Ginarsa, I Made and Soeprijanto, Adi and Purnomo, Mauridhi Hery (2008) Dokumen: Modeling of Chaotic Behavior in Power Systems using Recurrent Neural Networks. In: International Conference on Advanced Computational Intelligence and Its Application (ICACIA-2008) University of Indonesia September 1st – 2nd, 2008, 1-2 Sept 2008, Jakarta, Indonesia.
|
Text
Proceedings_ICACIA_2008_UI_Indonesia.pdf Download (540kB) | Preview |
Abstract
This paper presents the intensely studied route to chaotic oscillation in power systems. By using a three-bus simple power system, a route was found to cause chaotic behavior in the power systems which are evaluated, illustrated, and discussed in this study. Furthermore, chaotic behavior by using recurrent neural networks (RNN) and exact models are compared. In particular, we have proposed that RNN can be trained by using it on both the present input and past output, using back-propagation algorithm with adaptive learning rate and momentum. Performance of learning rate with momentum is better than learning rate without momentum. The appearance of chaotic behavior in a power system is already proven and can be modeled by using the RNN. A chaotic behavior is detected by a strange attractor (a chaotic attractor) in the phase-plane. The largest mean squared error (MSE) was observed to be 7.8296% obtained on the rotor speed at a disturbance of 1.7003 rad/sec. On the contrary, the least MSE was 0.0407% obtained on load voltage at disturbance 1.600 rad/sec.
| Item Type: | Conference or Workshop Item (Paper) |
|---|---|
| Keywords (Kata Kunci): | Power systems, chaotic behavior, recurrent neural networks (RNN), chaotic attractor, phase-plane. |
| Subjects: | T Technology > TK Electrical engineering. Electronics Nuclear engineering |
| Divisions: | Fakultas Teknik |
| Depositing User: | Dr. I Made Ginarsa S.T,M.T |
| Date Deposited: | 11 May 2023 04:27 |
| Last Modified: | 11 May 2023 04:27 |
| URI: | http://eprints.unram.ac.id/id/eprint/37636 |
Actions (login required)
 |
View Item |
