Centrifugal compressor surge control using nonlinear model predictive control based on LS-SVM

Abstract

This paper reports a surge control strategy for centrifugal compressor using nonlinear model predictive control based on Lease-Squared Support Vector Machine(LS-SVM) in order to increase efficiency of centrifugal compressor. The MISO nonlinear predictive models of compressor's discharge pressure and mass flow are developed by LS-SVM. In order to avoid surge, the conditions of anti-surge are chosen as the limiting conditions of the control objective and also the reference trajectory. The influence of actuator's response time is reduced by predict the compressor's input and output in advance. The operating point's fluctuate is avoided by the control objective which let the difference of the adjacent control input be the minimum. Simulation and experiment results show that the proposed control strategy can decrease the distance between surge line and control line at the same time avoid surge effectively, so compressor's efficiency is increased significantly.