Autocovariance-based plant-model mismatch estimation for linear model predictive control

Wang, Siyun; Simkoff, Jodie M.; Bâldea, Michael; Chiang, Leo H.; Castillo, Iván; Bindlish, Rahul; Stanley, David B.

doi:10.1016/j.sysconle.2017.03.002

Cited by 17 publications

(16 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Note thatΞ i are the coefficient matrices of the corresponding impulse response model (28) in the MPC. S i and Ξ i are step response and impulse response model coefficient matrix mismatches that arise due to parametric model mismatch a ij,l and b ij,l defined by (6), (7) for discrete time systems or k ij , ij , and ij defined by (17) for continuous time systems.…”

Section: Explicit Autocovariance-mismatch Relationmentioning

confidence: 99%

“…We first consider the nonparametric presentation of the plant in (1), as an infinite‐horizon step response (ISR) model (or alternatively a finite‐horizon step response (FSR) model) of the following form:

ISR: y (k) = \sum_{i = 1}^{+ \infty} S_{i} Δ u (k - i) + H_{o} e (k),

FSR: y (k) = \sum_{i = 1}^{N} S_{i} Δ u (k - i) + S_{N} u (k - N - 1) + H_{o} e (k)

with

Δ u (k) = u (k) - u (k - 1)

as the input vector, where N > 0 is the model horizon and S i are the step response model coefficient matrices that can be generated by the parametric models

G_{o} (z^{- 1}; θ)

in (1) or

G_{o} (s; θ)

in (21). For the details of computing coefficients { S i } from transfer functions, please refer to Reference 17. It should be noted that we are considering MIMO systems and thus S i are matrices.…”

Section: Problem Statementmentioning

confidence: 99%

“…In this section, we first follow the work in Reference 17 to establish an explicit nonlinear relation between the plant‐model mismatch and the autocovariance matrices of output signal y ( k ), where noise parameters will also play an important role. Then, we present the algorithm for carrying out the plant‐model coefficient mismatch estimation as well as the noise model parameter estimation simultaneously.…”

Section: Plant‐model Mismatch Estimationmentioning

confidence: 99%

“…Motivated by this, in our previous work we developed methods for plant‐model mismatch quantification using routine closed‐loop operating data. Our previous works have contributed to plant‐model mismatch estimation for parametric MPC models including those in transfer‐function 17 and state space form, 18 based on assumption that the exact noise model is known .…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Data‐driven plant‐model mismatch estimation for dynamic matrix control systems

Simkoff

Bâldea

et al. 2020

Intl J Robust & Nonlinear

Self Cite

View full text Add to dashboard Cite

This article addresses the plant-model mismatch estimation problem for linear multiple-input and multiple-output systems operating under the dynamic matrix control (DMC) implementation of model predictive control. An autocovariance-based method is proposed, aiming to identify parameter values that minimize the discrepancy between the theoretical autocovariance matrices derived from implementing the (explicit) DMC control law and the sampled autocovariance matrices calculated from operating data. We provide proof that the method results in unbiased estimates. A means for dealing with potential overfitting issues caused by the finite step response models used in DMC in practice is proposed. Several examples are presented to illustrated the theoretical developments. K E Y W O R D S autocovariance-based method, dynamic matrix control, model predictive control, noise model parameter estimation, plant model mismatch estimation

show abstract

Section: Explicit Autocovariance-mismatch Relationmentioning

confidence: 99%

ISR: y (k) = \sum_{i = 1}^{+ \infty} S_{i} Δ u (k - i) + H_{o} e (k),

FSR: y (k) = \sum_{i = 1}^{N} S_{i} Δ u (k - i) + S_{N} u (k - N - 1) + H_{o} e (k)

with

Δ u (k) = u (k) - u (k - 1)

as the input vector, where N > 0 is the model horizon and S i are the step response model coefficient matrices that can be generated by the parametric models

G_{o} (z^{- 1}; θ)

in (1) or

G_{o} (s; θ)

Section: Problem Statementmentioning

confidence: 99%

Section: Plant‐model Mismatch Estimationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Data‐driven plant‐model mismatch estimation for dynamic matrix control systems

Simkoff

Bâldea

et al. 2020

Intl J Robust & Nonlinear

Self Cite

View full text Add to dashboard Cite

show abstract

“…Also, iterative learning control, repetitive control, and run-to-run control, collectively referred to as learning-type adaptive control mechanisms, have been studied to address repetitive and run-based processes [59]. To determine plant-model mismatches for multiple-input multiple-output systems, an autocovariance-based estimation methodology for operation under model-based control has also been investigated in the literature [60]. ANNs have also been studied as adaptive control elements for the improvement of power plant performance [61].…”

Section: Setpoint Trackingmentioning

confidence: 99%

Development of Biomimetic-Based Controller Design Methods for Advanced Energy Systems

Mirlekar¹

View full text Add to dashboard Cite

Development of Biomimetic-based Controller Design Methods for Advanced Energy Systems Gaurav Mirlekar A biologically inspired optimal control strategy, denoted as BIO-CS, is proposed for advanced energy systems applications. This strategy combines the ant's rule of pursuit idea with multi-agent and optimal control concepts. The BIO-CS algorithm employs gradient-based optimal control solvers for the intermediate problems associated with the leader-follower agents' local interactions. The developed BIO-CS is integrated with an Artificial Neural Network (ANN)-based adaptive component for further improvement of the overall framework. In particular, the ANN component captures the mismatch between the controller and the plant models by using a singlehidden-layer technique with online learning capabilities to augment the baseline BIO-CS control laws. The resulting approach is a unique combination of biomimetic control and data-driven methods that provides optimal solutions for dynamic systems. The applicability of the proposed framework is illustrated via an Integrated Gasification Combined Cycle (IGCC) process with carbon capture as an advanced energy system example. Specifically, a multivariable control structure associated with a subsystem of the IGCC plant simulation in DYNSIM ® software platform is addressed. The proposed control laws are derived in MATLAB ® environment, while the plant models are built in DYNSIM ® , and a previously developed MATLAB ®-DYNSIM ® link is employed for implementation purposes. The proposed integrated approach improves the overall performance of the process up to 85% in terms of reducing the output tracking error when compared to stand-alone BIO-CS and Proportional-Integral (PI) controller implementations, resulting in faster setpoint tracking. Other applications of BIO-CS addressed include: i) a nonlinear fermentation process to produce ethanol; and ii) a transfer function model derived from the cyber-physical fuel cell-gas turbine hybrid power system that is part of the Hybrid Performance (HYPER) project at the National Energy Technology Laboratory (NETL). Other theoretical developments in this work correspond to the integration of the BIO-CS approach with Multi-Agent Optimization (MAO) techniques and casting BIO-CS as a Model Predictive Controller (MPC). These developments are demonstrated by revisiting the fermentation process example. The proposed biologically-inspired approaches provide a promising alternative for advanced control of energy systems of the future. iii Dedication || || Dedicated to Almighty God, my parents Leena and Vikas, my sister Devika, my paternal grandparents Chandrakant and Mangal, and maternal grandparents Vishwanath and Rajani iv Acknowledgments I would like to thank many individuals who impacted and influenced my life throughout my stay in the USA or back in my homeland India. There are too many to mention here, but I do sincerely thank you all. First of all, I would like to gratefully acknowledge, my Guru, Dr. Fernando V. Lima who gave me this great oppor...

show abstract