Related Experiment Videos
Deep learning framework for predicting measurement error drift in smart meter sensors under harsh coastal
Tianfu Huang1,2
1State Grid Fujian Marketing Service Center, Fuzhou, China.
Plos One
|August 5, 2026
Summary
Smart meter sensor reliability in coastal areas is improved by a new framework that predicts failures due to harsh marine conditions. This enhances accuracy and reduces downtime for critical infrastructure.
Area of Science:
- Electrical Engineering
- Materials Science
- Environmental Science
Background:
- Smart meter sensors in coastal environments face accuracy issues from salt fog, humidity, and temperature fluctuations.
- Corrosion and environmental stresses degrade critical components like current transformers and voltage dividers.
Purpose of the Study:
- To develop a robust reliability prediction framework for smart meter sensors in harsh marine atmospheres.
- To improve sensor accuracy and operational lifespan in challenging coastal environments.
Main Methods:
- Integration of an enhanced Transformer with multi-scale attention and a Bidirectional Long Short-Term Memory (BiLSTM) network.
- Utilizing particle swarm optimization for dynamic weight adjustment of model components.
- Introduction of novel Salt Fog Corrosion Index and Electrochemical Activity Factor for coastal-specific effects.
Main Results:
- The framework achieved a coefficient of determination of 0.944 and RMSE of 0.0121% on six years of field data.
- Demonstrated a 6.9% performance improvement over individual models.
- Identified load current-temperature interaction as the dominant drift mechanism (correlation 0.76).
Conclusions:
- The proposed framework accurately predicts sensor failures in harsh marine environments.
- Enables condition-based maintenance, reducing out-of-specification operation time by 85%.
- Offers practical solutions for enhancing sensor reliability beyond smart metering applications.