Research On The Predictive Modeling For The Carbon Content Model Of Pulverized Coal Of The Primary Air Duct Of Power Plant Boiler With Optimization Based On SVR

The coal-fired power generation is still dominated in China,which is determined by coal-dominated energy structure. The main energy is coal for coal-fired power generation. But the coal market changes frequently now, with the result that the coal becomes unstable for the power plant. As a result, the coal burning in power plant boilers is often inconsistent with the design coal, which will badly affect the safe operation and economic operation of the power plant boilers. The carbon content of pulverized coal can reflect the coal types in some extent. Therefore, for the instability problem of the coal types about boiler combustion, predictive modeling for the carbon content of pulverized coal of the duct pulverized will be researched in this thesis in the hope that will provide real-time online analysis of coal for reference. The main contents of the thesis are as follows.Firstly, the main factors of the measuring process are deeply analyzed for the carbon content of the pulverized coal, aiming at laying the foundation for predictive modeling of pulverized coal carbon content. Static electricity will be produced by friction and collision between the coal particles in the pneumatic conveying process,and the charged pulverized coal particles contain the carbon content information. Thus electrostatic technology is adopted to collect the electrostatic signal of coal particle.On this basis, the theoretical analysis and experimental data analysis are made for the relationship between the carbon content of coal powder and electrostatic signals.Meanwhile, the main factors of the pulverized coal carbon content in the measuring process are deeply analyzed by combining with correlation coefficient method and experimental data, in order to determine the main auxiliary variables of the model.Secondly, support vector regression(SVR) is employed in the research on the modeling of the carbon content of pulverized coal. The best normalization method and best kernel function are determined through comparative analysis for different normalization methods and kernel functions. The SVR model is trained and tested by using the field experimental data, and the obtained experiment results are comparedwith these obtained by using the BP-ANN model. The simulation results show that the SVR model is more accurate, holds better performance of promotion, and can better predict the carbon content of coal boiler of the duct pulverized.Finally, an improved generic algorithm is presented to optimize the prediction model of coal carbon content. Combining with cross-validation method, the optimization model of the carbon content of pulverized coal based on GA-SVR is explored and established. The training and testing error results charts, the comparing results charts between the true value and the predict value results, and the comparing results chart of a variety of performance indicators of the three models are obtained through simulation analysis. The results show that the improved GA-SVR optimization model has better prediction results than SVR model and BP-ANN model.