2network and BP networkalgorithm in this system is explored. After discussing sensor information fusion methodand Hybrid neural network method, the closed loop control system on CFBB combustionprocess is proposed.Then, it studies and introduces the dynamic energy balance equation, dynamicmaterial balance equation, dynamic carbon balance equation, dynamic oxygen balanceequation and evaporation area pressure dynamic balance equation of dense-phase area,transition area and dilute-phase area in the circulating fluidized bed boiler combustionsystem. After the study of vapor pressure, bed temperature, oxygen content and materiallayer different pressure controlled object site dynamic response curve, the mathematicalmodel and matrix equation on vapor pressure, bed temperature, oxygen content andmaterial layer different pressure are proposed with variation of coal feeding, primary air,secondary air, bed thickness and combustion rate disturbance.Furthermore, with combination of sensor-based information fusion and clusteringmethod, this paper researches on boiler control problems. It studies control problems of CFBB combustion system based on multi-sensor information fusion and hybrid neuralnetwork, meanwhile it designs boiler combustion clustering fusion control system basedon multi-sensor information fusion, ranging from data level to feature level and todecision level. And it has made a complete data fusion getting a thorough description ofoperating condition, and carried corresponding control strategy according to the operationfeatures of each group. The simulation experiment proves that this design of the controlsystem has strong feasibility and effectiveness, and the control system still has satisfactorycontrol effect especially in the case of failure of all sensors.Finally, to overcome excessive variables and unsatisfactory clustering effect inpulverized coal fired boiler running, this research has put forward fuzzy clustering methodbased on adaptive particle swarm and carried out fuzzy clustering on physical quantities ofprimary air, secondary air, oxygen content and exhaust gas temperature. The simulationexperiment proves that the clustering algorithm can effectively settle the problems broughtby initial value sensitivity and getting in local minimum in FCM.