A Supervised Learning based System Margin Assessment Considering Voltage Violation And Collapse

Abstract

This thesis presents a supervised learning-based regression model to derive operating points at voltage violation and voltage collapse in power systems. While previous studies used formula-based methods to derive voltage stability margins, such approaches can be computationally burdensome, particularly for large-scale systems, requiring considerable computation times for system margin derivation. Moreover, most studies focused on deriving voltage stability margins without considering the operating points where voltage violations occur. To address these issues, this paper proposes a regression model based on convolutional neural networks(CNN) to derive operating points at voltage violation and voltage collapse in power systems. The proposed model takes input data consisting of active power, reactive power, voltage magnitude, and phase angle and predicts the operating points in case of voltage violation and voltage collapse. The performance of the model is analyzed by comparing the results with those obtained from the Continuation Power Flow (CPF) algorithm. The IEEE 33-bus system is selected as the simulation system, assuming that distributed energy resources(DER) are integrated into the test system. A total of 3000 initial operating point data and corresponding operating point data for voltage violation and voltage collapse are generated for simulation purposes. The training data constitutes 70% of the total data, the validation data constitutes 15%, and the remaining data are used for testing the model. The simulation results demonstrate that the proposed model reduced the overall time required to derive voltage violation and collapse points while predicting relatively precise instability operating points.