Power System Transient Stability Prediction Based on GWO-SVM and Boosting Method

As various undesirable features such as intermittency, randomness and low inertia have been introduced into power system operation and control, it is imperative to develop efficient transient stability prediction (TSP) schemes. In view of the wide application of data mining technology, this paper proposes a data-driven method for TSP based on support vector machine (SVM) and ensemble learning. Firstly, grey wolf optimization (GWO) algorithm is introduced to select optimal hyperparameters of SVM. By approaching the selected position in every iteration, the constraint coefficient and balance factor are updated constantly and finally reach the fit values in search space. Moreover, an improved boosting method is applied to enhance the model performance, prediction results become more accurate and robust based on the combination of basic classifiers. Then, for the sake of filtering out the unreliable predictions, a trusted domain is set according to the distance between test sample and discriminant boundary. Finally, case studies on the IEEE 39-bus system illustrate the effectiveness of the proposed methodology.