Real-time Coherency Identification using a Window-Size-Based Recursive Typicality Data Analysis

TL;DR

This paper introduces a recursive, data-driven method for real-time coherency detection in power systems that determines the minimal observation window length based on typicality variance, without prior group assumptions.

Contribution

It proposes a novel recursive TDA-based approach that adaptively finds the minimal window size for coherency detection using only measurements, applicable to dynamic power system analysis.

Findings

Effective in the Kundur test system

Minimal window length depends on inter-area mode

Preserves measurement-only approach

Abstract

This work presents a data-driven analysis of minimal length necessary for coherency detection considering a recursive form of the typicality-based Data analysis (TDA). It proposes a methodology that encloses the observation of the variance of the typicality ({\tau} ) to asses the minimal window length necessary to determine the coherent buses, where the properties of the TDA approach and the groups of buses are iteratively calculated at every new data point sampled. Once the variance of each group reaches a certain value, the minimal window length is determined. Besides, this method preserves the TDA characteristics of using exclusively measurements, not requiring pre-determination of number of groups, group centers or cut-off constants. The method is applied to the well know 2-area Kundur test system, allowing to corroborate its effectiveness and draw conclusions regarding minimal window length dependence on the slowest inter-area mode.