Research And Application Of The Uncertainty Of The Coke Oven Flue Temperature Control Method

The coke oven heating-combustion process is a very complexindustrial process. The features of multi-variable, nonlinear, fuzzyuncertainty are the major control challenges in the coke oven burningprocess. Therefore, how to stabilize the flue temperature under influenceof the uncertain factors in the production process is the most criticalproblem for the tracking control of the flue temperature.On the basis of analysis of its production technology, the influencesof various uncertainty information on the coke oven heating-combustionprocess are analyzed qualitatively, such as the gas calorific value, andcoal moisture, and coal charging volume, and steam coal blending ratekinds proportion, and bunt-coke plan's program, and the accomplishmentof coke plan and so on. Then the fluctuation of the caloric value isemphatically studied in detail. The complex industrial process controlmethods for the coke oven heating-combustion process are studied for thecontrol problems existed. At last, the basic principle of the fluetemperature control on coke oven is proposed, and the general structureof the system is designed.The historical data of the coke oven production process is analyzedfirstly. Then the fuzzy Takagi-Sugeno method based on Gustafsson-Kessel is put forward, and the on-line fuzzy identification of coke ovenis realized. First, the premise parameters are fuzzy clustering with G-Kmethod. Second, the consequence is the identified with least-squaresubordination function to get the local models. Third, the subordinatedegree of all the local models are computed with membership functionsto get the global models. Finally, it is converted to the CARIMA model,so that it can be applied to the flue temperature control.Based on the CARIMA model, the constrained generalizedpredictive controller is designed. The characters between the parallelpredictors are made use of to identify the parameters of the furthestdistance output prediction expression in the predictor. The equivalence ofcontrol law between GPC and GPC is used to solve the parameters of theoptimal law, which avoids a great deal of computational cost and time cost on the solution of Diophantine equation. The parameters areconsidered, including prediction length, and control length, and stepgentle factor, and control weighting factor. The measure of improving thesystem performance is put forward. The stability and robust of controlstrategies are analyzed theoretically.The simulation results show that the fuzzy Takagi-Sugeno methodcan identify the coke oven well under different heating processes, and theidentification models have nice stability and robust. The system has goodstep response and periodic response characters with the control strategies.When the gas calorific value is fluctuant in stated range, the system is stillwith good anti-interference capability and robust, and steady fluetemperature. It provides an effective solution for coke oven heating-combustion process control.