1998 Fiscal Year Final Research Report Summary
Application of Artificial Intelligence for Modeling of Power Systems
| Project/Area Number |
09650328
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
電力工学・電気機器工学
|
|---|
| Research Institution | Kumamoto University |
|---|
Principal Investigator |
HIYAMA Takashi Faculty of Engineering, Professor, 工学部, 教授 (90040419)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KITA Toshihiro Faculty of Engineering, Instructor, 工学部, 助手 (20284739)
|
|---|
| Project Period (FY) |
1997 – 1998
|
| Keywords | Neural Networks / Nonlinear Systems / Governor-Turbine System / Stability Evaluation / Nose Curve / Dynamic Load Modeling |
|---|
| Research Abstract |
In this project, an artificial intelligence, especially artificial neural network, based new method has been proposed for the modeling of electric power systems. Study systems are modeled by using artificial neural networks based on the
measured real data. The proposed artificial neural networks are multi-layered ones with additional feedback loops from the output layer to the input layer with time delay. By using the proposed artificial neural networks, non-linear systems can be modeled quite accurately with relatively lower order non-linear difference equations. The proposed modeling method has been applied to the modeling of the load dynamics, the governor-turbine system foe a LNG thermal unit, and the dynamics between the real power and the system voltage on 500kV transmission lines. The accuracy of the proposed models have been demonstrated through comparison studies using actual measured data on the study systems. The comparison studies have also been performed between the proposed models and the conventional linear models. The proposed artificial neural network based models give quite accurate responses for given disturbances. In addition, the proposed models are robust ones, therefore, the models are available to some extent for the different situations from ones when the actual data were measured. By combining the proposed models with conventional models, more accurate stability analysis will be performed.
|
