Fast transient security evaluation of power systems by using pattern recognition techniques
A power system is a dynamic system. The reaction of a power network to the same set of disturbances is different for various initial equilibrium states. For a given set of contingencies, some of the initial equilibrium states are stable and some of them are unstable. The purpose of this dissertation is to identify if a given operating conditions of the system is stable (secure) or unstable (insecure) for certain disturbances by using real time data. The time required for on-line security analysis can be reduced if pattern recognition techniques are employed. The use of a pattern recognition technique in on-line transient security analysis of power systems is examined. Load magnitudes are treated as random variables with an assumed statistical distribution having a standard deviation of 10%. The simulation technique is applied, off-line, to check system security for the defined set of contingencies. For each initial system condition, the potentially good variables are identified. The number of variables is reduced and variables with the highest discriminatory power are identified. Two decision rules are then developed by using Generalized Square Distance and K-Nearest Neighbor classification techniques. Next, the performance of each classifier is evaluated by using two risk estimating techniques, Jackknife Risk Estimation and Independent Test Risk estimation. The best classifier is identified. Finally, using this classifier, a computer program is developed. This program is capable of predicting, on-line, the security and insecurity of the given power system for any initial system condition within the range defined for the training set. The important features of this program are its accuracy, speed, adaptability and up-dating scheme.
- OSTI ID:
- 6063219
- Resource Relation:
- Other Information: Thesis (Ph. D.)
- Country of Publication:
- United States
- Language:
- English
Similar Records
Fast transient stability calculation by pattern recognition method
Application of pattern recognition in transient security assessment