Power system transient stability assessment based on dimension reduction and cost-sensitive ensemble learning

In this paper, a new method for the transient stability assessment is presented based on dimension reduction and cost sensitive ensemble learning. The online transient stability assessment is constructed as a two-class classification problem. An online application framework is proposed based on offline training and online matching. In the framework, the dynamic simulation results which are the time series of power system variables are utilized as the training data. Considering the high dimensional characteristics of dynamic simulation results and to accelerate the training and prediction process, key feature variables are chosen by improved feature selection method. Furthermore, PCA method is used to reduce the dimension of the training data. The imbalanced sample distribution and unequal misclassification cost are considered. Therefore, the cost sensitive ensemble learning mechanism is put forward to enhance the performance of the transient stability assessment. Several evaluation indexes other than the accuracy are defined to guide the optimization of the cost penalty factor. Finally, simulations are conducted on the IEEE 10 machine 39 bus system. The results show that the proposed method not only reduces training and prediction time but also promotes accuracy.