Prediction Of Large-scale Solid Waste Incinerator Control Parameters Based On The Support Vector Machine And Image Processing

With the year-on-year increase in the yield of municipal solid waste in my country and the impact of related environmental protection policies,the number of various types of solid waste incinerators is also increasing.The current manual combustion control of solid waste incinerators is not conducive to maintaining stable combustion in the incinerator due to differences in personal experience and operating habits.Therefore,it is necessary to develop an accurate prediction model for the control parameters of the incinerator,so as to achieve the goal of automatic operation of the incinerator and ensure the stability of incineration in the incinerator.This paper took a 750 t/d large-scale municipal solid waste grate incinerator and a 130 t/h largescale biomass circulating fluidized bed incinerator as the research objects.Based on the support vector machine,classification prediction models for important control parameters of the incinerators such as grate speed,fan frequency were established.The model data was composed of DCS historical operating data and corresponding flame image data.The DCS parameters adopted Relief F algorithm supplemented by artificial filtering to achieve dimensionality reduction of the input features of each control parameter,and the flame image used image processing methods to extract relevant features.Model parameters optimization used GA and PSO to optimize 5 times each and took the average value.Through the test on the test set samples,it was found that the image features coupled with DCS features and pure DCS features as model input could enable each control parameter to achieve a prediction accuracy of more than 95%,which could provide a reference for the automatic adjustment of incinerator control parameters,and the prediction accuracy of the former was higher.Among them,the burning grate speed,the burn-out grate speed,the frequency of the primary fan and the frequency of the secondary fan of the grate incinerator averagely increased by 0.26%,0.02%,0.16%,and 0.51%respectively;the frequency of the primary fan,the frequency of the secondary fan,the frequency of the #1 induced draft fan and the frequency of the #2 induced draft fan of the fluidized bed incinerator averagely increased by 0.37%,0.11%,0.33%,and 0.28% respectively.The addition of image features was beneficial to improve the prediction performance of the model.Through the analysis of the importance of image features,several image features that had the greatest influence on the prediction of the control parameters of the two incinerator types could be obtained.For the grate incinerator,it was recommended that operators paid attention to gray average,flame area rate and horizontal and vertical coordinates of the flame centroid when performing combustion control.For the fluidized bed incinerator,it was recommended that the operator paid attention to the R channel standard deviation and the horizontal and vertical coordinates of the flame centroid when performing combustion control.Finally,this paper developed the flame monitoring and intelligent control software for a municipal solid waste grate incinerator and a biomass circulating fluidized bed incinerator.According to the research content,the prediction function of the setting value of the primary and secondary fan frequency was realized,and the actual engineering verification was carried out with the fluidized bed incinerator as the object.The result showed that the prediction accuracy of the frequency of the primary and secondary fan within a certain period of time was above 80%,which basically met the actual control requirements of the incinerator and had certain engineering application value.