Change search
ReferencesLink to record
Permanent link

Direct link
An implementation of recurrent neural networks for prediction and control of nonlinear dynamic systems
2003 (English)Independent thesis Advanced level (professional degree), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

Neural networks have been used in identification and control of nonlinear dynamic systems for decades. Using recurrent neural networks which are themselves dynamic systems makes is possible to achieve models superior both to linear models and feedforward neural networks. Recurrent neural networks are able to capture the true hidden dynamics of nonlinear systems. A software framework is developed which provides easy creation and training of arbitrary perceptron networks. An introduction to dynamic systems and system identification is given, as well as a short introduction to Kalman Filtering. The basics of neural networks in general and recurrent neural networks in particular are explained. Recurrent multilayer perceptrons trained by the extended Kalman filter method are evaluated in a series of experiments. First, network structures are evaluated empirically by using them to identify and predict chaotic time series. Identification performance on both real world systems and simulated systems is compared to results obtained using linear models and feedforward neural networks. An identified model of a simulated nonlinear dynamic system demonstrates the use of a recurrent neural network in a predictive control scheme.

Place, publisher, year, edition, pages
Keyword [en]
Technology, Recurrent neural networks, System identification, Extended, Kalman filter, Predictive control
Keyword [sv]
URN: urn:nbn:se:ltu:diva-46313ISRN: LTU-EX--03/119--SELocal ID: 3f3a292d-4e1c-4ced-a78c-a818ddd89387OAI: diva2:1019627
Subject / course
Student thesis, at least 30 credits
Educational program
Computer Science and Engineering, master's level
Validerat; 20101217 (root)Available from: 2016-10-04 Created: 2016-10-04Bibliographically approved

Open Access in DiVA

No full text

Search outside of DiVA

GoogleGoogle Scholar

Total: 1 hits
ReferencesLink to record
Permanent link

Direct link