Machine Learning Approach for Real-time Assessment of Voltage Stability Using Multiple Indicators Derived from Wide Area Synchrophasor Measurements
Voltage instability is one of the major causes of many blackouts such as Canada-United State blackout (2003), Sweden-Denmark blackout (2003), India blackout (2012), and Turkey blackout (2015). If reliable methods are available for online voltage stability assessment, operators can be warned and automated corrective actions can be initiated to prevent voltage collapse. Although, a large number of Voltage Stability Indices (VSIs) are reported in literature, they are not practically applicable for real-time monitoring or not sufficiently reliable under all operating conditions. This research proposal envisages the development of Composite Voltage Stability Indices (CVSIs) combining the strengths of previously proposed Voltage Stability Indices computable from wide area synchrophasor measurements. Advanced machine learning techniques will be applied to derive CVSIs. It is expected that such a CVSI would be more reliable and applicable under wide range of conditions. The machine learning based algorithms for calculating CVSIs would be trained using the data generated through offline simulation and then tested using synchrophasor data generated through real-time simulations performed on RTDS real-time simulator. The proposed research will provide training opportunities for one M.Sc. student, while demonstrating and testing the synchrophasor capabilities of the simulators developed by the partner organization.