Digital Twin for Transformer Moisture Management: Closing the Loop from Sensing to Action

DNSP’s pioneering implementation of a full‑cycle Digital Twin (DT) framework for moisture management in power transformers. Moisture remains one of the most critical factors affecting the condition of oil‑paper insulation, influencing dielectric strength, ageing rate, and overall transformer reliability. Conventional offline moisture assessment, typically based on periodic sampling, has long been recognised as insufficient, often missing transient moisture events and providing limited insight into dynamic moisture behaviour. The goal of this paper is to describe an end‑to‑end DT solution that overcomes these limitations by integrating real‑time sensing, advanced analytics, and automated intervention into a unified operational framework.

The DT architecture deployed by the DNSP combines carefully selected instrumentation, a cloud‑based digital platform, and physics‑guided machine learning (PGML) algorithms to create a continuously updated representation of each transformer’s moisture state. The system incorporates dual‑probe moisture and temperature sensors, breather monitoring, and permanently connected dehydration units. Data is streamed through an edge gateway into

Gemini+, the DNSP’s in‑house dual‑probe monitoring system, where it is processed, visualised, and used to drive both diagnostic insights and automated control actions.

A key analytical component of the DT is the Moisture Cloud (MC), a novel visualisation method recently recognised in emerging industry standards. The MC enables intuitive interpretation of moisture–temperature hysteresis and supports accurate estimation of water‑in‑paper content, water activity, and solubility coefficients. PGML techniques are used to determine solubility parameters, infer water‑in‑paper activity, and detect sensor drift or anomalies. These methods ensure robust moisture estimation even when one sensor channel degrades, maintaining continuity of assessment and control.

The DT framework closes the loop from sensing to action by integrating permanently connected molecular‑sieve dehydration units. When elevated moisture levels are detected, the system automatically optimises dry‑out cycles through a machine‑learning‑based control algorithm that balances moisture removal with mechanical integrity considerations. The dehydration units report sorbent saturation, water‑removal rates, and operational status to the DT, enabling predictive maintenance and timely desiccant replacement.

Field deployments demonstrate that the DT enables early detection of moisture ingress events that would not be captured by periodic sampling, supports targeted maintenance of breathers and seals, and delivers sustained moisture reduction through continuous dry‑out. The approach aligns with recent IEEE and IEC standards emphasising real‑time monitoring and provides utilities with a scalable, data‑driven method for improving transformer reliability, extending asset life, and reducing operational risk.

By integrating real‑time monitoring, predictive diagnostics, and automated intervention, the

Digital Twin framework presented in this paper offers a practical and effective model for modern moisture management in distribution networks.