Research On Data Modeling Method Based On Partial Least Squares And Neural Network

The measurement of Boiler unit key parameters is very important to the safeoperation of the unit, But its characteristic is very complex, is influenced by such as theboiler load, a total of wind pressure, secondary air total wind wind pressure and otherfactors, is difficult to estimation with formula. It is generally determined with the actualtest method, and the predicted results explore ways to reduce the fly ash carbon contentby the predicted results. Due to large operation data, working condition of prediction islimited, there exists the influence on boiler combustion characteristics, which causes thedifficulty in data analysis. Therefore, we need to establish the relationship between flyash carbon content and operation parameters to guide the optimization operation ofboiler.This paper find out the each variable which influence fly ash carbon content basedon thermal power unit operation data. Take concrete analysis of each variable to get rid ofthe variables which influence fly ash carbon content. Sift the optimal modeling dataaccording to thermal power unit operation data. Have developed the prediction model offly ash carbon content based on partial least squares and neural network. Take predictionfly ash carbon content on the selected data, and compare it with the actual operation offly ash carbon content, the result shows that the prediction is accuracy. Validate theaccuracy of the model by using test data. At the same time, studied the calculationprocess of partial least square method and neural network, write the related calculationprogram in matlab language. The research results shows that it is reasonable andaccuracy that prediction model of fly ash carbon content using partial least squares andneural network. However, there still exists certain deviation between prediction resultsand the actual result, the specific reason remains to be further analysis.