Please use this identifier to cite or link to this item:
http://acervodigital.unesp.br/handle/11449/75500
- Title:
- Artificial neural network model of discharge lamps in the power quality context
- Universidade de São Paulo (USP)
- Universidade Estadual Paulista (UNESP)
- 2195-3880
- 2195-3899
- This paper presents a methodology for modeling high intensity discharge lamps based on artificial neural networks. The methodology provides a model which is able to represent the device operating in the frequency of distribution systems, facing events related to power quality. With the aid of a data acquisition system to monitor the laboratory experiment, and using $$\text{ MATLAB }^{\textregistered }$$ software, data was obtained for the training of two neural networks. These neural networks, working together, were able to represent with high fidelity the behavior of a discharge lamp. The excellent performance obtained by these models allowed the simulation of a group of lamps in a distribution system with shorter simulation time when compared to mathematical models. This fact justified the application of this family of loads in electric power systems. The representation of the device facing power quality disturbances also proved to be a useful tool for more complex studies in distribution systems. © 2013 Brazilian Society for Automatics - SBA.
- 1-Jun-2013
- Journal of Control, Automation and Electrical Systems, v. 24, n. 3, p. 272-285, 2013.
- 272-285
- Artificial neural networks
- Computational model
- Discharge lamp
- Electrical system simulation
- Power quality
- Artificial neural network models
- Data acquisition system
- Distribution systems
- Electrical systems
- High intensity discharge lamps
- Laboratory experiments
- Power quality disturbances
- Computer simulation
- Discharge lamps
- Electric power systems
- Facings
- Local area networks
- Mathematical models
- MATLAB
- Neural networks
- http://dx.doi.org/10.1007/s40313-013-0027-0
- Acesso restrito
- outro
- http://repositorio.unesp.br/handle/11449/75500
There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.