Online monitoring of power system stability is essential for its reliability, especially if it includes a large amount of alternative energy sources. Each type of power system stability is unique, requiring a specialized approach for its assessment. In this thesis a new approach to small-signal stability assessment is presented, which does not require real time calculation of eigenvalues or eigenvectors. Considering that both are strongly dependent on power system operating conditions (list of operational power plants, their power infeed, power flow, topology), which are reoccurring to a certain extent, the proposed concept relies on a data base consisting of numerous foreseen/past operating conditions that have been analysed in advance. Screening for similarities within the database is something that is performed in real time. The core od similarity screening is a decision tree technique, which splits the original data in two layers, according to the list of operating power plants and topology. Only then screening of the database for most similar operating conditions is done by help of principal component analysis and k-d trees. In the thesis the efficiency and reliability of the method are tested on the IEEE 39 bus system. The speed and accuracy using different input values is assessed and compared to the speed and accuracy of the existing method. The possibility of combining both methods is also presented, thus using advantages of both methods and avoiding their deficiencies.
|
