The tool, described in a study published in Results in Engineering, is called the Probabilistic Modern Stability Assessment Index, or P-MSAI. It was developed by Ahmed H.A. Adam, Salah Kamel, José Luis Dominguez-Garcia and Khaled Eltag, and it extends an existing high-fidelity voltage stability metric known as the Modern Stability Assessment Index (MSAI) into the realm of probability and risk. The core insight is deceptively simple: every measurement a control room receives — active power, reactive power, voltage magnitude, voltage angle — carries uncertainty, and a stability assessment that ignores that uncertainty can be dangerously misleading.

Voltage stability refers to a power system’s ability to maintain acceptable voltage levels across all its buses under both normal and stressed conditions. When a transmission corridor approaches voltage collapse, the consequences can be catastrophic. Historical cascading failures, including the 1987 Tokyo blackout and more recent collapses in Bangladesh and Pakistan, demonstrate that voltage instability is often a precursor to widespread service interruptions and severe economic losses. For decades, engineers have relied on line-based stability indices — scalar values computed from power-flow solutions — to gauge how close each transmission line is to its collapse threshold, conventionally set at unity.

The problem, the researchers argue, is that nearly all of these indices are deterministic point estimators. Classical formulations such as the Lmn index, the Fast Voltage Stability Index (FVSI), the Line Stability Factor (LQP), and their successors rely on simplifying assumptions: some neglect line resistance, others ignore shunt admittance, voltage-angle separation, or the direction of power flow. Even the MSAI, which uses the full ABCD-parameter model of a transmission line and accounts for series impedance, shunt admittance, angle differences and both active and reactive power flows, produces only a single number from a single operating point. Two lines with identical deterministic index values could carry radically different levels of risk if one operates amid noisy measurements and volatile loads while the other sits in a calm, well-instrumented corner of the grid.

To close this gap, the team turned to the First-Order Second-Moment (FOSM) method, a classical uncertainty-propagation technique from reliability engineering. The approach treats the four key operating variables — receiving-end active power, receiving-end reactive power, sending-end voltage magnitude and voltage-angle difference — as random variables with specified means and variances. A first-order Taylor expansion of the MSAI function around the mean operating point, combined with a rigorously derived closed-form gradient vector, allows the variance of the stability index to be computed analytically. Crucially, this requires only one deterministic power-flow solution per operating state, rather than the thousands of repeated simulations demanded by Monte Carlo methods — a difference the authors estimate at three to four orders of magnitude in computational cost, making the method viable for real-time contingency screening.

From the propagated mean and variance, the framework generates a suite of operationally meaningful risk metrics. The expected value of the index represents the nominal stability level. The variance quantifies sensitivity to uncertainty in power injections and measurements. The collapse probability — computed under a Gaussian approximation of the output — estimates the likelihood that the index exceeds its critical threshold of unity. And a 95 percent confidence upper bound provides a conservative security indicator: mathematically, requiring this bound to remain below unity is exactly equivalent to keeping the collapse probability at or below 5 percent, a threshold operators can adjust to match their risk appetite. The framework also defines a risk-constrained loading margin, the maximum load increase a bus can tolerate before its collapse probability breaches the prescribed limit.

The researchers validated the method on the IEEE 30-bus, 57-bus and 118-bus benchmark systems under base-case conditions, heavy active-power loading, heavy reactive-power loading, combined apparent-power stress, and N-1 and N-2 contingency scenarios. Under nominal loading, all indices agreed that the systems were secure, with collapse probabilities effectively zero. But under stress, the probabilistic layer revealed what deterministic numbers concealed. In the 30-bus system under heavy reactive loading, for example, a line with a deterministic MSAI of only 0.800 showed a mean probabilistic index of 0.955 and a collapse probability of 3.67 percent — an early warning invisible to any point estimator. Conversely, some lines whose deterministic values marginally exceeded unity exhibited collapse probabilities well below the 5 percent threshold, indicating that the violation was marginal once realistic uncertainties were accounted for.

The contingency-ranking results are perhaps the most striking. Traditional ranking sorts outages by the maximum post-contingency deterministic index, but the P-MSAI ranks them by maximum collapse probability. In the 118-bus system under double-line outages, the simultaneous loss of lines 96–97 and 75–74 emerged as the most severe event, driving line 49–69 to a deterministic MSAI of 0.9354 and a 95 percent confidence bound of 0.9508. In several cases, the probabilistic ranking reversed the ordering produced by deterministic screening: an outage that looked moderate on average could carry amplified uncertainty sensitivity near the stability boundary, while a nominally worse-looking outage might pose less actual risk. Such reversals, the authors note, mean that deterministic methods can systematically underestimate the contingencies that matter most.

The method has honest limitations, which the authors characterize explicitly. Because FOSM relies on local linearization, its accuracy is highest under moderate uncertainty and smooth operating conditions; near collapse, where the index response becomes strongly nonlinear, Monte Carlo validation or higher-order methods may be needed. The Gaussian output approximation, the diagonal covariance assumed in the benchmark studies, and the absence of explicit renewable-penetration scenarios in the case studies all leave room for future refinement. The framework can accommodate full covariance matrices derived from historical SCADA records, phasor measurement unit data and forecast-error statistics, and renewable variability can be represented through uncertain power injections at the steady-state level.

What the P-MSAI ultimately offers is a shift in paradigm: from asking whether a line looks stable to asking how confident we can be that it is stable. As decarbonization pushes grids to operate with thinner margins and noisier inputs, tools that quantify risk rather than merely proximity to collapse could become indispensable to the control rooms keeping the lights on. The study suggests that the era of the single-number stability verdict is drawing to a close, replaced by a richer, uncertainty-aware picture of exactly where the next blackout might begin.

Subject of Research: Probabilistic voltage stability assessment and contingency ranking in power transmission systems using the first-order second-moment method

Article Title: Probabilistic modern stability assessment index based on the first-order second-moment method for voltage stability assessment and contingency ranking

Article References: Adam, A. H., Kamel, S., Dominguez-Garcia, J. L., & Eltag, K. (2026). Probabilistic modern stability assessment index based on the first-order second-moment method for voltage stability assessment and contingency ranking. Results in Engineering, 32, Article 113171. https://doi.org/10.1016/j.rineng.2026.113171

Image Credits: AI Generated

DOI: 10.1016/j.rineng.2026.113171

Keywords: voltage stability, power grids, FOSM method, uncertainty quantification, contingency ranking, renewable energy, voltage collapse, transmission lines, risk assessment, power flow, MSAI, smart grid