Research On Combustion System Modeling And Performance Optimization Of 1000MW Double Tangential Circle Coal Fired Boiler

It is of great significance to improve the operation efficiency of thermal power plants and reduce the use of fossil energy for carbon peak and carbon neutralization.The performance of combustion system in thermal power plant directly affects the power generation efficiency and operation cost.If the combustion system can be described quickly and accurately,which has nonlinear,large inertia,time delay characteristics,we can provide real-time operation adjustment suggestions from the aspects of economy and reliability,so as to optimize the performance of combustion system and improve the operation efficiency of power plant.Taking the combustion system of a 1000 MW Double Tangential pulverized coal boiler as the research object,this paper analyzes the structure and operation parameters of the combustion system,selects the relevant influencing parameters of the boiler thermal efficiency,and applies the Spearman and Pearson correlation coefficient method to determine the load,calorific value of coal,ambient temperature,temperature,etc The main steam pressure and other 13 strong correlation parameters are used as the input parameters of fast modeling.In order to ensure the accuracy of combustion system modeling,the steady-state condition data screening and condition boundary division are carried out for the system operation data.A method of steady-state condition identification based on filtering method is proposed to reduce the influence of system time delay and thermal inertia.Experiments show that the accuracy of the method is 97.22%.The binary Kmeans clustering algorithm is used to divide the steady-state condition data.The elbow method and the contour coefficient method are used to determine the optimal K value.The effectiveness of the algorithm is demonstrated by comparing the isometric method.Then,a boiler thermal efficiency model based on LSTM neural network is built by using keras3.0 architecture.Through the steady-state data modeling experiments of different structural parameters,a three-layer network structure with 5 min time step,29 hidden layer nodes and Adam optimizer with learning rate of 0.005 is determined.Finally,the optimal parameters of PSO are determined,that is,inertia weight is0.6,individual learning factor and social learning factor are 1.0.Particle swarm optimization(PSO)was used to optimize the input parameters of the above 13 models with the objective of minimizing the total amount of heat loss of exhaust gas and incomplete combustion loss of solids.In the full operation condition optimization test of 1000 MW Double Tangential pulverized coal fired boiler,the average heat loss can be reduced by 1.095%,and the maximum can be reduced by 2.07%.And in the process of field test,the program runs rapidly.