Research On Optimization Operation Of Boiler Combustion System In Power Station And Its Application

Compared with developed countries, the primal problem with which the power Utility boilers are one of the three main frames of thermal power stations. At present the thermal utility boilers in China give priority to sub-critical and supercritical and large capacity boilers, compared with developed countries there is a larger difference with performance indexes because of the installation itself and performance management, mainly manifested by high coal consumption and low thermal efficiency. To increase thermal efficiency of the coal fired power plant boilers, save limited coal resources and reduce pollution produced in the course of coal consumption are urgent to realize continuable development of energy resources in China as well as a problem to be overcome by electrically scientific and technical personnel.The article's subject investigated is the utility boiler combustion system, and optimized theory and application of the boiler combustion system are discussed. The utility boiler installation is huge with variables, big lagging and nonlinearity. Because of highly complicated combustion process in the boiler it is unable to set up a model for combustion by use of theoretical method. Advanced artificial intelligent technology of neural network is introduced to the system, and a model for the boiler combustion system is set up by use of RBF neural network according to historical data in the boiler combustion process, and optimized operating mode with different working parameters are found in the model by use of non linear optimizing technology so that we can give guidance to boiler combustion adjustment and realize optimized operation of the boiler combustion system.Aiming at present situation that it is unable to calculate online thermal efficiency, a kind of practical methods based upon fitting formula is put forward in the article. The requirements may be met to optimize operations of the utility boiler combustion system by use of operating control parameters such as smoke emission temperature, fuel economizer outlet oxygen capacity, reference temperature, fly ash carbon content, slagcarbon content, and generating unit load through industrial analysis of coal consumption such as low heat value and ash content as well as through online calculation of thermal efficiency of the boiler without any analysis of coal elements.hi Order to refine the model to increase both precision of the model and transparency of modeling process the boiler consumption system is divided into two stages in the article. At first it is to respectively set up mathematical model among fly ash carbon content, smoke emission temperature and relevant operating parameters by use of modular RBF neural network; secondly the neural network's output capacity appears as output of the boiler's thermal efficiency calculated to set up a modular integrated model for the neural network of the coal fired power plant boilers. Thus those that can be mathematically described in the model for consumption is expressed with functions, and those that cannot be mathematically described is expressed with neural network to increase transparency of the black box model.hi the article a kind of online model for learning of RBF network and learning algorithm is put forward. Dormant nodes at implicit level will be activated, working nodes at implicit level dynamically added, and new learning sample will be taken as the center of newly added working nodes for online learning of connection weight value when there is new learning sample coming into existence or there are characteristic changes in the course of the boiler's operation. Such model may make an online study of the rule among the boiler's fly ash carbon content, smoke emission temperature, and all kinds of adjustment parameters. According to up-to-date data in consumption process of the boiler the mathematical model for consumption may be modified and optimized online so that the model may be increasingly expanded and improved as it goes by. hi the meanwhile it may be guaranteed that the model may match varied features of the boiler, enough precision of the integrated model for RBF neural network of the modular boiler consumption system may be kept all the time so that the optimized consumption may be effective for a long period of time.On the basis of the integrated model for RBF neural network of the modular boiler combustion system we adopt genetic algorithm based upon real code to optimize. Aiming at the features of RBF neural network the algorithm is improved, and the center of RBF concealed nodes is applied to heuristic optimization. It is proved by experiments...