Prediction of transient stability status and coherent generator groups

Abstract
The prediction of transient stability status and instability scenarios such as coherent generator groups has become extremely important for the improvement of power system performance in the event of large disturbances. This research workis aimed at developing various schemes for providing advance information on the stability status of power systems following a transient disturbance. The work also focuses on predicting coherent generator groups that are likely to be formed when a system is predicted to become transient unstable. Power System Simulator for Engineers (PSSE) software is employed to model a test system and carry out simulations. MATLABĀ® software is used for the analysis and development of the prediction schemes. Three schemes for predicting transient stability status and one scheme for predicting coherent generator groups have been developed. Rotor speed deviation following a disturbance was found to be an excellent input parameter for the schemes. The first proposed method for predicting transient stability status employs the Daubechies 4 mother wavelet to decompose rotor speed deviation data obtained for three consecutive cycles after the tripping of a line or bus. The wavelet entropies subsequently obtained are then used as input to an algorithm which predicts the transient stability status of a power system. The prediction accuracy of this method was found to be 91.2%. The second transient stability status prediction schemeuses the sum of the maximum rotor speed deviations of system generators obtained within the first cycle after the tripping of a bus or line following a transient disturbance for its prediction. The obtained sum is then used as input data to a trained multilayer perceptron neural network (MLPNN) which is used for the prediction. This method was found to be 100% accurate. The third scheme developed is anchored onstability status prediction of each generator based the maximum rotor speed deviation (MSD) of that generator, obtained within the first cycle after the tripping of a line or bus. The MSD of each generator has also been used as input to a trained MLPNN assigned to that generator forits transient stability prediction. The method has been found to return 98.05% accuracy in generator transient stability predictions.The final scheme developed is aimed at identifying coherent generator groups only after successful prediction of system transient instability. This final scheme has been realized via two MLPNNs each of which is equipped with three input neurons. Herein, the input data to the three input neurons are: (1) MSD of a reference generator in a predicted coherent group, (2) MSD of a generator to be placed in a group, and (3) the difference between the two MSDs. The proposed coherency prediction scheme has been tested exhaustively and found to be90.43% accurate. The proposed schemes which rely principally on MSD are computationally fast and efficient in their respective predictive capabilities; and can therefore be easily implemented on-line in real systems with the availability of phasor measurement units.
Description
A thesis submitted to the Department of Electrical and Electronic Engineering, Kwame Nkrumah University of Science and Technology, in partial fulfilment for the degree of Doctor of Philosophy.
Keywords
Citation