A Data-Driven Method for Approximation of Dynamic Security Region Boundary Based on Kernel Fuzzy Membership

With the gradually increasing scale and complexity of the power grid, the dynamic characteristics of the power system become more and more complicated, leading to some anomalous data appearing in the operation analysis with the dynamic security region (DSR) method. In recent years, methods like support vector machine (SVM) have been gradually applied in approximating the linear boundary of the DSR due to the high accuracy and strong data mining ability. However, the approximating accuracy of traditional SVM methods is easily disturbed by the noise problem. Therefore, this paper proposes a fuzzy support vector machine (FSVM) based on kernel fuzzy membership, which can ensure the approximating accuracy of the DSR under the disturbance of the noise. Taking the two spatial relationships between samples into account, this paper uses distance fuzzy membership and compactness fuzzy membership to construct a more effective kernel fuzzy membership. Additionally, the accuracy and recall rate are used as the evaluation indicators to judge the approximation effect. The proposed method is tested on the IEEE 9 system, and the results show that the proposed method is feasible and effective with high accuracy.