An Experimental Evaluation of an Adaptive MPC Scheme based on Output Error Models Parameterized using Orthonormal Basis Filters∗. Issue 1 (2016)

• Record Type: Journal Article
• Authors: Patil, Prashant V.; Choudhary, Rahul; Patel, Nitinkumar; Patwardhan, Sachin C.
• Is Part Of: IFAC-PapersOnLine. Volume 49:Issue 1(2016)
• Page Start: 664
• Page End: 669
• Publication Date: 2016
• Subjects: Model Predictive Control -- Adaptive Control -- Orthonormal Basis Filters; Automatic control -- Periodicals; 629.805
• DOI: 10.1016/j.ifacol.2016.03.132
• Languages: English
• ISSNs: 2405-8963

Abstract

Abstract: Over the last three decades, Model Predictive Control (MPC) has emerged as an important tool for handling constrained multivariable control problems. Most of the commercially available MPC schemes employ linear prediction models, which can quickly become obsolete and this results in deterioration of the closed loop performance. A possible remedy to this problem is to employ adaptive MPC (AMPC) scheme in which the model parameters are updated online. However, the parameter drift leading to instability is an important concern while developing an AMPC scheme. In this work, it is proposed to develop a Multiple Input Multiple Output(MIMO) AMPC scheme based on multiple Multiple Input Single Output(MISO) models that have output error (OE) structure and are parameterized using orthonormal basis filters (OBF). The poles of the OBF are estimated offline from a batch of perturbation data. In the model update component of the proposed AMPC formulation, the OBF poles are kept fixed and only the Fourier coe¢cients are updated online using the recursive least squares method. This measure avoids the transgression of the poles outside the stable region during the transients. Efectiveness of the proposed AMPC scheme is demonstrated by conducting experimental studies on the benchmark quadruple tank system. Analysis of the experimental results reveals that the proposed AMPC formulation provides a promising approach for controlling moderately nonlinear systems.