AI-Powered Robotics for Reliable, Stable, and Robust Operation and Maintenance of Power Systems: Architecture, Unified Modeling, and Quantitative Evaluation
Year: 2026
Type: preprint
Abstract: Robotic inspection and artificial intelligence (AI) are increasingly used in power-system operation and maintenance (O&M), yet they are usually evaluated in isolation—by image-detection accuracy or robot mobility—rather than by their effect on the quantities utilities are regulated on. This paper proposes AIRO-PS, a standards-aligned six-layer architecture that integrates aerial, line-mounted, substation, and live-line robots with edge AI, zoned communications, and a network digital twin, together with a unified modeling framework that maps each robotic or AI capability to reliability, stability, and robustness parameters: pre-failure detection probability, effective failure rate, restoration-time components, dynamic thermal rating, critical-clearing-time margin, and guaranteed control error. The framework is evaluated through eight coupled simulation studies and a techno-economic appraisal. On an 11-kV test feeder, a 6000-year sequential Monte Carlo simulation shows that AI-robotic O&M reduces SAIFI by 67.3%, SAIDI by 83.8%, and expected energy not supplied by 84.1% relative to conventional O&M, and lowers the 90th-percentile annual SAIDI by 77.1%. Robot-sensed dynamic line rating raises mean ampacity by 36.8% under hot-climate conditions, a multilayer-perceptron transient-stability classifier reaches 98.85% accuracy, Bayesian prognostics with monthly robotic data cut remaining-useful-life error by 86%, a Q-learning maintenance policy performs within 1.5% of the average-cost optimum, and sliding-mode stand-off control reduces unmanned-aerial-vehicle tracking error near live conductors by 89%. A 250-feeder deployment yields an internal rate of return of 31.1%.
Source: Research Square
Author: Mohammad Ali
Institution: Atlantic International University
Language: English
Subfield: Safety, Risk, Reliability and Quality
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7214587045
Cites:
Cited by:
Related to: