An AI-driven hybrid CNN–SVM framework for real-time fault diagnosis in power metering secondary circuits

The reliable operation of power metering secondary circuits is essential for accurate energy measurement and safe power-system management. Conventional diagnosis depends heavily on manual inspection and fixed rules and therefore struggles with nonlinear, transient, and composite faults. This study proposes an artificial intelligence-driven dual-path framework that combines a lightweight one-dimensional convolutional neural network (1D-CNN) with a support vector machine (SVM). Current, voltage, active power, reactive power, phase angle, and harmonic-distortion signals are aligned, denoised, segmented, normalized, and represented by statistical and wavelet-packet features. Unlike serial CNN–SVM schemes, the CNN branch learns local patterns from normalized signal windows in parallel with an SVM branch operating on physically interpretable statistical features. Their calibrated probabilities are integrated by sample-wise confidence-adaptive fusion. On a controlled platform containing 500 independently acquired records in eight operating classes, grouped evaluation yielded 96.5% ± 1.3% accuracy, a 2.1% misjudgment rate, and a 0.42 s mean response time under standard conditions. Comparisons with SVM, decision tree, 1D-CNN, BiGRU, CNN–LSTM, and a lightweight Transformer, together with ablation and perturbation tests, support the contribution of the parallel branches and adaptive fusion. The present evidence establishes a reproducible edge-oriented proof of concept; multi-site field validation remains necessary before operational deployment.

Authors

Institutions

Publication Details

Journal
Discover Applied Sciences
Published
2026-11-16
DOI
https://doi.org/10.1007/s42452-026-09921-2
Primary Topic
Power Systems Fault Detection
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

An AI-driven hybrid CNN–SVM framework for real-time fault diagnosis in power metering secondary circuits

Chao Zhang, Fei Liang, Rui Tian, Yinzhe Xu et al.
4 links
Discover Applied Sciences
Power Systems Fault Detection
article

An AI-driven hybrid CNN–SVM framework for real-time fault diagnosis in power metering secondary circuits

Chao Zhang, Fei Liang, Rui Tian, Yinzhe Xu, Yu Yan, Xu Chen
article en

Abstract

The reliable operation of power metering secondary circuits is essential for accurate energy measurement and safe power-system management. Conventional diagnosis depends heavily on manual inspection and fixed rules and therefore struggles with nonlinear, transient, and composite faults. This study proposes an artificial intelligence-driven dual-path framework that combines a lightweight one-dimensional convolutional neural network (1D-CNN) with a support vector machine (SVM). Current, voltage, active power, reactive power, phase angle, and harmonic-distortion signals are aligned, denoised, segmented, normalized, and represented by statistical and wavelet-packet features. Unlike serial CNN–SVM schemes, the CNN branch learns local patterns from normalized signal windows in parallel with an SVM branch operating on physically interpretable statistical features. Their calibrated probabilities are integrated by sample-wise confidence-adaptive fusion. On a controlled platform containing 500 independently acquired records in eight operating classes, grouped evaluation yielded 96.5% ± 1.3% accuracy, a 2.1% misjudgment rate, and a 0.42 s mean response time under standard conditions. Comparisons with SVM, decision tree, 1D-CNN, BiGRU, CNN–LSTM, and a lightweight Transformer, together with ablation and perturbation tests, support the contribution of the parallel branches and adaptive fusion. The present evidence establishes a reproducible edge-oriented proof of concept; multi-site field validation remains necessary before operational deployment.

Discover Applied Sciences
State Grid Corporation of China (China) (CN)
Industry, innovation and infrastructure
Power Systems Fault Detection
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon