Computational issues in nonlinear predictive control

Sistu, Phani B.; Gopinath, Ravi; Bequette, B. Wayne

doi:10.1016/0098-1354(93)80027-k

Cited by 53 publications

(23 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…We will then focus on the details required to obtain the next manipulated variable move (prediction step). A comparison of possible methods to solve the prediction equations is provided by Sistu et al [7]. A discussion of many issues in NMPC is provided by Bequette [2].…”

Section: Mpc Overviewmentioning

confidence: 99%

“…In implementation, this involves linearizing the nonlinear model at every sampling instant. Sistu et al [7] have discussed the computational issues in using a full nonlinear model as compared to using a linearized model, in the prediction phase. For the drug delivery model, this is not straightforward because of the different time scales associated with the hemodynamic and pharmacodynamic states.…”

Section: Linearizing the Nonlinear Modelmentioning

confidence: 99%

See 1 more Smart Citation

The Design of Nonlinear Predictive Controllers: Application to a Drug Infusion System

Gopinath

Bequette

1995

Methods of Model Based Process Control

View full text Add to dashboard Cite

The importance of a number of practical issues when implementing nonlinear model predictive control on a drug infusion system are shown. The motivation for this system is a patient under congestive heart failure (heart attack) conditions. The control objective is to adjust the infusion rate of drugs to maintain the patient's blood pressure and blood flowrate at desired values. A selective linearization technique is introduced to reduce the computation time for the model prediction step. A rule-based override procedure is developed to initialize the drug delivery rates to avoid unacceptably slow responses and operation in the 'forbidden zone'. The performance of the multirate rule-based override model predictive control is illustrated through simulation examples. MPC OverviewConceptual simplicity is perhaps one of the prime reasons for the success of model predictive control (MPC) for multi variable systems. Although MPC has been presented in a number of papers in this workshop, we briefly review the technique for continuity of discussion. The MPC approach for a SISO system, sampled at a constant single-rate, is shown in Figure 1. The basic idea is to find a set of m control moves which will minimize an objective function (usually a least squares error in desired setpoint response) evaluated over a prediction horizon of p time-intervals. Constraints include minimum, maximum and rate-of-change values of the manipulated variable. Although a set (horizon) of control moves is obtained in this optimization procedure, usually only the first control move is implemented on the process. The measured output at the end of the time step is used to compensate for plant/model mismatch, and the optimization problem is re-solved after shifting forward one time step (moving horizon formulation).When implementing nonlinear model predictive control (NMPC) a number of critical decisions must be made, including:• What type of dynamic model will be used for output prediction? • How will the modeling equations be solved?• How will the model be cO.i'ected for plant/model mismatch? 607 R. Berber (eeL). Methods of Model Based Process

show abstract

Section: Mpc Overviewmentioning

confidence: 99%

Section: Linearizing the Nonlinear Modelmentioning

confidence: 99%

The Design of Nonlinear Predictive Controllers: Application to a Drug Infusion System

Gopinath

Bequette

1995

Methods of Model Based Process Control

View full text Add to dashboard Cite

show abstract

“…The advantage of this approach is that a nonlinear model predictive controller is constructed through several local linear model predictive controllers which usually have analytical solutions. Hence, the control actions of this nonlinear model predictive controller can be obtained analytically avoiding the time consuming numerical search procedures and the uncertainty in convergence to the global optimum which are typically seen in conventional nonlinear model based predictive control strategies [6,30,31].…”

Section: Neuro-fuzzy Network Model-based Predictive Controlmentioning

confidence: 99%

Long Range Predictive Control of Nonlinear Processes Based on Recurrent Neuro-Fuzzy Network Models

Zhang¹,

Morris²

2000

Neural Computing & Applications

View full text Add to dashboard Cite

A recurrent neuro-fuzzy network-based nonlinear long range model predictive control strategy is proposed in this paper. The process operation is partitioned into several fuzzy operating regions. Within each region, a local linear model is used to model the process. The global model output is obtained through the centre of gravity defuzzification. Based upon a neuro-fuzzy network model, a nonlinear model-based predictive controller can be developed by combining several local linear model-based predictive controllers which usually have analytical solutions. This strategy avoids the time consuming numerical optimisation procedure, and the uncertainty in convergence to the global optimum which are typically seen in conventional nonlinear modelbased predictive control strategies. Furthermore, control actions obtained based on local incremental models contain integration actions which can naturally eliminate static control offsets. The technique is demonstrated by an application to the modelling and control of liquid level in a water tank.

show abstract

“…Integration of the model from time step k-1 to current time step k is represented by: since the states are not corrected with feedback measurements. A detailed comparison of the computation times for the various implementation strategies, including different optimization codes and numerical integrators, is performed by Sistu et al (1993). There are signifi cant computational savings to the NLQDMC approach of Garcia (1984), particularly if analytical derivatives are used to in the model linearization.…”

Section: Model Predictive Control Using Fundamental Modelsmentioning

confidence: 99%

Non‐Linear Model Predictive Control: A Personal Retrospective

Bequette

2007

Can J Chem Eng

View full text Add to dashboard Cite

An overview of non-linear model predictive control (NMPC) is presented, with an extreme bias towards the author's experiences and published results. Challenges include multiple solutions (from non-convex optimization problems), and divergence of the model and plant outputs when the constant additive output disturbance (the approach of dynamic matrix control, DMC) is used. Experiences with the use of fundamental models, multiple linear models (MMPC), and neural networks are reviewed. Ongoing work in unmeasured disturbance estimation, prediction and rejection is also discussed.On présente un aperçu général du contrôle prédictif par modèles non linéaires (NMPC), en mettant l'accent en particulier sur les expériences des auteurs et les résultats publiés. Les défi s incluent des solutions multiples (à partir des problèmes d'optimisation non convexes), ainsi que la divergence entre les sorties de modèle et d'installation lorsque la perturbation de sortie additive constante (la méthode du contrôle de matrice dynamique, DMC) est utilisée. Les expériences avec les modèles fondamentaux, les modèles linéaires multiples (MMPC) et les réseaux neuronaux sont examinées. Le travail actuellement mené sur l'estimation, la prédiction et le rejet des perturbations non mesurées est également examiné.

show abstract

Computational issues in nonlinear predictive control

Cited by 53 publications

References 9 publications

The Design of Nonlinear Predictive Controllers: Application to a Drug Infusion System

The Design of Nonlinear Predictive Controllers: Application to a Drug Infusion System

Long Range Predictive Control of Nonlinear Processes Based on Recurrent Neuro-Fuzzy Network Models

Non‐Linear Model Predictive Control: A Personal Retrospective

Contact Info

Product

Resources

About