Transient Stability Analysis Method of Power System Based on Multi-source Data Drive

Purposes At present, the data driven method represented by deep learning has been widely used in power transient stability analysis. However, the existing transient stability models for researching data driving have some problems, such as limited generalization ability and insufficient model accuracy, when facing small samples, weak samples, and other actual scenarios. In order to improve the expression ability of the model, a refined transient stability assessment method is proposed in this paper according to operation data and fault data. Methods First, four fault information characteristics, namely fault time, fault location, disturbed line, and load level, are constructed according to the transient stability mechanism model of power system. Then, two feature fusion methods, parallel fusion and serial fusion, are proposed to realize the unified expression of operation features and fault features. The influence of multi-source feature fusion on transient stability analysis model is analyzed in depth. Findings The experimental results of the New England system example show that the transient stability analysis method based on multi-source data hybrid drive is conducive to improving the accuracy of the transient stability assessment model, and still has a high accuracy in practical scenarios such as small samples and weak samples.