Latency-Aware Deep Learning Benchmark for Real-Time Cyber-Physical Attack and Fault Classification in Inverter-Dominated Power Grids

2026-05-17

AI Summary

This study addresses the challenge of real-time, sub-cycle anomaly detection—including physical faults and cyberattacks—in inverter-dominated power grids by proposing a latency-aware deep learning evaluation framework. Leveraging high-fidelity time-domain signals generated from industrial-grade electromagnetic transient simulations, the work systematically assesses the real-time classification performance of eight neural architectures—from MLPs to Transformers—under streaming inference conditions. For the first time, it establishes a reproducible AI benchmark capable of sub-cycle response, revealing a significant gap between model decision latency and actual end-to-end deployment latency, which typically ranges from 50 to 90 milliseconds (exceeding three fundamental power cycles). Although all models achieve classification responses under 15 milliseconds, the findings underscore a critical disparity between current AI capabilities and the stringent latency requirements of protective relaying, offering vital guidance for AI-driven protection applications.

Abstract

This work introduces a latency-aware benchmarking framework for evaluating deep learning models in power system anomaly detection using high-fidelity, time-domain signals generated from an industry-grade electromagnetic transient simulator. Eight neural network architectures, ranging from MLPs to Transformers, were systematically evaluated on streaming datasets representing both physical faults and cyber-attacks in inverter-dominated networks. All models successfully classified two representative multi-event sequences in real time with sub-cycle response times below 15 ms. However, although classification decisions occurred within one cycle, the end-to-end inference latency consistently exceeded three cycles, ranging from 50 to 90 ms. These results highlight a critical gap between algorithmic capability and protection-grade deployment, pointing to the need for further optimization and hardware acceleration. The findings establish a reproducible benchmark for sub-cycle anomaly detection and provide guidance for transitioning machine learning methods from research prototypes to real-world protection applications.

Problem

• real-time
• cyber-physical attack
• fault classification
• inverter-dominated power grids
• latency-aware

Innovation

• latency-aware benchmarking
• real-time anomaly detection
• inverter-dominated grids
• deep learning for power systems
• sub-cycle classification