Outage Identification from Electricity Market Data: Quickest Change Detection Approach
Milad Hoseinpour, Shubhanshu Shekhar, Vladimir Dvorkin · 2026
Read Outage Identification from Electricity Market Data: Quickest Change Detection Approach by Milad Hoseinpour, Shubhanshu Shekhar, Vladimir Dvorkin (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-13605.txt.
Introduction
Milad Hoseinpour, Shubhanshu Shekhar, and Vladimir Dvorkin work on spotting power grid faults fast in 2026. When part of the grid fails, traders face big money risk at once. Quick action to hedge that risk needs quick proof that a fault has in fact hit. The team builds its test on public market data, not on grid sensors.
Their method rests on quickest change detection, a branch of stats for live data streams. It weighs how well new data fits life before a fault versus after one. That ratio helps the test flag a shift as soon as the proof is strong. Fast, firm alerts could cut losses when the grid breaks.
Catalog record
| CREATOR | Milad Hoseinpour, Shubhanshu Shekhar, Vladimir Dvorkin |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 40 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-13605
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Outage Identification from Electricity Market Data: Quickest Change Detection Approach Milad Hoseinpour, Shubhanshu Shekhar, Vladimir Dvorkin Abstract Power system outages expose market participants to significant financial risk unless promptly detected and hedged. We develop an outage identification method from public market signals grounded in the parametric quickest change detection (QCD) theory. Parametric QCD operates on stochastic data streams, distinguishing pre- and post-change regimes using the ratio of their respective probability density functions. To derive the density functions for normal and post-outage market signals, we exploit multi-parametric programming to decompose complex market signals into parametric random variables with a known density. These densities are then used to construct a QCD-based statistic that triggers an alarm as soon as the statistic exceeds an appropriate threshold. Numerical experiments on a stylized PJM testbed demonstrate rapid line outage identification from public streams of electricity demand and price data. I Introduction Power grids are facing increasing threats from extreme weather and aging infrastructure. Slow anomaly detection incurs enormous costs: power outages alone cost the U.S. economy as much as $150 billion annually [ 32 ] . The outages caused by the 2021 Winter Storm Uri led to billions in damages in Texas, with days-long blackouts, and surging prices up to $9,000/MWh [ 11 ] . Early warning signs appeared in public price signals, yet delayed detection and hedging drove major utilities into bankruptcy [ 18 ] . While subsequent reforms mandated faster outage reporting by ERCOT [ 10 ] , market participants (utilities, traders, etc.) require independent detection to act preemptively rather than await official reports. System operators across the U.S. publish information on marginal prices, aggregated generator and load statistics, binding constraints, along with other operational and market data. This has given rise to modern data analytics companies, such as Grid Status [ 13 ] and Electricity Maps [ 8 ] , which provide real-time access to market data streams, leading to the question: How to leverage public operational and market data for independent and fast anomaly detection in power systems? The theory of Quickest Change Detection (QCD) offers the right formalism to answer this question [ 34 ] . QCD involves continuously monitoring a data stream, e.g., electricity prices, to raise an alarm as soon as an abrupt distributional change is detected. A Cumulative Sum ( CuSum ) is a powerful QCD statistic that accumulates evidence over time and compares it against a threshold, providing a theoretically optimal framework for detecting abrupt distributional changes [ 23 ] . These properties have led to the development of numerous variants of CuSum informed by the network topology [ 6 , 26 , 22 , 24 ] . However, topology alone may not fully characterize the cause of distributional shifts. This is particularly relevant to electricity markets governed by optimization engines with specific objectives and constraints. For example, a transmission line outage in power systems triggers an economic re-dispatch optimization, whose solution must respect generation cost structure, along with power flow and generation limits in the post-outage topology. This raises the question: How to integrate optimization objectives and constraints within the design of CuSum statistics for faster outage detection? Contributions : We propose a CuSum -based method for detection and identification of critical events (e.g., line or generator outage) in power systems using publicly available market data streams. Our method is based on the statistical relationship between observable market data (e.g., demand, prices, and aggregated dispatch statistics) and the underlying market-clearing structure, that is leveraged in a CuSum scheme accumulating evidence of structural changes in real-time. We contribute to anomaly detection in power systems by: 1. Developing probabilistic models of real-time market data streams in normal and post-outage states using multi-parametric programming [ 33 , 27 , 28 ] . The models decompose stochastic signals into simpler streams that exhibit distinct characteristics in normal and post-outage states, facilitating outage detection and identification. 2. Designing a CuSum -based statistic that leverages these models to identify structural changes, such as line and generator outages, from streams of market-clearing data. The proposed statistic tests whether the observed stream originates from the normal or post-outage model, accumulating statistical evidence when an outage occurs and remaining quiescent otherwise. 3. Identifying and exploring detection thresholds for the proposed statistic, enabling the control of false alarm rates and detection delays. We apply the method to line outage identification using electricity demand and prices from the stylized PJM testbed. Related Work : Outage detection is central to power system operations. A substantial body of work has focused on outage identification from streams of Phasor Measurement Unit (PMU) data, with early contributions relating observed streams to those induced by models under varying topologies [ 30 , 31 ] . The combinatorial nature of line outage identification subsequently led to mixed-integer formulations, as in [ 9 ] . The computational burden was later alleviated by framing the problem as sparse recovery problem in [ 36 ] and by leveraging related methods from the signal processing domain [ 12 ] . More recently, detection has been offloaded to machine learning models, e.g., based on convolutional neural networks [ 19 ] and decision trees [ 14 ] . Despite strong empirical validation, these methods lack awareness of the fundamental performance metrics in stochastic settings: detection delay and false alarm rate. The two are intrinsic to QCD schemes, which have been applied to line outage detection from PMU measurements in both quasi-steady-state [ 6 ] and transient [ 26 ] settings. Existing QCD schemes, however, rely on PMU measurements available only to system operators, rendering them unsuitable for independent analytics from streams of market-clearing data. The use of public market data for inference was explored in [ 16 , 17 ] using electricity prices for topology identification. In a similar spirit, inverse optimization techniques have been applied to recover market structure, including transmission constraints [ 3 ] , supply curves [ 21 , 20 ] , and electricity demands [ 7 ] . While able to extract structural information, these methods assume static market structure. Hence, they do not provide definitive outage detection, as they fail to distinguish between pre- and post-outage samples in data streams. When applied to outage identification (e.g., inferring that a faulted line has zero transfer capacity), the inertia induced by pre-outage samples leads to significant detection delays. By contrast, the proposed QCD statistic posits nominal and candidate post-outage market structures (e.g., one per line outage) and acts on public market signals to raise an alarm as soon as possible after a structural change occurs. We focus on parametric QCD [ 35 ] , which requires explicit probabilistic densities of observed signals. This is not restrictive for market-clearing inputs such as electricity demand and renewable power generation, which are known to be well-approximated by parametric distributions [ 25 ] . However, the densities of dispatch statistics and electricity prices are not readily available, as market-clearing optimization is a nonlinear transformation of input parameters. To address this challenge, we exploit multi-parametric programming [ 33 , 27 , 28 ] , which decomposes market-clearing optimization into critical regions, each admitting an affine mapping from inputs to market outcomes. The affine structure of critical regions permits closed-form derivation