Enhanced-Online-Random-Forest Model for Static Voltage Stability Assessment Using Wide Area Measurements

Abstract

Application of data mining based methods in online voltage stability assessment has attracted vast attentions in recent years. To account for significant system changes, most of the data mining based methods reconstruct an entire model based on the updated training database. Instead of entirely rebuilding a model in offline mode, this paper presents a novel online learning framework for monitoring the voltage stability of a transmission grid using wide area measurements. A new enhanced online random forest model based on the drift detection and online bagging techniques is proposed. It enables to online update the trees involving tree growth and/or tree replacement. The trees in the forest are then combined via a weighted majority voting, which makes the decision model better adapted to system changes. The framework was first demonstrated on the IEEE 57-bus system, and then applied to a practical power system, the Taiwan power (Taipower) system composed of 1821 buses. In addition to accuracy-based measures, robustness and speed of the proposed framework were also validated. Extensive studies demonstrate that the proposed framework is able to provide reliable and accurate online voltage stability assessment.