A Data-Driven approach for oscillatory stability assessment misclassification restriction using Neyman-Pearson umbrella classifiers

Abstract

With the gradual growth of the scale of power systems, the system operating conditions are becoming increasingly complex, and the requirements for oscillatory stability assessment (OSA) are also increasing. Considering that the misclassification of OSA may lead to serious consequences, this article proposes a data-driven approach to restrict the misclassification of OSA for industrial applications. The approach includes a feature extraction procedure and an OSA model. The former is constructed on the basis of Pearson correlation coefficient (PCC) and partial mutual information (PMI), and the latter is constructed on the basis of the Neyman-Pearson umbrella (NPU) algorithm. In the feature extraction procedure, the relevancies between operating variables and the oscillatory stability indicator (OSI) are detected through PCC and PMI, and the pivotal features highly correlated with the OSI are extracted. In the OSA model, the first category error in OSA can be efficiently restricted by regulating the first category error threshold of the NPU classifiers. Additionally, a process for model updating is designed to enhance the applicability to various operating conditions for the approach. The performance of the OSA approach is verified by a series of tests on a 23-bus system and a practical 1648-bus system.