Voltage Sag Stochastic Estimation Method Based on Two-Dimensional Trajectory of Multiple Monitoring Parameters for Distributed Generators

In distribution networks, the occurrence of voltage sag is widely recognized as one of the most disconcerting power quality phenomena. The assessment of the different levels of sag at diverse nodes is of considerable significance for the effective mitigation of voltage sag. Amidst the ongoing proliferation of distributed generators (DG), a discernible transformation in the fault attributes of distribution networks is taking place. Traditional voltage sag assessment methods based on stochastic fault simulation face several challenges: in accordance with prevailing grid standards, it is imperative for DG to possess ability against low voltage ride-through (LVRT) events, rendering them no longer equivalent to constant voltage sources during faults, making traditional assessment methods inadequate due to accuracy issues. Moreover, different units have varying LVRT strategy parameters that are difficult to obtain, making it challenging to establish accurate equivalent models of power sources. To address these challenges, this paper proposed a voltage sag stochastic estimation method based on the two-dimensional trajectory of multiple monitoring parameters for DG. Firstly, a two-dimensional trajectory was constructed based on various monitoring data, and a convolutional neural network model was utilized to automatically identify key control parameters such as reactive compensation coefficients during LVRT processes. Based on the parameter identification results, an equivalent model of DG under fault conditions was established. Subsequently, voltage sag stochastic estimation was achieved based on the Monte Carlo method. The proposed methodology's efficacy was substantiated by the outcomes of the simulation. The findings demonstrated that the proposed method accurately identified different unit control strategy parameters and achieved significantly improved accuracy in voltage sag stochastic estimation compared to traditional methods.