A research team led by Tunan Chen, Renli Cheng, Jiangnan Li, Mingyu Ou, and Ning Zhou has introduced an alert-ranking framework named AERIS to help power system operators detect extreme intermittency intervals early enough to intervene. Writing in Discover Applied Sciences, the authors address an operational challenge in grids with high shares of renewable power, where control room personnel must determine not merely the magnitude of an upcoming deviation but whether current conditions indicate an imminent extreme event. The researchers formulated this task as a forward-looking screening problem, designing AERIS around a temporal variational risk representation that learns a future stress score alongside an extreme-event label.
To evaluate the framework under realistic grid conditions, the authors tested AERIS on one year of open quarter-hourly operational data provided by Elia, the transmission system operator in Belgium. The screening label used to mark extreme intervals was validated directly against imbalance-price signals published by Elia.
When constrained to a tight two percent alert budget, AERIS achieved the highest precision and recall among all twenty-seven evaluated methods, recording a precision of 44.7 percent and a recall of 69.4 percent. The framework also provided the most accurate ranking of system transitions into the extreme operating regime. The comparative advantage of AERIS proved most pronounced under conditions simulating real-time operational constraints. When all incoming input data arrived with a delay of one interval, AERIS outperformed tuned boosting regressors by a margin ranging from 5.9 percent to 43.8 percent, while also outperforming a baseline persistence rule by 22 percent.
The alert-ranking framework maintained its predictive advantage across extended operating horizons of 45 and 60 minutes ahead, outperforming both the persistence rule and the tuned boosting regressors across these longer lead times. At these multi-step horizons, AERIS retained 38 percent of its initial one-step predictive skill, compared to approximately 30 percent retention for the tuned boosting regressors. Computational tests demonstrated that AERIS requires 0.09 milliseconds to score a single operational interval on a central processing unit, and because the framework relies entirely on routinely published operator data, the authors conclude it can serve directly as an operational screening layer within existing control rooms.
Subject of Research: Forward-looking screening framework for extreme renewable energy intermittency in power systems
Article Title: Future extreme intermittency screening and decision support for high-penetration renewable power systems
Article References: Chen, T., Cheng, R., Li, J., Ou, M., & Zhou, N. (2026). Future extreme intermittency screening and decision support for high-penetration renewable power systems. Discover Applied Sciences. https://doi.org/10.1007/s42452-026-09784-7
Image Credits: AI Generated
DOI: 10.1007/s42452-026-09784-7
Keywords: renewable energy, power systems, grid stability, extreme intermittency, machine learning, decision support, transmission system operators, variational risk representation, Future, extreme, intermittency, screening
News Source: Denise Maddox. (November 11, 2026). AERIS Framework Screens Extreme Intermittency in High-Penetration Renewable Power Systems. Scienmag.



