Voltage sags state estimating in weakly metered power grid with improved Monte Carlo simulation and data-driven LS-SVM modeling

Abstract

With the widespread integration of new energy sources and power electronic equipment, power quality (PQ) issues have become increasingly severe in power system, and PQ data has become a foundation for safe operation of power grid. However, due to economic constraints, a large number of distribution network nodes are not equipped with voltage sag or PQ monitoring devices. To address issues of difficulties in detecting sags at non-monitor nodes, a voltage sags state estimating in weakly metered power grid is proposed with Monte Carlo simulation and data-driven least squares support vector machine (LS-SVM) modeling. First, the framework of the voltage sag estimation system is developed. Then, the spatio-temporal correlation of voltage sags is modeled based on probability distributions, and an improved adaptive Monte Carlo simulation is employed to generate spatio-temporally sag scenarios. Finally, a data-driven LS-SVM model is established for the estimation of voltage sags deep at non-monitor nodes. Simulation results with the IEEE 33-node network demonstrate that the proposed method can effectively estimate voltage sag parameters at non-monitor nodes, achieving an average estimation accuracy of 88.6.%. This can provide effective support for wide-area monitoring and management of voltage sags in complex distribution grids with sparse metering.