Research On Modeling And Optimal Control Of Combustion And Denitrification Process In Coal-fired Power Station Boiler

In recent years,with the booming economy of China,the demand for electricity is increasing,and the main form of domestic power supply energy is still coal-fired power generation.The process of coal-fired power generation produces pollutants such as nitrogen oxides,which brings great impact and challenges to the living environment of human society.In response to the national energy conservation and emission reduction and green development strategy of the power industry,it is imperative to achieve ultra-low emissions from coal-fired power stations.Up to now,great results have been achieved by adding or modifying hardware equipment such as denitrification devices to achieve low nitrogen emissions,but there is still room for optimization by means of software such as artificial intelligence algorithms to achieve modeling and optimal control of power plant boilers.Therefore,in view of the complex characteristics of the boiler operation process of coal-fired power stations,such as multi-parameter,non-linear and multi-conditions,it is of great significance and value to carry out research on modeling and optimal control of the combustion and denitrification process of power station boilers in this thesis.The main research works of this thesis are as follows:（1）For the complex characteristics of power station boiler combustion process with multi-parameter coupling and multi-variable working conditions,the most relevant feature variables of the boiler combustion process model are firstly screened based on the correlation analysis method and random forest feature selection algorithm.Then,the Mini Batch KMeans clustering algorithm is applied to classify the dataset into boiler working conditions clusters.Next,a Stacking fusion framework prediction model（referred to as,Stacking-XRLGL）based on XGBoost（X）,Random Forest（R）,Light GBM（L）,GRU（G）,and linear regression（L）is established to achieve accurate prediction of boiler chamber outlet NO_Xconcentration and power supply coal consumption.Finally,based on the deep reinforcement learning of the DDPG algorithm,the optimal control of controllable operating parameters of the boiler combustion process is achieved.When the optimization objective is to reduce the NO_Xemission concentration at the furnace outlet,the outlet NO_Xconcentration at the combustion stage decreases from 227.697 mg·m^-3to 216.840 mg·m^-3after the optimization,but the economic cost of the combustion process increases by 1.48%.When the optimization goal is to reduce the cost of the boiler combustion process,the economic cost can be reduced by 1%to 3%,but the NO_Xemission concentration at the furnace outlet increases by 4.95%.（2）For the SCR power plant denitrification process with multi-variable working conditions,Firstly,the data of the characteristic variables most relevant to the denitrification process model are collected from the DCS system based on expert experience.Then,the dataset is clustered into denitrification working conditions based on Mini Batch KMeans clustering algorithm.Next,the Stacking-XRLGL multi-model fusion algorithm is used to achieve accurate prediction of NO_Xemission concentration at the SCR denitrification outlet.Finally,based on the deep reinforcement learning of the DDPG algorithm,the optimal value of the controllable operating parameters of the denitrification process is sought under the constraint that the denitrification efficiency is in the range of[85%,95%]with the optimization objective of reducing the cost of the denitrification process,and the economic cost of the denitrification stage is reduced by about 6.74%after the optimization.（3）Based on the research content（1）and（2）,the boiler combustion and denitrification processes of power station boilers are firstly combined to construct a boiler combustion and denitrification synergistic total economic cost minimization model.Then,based on the deep reinforcement learning of the DDPG algorithm,the optimal values of each controllable operating parameter of the power plant boiler synergistic system are sought with the optimization objective of reducing the total cost（total cost=combustion process cost+denitrification process cost）,under the constraints of meeting the national emission standard for NO_X(below 50 mg·m^-3)and the denitrification efficiency in the range of[85%,95%].Finally,the experimental optimization control results in a total cost reduction of 1%to 3%and a total economic cost reduction of 0.89%and 2.26%compared to the boiler combustion single-stage and SCR denitrification single-stage optimization,respectively.