Dimensionality Reduction and Early Event Detection Using Online Synchrophasor Data

This paper proposes a novel approach to utilizing large online synchrophasor data for early event detection in power systems. Based on principal component analysis (PCA), a linear basis of the massive online phasor measurement unit (PMU) data can be extracted to reduce the dimensionality. Using the linear basis with much reduced dimensionality, an early event detection algorithm is proposed. This algorithm is capable of predicting the changes of system operating conditions by comparing the error between PCA-projected and actual values from a few selected locations. Numerical case studies based on both PSS/E simulation and actual PMU data from Electric Reliability Council of Texas are conducted to demonstrate the efficacy of this algorithm.