A Lightweight, Transferable, and Self-Adaptive Framework for Intelligent DC Arc-Fault Detection in Photovoltaic Systems

2026-04-16 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Internet of Things (IoT) & Connected Devices · Depth: Expert, extended

Summary

A new lightweight, transferable, and self-adaptive learning-driven framework (LD-framework) has been developed for intelligent DC arc-fault detection in residential photovoltaic (PV) systems. This framework addresses challenges like spectral interference from inverters, hardware heterogeneity, and long-term operating condition drift. It comprises three components: LD-Spec, a spectrum-based neural network for efficient on-device inference and arc discrimination; LD-Align, which performs cross-hardware representation alignment to ensure robust detection across different inverter platforms; and LD-Adapt, a cloud–edge collaborative self-adaptive mechanism for detecting novel operating regimes and performing controlled model evolution. Extensive hardware experiments with over 53,000 labeled samples demonstrated near-perfect detection, achieving 0.9999 accuracy and a 0% false-trip rate across diverse nuisance-trip-prone conditions. Cross-hardware transfer was achieved with only 0.5%–1% labeled target data, and field adaptation recovered detection precision from 21% to 95% under unseen conditions.

Key takeaway

For MLOps Engineers deploying safety-critical systems like PV arc-fault detectors, consider adopting a lifecycle-aware, adaptive framework. Your systems will benefit from on-device spectral processing for efficiency, cross-hardware alignment to minimize retraining costs, and a cloud-edge self-adaptation mechanism to maintain reliability against long-term operational drift and unseen conditions. This approach ensures sustained high performance and reduces false alarms in dynamic real-world environments.

Key insights

A self-adaptive, learning-driven framework ensures robust DC arc-fault detection in PV systems despite hardware and environmental variability.

Principles

• Spectral features enable robust arc discrimination.
• Minimal target data suffices for cross-hardware transfer.
• Cloud-edge collaboration supports model self-evolution.

Method

The LD-framework uses LD-Spec for on-device spectral detection, LD-Align for cross-hardware feature alignment with mixed-domain fine-tuning, and LD-Adapt for cloud-device coordinated self-adaptation via novelty detection and two-stage model evolution.

In practice

• Use compact CNNs for on-device spectral analysis.
• Employ source replay to prevent catastrophic drift during adaptation.
• Implement canary deployments for safe model updates.

Topics

• DC Arc-Fault Detection
• Photovoltaic Systems
• Spectral Representation Learning
• Cross-Hardware Domain Adaptation
• Self-Adaptive Model Updating

Best for: MLOps Engineer, AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Outage Detection in Self-Healing Smart Grids Using Reinforcement Learning with Spectral Graph Neural Networks — Spectral GNNs in RL enable real-time, scalable outage management in smart grids by capturing global network topology.
• Enabling Predictive Maintenance in District Heating Substations: A Labelled Dataset and Fault Detection Evaluation Framework based on Service Data — A new public dataset and evaluation framework enable reproducible early fault detection in district heating substations.
• SOLAR: Communication-Efficient Model Adaptation via Subspace-Oriented Latent Adapter Reparametrization — SOLAR compresses PEFT adapters by reparameterizing updates using foundation model singular vectors, reducing communication costs.
• Safe Deep Reinforcement Learning for Building Heating Control and Demand-side Flexibility — A DRL framework with an adaptive safety filter optimizes building heating for energy efficiency and grid flexibility.
• Step-adaptive multimodal fusion network with multi-scale cloud feature learning for ultra-short-term solar irradiance forecasting — A multi-source fusion model with adaptive compensation improves ultra-short-term solar irradiance forecasting accuracy.
• FUSE: Frequency-domain Unification and Spectral Energy Alignment for Multi-modal Object Re-Identification — FUSE improves multi-modal ReID by leveraging frequency-domain analysis for spectral disentanglement and energy alignment across modalities.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.