Research On The Process Modeling Of Boiler Combustion Based On Bayesian Network And Fuzzy Neural Network

With the significant development of computer, the digital technology has been applied electric power systems widely. How to analyze, explore and discover useful knowledge rapidly and efficiently from these data becomes even more important. With the restructuring of power industry, power plants of coal-fired have been trying their best to improve the efficiency, reduce generate electricity cost and decrease environmental pollution. Based on combustion process of the power plant boiler, the network modeling approach of combustion system is analyzed and studied.For data mining with small samples, the Bayesian network has a strong ability to deal with the problem with uncertainty by combining priori knowledge and sample information. Based on solid theoretical basis, expressing of knowledge framework and strong reasoning power, the model on combustion efficiency is established.On the contrary, for the massive data during the operation of coal-fired boiler, the traditional RBF network and fuzzy neural network based RBF function network are applied. The principles, learning algorithms, network architecture and the robustness analysis to noise are also discussed. Still with the combustion in power plant as the example, based on the characteristics of f massive volume, process uncertainty, dynamic and diversity, high dimension and the strong correlation of the data from DCS and SIS system, the outliers are eliminated in preprocess. Then the stead state is determined and then the model is established and the useful information implicit in the data are explored.