Skip to Main content Skip to Navigation
Theses

State and parameter estimation, and identifiability of quasi-LPV models

Abstract : Two problems relevant to the model-based approaches to fault diagnosis and degradation estimation in commissioned buildings are investigated in this thesis: adaptive observers for state and parameter estimation, and parameter identifiability. The system models considered are the quasi-LPV models with affine parameterization. Using the Takagi-Sugeno (T-S) polytopic approach, two observer designs, one for continuous-time models and another for discrete-time models are provided. Both models use a Luenberger structure for the state estimation part and deploy the Lyapunov design approach. An innovative non-linear estimation model is obtained through the design process for the continuous-time parameter estimation whereas a proportional-integral (PI) structure is used for discrete-time. A brief third contribution is a decoupled state and parameter estimation that makes use of the parity-space approach and realized using a finite memory observer strategy. For the fourth contribution of parameter identifiability, a parity-space formulation using null-space computation is used for the elimination of states of the model from which the exhaustive summary of the model is extracted and the identifiability of the model verified. All the results are illustrated using examples
Document type :
Theses
Complete list of metadatas

Cited literature [134 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-01862671
Contributor : Abes Star :  Contact
Submitted on : Monday, August 27, 2018 - 3:39:05 PM
Last modification on : Friday, May 17, 2019 - 11:40:41 AM
Long-term archiving on: : Wednesday, November 28, 2018 - 3:55:15 PM

File

DDOC_T_2018_0059_SRINIVASARENG...
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01862671, version 1

Collections

Citation

Krishnan Srinivasarengan. State and parameter estimation, and identifiability of quasi-LPV models. Automatic. Université de Lorraine, 2018. English. ⟨NNT : 2018LORR0059⟩. ⟨tel-01862671⟩

Share

Metrics

Record views

275

Files downloads

597