Nonlinear Model Predictive Control of MIMO System with Relevance Vector Machines and Particle Swarm Optimization

This paper demonstrates control accuracy and computational efficiency of nonlinear model predictive control (NMPC) strategy which utilizes a probabilistic sparse kernel learning technique called Relevance vector regression (RVR) and particle swarm optimization with controllable random exploration velocity (PSO-CREV). An accurate reliable nonlinear 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 predictive model and an efficient and fast optimization algorithm. To compare the performance, model predictive control (MPC) using a deterministic sparse kernel learning technique called Least squares support vector machines (LS-SVM) regression is done on a highly nonlinear distillation column with severe interacting process variables. SVR based MPC shows improved tracking performance with very less computational effort which is much essential for real time control.