Ramprasad Yelchuru Sigurd Skogestad Henrik Manum

Slides:



Advertisements
Similar presentations
Ramprasad Yelchuru, Optimal controlled variable selection for Individual process units, 1 Optimal controlled variable selection for individual process.
Advertisements

1 Name, title of the presentation Master project: Control structure design for a sequence of distillation columns Tor Anders Marvik Supervisor: Sigurd.
MS&E 211 Quadratic Programming Ashish Goel. A simple quadratic program Minimize (x 1 ) 2 Subject to: -x 1 + x 2 ≥ 3 -x 1 – x 2 ≥ -2.
Control Structure Selection for a Methanol Plant using Hysys/Unisim
Advanced Controls Technology An Industrial and Academic Perspective on Plantwide Control James J. Downs Eastman Chemical Company Sigurd Skogestad Norwegian.
Separating Hyperplanes
1 CONTROLLED VARIABLE AND MEASUREMENT SELECTION Sigurd Skogestad Department of Chemical Engineering Norwegian University of Science and Technology (NTNU)
1 Effective Implementation of optimal operation using Self- optimizing control Sigurd Skogestad Institutt for kjemisk prosessteknologi NTNU, Trondheim.
1 Logic-Based Methods for Global Optimization J. N. Hooker Carnegie Mellon University, USA November 2003.
Support Vector Machines (SVMs) Chapter 5 (Duda et al.)
280 SYSTEM IDENTIFICATION The System Identification Problem is to estimate a model of a system based on input-output data. Basic Configuration continuous.
1 Introduction to Linear and Integer Programming Lecture 9: Feb 14.
Unconstrained Optimization Problem
Support Vector Regression David R. Musicant and O.L. Mangasarian International Symposium on Mathematical Programming Thursday, August 10, 2000
Trading Convexity for Scalability Marco A. Alvarez CS7680 Department of Computer Science Utah State University.
1 1 Slides by John Loucks St. Edward’s University Modifications by A. Asef-Vaziri.
1 Operation of heat pump cycles Jørgen Bauck Jensen & Sigurd Skogestad Department of Chemical Engineering Norwegian University of Science and Technology.
1 Lecture 4 Maximal Flow Problems Set Covering Problems.
Control of floor heating process Siri Hofstad Trapnes Supervisors: Sigurd Skogestad and Chriss Grimholt Direct heating in the floor and room Keep the temperature.
Ramprasad Yelchuru, MIQP formulations for optimal controlled variables selection in Self Optimizing Control, 1/16 MIQP formulation for optimal controlled.
Chapter 4 - Linear Programming: Computer Solution Excel Solver
GHGT-8 Self-Optimizing and Control Structure Design for a CO 2 Capturing Plant Mehdi Panahi, Mehdi Karimi, Sigurd Skogestad, Magne Hillestad, Hallvard.
1 Self-optimizing control: Simple implementation of optimal operation Sigurd Skogestad Department of Chemical Engineering Norwegian University of Science.
Developing models for optimization Lec 5 week 6. Example (1) Classify the following model.
Optimization for Operation of Power Systems with Performance Guarantee
Mathematical Models & Optimization?
Chapter 6 Linear Programming: The Simplex Method Section 3 The Dual Problem: Minimization with Problem Constraints of the Form ≥
1 Practical plantwide process control. Extra Sigurd Skogestad, NTNU Thailand, April 2014.
1 E. S. Hori, Maximum Gain Rule Maximum Gain Rule for Selecting Controlled Variables Eduardo Shigueo Hori, Sigurd Skogestad Norwegian University of Science.
Barnett/Ziegler/Byleen Finite Mathematics 11e1 Learning Objectives for Section 6.3 The student will be able to formulate the dual problem. The student.
1 Self-optimizing control Theory. 2 Step S3: Implementation of optimal operation Optimal operation for given d * : min u J(u,x,d) subject to: Model equations:
1 Active constraint regions for economically optimal operation of distillation columns Sigurd Skogestad and Magnus G. Jacobsen Department of Chemical Engineering.
3) OBJECTIVE FUNCTION & DISTURBANCES Objective function: Assuming product prices are the same, p D = p S = p B and (p-p F ) = p’, with F given and Q =
1 Optimization of LNG plants: Challenges and strategies Magnus G. Jacobsen Sigurd Skogestad ESCAPE-21, May 31, 2011 Porto Carras, Chalkidiki, Greece.
1 From process control to business control: A systematic approach for CV-selection Sigurd Skogestad Department of Chemical Engineering Norwegian University.
1 Self-optimizing control From key performance indicators to control of biological systems Sigurd Skogestad Department of Chemical Engineering Norwegian.
STATIC ANALYSIS OF UNCERTAIN STRUCTURES USING INTERVAL EIGENVALUE DECOMPOSITION Mehdi Modares Tufts University Robert L. Mullen Case Western Reserve University.
Control Structure Design: New Developments and Future Directions Vinay Kariwala and Sigurd Skogestad Department of Chemical Engineering NTNU, Trondheim,
1 Effective Implementation of optimal operation using Self- optimizing control Sigurd Skogestad Institutt for kjemisk prosessteknologi NTNU, Trondheim.
Massive Support Vector Regression (via Row and Column Chunking) David R. Musicant and O.L. Mangasarian NIPS 99 Workshop on Learning With Support Vectors.
1 Self-optimizing control From key performance indicators to control of biological systems Sigurd Skogestad Department of Chemical Engineering Norwegian.
Approximation Algorithms based on linear programming.
Computational Approach for Adjudging Feasibility of Acceptable Disturbance Rejection Vinay Kariwala and Sigurd Skogestad Department of Chemical Engineering.
Support Vector Machines (SVMs) Chapter 5 (Duda et al.) CS479/679 Pattern Recognition Dr. George Bebis.
Coordinator MPC with focus on maximizing throughput
Implementation of a MPC on a deethanizer
deterministic operations research
Economic Plantwide Control:
Part 4 Nonlinear Programming
Recent Development of Global Self-Optimizing Control
Glenn Fung, Murat Dundar, Bharat Rao and Jinbo Bi
Geometrical intuition behind the dual problem
IBM CPLEX Global Non-Convex MIQP
Sigurd Skogestad Institutt for kjemisk prosessteknologi
Example regulatory control: Distillation
Plantwide control: Towards a systematic procedure
PLANTWIDE CONTROL Sigurd Skogestad Department of Chemical Engineering
CONTROLLED VARIABLE AND MEASUREMENT SELECTION
Outline Control structure design (plantwide control)
Implementation of a MPC on a deethanizer
Plantwide control: Towards a systematic procedure
Sigurd Skogestad Institutt for kjemisk prosessteknologi
Economic plantwide control: A systematic approach for CV-selection
Example regulatory control: Distillation
Linear Programming Problem
Example regulatory control: Distillation
Example regulatory control: Distillation
Optimal measurement selection for controlled variables in Kaibel Distillation Column: A MIQP formulation Ramprasad Yelchuru (PhD Candidiate) Professor.
Example “stabilizing” control: Distillation
Outline Control structure design (plantwide control)
Presentation transcript:

Ramprasad Yelchuru Sigurd Skogestad Henrik Manum MIQP formulation for controlled variable selection in Self Optimizing Control Ramprasad Yelchuru Sigurd Skogestad Henrik Manum MIQP - Mixed Integer Quadratic Programming 1

Motivation L,V: Steady-state degrees of freedom T1, T2, T3,…, T41 Tray temperatures qF L,V: Steady-state degrees of freedom Want to minimize cost J Which two individual measurements or measurement combinations should be selected as controlled variables (CVs) to minimize the cost J? y = temperatures; H = selection/combination matrix c = H y, H=?

Ref: Halvorsen et al. I&ECR, 2003 Mathematically Optimal steady-state operation Controlled variables, cs = constant + - K H y c u u J d Loss Loss is due to (i) Varying disturbances (ii) Implementation error in controlling c at set point cs At optimum Ju = 0 Ref: Halvorsen et al. I&ECR, 2003 Kariwala et al. I&ECR, 2008 3

Objective function unaffected by D. So can choose freely. Non-convex optimization problem D : any non-singular matrix Objective function unaffected by D. So can choose freely. We made H unique by adding a constraint as Convex optimization problem Global solution subject to Problem is convex in decision matrix H

Problem is convex QP in decision vector Vectorization subject to Problem is convex QP in decision vector

Controlled variable selection st. Optimization problem : Minimize the average loss by selecting H and CVs as (i) best individual measurements (ii) best combinations of all measurements (iii) best combinations with few measurements

We solve this MIQP for n = nu to ny MIQP Formulation Big M approach We solve this MIQP for n = nu to ny

Case Study : Distillation Column T1, T2, T3,…, T41 Tray temperatures qF Binary Distillation Column LV configuration 41 Trays Level loops closed with D,B 2 MVs – L,V 41 Measurements – T1,T2,T3,…,T41 3 DVs – F, ZF, qF *Compositions are indirectly controlled by controlling the tray temperatures

Case Study : Distillation Column Data Results Controlled variables (c) Optimal individual measurements Loss2 = 0.0365 Optimal 4 measurement combinations Loss4 = 0.0164

Case Study : Distillation Column 1010 100 Comparison with customized Branch And Bound (BAB)* MIQP is computationally more intensive than Branch And Bound (BAB) methods (Note that computational time is not very important as control structure selection is an offline method) MIQP formulations are intuitive and easy to solve * Kariwala and Cao, IEEE Trans. (2010)

Comparison of MIQP, Customized Branch And Bound (BAB) methods Customized BAB methods can handle only monotonic cost functions (J) Customized routines are required BABs require a deeper understanding of the customized routines to solve problems with structural constraints Computationally faster than MIQPs as they exploit the monotonic properties efficiently Monotonicity of the measurement combinations needs to be checked before using PB3 for optimal measuement subset selections MIQP formulations can accommodate wider class than monotonic functions (J) MIQPs are solved using standard cplex routines MIQPs are simple and are easy to incorporate few structural constraints MIQPs are computationally intensive than BAB methods Single MIQP formulation is sufficient for the described problems

MIQP formulation with structural constraints If the plant management decides to procure only 5 temperature sensors T1,T2,…,T10 qF T11,T12,…,T35 T36,T37,…,T41 Tray temperatures

Conclusions In self optimizing control the following problems are solved with an MIQP based formulation for optimal controlled variables selection. Selection of optimal individual measurements as control variables (CVs) Selection of CVs as combinations of all measurements Selection of CVs as combinations of optimal measurement subsets MIQPs are more simple, intuitive and are easy compared to customized Branch and Bound methods