Finite-Sample Probabilistic Safety Certification for AI-Based Grid-Edge Coordination
On the map: Offline RL and imitation learning from data → Safe offline RL
Abstract (arXiv)
Coordinating large population of flexible grid-edge devices can alleviate the need for time-consuming and capital-intensive network upgrades, and AI-based control methods such as multi-agent reinforcement learning or imitation learning are promising in their real-time decision scalability. However, system operators still need an independent and rigorous way to decide whether a given AI system is safe enough for deployment. This paper develops a finite-sample probabilistic safety certification framework for black-box AI decision models in closed-loop grid operation. The central idea is to reduce the complete input--AI--grid evaluator workflow to a binary unsafe outcome under an operator-defined safety specification, and then use exact binomial inference to certify the corresponding unsafe operation probability. Given a set of held-out calibration scenarios, the framework returns the tightest one-sided upper certificate and an accept/reject deployment criterion that controls the probability of false safety certification. Because the certification is for the calibration distribution that may deviate from the future operation, we further combine the nominal certificate with physically interpretable sample-space adversarial attacks, a concept widely used in AI to investigate the fragility of AI models. Case studies on grid-edge flexibility coordination with 1{,}000-agent AI models (independent parameters) verify the finite-sample safety guarantee and the value of integrating adversarial attacks into a rolling-window training-certification-deployment flow.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Safe offline RL, one of 13 directions in Offline RL and imitation learning from data. Builds on 21 of its 43 references that are in the map. First posted on arXiv 23 Sep 2026.
Builds on
21 of 43 references.
References · 21
- 2014Explaining and Harnessing Adversarial Examplescited by 7,287
- 2015Risk-Constrained Reinforcement Learning with Percentile Risk Criteriacited by 320
- 2021Beta-CROWN: Efficient Bound Propagation with Per-neuron Split Constraints for Complete and Incomplete Neural Network Robustness Verificationcited by 224
- 2020Learning Optimal Power Flow: Worst-Case Guarantees for Neural Networkscited by 28
- 2019Verification of Neural Network Behaviour: Formal Guarantees for Power System Applicationscited by 26
- 2025The 6th International Verification of Neural Networks Competition (VNN-COMP 2025): Summary and Resultscited by 26
- 2024A Review of Safe Reinforcement Learning Methods for Modern Power Systemscited by 11
- 2023Centralised rehearsal of decentralised cooperation: Multi-agent reinforcement learning for the scalable coordination of residential energy flexibilitycited by 5
- 2023GPU-Accelerated Verification of Machine Learning Models for Power Systemscited by 4
- 2026TorchLean: Formalizing Neural Networks in Leancited by 4
- 2017Comprehensive Modeling of Three-Phase Distribution Systems via the Bus Admittance Matrixcited by 3
- 2023Home Electricity Data Generator (HEDGE): An open-access tool for the generation of electric vehicle, residential demand, and PV generation profilescited by 3
9 more
- 2023Self-Supervised Learning for Large-Scale Preventive Security Constrained DC Optimal Power Flowcited by 3
- 2024Scalable Exact Verification of Optimization Proxies for Large-Scale Optimal Power Flowcited by 2
- 2025Distributionally Robust Predictive Runtime Verification under Spatio-Temporal Logic Specificationscited by 2
- 2026Trustworthiness Layer for Foundation Models in Power Systems: Application to N-k Contingency Screeningcited by 2
- 2026Decision-calibrated prediction sets for robust power system operationscited by 2
- 2023Safeguarding Learning-based Control for Smart Energy Systems with Sampling Specificationscited by 1
- 2024Neural Network Certification Informed Power System Transient Stability Preventive Control with Renewable Energycited by 1
- 2026Verification and Validation of Physics-Informed Surrogate Component Models for Dynamic Power-System Simulationcited by 1
- 2026GradMAP: Gradient-Based Multi-Agent Proximal Learning for Grid-Edge Flexibilitycited by 1
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026Black-Box Reliability Certification for AI Agents via Self-Consistency Sampling and Conformal Calibration
- 2024Towards Guaranteed Safe AI: A Framework for Ensuring Robust and Reliable AI Systems
- 2026Position: A Three-Layer Probabilistic Assume-Guarantee Architecture Is Structurally Required for Safe LLM Agent Deployment
- 2020Safety Considerations in Deep Control Policies with Safety Barrier Certificates Under Uncertainty
- 2025Feasibility-Aware Pessimistic Estimation: Toward Long-Horizon Safety in Offline RL
- 2026ROBOGATE: Adaptive Failure Discovery for Safe Robot Policy Deployment via Two-Stage Boundary-Focused Sampling ↗
- 2026When Safety Geometry Collapses: Fine-Tuning Vulnerabilities in Agentic Guard Models
- 2025Evaluating LLM Agent Adherence to Hierarchical Safety Principles: A Lightweight Benchmark for Probing Foundational Controllability Components