Development of a Digital Twin-Based Transformer Partial Discharge Diagnostic System

Gas-insulated switchgear (GIS) enables relatively straightforward PD signal detection and fault localization due to its simple internal structure. In contrast, power transformers face significant challenges in PD detection and fault localization because their complex internal components and mixed insulating media distort and attenuate UHF signals. To address these challenges, a digital twin-based transformer PD diagnostic system has been developed to enable accurate signal analysis and fault tracking while considering the complex internal environment of transformers.

This study implements a simulation function for PD signal propagation paths and attenuation in a virtual environment based on transformer design data (internal geometry, numerical parameters, and material properties) derived from 3D modelling software. By automatically extracting existing 3D model data via an Open API, the internal geometry and material characteristics of the transformer are incorporated into the simulation. Furthermore, a 3D-based simulation considering arbitrary placement of PD sources and UHF sensors was developed, enabling integrated implementation of optimal sensor placement, PD simulation and validation, and defect source localization.

A simulation-based optimization algorithm was developed to determine the optimal placement of UHF sensors by considering internal PD signal attenuation, the number of installable sensors, and placement constraints, thereby maximizing sensor detection coverage.

To model signal transmission, a simulation-based framework identifies feasible transmission paths for PD signals and analyzes UHF propagation characteristics influenced by internal structures and materials. Medium-dependent parameters such as propagation velocity and attenuation rate are mapped as weights to construct an M-Matrix that models signal attenuation, delay, and blocking effects. The M-Matrix is further utilized to evaluate post-installation sensor reliability by comparing measured signal amplitudes with simulated reference signals.

PD source localization is performed using the M-Matrix–based simulation framework and a pre-mapped localization database (Find Table). Measured signals are compared with simulated signal characteristics to estimate PD location and localization confidence. Unlike conventional approaches that typically require at least four sensors in 3D space, the proposed method enables accurate localization with only three sensors and delivers a 20% accuracy improvement.

To validate the performance of the digital twin-based transformer PD diagnostic system, simulations and field applications were conducted. Using 3D transformer design data, attenuation between PD sources and UHF sensors was simulated and validated through comparison with measured signals. The system was further applied to an operating transformer to verify UHF sensor reliability and PD location estimation under actual discharge conditions.