الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 132 |  |  |
| | | from | 132 | | |
Abstract
The power system conditions always keep on changing because of the continuous variation nature of the system loads over time and also because of being subjected to several disturbances. According to the size of these disturbances, the power system maintains its stability. As a result of this disturbances, the power system may operate on the verge of stability. The protective relays within the power system should offer the required flexibility and adapt its setting to help maintain the system stability. Many adaptive techniques have been introduced so that the power system protective schemes always offer the required protection for the system elements thus increasing its reliability.
It is highly important to use a dependable algorithm that can evaluate the system conditions, and define the current status of the power system. The data mining-based techniques are distinctive as they are highly dependable and accurate. The most frequently used data mining algorithm for the system evaluation process is the decision trees (DT).
This thesis is interested in the data mining techniques that is used in evaluating the status of the power system. It presents a data mining model depending on support vector machines (SVM) that is built for classifying the system condition after analyzing the data coming from the system. This model is responsible for triggering on and off a protective methodology according to the system status whether it is normally stable; referred to as safe, or on the verge of stability; referred to as stressed.
The thesis offers two comparisons between two different data mining models; one of the models depends on DT which is the most widely data mining technique in power system, and the other model uses SVM technique presented in this thesis. The two comparisons depend on training and testing the two models using two different databases; data base-1 and data base-2. Data base-1 was used in a previous study, while data base-2 was generated from IEEE 30-bus test system. It was built by solving load flow analysis problem for IEEE-30 bus test system several times at different loading conditions. For each individual case a load flow analysis is performed, the system parameters at each bus of the IEEE-30 bus test system such as voltage magnitude, current, active and reactive powers are recorded in the database. The result of the two comparisons shows that the SVM model proves to have higher correct rate in evaluating and predicting the system conditions than the decision trees model. Using a higher correct rate model in determining the system conditions helps enhancing the power system reliability.