Research On Modeling For Electrode Regulating System Of Electrical Smelting Furnace For Magnesia

Electrical smelting furnace for magnesia is a controlled object with strong nonlinearity, time variation and strong coupling, and is a kind of violent nonlinear change load by electrical characteristics, which is difficult to build its mathematics model. In this thesis, it is studied in mechanism modeling and artificial intelligent method identification separately, under the background of researching the electrode regulating system of three phase electrical smelting furnace for magnesia which is widely used in electrical smelting magnesia areas. The research lays a model foundation to the simulation experience of intelligent control in electrical smelting furnace for magnesia, which relies on "Project 985" science and technology innovation flat of industrial process integrated automation. Also, it provides theory foundation for the further study on researching inherent galvanothermy physics of electrical smelting furnace for magnesia, optimizing the smelting process, improving the quality of power supply, and researching advanced control method. The main jobs in this thesis are concluded as follows:(1) According to the control strategy of the regulating system which is studied, the mathematical model of the AC arc and three-phrase power supply system are built firstly. With analyzing each working mechanism of the electrode regulating system individually, combining with the arts and crafts of smelting process and reasonable hypothesis under the related theoretical knowledge of industrial furnace, the complicit function equations between arc length and arc current are given. Through the simulation in Simulink, the trend of model is validated right.(2) In support of measured data, the dynamic model of Elman regression neural network is established secondly. This model makes up the deficiency of mathematical model which ignores the consideration of the random disturbance. Matlab neural network toolbox is used to build and test the model. And a right approximation effect is gotten as a result.(3) A hybrid model on the basis of mathematical model and compensation neural network model is discussed at the end of the thesis. As mechanism model is usually built on some assumed conditions, and pure neural network model is lack of related physical meanings, the research on hybrid modeling method is meaningful. The simulation results prove that the hybrid method is of simple structure and high approximation accuracy.