Cascading Event Analysis in Distribution Networks Based on High-Dimensional Factor Models

Cascading event detection is essential for situational awareness and the secure operation of distribution networks. In this paper, based on high-dimensional factor models (HDFMs), an approach is proposed for the decomposition and spatial localization of cascading events. The HDFM divides the raw online monitoring data into factors (spikes, indicating event signals) and residuals (a bulk, indicating noises or normal fluctuations). The estimated number of factors is employed as the indicator to detect the occurrence of subevents. In addition, the autoregressive rate of residuals measures the changes in the temporal correlation of noises to track the system's operating state. Case studies verify the proposed approach.