Neyman-Pearson Umbrella Algorithm-Based Static Voltage Stability Assessment With Misclassification Restriction: An Integrated Data-Driven Scheme

Abstract

The online static voltage stability assessment (VSA) has attracted more attention due to the complexity of the operating conditions of modern power systems. Considering that the first category error of VSA misclassification has more serious influence than the second category error, an integrated data-driven scheme for real-time VSA with misclassification restriction is proposed to make it more practical for industrial applications. In the scheme, bagging nearest-neighbor prediction independence test (BNNPT) is employed in the feature selection process, and the Neyman-Pearson umbrella (NPU) algorithm is adopted to construct the VSA model. Benefiting from the ability of BNNPT to explore the correlations between variables, the crucial variables strongly related to the voltage stability index (VSI) are selected from the operation data. For the VSA model, an efficient model updating strategy is designed to achieve fast VSA. Moreover, the VSA model can satisfy different practical VSA demands through adjusting the first category error threshold and the circular splitting mechanism. When the PMU data is sent to the VSA model, the assessment results can be calculated by the model quickly. The tests on a 23-bus test system and a practical 7917-bus system illustrate the satisfactory performance.