Study On Modeling And Optimization Of A Desulphurization System In Coal-fired Power Plant Based On CMAC

The rapidly and widely built environmental protection facilities in coal-fired power plants have exposed problems such as large design margins and excessive transformation,resulting in energy waste in related systems.The deficient adjustment of traditional PID technology for complex systems,the rapid popularization and application of the intelligent data modeling technology in the power plant and the insufficient studies on the optimization model,all of which give rise to that many scholars are still committed to the process optimization of related systems based on operation data.Based on 2×44640 sets of data,artificial intelligence algorithms are used to carry out three aspects of research for wet flue gas desulfurization(WFGD)system,which are prediction model building,model’s error analysis and structure determination and operating condition optimization.Based on 44640 sets of data,cerebellar model articulation controller(CMAC)is applied to build prediction model for WFGD with desulphurizing ratio and the economic cost two objectives.In the model building process,the grey relation entropy analysis and uniform design method are used to screen the input variables and study the model parameters separately.Traditional regression analysis and proposed location number analysis(LNA)method are adopted to analyze output errors of experiment group and predict the results of test group.Results show that the LNA demonstrates good fitting and predictive capabilities.Based on this,the method is used to determine the structural parameters of the prediction model with an average error less than 1%.In the final model validation process,with the determined structure,95% and 5% of the processed data is used for modeling and validation separately.In order to research on the error analysis and structure determination of CMAC as well as further study LNA method,based on another 44640 sets of data,4 experimental groups and 1 prediction group are designed under uniform design principles.To analyze the results,the error analysis methods of multiple linear regression,multiple quadratic regression,LNA,and actual address number analysis methods are attempted and compared.Results demonstrate LNA is very suitable for error analysis and prediction of CMAC,which has the advantages of wide prediction range,high precision,and simple parameter adjustment.In addition,the limit of the LNA method is also performed.With the model building and error analysis research completed,genetic algorithm and artificial bee colony algorithm are used to optimize the specific working condition.The results show that,compared with the optimal operating state of the original data,the reduction in energy consumption under the optimal operating parameters of genetic algorithm and artificial bee colony algorithm are 30.38% and 58.47% respectively.