Collector Pressure Control On Data Mining

Coke is the main raw material of the iron industry.During the coking production process,a large amount of raw gas will be produced.The raw gas not only contains certain harmful gases,but also contains recyclable energy.Therefore,in the process of coking gas,it is necessary Recycle raw gas.In the raw gas recovery system,the gas collector is an important part of the export system.The stability of the gas collector pressure can effectively improve the quality of coal production and reduce environmental pollution.In the traditional control methods currently in use,due to the coupling and nonlinear characteristics of the pressure of the gas collector,the control effect cannot be achieved in an ideal state.The development of data mining technology can use the actual data collected on the spot to find hidden Laws and enhancing the control effect of gas collector pressure are of great significance to the steel industry.Aiming at the control problem of manifold pressure,this paper attempts to integrate various data mining technologies such as rough set,attribute specification and support vector machine into the manifold pressure control,and designs a multi-link intelligent control scheme from "pretreating-attribute extraction-pressure control".Firstly,in view of the complexity of the data and the diversity of expression,the data is preprocessed and the data is normalized to make all the data have the same weight.Secondly,based on the preprocessed data,the difference between the information entropy of decision attribute and the conditional information entropy of decision attribute and the conditional information entropy of decision attribute is calculated by using the attribute reduction of rough set theory,and the main parts of all factors affecting the pressure of the collector are mined out.Finally,the fusion of smoothing model and Least Square Support Vector model was used to smooth the data first to reduce the influence of noise data on the experiment.LS-SVM(Least Square Support Vector Machine SVM)was used to improve the nonlinear approximation ability and realize the control effect of the pressure of the collector.Through experimental comparison,the prediction result of the smooth support vector model is compared with the prediction result of the support vector machine and the prediction result of the BP neural network.The root mean square error is compared.The prediction effect of the smooth support vector machine model is better in the root mean square error.Smaller,the control effect is more precise.It provides a more effective control method for coking gas production.Figure 24;Table 9;Reference 59...