Research On Parameters Identification Maths For Arc Furnace Regulation System

The three-phase AC arc furnace in large-scale steelmaking is increasing day by day,and the part it needs to control is the electrode regulating system.The electrode regulating system consists of two parts,one is hydraulic system and the other is power supply system.In order to control the electric arc furnace better,it is very important to identify the electrode regulating system of the electric arc furnace better.For a complex industrial system such as three-phase electric arc furnace,the mechanism model can not be deduced by simple mathematical and physical theorems,but the parameters of the model can only be obtained by data method,that is,system identification method.The electrode regulating system of electric arc furnace can be regarded as the parallel connection of three Hammerstein-Wiener models.The model consists of three Hammerstein modules(static dead-time saturated non-linear module and third-order dynamic linear module in series)and a three-input three-output non-linear module(coupled arc module).Firstly,each module of the identification system is introduced,and then introduced the working principle of the arc electrode regulating system.Then the hydraulic system and the arc model structure are established respectively.Secondly,for the electrode regulating system of single variable electric arc furnace,the over-parameterized model of single variable system is established firstly,and the idea of multiinformation is introduced to solve the problem of insufficient utilization of data in the process of system identification.Then the stochastic gradient identification method suitable for multivariable system and online identification is adopted to identify parameters.Lastly,based on the study of single variable system of arc furnace,this paper extends the method to multi-variable system.Firstly,the over-parametric model of multivariable system is established,and the idea of multi-information combined with stochastic gradient identification method is extended to multivariable system.By increasing the length of information,the data are fully utilized and the identification accuracy is greatly improved.Finally,the effectiveness of the proposed method is proved by computer simulation.