Transient Stability Assessment of Power Systems Based on KPCA and Gaussian Process

This paper presents a new method for transient stability assessment of power systems using kernel principal component analysis (KPCA) and Gaussian process (GP). Considering the possible real-time information provided by PMU, a group of system-level classification features are firstly extracted from the power system operation condition to construct the original feature set. Then KPCA is used to reduce the dimension of input space, and GP is employed to build a TSA model. Furthermore, the classification accuracy and generalization performance of the GP model are improved by combining existing single covariance functions to make new composite ones. The proposed method can overcome the disadvantages that many of the current machine learning methods usually suffer from, such as overfitting, difficulty in parameter selection and prediction with no probability interpretation. The effectiveness of the proposed method is validated by the simulation results on the New England 39-bus test system. Key-Words: Transient stability assessment, KPCA, Gaussian process, composite kernel function, machine learning, phasor measurement unit