Research On NO_x Emission Prediction Method Of Power Plant Based On Optimized LSSVM

At present,the national standards for pollutant emissions from power plants are gradually increasing,and the research and model establishment of nitrogen oxide emis-sions from power plants is the focus of it.However,complex nonlinear systems cannot be modeled by traditional prediction models with low prediction accuracy and poor generalization ability.In this paper,an improved Fruit Fly Optimization Algo-rithm(FOA),an improved mature and stable technology,is used to optimize the Least Squares Support Vector Machine(LSSVM),and on this basis,a prediction algorithm is designed.Verified by the data of a power plant unit,the model proposed in the article improves the prediction accuracy of NO_x emissions.The main research work of the paper is listed below:(1)The research situation of domestic and foreign boiler flue gas NO_xemission prediction models is analyzed,and the NO_x emission characteristics of power plant boilers is studied in this paper.(2)A combination of sample space,Principal Component Analysis(PCA)and segmented fitting methods are used to process the data.After the data is normalized,PCA is used to find the principal components of the sample subspace and calculate the contribution rates of different principal components.Then the principal component with a larger contribution rate is selected to replace the initial input variable.Finally,after the modeling is completed,the data of each sample space is processed by the method of segmented fitting to obtain the complete output.(3)Three optimization algorithms are studied:Genetic Algorithm(GA),Particle Swarm Optimization(PSO),and FOA.The idea of imitating fruit flies to find food sources and share information is used by FOA to get the final result,but FOA is easier to find the optimal value in a local range than the entire search range.In order to im-prove the prediction accuracy,this paper studies the method of improving FOA,the purpose is to improve its global search ability to enhance the optimization of LSSVM parameters.(4)A combined prediction modeling method of LSSVM based on PCA to find principal components and improved FOA optimization is proposed.Firstly,the sample space is divided according to the size of the NO_x value.Secondly,PCA is used to reduce the dimensionality of each subspace,and the principal components whose cu-mulative contribution rate is higher than 90%are regarded as the input of LSSVM.Then MFOA is used as a method to optimize the hyper-parameters of the kernel func-tion width and penalty factor of LSSVM,and then MFOA-LSSVM is used to establish the prediction model of each subspace,and finally the final output of the model is obtained by segmented fitting.Finally,the final output of the model is obtained by segmented fitting.After theoretical research,this paper uses the proposed prediction method to sim-ulate the experimental boiler and prove its effectiveness.Finally,other prediction mod-els are used as a comparison to achieve the goal of verifying the superiority of the proposed method.The comparison results show that the prediction algorithm based on PCA-MFOA-LSSVM can effectively realize the boiler NO_x emission prediction mod-eling,and has high-precision and efficient prediction performance.