C. Ramos

Non-linear robust identification using evolutionary algorithms

This work describes a new methodology for robust identification (RI), meaning the identification of the parameters of a model and the characterization of uncertainties. The alternative proposed handles non-linear models and can take into account the different properties demanded by the model. The indicator that leads the identification process is the identification error (IE), that is, the difference between experimental data and model response. In particular, the methodology obtains the feasible parameter set (FPS, set of parameter values which satisfy a bounded IE) and a nominal model in a non-linear identification problem. To impose different properties on the model, several norms of the IE are used and bounded simultaneously. This improves the model quality, but increases the problem complexity. The methodology proposes that the RI problem is transformed into a multimodal optimization problem with an infinite number of global minima which constitute the FPS. For the optimization task, a special genetic algorithm (@e-GA), inspired by Multiobjective Evolutionary Algorithms, is presented. This algorithm characterizes the FPS by means of a discrete set of models well distributed along the FPS. Finally, an application for a biomedical model which shows the blockage that a given drug produces on the ionic currents of a cardiac cell is presented to illustrate the methodology.

Robust constrained receding-horizon predictive control via bounded data uncertainties

The main objective of this work consists of obtaining a new robust and stable Model Predictive Control (MPC). One widely used technique for improving robustness in MPC consists of the Min-Max optimization, where an analogy can be established with the Bounded Data Uncertainties (BDU) method. The BDU is a regularization technique for least-squares problems by taking into account the uncertainty bounds. So BDU both improves robustness in MPC and offers a guided way of tuning the empirically tuned penalization parameter for the control effort in MPC due to the duality that the parameter coincides with the regularization one in BDU. On the other hand, the stability objective is achieved by the use of terminal constraints, in particular the Constrained Receding-Horizon Predictive Control (CRHPC) algorithm, so the original CRHPC-BDU controller is stated, which presents a better performance from the point of view of robustness and stability than a standard MPC.

Predictive LPV control of a liquid-gas separation process

The problem of controlling a liquid-gas separation process is approached by using LPV control techniques. An LPV model is derived from a nonlinear model of the process using differential inclusion techniques. Once an LPV model is available, an LPV controller can be synthesized. The authors present a predictive LPV controller based on the GPC controller [Clarke D, Mohtadi C, Tuffs P. Generalized predictive control - Part I. Automatica 1987;23(2):137-48; Clarke D, Mohtadi C, Tuffs P. Generalized predictive control - Part II. Extensions and interpretations. Automatica 1987;23(2):149-60]. The resulting controller is denoted as GPC-LPV. This one shows the same structure as a general LPV controller [El Gahoui L, Scorletti G. Control of rational systems using linear-fractional representations and linear matrix inequalities. Automatica 1996;32(9):1273-84; Scorletti G, El Ghaoui L. Improved LMI conditions for gain scheduling and related control problems. International Journal of Robust Nonlinear Control 1998;8:845-77; Apkarian P, Tuan HD. Parametrized LMIs in control theory. In: Proceedings of the 37th IEEE conference on decision and control; 1998. p. 152-7; Scherer CW. LPV control and full block multipliers. Automatica 2001;37:361-75], which presents a linear fractional dependence on the process signal measurements. Therefore, this controller has the ability of modifying its dynamics depending on measurements leading to a possibly nonlinear controller. That controller is designed in two steps. First, for a given steady state point is obtained a linear GPC using a linear local model of the nonlinear system around that operating point. And second, using bilinear and linear matrix inequalities (BMIs/LMIs) the remaining matrices of GPC-LPV are selected in order to achieve some closed loop properties: stability in some operation zone, norm bounding of some input/output channels, maximum settling time, maximum overshoot, etc., given some LPV model for the nonlinear system. As an application, a GPC-LPV is designed for the derived LPV model of the liquid-gas separation process. This methodology can be applied to any nonlinear system which can be embedded in an LPV system using differential inclusion techniques.

Nonlinear robust identification using multiobjective evolutionary algorithms

In this article, a procedure to estimate a nonlinear models set (Θp) in a robust identification context, is presented. The estimated models are Pareto optimal when several identification error norms are considered simultaneously. A new multiobjective evolutionary algorithm

LQR Robusto Mediante Incertidumbre Acotada en los Datos

\epsilon\nearrow - MOEA

LP-DMC CONTROL OF A CHEMICAL PLANT WITH INTEGRAL BEHAVIOUR

has been designed to converge towards Θ

Evolutionary auto-tuning algorithm for PID controllers

_{P}^{\rm \star}

Nonlinear robust identification with ϵ-GA: FPS under several norms simultaneously

, a reduced but well distributed representation of ΘP since the algorithm achieves good convergence and distribution of the Pareto front J(Θ). Finally, an experimental application of the

PRINCIPAL COMPONENT GPC WITH TERMINAL EQUALITY CONSTRAINT

\epsilon\nearrow - MOEA

Model-based predictive control of greenhouse climate for reducing energy and water consumption

algorithm to the nonlinear robust identification of a scale furnace is presented. The model has three unknown parameters and l∞ and l1 norms are been taken into account.