Applications of Trend-Filtering to Bulk PMU Time-Series Data for Wide-Area Operator Awareness

Abstract

Extraction of dominant trend-component forms the backbone of time-series analysis and forecasting. In power system operation, knowledge of real-time trends in network variables such as bus voltage magnitude, angular separation, line loading is helpful to build situational awareness. In this work, a multivariate trend filtering scheme is presented to detect, quantify and extrapolate real-time trends in measurement data. The multivariate formulation is necessary so as to provide single framework for wide-area visualization of the network's state. Using the method, two control center applications are demonstrated, namely, monitoring of line loading and frequency-data event detection. Performance of the scheme is validated on field data obtained from the wide-area measurement system implemented on the Indian grid. Results indicate application potential in areas such as dynamic security, stability assessment in addition to general network diagnostics.