Physics-Informed Machine Learning for Transformer Condition Monitoring -- Part II: Physics-Informed Neural Networks and Uncertainty Quantification

2025-12-20

AI Summary

Conventional transformer condition monitoring lacks physical interpretability and insufficient uncertainty quantification under small-sample regimes. Method: This paper proposes a Bayesian physics-informed neural network (Bayesian PINN)-driven thermo-aging coupled modeling framework. It integrates Bayesian deep learning with partial differential equation (PDE) constraints governing heat conduction and insulation aging, establishing a spatiotemporal multi-physics model that jointly enforces physical consistency and quantifies epistemic uncertainty. Variational inference enables efficient posterior approximation, yielding interpretable, confidence-aware health assessments from sparse monitoring data. Contribution/Results: Evaluated on real-world transformer datasets, the framework reduces thermal field prediction error by 37% and improves the coverage of 95% confidence intervals for aging trend estimation to 92%. These advances significantly enhance robustness and reliability in early fault detection.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems

Abstract

The integration of physics-based knowledge with machine learning models is increasingly shaping the monitoring, diagnostics, and prognostics of electrical transformers. In this two-part series, the first paper introduced the foundations of Neural Networks (NNs) and their variants for health assessment tasks. This second paper focuses on integrating physics and uncertainty into the learning process. We begin with the fundamentals of Physics-Informed Neural Networks (PINNs), applied to spatiotemporal thermal modeling and solid insulation ageing. Building on this, we present Bayesian PINNs as a principled framework to quantify epistemic uncertainty and deliver robust predictions under sparse data. Finally, we outline emerging research directions that highlight the potential of physics-aware and trustworthy machine learning for critical power assets.

Problem

• Integrating physics with machine learning for transformer monitoring
• Applying PINNs to thermal modeling and insulation aging
• Quantifying uncertainty in predictions using Bayesian PINNs

Innovation

• Physics-Informed Neural Networks model thermal aging
• Bayesian PINNs quantify uncertainty with sparse data
• Integrates physics and machine learning for transformers