Intelligent Diagnosis Method for Transformer Measurement Error Based on Multi-Source Sensor Data Fusion and Causal Path Optimization

Abstract

An intelligent diagnosis and calibration model that integrates multi-source sensing data is constructed to address the measurement errors caused by multi-source interference in transformers. The system integrates multidimensional sensing information such as current, voltage, temperature, and vibration through a weighted feature fusion mechanism, constructs a DAG to represent the causal relationship between key interference variables, and embeds a path scoring and optimization algorithm based on dynamic programming to improve the real-time and accuracy of fault chain identification. The model is deployed at the edge on an ARM architecture embedded platform, with lightweight structure and engineering feasibility. The measured data comes from 110kV and 220kV substations. The experimental results show that the recognition accuracy reaches 96.2%, the average response time is 275ms, and the computational resource utilization rate is 29.6%. It exhibits good robustness and output stability in complex scenarios such as electromagnetic interference, temperature fluctuations, and load disturbances. This model provides a feasible path and deployment basis for achieving high-precision metering and real-time intelligent operation and maintenance in modern power systems.