Identifying Useful Statistical Indicators of Proximity to Instability in Stochastic Power Systems

TL;DR

This paper develops a semi-analytical method to identify reliable statistical indicators, like autocorrelation and variance, for early detection of instability in stochastic power systems, considering measurement noise and specific system changes.

Contribution

It introduces a fast semi-analytical approach to compute variance and autocorrelation in power systems and clarifies conditions for their effective use as early warning signs.

Findings

Variance increase indicates local stress points.

Autocorrelation growth signals system-wide instability.

Filtering enhances detection reliability under noise.

Abstract

Prior research has shown that autocorrelation and variance in voltage measurements tend to increase as power systems approach instability. This paper seeks to identify the conditions under which these statistical indicators provide reliable early warning of instability in power systems. First, the paper derives and validates a semi-analytical method for quickly calculating the expected variance and autocorrelation of all voltages and currents in an arbitrary power system model. Building on this approach, the paper describes the conditions under which filtering can be used to detect these signs in the presence of measurement noise. Finally, several experiments show which types of measurements are good indicators of proximity to instability for particular types of state changes. For example, increased variance in voltages can reliably indicate the location of increased stress, while growth of autocorrelation in certain line currents is a reliable indicator of system-wide instability.