Weather-Aware Fault Prediction and Budgeted Sensing for Power Distribution: A Hybrid RL and Quantum Optimization Approach
Year: 2025
Type: preprint
Abstract: Power distribution networks face increasing challenges from weather-related outages, demanding intelligent fault prediction and optimal sensor deployment strategies. This paper introduces a novel hybrid classical-quantum framework for weather-aware fault prediction and budgeted sensor placement in distribution networks. We formulate the sensor placement problem as a stochastic budgeted max-coverage optimization and propose three solution approaches: (1) reinforcement learning (RL) formulation as a Markov decision process with rewards based on marginal outage cost reduction, (2) quantum optimization via QAOA and quantum annealing using Ising model mappings, and (3) classical baselines including greedy submodular approximation and Bayesian optimization. For fault prediction, we develop multimodal fusion models combining IoT sensor streams with high-resolution weather forecasts from India Meteorological Department (IMD) data. Our experimental evaluation on a realistic power distribution simulator demonstrates significant improvements: the RL approach achieves 23.4% reduction in expected outage costs compared to random placement, while quantum methods provide competitive solutions for moderatesized subproblems with 15-20% cost reduction. The hybrid framework shows promise for scalable deployment in real power distribution networks, with practical implications for resilient smart grid infrastructure.
Institution: Malla Reddy University
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4414364895
Cites:
Cited by:
Related to: