Nonlinear model predictive control of MIMO system with Relevance vector machines and Particle swarm optimization

Abstract

This paper demonstrates control accuracy andcomputational efficiency of nonlinear model predictive control (NMPC) strategywhich utilizes a probabilistic sparse kernel learning technique calledRelevance vector regression (RVR) and particle swarm optimization withcontrollable random exploration velocity (PSO-CREV). An accurate reliablenonlinear model is first identified by RVR with a radial basis function (RBF)kernel and then the optimization of control sequence is speeded up by PSO-CREV.An improved system performance is guaranteed by an accurate sparse predictivemodel and an efficient and fast optimization algorithm. To compare theperformance, model predictive control (MPC) using a deterministic sparse kernellearning technique called Least squares support vector machines (LS-SVM)regression is done on a highly nonlinear distillation column with severeinteracting process variables. SVR based MPC shows improved tracking performancewith very less computational effort which is much essential for real timecontrol.