Generator control action classification based on localized voltage measurements

Voltage disturbances in power systems are caused by various events such as faults or control actions. It is important to accurately identify them for fast remedial response and damage prevention. Recently, data mining techniques have been applied in power system research, usually from the perspective of the entire system. The analyzed data include measurements from transmission buses, substations, and other major components in the network. This big data analysis is useful and enables powerful insight but studies on small, local areas of the larger system needs to be explored further. By focusing on single components such as generators, we can provide concrete and detailed insights to individual operators. For instance, direct impact on the local area can be studied after an event - allowing for better understanding of its propagation effects throughout the entire system. This provides “zoomed-in” situational awareness on a single component and its local area, aiding in event detection and protection efforts. In this paper, we use localized data to classify generator control actions based on voltage dip information measured at generator terminals and at neighboring buses. We perform root cause analysis using the measurement data. By using single-level one dimensional wavelet decomposition and Support Vector Machine methodologies, we have developed an algorithm, Generator Control Action Classification (GCAC) that examines the voltage disturbances and classifies the cause as a certain control action at a particular bus. The method is tested on simulated data from power system simulation software, PowerWorld, and training and testing data is obtained using PowerWorld case studies. The impact of training data availability from different generators while classifying different generator control actions is also studied. Promising results of high accuracy are achieved and discussed.