ISA MOD UNLEASH-2004 Models Unleashed Virtual Plant and Model Predictive Control Applications A Pocket Guide.pdf

《ISA MOD UNLEASH-2004 Models Unleashed Virtual Plant and Model Predictive Control Applications A Pocket Guide.pdf》由会员分享，可在线阅读，更多相关《ISA MOD UNLEASH-2004 Models Unleashed Virtual Plant and Model Predictive Control Applications A Pocket Guide.pdf（180页珍藏版）》请在麦多课文档分享上搜索。

1、Models Unleashed: Virtual Plant and Model Predictive Control ApplicationsA pocket guideMcMillan2004.book Page i Wednesday, September 17, 2003 11:24 AMMcMillan2004.book Page ii Wednesday, September 17, 2003 11:24 AMModels Unleashed: Virtual Plant and Model Predictive Control ApplicationsA pocket guid

2、eby Gregory K. McMillanand Robert A. CameronMcMillan2004.book Page iii Wednesday, September 17, 2003 11:24 AMCopyright 2004ISAThe Instrumentation, Systems and Automation Society67 Alexander DriveP.O. Box 12277Research Triangle Park, NC 27709All rights reserved. Printed in the United States of Americ

3、a.1098765432ISBN 1-55617-857-3No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, elec-tronic, mechanical, photocopying, recording or otherwise, without the prior written permission of the publisher.Library of Congress Cataloging-in-Publi

4、cation Data Data is in Progress ISA wishes to acknowledge the cooperation of those manufac-turers, suppliers, and publishers who granted permission to reproduce material herein. The Society regrets any omission of credit that may have occurred and will make such corrections in future editions.McMill

5、an2004.book Page iv Wednesday, September 17, 2003 11:24 AMNoticeThe information presented in this publication is for the general education of the reader. Because neither the author nor the publisher has any control over the use of the informa-tion by the reader, both the author and the publisher dis

6、claim any and all liability of any kind arising out of such use. The reader is expected to exercise sound professional judgment in using any of the information presented in a particular applica-tion. Additionally, neither the author nor the publisher have investigated or considered the effect of any

7、 patents on the abil-ity of the reader to use any of the information in a particular application. The reader is responsible for reviewing any possi-ble patents that may affect any particular use of the informa-tion presented. Any references to commercial products in the work are cited as examples on

8、ly. Neither the author nor the publisher endorses any referenced commercial product. Any trademarks or tradenames referenced belong to the respective owner of the mark or name. Neither the author nor the publisher makes any representation regarding the availability of any referenced commercial produ

9、ct at any time. The manufacturers instruc-tions on use of any commercial product must be followed at all times, even if in conflict with the information in this publi-cation.McMillan2004.book Page v Wednesday, September 17, 2003 11:24 AMMcMillan2004.book Page vi Wednesday, September 17, 2003 11:24 A

10、MThis guide is dedicated to the memory ofMark Mennen,who in two years did more to unleash models for maximum benefitsthan most have done in a lifetime.McMillan2004.book Page vii Wednesday, September 17, 2003 11:24 AMMcMillan2004.book Page viii Wednesday, September 17, 2003 11:24 AMTABLE OF CONTENTS

11、ixTable of ContentsAbout the Authors. xiPreface. xiiiChapter 1.0Simulation .1Overview 1Procedure8Examples 24References.32Chapter 2.0Identification .33Overview 33Procedure50Examples 56References.70Chapter 3.0Setup.71Overview 71Procedure86Examples 88References.96McMillan2004.book Page ix Wednesday, Se

12、ptember 17, 2003 11:24 AMx TABLE OF CONTENTSChapter 4.0Tuning99Overview 99Procedure111Examples 117References.141Appendix AWebsite (www.isa.org) Contents 143Appendix BCondition Number 147Appendix CGlossary.151Appendix DModern Myths .155Appendix ESpirited Explanation.159McMillan2004.book Page x Wednes

13、day, September 17, 2003 11:24 AMABOUT THE AUTHORS xiAbout the AuthorsGregory K. McMillan is a retired Senior Fellow from Solutia Inc. During his 33 year career with Monsanto Company and its spin off Solutia Inc., he specialized in improving loop perfor-mance, controller tuning, valve dynamics, oppor

14、tunity assessments, dynamic simulation, fermentor control, pH control, and reactor con-trol. Greg is the author of numerous articles and a dozen or so books, his most recent being the ISA bestseller for 2001 titled Good Tuning A Pocket Guide and the ISA bestseller for 2003 titled Advanced Control Un

15、leashed. He has contributed to several handbooks, is the editor of the Process/Industrial Instrumentation and Controls handbook, and a columnist for Control Magazine. He is one of InTechs 50 most influential industry innovators for advancing automation and con-trol technologies.Greg is an ISA Fellow

16、 and received the ISA “Ker-mit Fischer Environmental” Award for pH con-trol in 1991, the Control Magazine “Engineer of the Year” Award for the Process Industry in 1994, and was one of the first inductees into the Control Magazine “Process Automation Hall of Fame” in 2001. He received a B.S. from Kan

17、sas University in 1969 in Engineering Physics and a McMillan2004.book Page xi Wednesday, September 17, 2003 11:24 AMxii ABOUT THE AUTHORSM.S. from University of Missouri Rolla in 1976 in Electrical Engineering (Control Theory). Presently, Greg is a Professor at Washington Uni-versity in Saint Louis,

18、 Missouri and is a consult-ant through EDP Contract Services in Austin, Te x a s .Robert A. Cameron is a process control engi-neer with 18 years experience, working for Bailey Controls and Monsanto/Solutia. His areas of interest and expertise are control system pro-gramming and optimization, instrum

19、entation, and advanced control implementation. Bob is an ISA member and co-recipient of the 2001 R.N. Pond award (best ChemPID paper of the year) for the Expo2001 paper “Constrained Multivariable Predictive Control of Plastic Sheets.” He received a B.S. in Chemical Engi-neering from Clarkson Univers

20、ity in 1985.McMillan2004.book Page xii Wednesday, September 17, 2003 11:24 AMPREFACE xiiiPrefaceThe process industry is increasingly recognizing the value of models and opportunity they offer to capture and exploit plant knowledge. As plants are pushed to operate at maximum effi-ciency, these models

21、 can be the ticket to ride plant constraints for fun and profit. Such mod-els can be as small as a first-order model obtained from step tests or as large as a first-prin-ciple model that is based on mass and energy balances. This pocket guide is intended to pro-vide users with a concise presentation

22、 of the concepts, procedures, and examples they will need to construct and apply both types of mod-els using “state-of-the-art” software for simula-tion and model predictive control. The guide contains details, data, and test results that are not yet available in the literature. The user need not ha

23、ve an advanced degree to get the most out of this guide. Building on the knowledge and objectives of Advanced Control Unleashed, Models Unleashed is designed to enable the engineers closest to the application to exploit their experi-ence by embedding it in a model and a control system.McMillan2004.b

24、ook Page xiii Wednesday, September 17, 2003 11:24 AMMcMillan2004.book Page xiv Wednesday, September 17, 2003 11:24 AMSIMULATION 11.0SimulationOverviewProcess control deals with change. If process conditions were constant there would be no need for a control system. In a plant, the operat-ing conditi

25、ons continually fluctuate, primarily because of changes in raw materials, production rate, product mix, equipment performance, foul-ing, catalyst, utilities, and ambient conditions. Except for production rate changes, most of these disturbances, as well as their effect on key process variables, are

26、not measured on line. Control systems do not eliminate this variabil-ity, but they can transfer it from a controlled variable to a less important manipulated vari-able. Control systems can also move the process to a more optimum operating point.Unfortunately, control loops that suffer from stick-sli

27、p, incorrect tuning, and interaction may actually increase the variability. The feedback and feedforward settings of most PID controllers are tuned by a trial-and-error method and thus reflect more the personal preferences of the McMillan2004.book Page 1 Wednesday, September 17, 2003 11:24 AM2 SIMUL

28、ATIONtuner than any knowledge of the process dynam-ics and objectives. Plants rarely try to decouple PID controllers, but when they do it is primarily based on just some estimated steady state gains.While model predictive control (MPC) cannot eliminate the stick-slip problem, it can effec-tively add

29、ress the issues of tuning and interac-tion while it simultaneously handles process limits and optimizes process objectives. An MPC systems control actions are derived from an experimental model that is obtained by rigor-ous plant testing, which takes into account the interactions and the response of

30、 constraints. If this model is accurate, MPC requires less tuning than PID control. Furthermore, the main MPC tuning parameter is designed to set the amount of variability that is transferred from the con-trolled variables to the manipulated variables. The embedding of this dynamic model enables MPC

31、 to simply trade off between performance and robustness, maintain an allowable degree of variability in both the controlled and manipu-lated variables, decouple interrelated loops, opti-mize set points, and honor constraints.McMillan2004.book Page 2 Wednesday, September 17, 2003 11:24 AMSIMULATION 3

32、MPC uses an experimental dynamic model that is obtained by making steps in the manipulated and disturbance variables, identifying either matrix coefficients directly or the parameters, such as process gain, time delay, and time lag, so as to predict a trajectory from previous changes in the manipula

33、ted and disturbance variables. The models are linear, and the effects are com-bined by linear superposition. Thus, the knowl-edge of the future that MPC provides excludes the effect of nonlinearities and unmeasured upsets. These unknowns are addressed in the present by biasing the trajectory by a fr

34、action of the difference between the predicted and actual value. The old adage that you can only control what you know still applies. A first-principle dynamic model can quantify nonlinearities and unmeasured process condi-tions. Until recently, these models required that hundreds of differential eq

35、uations be set up and numerically integrated. However, process flow diagram (PFD) models that are used for process design can now be made dynamic by using tech-nically advanced simulation software. This guidebook will explore how to develop and McMillan2004.book Page 3 Wednesday, September 17, 2003

36、11:24 AM4 SIMULATIONapply dynamic PFD models to improve the capability of MPC. Consider what opportunities would open up if the conditions, properties, and compositions of each stream in the PFD that was used to design the plant were updated dynami-cally and displayed. The previously unknown upsets

37、could become disturbance variables, and users could add compositions and yields that are real indicators of product quality and process performance as controlled variables to the MPC. The dynamic PFD model could be driven to explore nonlinearities and new operating regions as well as step-tested to

38、identify the parameters for the MPC experimental models. At a mini-mum, the insight and knowledge users gain from exploring the dynamics and pathways of vari-ability would improve the design and justifica-tion of MPC systems.Historically, first-principle dynamic models have been severely limited to

39、a few unit operations and mostly static estimates of a few physical properties, such as specific heat, density, and boiling point. These models were programmed by setting up, sorting, and numerically integrating the differential equations for McMillan2004.book Page 4 Wednesday, September 17, 2003 11

40、:24 AMSIMULATION 5accumulating energy and material within a volume. These models typically could only be run and maintained by the programmer, who was one of an elite handful of specialists in the process industry.Just before the turn of the new century, graphi-cally configured dynamic PFD models wi

41、th extensive physical property packages appeared that greatly expanded the scope of the potential applications and the models users. Some “state-of-the-art” software offers users the ability to switch a steady state PFD model into the dynamic mode. Steady state models with comprehensive physi-cal-pr

42、operty data packages have been used to design processes since the 1960s. As shown in figure 1-1, these models all assume that the accu-mulation, generation, and consumption of mass and energy are zero, so the outputs from the vol-ume can be calculated from the inputs by an iterative procedure. The b

43、oundary could be the whole or subdivided volume of a piece of pro-cess equipment, such as a heat-exchanger pass or column tray or a section of piping. These are classified as “lumped parameter models” since McMillan2004.book Page 5 Wednesday, September 17, 2003 11:24 AM6 SIMULATIONthey do not involv

44、e partial differential equa-tions. Since the accumulation, generation, and consumption are all zero, steady state models cannot be used to simulate batch operations, startups, shutdowns, transitions, reaction kinet-ics, crystal growth or attrition, and cell birth, growth, or death. With respect to p

45、rocess dynamics, successive runs could be made to show the change in the process variable within the volume for a specific change in a disturbance or manipulated variable. In fact, a steady state model excels at this capability since it has the quality and complexity of detail thats needed to reveal

46、 process relationships and interactions. However, some steady state models require hours or days to converge on a solution. In fact, for large changes, they may never converge. These realities can render them an ineffective tool for identifying gains and exploring new operating regions. Dynamic mode

47、ls can move to a drastically different set of operating conditions that in a steady state model would have caused severe, if not fatal, convergence problems. Also, dynamic models are needed to show process time delays and time lags. However, untilMcMillan2004.book Page 6 Wednesday, September 17, 200

48、3 11:24 AMSIMULATION 7Figure 1-1 Lumped Parameter ModelAccumulation, Generation,and Consumption ofMaterial and EnergyInputsOutputsSubsystemBoundaryRrecycleInsteadystatemodels, the accumulation, generation, andconsumptionarezero. Valve size, pressuredrop,andpositionhavenoeffectonflow. Theoutputsareca

49、lculatedfrom inputs. For recyclestreams,the programiterates untilthe outputandthe inputofthe recycleblock are withina tolerance spec.McMillan2004.bookPage7 Wednesday,September 17,200311:24 AM8 SIMULATIONrecently, developing dynamic models was a sepa-rate and intensive effort that involved program-ming hundreds to thousands of differential equations. Now that dynamic PFD models can be built on the knowledge stored in the steady state model, we have the best of both worlds.ProcedureThe procedure presented in this section describes how to create the high-fidelity virtual