Power System Stability Assessment with Supervised Machine Learning

Power system stability assessment has become an important area of research due to the increased penetration of photovoltaics (PV) in modern power systems. This work explores how supervised machine learning can be used to assess power system stability for the Western Electricity Coordinating Council (WECC) service region as part of the Data-driven Security Assessment for the Multi-Timescale Integrated Dynamics and Scheduling for Solar (MIDAS) project. Data-driven methods offer to improve power flow scheduling through machine learning prediction, enabling better energy resource management and reducing demand on real-time time-domain simulations. Frequency, transient, and small signal stability datasets were created using the 240-bus and reduced 18-bus models of the WECC system. Supervised machine learning was performed to predict the system’s frequency nadir, critical clearing time, and damping ratio, respectively. In addition to varying algorithm hyperparameters, experiments were performed to evaluate model prediction performance through various data entry methods, data allocation methods during model development, and preprocessing techniques.

This work also begins analysis of Electric Reliability Council of Texas (ERCOT) grid behavior during extreme frequency events, and provides suggestions for potential supervised machine learning applications in the future. Timestamped frequency event data is collected every 100 milliseconds from Frequency Disturbance Recorders (FDRs) installed in the ERCOT service territory by the Power Information Technology Laboratory at the University of Tennessee, Knoxville. The data is filtered, and the maximum Rate of Change of Frequency (ROCOF) is calculated using the windowing technique. Trends in data are evaluated, and ROCOF prediction performance is verified against another ROCOF calculation technique.