The Linear Parameter-Varying (LPV) paradigm represents a naturalextension of the classical Linear Time-Invariant (LTI) framework.By virtue of the so-called scheduling signal, LPV modelscan accurately describe the behavior of a large class of timevaryingand nonlinear dynamical systems, while preservingthe linearity between input and output signals. Thus, LPVmodels are of great practical significance for modeling andcontrol of many industrial systems, as the simplicity of theLTI theory is retained. This thesis is concerned with datadrivenmodeling of LPV systems, addressing some of thewell-known issues pertaining to LPV model identification.In particular, the main research questions addressed in thiscontribution include selection of the model structure, dealingwith noisy measurements of the scheduling signal, identificationof the plant model from closed-loop data, and identificationof linear fractional representations of LPV modelswhich are especially suitable for controller synthesis. Theproposed methodologies mainly fall under the framework ofparametric and non-parametric LPV identification using machinelearning techniques and provide a set of tools for automatedselection of LPV modeling parameters. Furthermore,as an alternative to the conventional parametric and nonparametricapproaches, a novel algorithm based on mixedintegerprogramming is devised for identifying LPV modelsthrough piecewise affine regression. The framework of thisalgorithm is applied to the problem of energy disaggregationwith a real-world benchmark dataset.
Towards automated data-driven modeling of linear parameter-varying systems / Mejari, M.D.. - (2018 Jul 06). [10.13118/mejari-manas-dilip_phd2018]
Towards automated data-driven modeling of linear parameter-varying systems
Mejari, Manas Dilip
2018
Abstract
The Linear Parameter-Varying (LPV) paradigm represents a naturalextension of the classical Linear Time-Invariant (LTI) framework.By virtue of the so-called scheduling signal, LPV modelscan accurately describe the behavior of a large class of timevaryingand nonlinear dynamical systems, while preservingthe linearity between input and output signals. Thus, LPVmodels are of great practical significance for modeling andcontrol of many industrial systems, as the simplicity of theLTI theory is retained. This thesis is concerned with datadrivenmodeling of LPV systems, addressing some of thewell-known issues pertaining to LPV model identification.In particular, the main research questions addressed in thiscontribution include selection of the model structure, dealingwith noisy measurements of the scheduling signal, identificationof the plant model from closed-loop data, and identificationof linear fractional representations of LPV modelswhich are especially suitable for controller synthesis. Theproposed methodologies mainly fall under the framework ofparametric and non-parametric LPV identification using machinelearning techniques and provide a set of tools for automatedselection of LPV modeling parameters. Furthermore,as an alternative to the conventional parametric and nonparametricapproaches, a novel algorithm based on mixedintegerprogramming is devised for identifying LPV modelsthrough piecewise affine regression. The framework of thisalgorithm is applied to the problem of energy disaggregationwith a real-world benchmark dataset.| File | Dimensione | Formato | |
|---|---|---|---|
|
Mejari_phdthesis.pdf
Accesso aperto
Tipologia:
Tesi di dottorato
Licenza:
Creative commons
Dimensione
2.11 MB
Formato
Adobe PDF
|
2.11 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


