Online Prediction Model Based on Adaptive Recursive Least Squares Support Vector Machine for Silicon-Manganese Alloy Composition

摘要

A new online prediction model based on Adaptive Recursive Least Squares Support Vector Machine (ARLSSVM) is presented in this paper, and applied to predict silicon-manganese alloy composition in a 30MVA submerged arc furnace smelting process. By using Recursive Least Squares Support Vector Machine (RLSSVM) regression algorithm, it avoids the difficulty of solving high-dimensional inverse matrix and improves the calculated speed, making the model update rapidly. By using the adaptive increased and decreased memory learning algorithm, it not only improves the dynamic tracking performance of the model, but also enhances its accuracy. The simulation results show its effectiveness.