Research On Soft Measurement Method Of Power Plant Denitration System

Thermal power is the main form of power generation in China.The NOx emissions from the thermal power unit boilers can cause environmental pollution.To achieve increasingly stringent environmental indicators,while taking into account the benefits of power plants,precise denitrification control is very important.At present,the thermal power unit is widely used for SCR denitrification.The system needs to measure the NOx concentration to carry out the corresponding ammonia denitration control.Because the smoke measuring device used in the power station has a delay in measuring the NOx concentration,the control action of the SCR denitration system is not timely and accurate.Therefore,it is important to achieve accurate and real-time measurement of NOx concentration of the power station.In this paper,the emission of the denitration system inlet and outlet is taken as the research object,and the concentration soft measurement model is established for the delay of the measurement analyzer to achieve accurate and rapid measurement of emissions.Firstly,the SCR denitration process and the principle of furnace generation are analyzed,and the influencing factors affecting the inlet and outlet concentrations of the SCR denitration reactor are obtained.Then,the field data is preprocessed and the improved mutual information algorithm is used to screen the optimal auxiliary variables.Considering the maximum correlation between the auxiliary variable and the dominant variable,the minimum redundancy between the auxiliary variables,the filtered variable is used as the input variable of the soft measurement model.The inlet concentration and outlet concentration of the SCR denitration reactor after the delay time Ts were used as the output variables of the soft measurement model at the inlet and outlet,respectively.Finally,the soft-measurement model of NOx mass concentration based on PSO-ELM and PSO-SVM is established by optimizing the input weight,implicit layer offset and SVM penalty factor and kernel function parameters of the ELM by PSO algorithm.The results show that both PSO-ELM and PSO-SVM models can accurately predict NOx.Through data preprocessing and MMIFS screening of optimal auxiliary variables,the complexity of the model can be effectively reduced and the generalization ability of the model can be improved.ELM,SVM is optimized by PSO algorithm and has higher prediction accuracy.The PSO-ELM model is faster and more accurate,and can also be applied to larger data samples.The overall performance of the PSO-ELM model is better.