1 / 104

Data-Based Modelling for Control

Data-Based Modelling for Control. Paul M.J. Van den Hof. www.dcsc.tudelft.nl/~pvandenhof/publications. 2006 IEEE Workshop Advanced Process Control Applications for Industry (APC2006), Vancouver, Canada, May 8-10 2006. Contents. Introduction. Basic facts on system identification.

camden
Download Presentation

Data-Based Modelling for Control

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Data-Based Modelling for Control Paul M.J. Van den Hof www.dcsc.tudelft.nl/~pvandenhof/publications 2006 IEEE Workshop Advanced Process Control Applications for Industry (APC2006), Vancouver, Canada, May 8-10 2006.

  2. Contents • Introduction • Basic facts on system identification • Example from a MSW incineration plant • Models for control • Model uncertainty and model validation • Basis functions model structures • Cheapest experiments • Discussion and prospects

  3. “obtaining process models is the single most time- consuming task in the application of model-based controllers” (Ogunnaike, An Rev Control, 1996; Hjalmarsson, Automatica, 2005) Introduction Costs distribution in an advanced process control project: • Feasibility study 10% • Pre-tests 10% • Model identification 40% • Controller tuning 15% • Commissioning and training 25% (Zhu, IFAC SYSID, 2006) Intro Sysid ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  4. Which kind of models to consider? First principles / rigorous models Process design; planning and scheduling; off-line • large number of equations (PDE,ODE,DAE) • high computational complexity • question of validation • nonlinear Data-based models • compact model structures • computational feasible • validated by data • often linearized Advanced control; on-line operations; on-line For advanced process control data-based models seem dominant Intro Sysid ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  5. On-line use of first principle models • State dimension >> • f and h nonlinear • For monitoring/diagnosis problems, state variables have clear • physical interpretation, which has to be retained • Full models in general too complex for on-line evaluations • Input-output model reduction destroys the state structure • State-based model reduction techniques (POD,…) only help computationally in the case of linearf and h • The (nonlinear) mappings have to be approximated/simplified Intro Sysid ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  6. Model structures Black box Emphasis, for the moment Well sorted out in linear case Not mature in nonlinear case Physics-based Problem of accurate parametrization (where to put the unknowns?) Identifiability Data-based models (identification) • Relatively easily obtained • Model costs are related to experiments on the plant Intro Sysid ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  7. “Here is a dynamical process with which you are allowed to experiment (preferably cheap). Design and implement a high-performance control system”. • Issues involved: • Experiment design • Modelling / identification • Characterization of disturbances and uncertainties • Choice of performance measure • Control design and implementation • Performance monitoring Intro Sysid ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  8. t 1.4 p 1.2 M p ± 1% 1 90% 0.8 0.6 y(t) 0.4 0.2 10% 0 t r t s time Classical experiments for finding control-relevant dynamics • Ziegler/Nichols tuning rules • for PID-controllers • Relay feedback: amplitude and frequency at -180° phase Intro Sysid ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  9. Ad-hoc simple cases to be extended to • general methodology for model-based control, • including issues of robustness induced by • model uncertainties Intro Sysid ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  10. Identification for Control (1990-…) • Basic principles for identifying models, well sorted out • Relation with control through • Certainty equivalence principle: • “Controller based on exact model is suited for • implementation on the plant” However: • Identification had been extended to identify • approximatemodels • Control design had been evolved to robust control • taking account of model uncertainties Intro Sysid ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  11. Data Model Feedback control system Feedback control system disturbance reference input + output Model Controller controller process controller process - Experiments: Intro Sysid ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  12. r + C u G0 y - When is a model suitable for control? For a given controller C: r Ĝ + C u y - Designed loop Achieved loop • Both loops should be “close” (r  y): should be small • Disturbance effects on y should be similar Intro Sysid ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  13. wb When is a model suitable for control? plant model1: accurate for w<wb model2: accurate for wwb Model quality becomes dependent on control bandwidth (to be designed) Intro Sysid ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  14. Control bandwidth is based on model + .. exp. design Experiment Experiment Experiment data If models are uncertain/approximate due to limited experiment, achievable performance needs to be discovered Identificatie Identification model Control design Regelaarontwerp controller ! modelling for control is learning (Schrama, 1992; Gevers, 1993) Implementatie Implementatie Implementation evaluation Intro Sysid ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  15. high-order model high-order controller experiment data low-order model low-order controller From experiment to control: validation and uncertainty • Current opinion: • Extract all information from data, but • Keep experiments simple Intro Sysid ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  16. Identification Control 1990 - : Schrama, Gevers, Bitmead, Anderson, Åström, Rivera, …. • control-relevant • nominal model • nominal control 1994 - : Hakvoort, de Vries, Ninness, Bitmead, Gevers, Bombois, … • nominal model + • uncertainty bound • nominal control + • stab/perf robustness • control-relevant • model uncertainty set • robust control; worst-case • performance optimiz. 1997 - : de Callafon & vdHof,Douma • design of “cheap” • experiments for id of • uncertainty sets 2002 - : Bombois, Gevers,Hjalmarsson, vdHof, • control under performance • guarantees Development trend: Intro Sysid ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  17. Contents • Introduction • Basic facts on system identification • Example from a MSW incineration plant • Models for control • Model uncertainty and model validation • Basis functions model structures • Cheapest experiments • Discussion and prospects

  18. Basic facts on system identification Identification of parametric models through prediction error identification (open-loop) Data generating system: Predictor model: e is realization of stochastic white noise process From measured data {u(t),y(t) }, t=1,..,N to estimated model Intro System ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  19. {u(t),y(t) }, t=1,..,N fractions of polynomials Convex or non-convex optimization Prediction error framework: (Ljung, 1987) Intro System ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  20. 2. If and u is sufficiently exciting then provided that G and H are parametrized independently. Asymptotic variance typically dependent on (frequency-dependent noise to signal ratio) Classical consistency results 1. If and u is sufficiently exciting then Intro System ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  21. Since parameter estimates are asymptotically normally distributed (cental limit theorem), the variance expression can be converted to parameter confidence regions, e.g. 3s-bounds (99.7%). Using the mappings • the uncertainty bounds • can be converted to • frequency response • step response • etc. Intro System ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  22. Computational issues 1. In general situation: non-convex optimization(with risk of local minima) 2. Convex optimization if prediction error is affine in the parameters:property of model structure: FIR: ARX: ORTFIR: with A,B,Fpolynomials in q-1 : Intro System ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  23. 1. If then 2. If and (fixed) then Design variables in general case: model structure Characterization of asymptotic estimate Limiting parameter estimate: i.e. minimizing the power in the weighted residual signal Substituting the expressions from the signal block diagram delivers Intro System ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  24. Same approach can be followed (direct method), on the basis of measurements u(t),y(t) Consistency result: provided that and either: • r is sufficiently exciting, or • C is sufficiently complex (high order / time-varying) Accurate noise modelling is necessary for identifying G Closed-loop situation Intro System ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  25. In direct closed-loop identification, possibilities for separately identifying G0 and H0 are lost. In a MIMO situation this happens already when a single loop is closed: Intro System ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  26. Asymptotic variance in closed-loop identification where now because of the closed-loop. Writing a simple analysis leads to reference part noise part (only the reference part of the input signal contributes to variance reduction) Intro System ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  27. 2. Two-stage method • Identify transfer r  u • Simulate • Identify G0 as transfer Input signal u is denoised Alternative indirect methods When focussing on plant model only Several options, among which: 1. Indirect Method • Identify transfer r  y • Retrieve plant model, with knowledge of C Intro System ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  28. Properties indirect methods 1. If then 2. If then provided that r is sufficiently exciting and C is linear General expression for the asymptotic estimate (with slight variations) Closed-loop properties of the plant are approximated. Note that: separate identification of G0 and H0 is possible. Intro System ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  29. |.| T 0 10 S -1 10 -2 10 -1 0 ω 10 10 red blue Low frequencies are hidden; frequencies around bandwidth are emphasized Intro System ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  30. Alternative closed-loop ID methods • IV methods, using r as instrumental variable • Coprime factor identification (related to gap, nu-gap metric) • Dual-Youla identification Intro System ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  31. Wrap-up PE identification • Mature framework for system ID • Open-loop and closed-loop data can be handled • Stochastic noise framework • Extensions to multivariable situation available • Analysis is available but mainly for infinite data • Analysis much more explicit than e.g. for subspace ID /state-space models approximate models – design variables Intro System ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  32. Contents • Introduction • Basic facts on system identification • Example from a MSW incineration plant • Models for control • Model uncertainty and model validation • Basis functions model structures • Cheapest experiments • Discussion and prospects

  33. Municipal Solid Waste Combustion (Martijn Leskens) IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  34. (Nolinear) Model Predictive Control of MSWC Plants • Aim: NMPC of furnace and boiler part of MSWC plant: NMPC requires good dynamic model of MSWC plant  MODEL VALIDATION Simulation results IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  35. Closed-loop identification of MSWC plants • Closed-loop experimental configuration typically encountered in MSWC plants: IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  36. Experimental model is fit in the same i/o structure as the first principles model “PARTIAL” closed-loop identification: u1 = “open-loop” inputs y1 = “open-loop” outputs Etc. IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  37. Goal • Identification of linear models in 2 working points:Tprim = 70, 120 °C • Use these models to validate/calibrate a simple first-principles model Identification setup • RBS excitation of all controlled inputs • Closed-loop identification with (indirect) two-stage method • Use of high-order ARX models and model-reduction • Enforcements of static gains to improve low-frequent behaviour • Sample time of 1 minute • Identified model validated through correlation tests • 8 scalar transfers identified with order between 2 – 5. Simplified physical model (5th order NL) tuned to identified models. IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  38. Considerable disturbances on output data: dashed is measured data, solid is simulated data IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  39. Estimated model (dashed) and NL-physical model (solid) upon excitation of the waste inlet IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  40. Estimated model (dashed) and NL-physical model (solid) upon excitation of the primary air flow IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  41. Results • Identification and validation results for Tprim = 70 (I): good to very good agreement: Responses on step from waste inlet flow IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  42. Results • Identification and validation results for Tprim = 120 (I): moderate to reasonable agreement: Responses on step from waste inlet flow IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  43. Results • Closed-loop identification strategy is ‘easily’ applicable in an industrial setting and works well • Fitting of first-principles model is still rather ad-hoc • Models are accurate enough for model-based MPC IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  44. Contents • Introduction • Basic facts on system identification • Example from a MSW incineration plant • Models for control • Model uncertainty and model validation • Basis functions model structures • Cheapest experiments • Discussion and prospects

  45. wb How poor can models be? plant model1: accurate for w<wb model2: accurate for wwb IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  46. Controlled with 5th order controller, with I-action, bandwidth 0.5 rad/s Model quality becomes dependent on control bandwidth (to be designed) IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  47. In general terms: • Need for a structured way to measure control relevance of models, and • methods to identify them from data What looks like a good model in open-loop may be poor in closed-loop and vice versa • Rule of thumb: models need to be accurate around control bandwidth IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  48. r + C u G0 y - When is a model suitable for control? For a given controller C: r Ĝ + C u y - Designed loop Achieved loop Performance measure for model quality could be: The power of the difference signal: In frequency domain: IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  49. Indirect closed-loop ID delivers: Can this performance measure be minized through identification? Requested: Conclusion: A C-relevant model is identified by indirect closed-loop ID, by choosing IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

  50. Can this be achieved by open-loop identification? OL-expression (OE-case): Required integrand: This requires: which is unfeasible because of lack of knowledge of IntroSystem ID MSW-example Models for control Uncertainty&validation Basis functions Experiment design Discussion

More Related