Investigation Of Performance Of Parameter Estimation Method In System Identification

Mohd Rafie, Nur Afifah Faizah (2016) Investigation Of Performance Of Parameter Estimation Method In System Identification. Project Report. UTeM, Melaka, Malaysia. (Submitted)

Text (24 Pages) Investigation Of Performance Of Parameter Estimation Method In System Identification.pdf - Submitted Version Download (385kB)

Abstract

System identification aims to develop mathematical models for dynamical systems using measured input and output signals. Model structure selection is one of the important steps in a system identification process. Several important criteria for a desirable model structure include its accuracy in model output, model residuals, final prediction error (FPE) and loss function. For this project, linear model was used that is ARX model. Dynamic models are processed before attempting model parameter or structure estimation procedures. This project was carried out using the Graphical User Interface (GUI) in MATLAB application. This project explores the performance of difference parameter estimation method in identification. Parameter estimation is one of the most important things that need to be considered in accurately describing system behavior through mathematical models such as statistical probability distribution functions, parametric dynamic models, and data-based models. The effectiveness of least square and instrumental variable (IV) estimator are investigated. This project brief overview of the system identification process. Vital to pay consideration on all parts of this methodology, from the experiment design to model validation in order to get best results. The usefulness and applicability of both methods are discussed based on the results of model output, model residuals, final prediction error (FPE) and loss function. In conclusion, this investigation found that instrumental variable performs slightly better because the noise is purposely made correlated with input. In opposition, least square is known be better when noise is uncorrelated. This prove the theoretical that instrumental variable method can balance the impacts of a more general class of noise signals while least square method can minimizes the sum of the squared prediction errors between the prediction model and the output data.

Actions (login required)

View Item