Power consumption prediction of submerged arc furnace based on multi-input layer wavelet neural network

Two-input layer wavelet neural network prediction model is established, through the analysis of the smelting process of submerged arc furnace and combining of wavelet analysis and neural network theory, used to predict the power consumption of the submerged arc furnace timely. The input variables are not input in one layer, but in different layers according to their action sequences,thereby reducing the scale of the network. Then the genetic algorithm (GA) is used to optimize the weights of neural network, thus achieve the purpose of global optimization and fast convergence speed. The validity of the method mentioned can be proved by simulation and the experiment result.

I. Introduction

The smelting process of submerged arc furnace is a complex process with the characteristics of multivariate, strong coupling, strong uncertainties. There are strong interaction and many interfering factors between input variables and output variables, especially, time varying, nonlinear, it is difficult to accurately describe with mathematical models, thus, the conventional methodologies for optimal control based on mathematical model is difficult to use. Some critical technical parameters can not be measured online, and large fluctuation is shown on the boundary conditions for production. So, it is technical difficulties of smelting process submerged arc furnace that how to set up a complex industrial process model using smart way and modeling method so that we can amend plans and control instructions in advance and reduce deviations^[2]–[3].