Development of neural network based techniques for power system transfer capability calculations and reliability evaluation /
Today transmission networks are used in a manner and to an extent not contemplated when they were planned and designed. Transfer capability, acting as an indictor of the capability of a transmission network to reliably move electric power from one area of supply to another of need, is used by system...
| Main Author: | |
|---|---|
| Format: | Thesis Book |
| Language: | English |
| Published: |
[Place of publication not identified] :
[publisher not identified] ;
2000.
|
| Subjects: | |
| Online Access: | http://proxy.library.tamu.edu/login?url=http://proquest.umi.com/pqdweb?did=731990241&sid=1&Fmt=2&clientId=2945&RQT=309&VName=PQD |
| Summary: | Today transmission networks are used in a manner and to an extent not contemplated when they were planned and designed. Transfer capability, acting as an indictor of the capability of a transmission network to reliably move electric power from one area of supply to another of need, is used by system planners to indicate a system's strength and by system operators to determine current ability to transfer power. On the other hand, with the system growing in size and complexity, power system reliability is becoming an increasingly important and complex issue. New techniques need to be developed for power system transfer capability studies and power system reliability evaluation. This dissertation presents neural network based techniques for power system transfer capability studies and power system reliability evaluation. Multi-Layer Feed-Forward Neural Network, which has strong ability to model complex nonlinear relationships between inputs and outputs, is used to develop the techniques for transfer capability studies. Two neural network based techniques are developed to speed up the Monte Carlo simulation for loss of load probability calculation. The first one is based on the Self-organizing Map, which has the ability to project input signals nonlinearly from a high-dimensional space to a 2-dimensional space while maintaining their original topological relationship. The other is based on the Learning Vector Quantization (LVQ), which has the strong classification ability that divides the data space into quantization regions whose borders are hyper-planes. A modified IEEE 30-bus system is used to demonstrate the effectivess of the proposed Multi-Layer Feed-Forward Neural Network based techniques for transfer capability studies. Results have shown that the method is capable of reflecting variation in load levels and status of generation and transmission system accurately. The IEEE Reliability Test System is used to test the proposed methods based on Self-organizing Map and Learning Vector Quantization combined with Monte Carlo simulation. It has been shown that both methods can quickly estimate the value of loss of load probability with reasonable accuracy. Compared to the straight Monte Carlo simulation, the computational time can be greatly reduced by using these two neural network based reliability evaluation techniques. |
|---|---|
| Item Description: | Vita. "Major Subject: Electrical Engineering". |
| Physical Description: | xi, 96 leaves : illustrations ; 28 cm. Issued also on microfiche from University Microfilm Inc. |
| Bibliography: | Includes bibliographical references (leaves 89-95). |