A three-stage random forest integrating fuzzy matrix method in low frequency oscillation classification of power system

Abstract

In view of the problems that the early warning process of low frequency oscillation in power system is vulnerable to the influence of complex grid environment. The classification speed is slow due to the large amount of processing data in the classification process. A three-stage random forest based on a fuzzy matrix method is proposed in this paper to improve the accuracy and the classification speed of low frequency oscillation early warning in power system. Firstly, the fuzzy matrix comprehensive evaluation is carried out by PMU data, and the evaluation score S will be obtained to determine whether low-frequency oscillation occurs and makes a quick warning. Then, the data is processed by Synchronous Wavelet Transform (SWT), and the damping ratio and attenuation factor of the data are obtained. Furthermore, Random Forest 2(RF 2) and RF 3 are used to judge the type of low frequency oscillation. Finally, simulation results show that the comprehensive fuzzy matrix improves the accuracy of low-frequency oscillation early warning, and the three-stage classification method reduces the amount of data processing and improves the classification speed and stability.