Multi-agent defense protects energy forecasts from sensor attacks
Detect, Repair and Audit: A Hierarchical Multi-Agent Defense Against Adversarial Attacks in Energy Forecasting
Cryptography and Security
Summary
Electricity demand forecasts help energy companies plan how much power to generate and distribute. The authors found that these forecasts can be tricked by bad sensor data, causing costly mistakes. They created a three-part defense system with smart agents that detect bad data, fix it quickly, and check for problems with human help. Their system works well to stop attacks and runs fast enough for real-world power control centers.
What this means in practice
- •For energy grid operators: Automatically detect and correct compromised sensor data to maintain accurate power demand forecasts under attack.
- •For industrial control center teams: Audit virtual power plant forecasts using a human-validated language model agent to ensure trustworthy operations.
Authors
Zewei Shi, Ruoxi Sun, Minhui Xue, Feng Liu, Hua Ma, Xingliang Yuan, Surya Nepal
Abstract
Accurate electricity-demand forecasting is essential for the modern energy infrastructure. Recently, Times-Series Large Language Models (TS-LLMs) have emerged as a promising approach to forecasting more accurate demand from historical consumption data in zero- and few-shot settings, helping energy operators effectively plan supply and allocate energy resources. However, these models are vulnerable to compromised sensor inputs that can distort forecasts and lead to erroneous market orders or unnecessary load shedding. Existing defenses focus primarily on end-device forecasting, overlooking the distinct data-drift, latency, and audit requirements across energy infrastructure. To address this gap, we extract from major energy standards and regulatory frameworks a three-layer conceptual model that captures the security postures for each layer. Guided by this model, we develop a Multi-Agent Reinforcement Learning (MARL) detector coupled with signal repair at local control centers, designed to mitigate distributional data drift through complementary feature-based detection and clean-reference-calibrated repair while meeting latency budgets. We further deploy an evidence-grounded LLM auditing agent with human-in-the-loop validation for audit-critical Virtual Power Plants (VPPs). In our evaluation, the best MARL configuration achieves an F1 score above 0.91 across all dataset-attack pairs, while signal repair reduces adversarial-to-clean input Relative Squared Error (RSE) by up to 83%. The few-shot auditor detects 77.78% of attacks in the Sherlock dataset. On the evaluated hardware, detection and repair require 7.95 ms on average, with a worst-observed latency of 14.50 ms, remaining within the local-center latency budget informed by IEEE 1646.