DGA Monitoring for Transformer Fleets: From Single-Asset Tests to Fleet Condition Intelligence

Interpreting the dissolved gas analysis of one transformer is a well-understood engineering task. Doing it across a fleet of hundreds — with different ages, duty cycles, laboratories, and monitor types — is a different problem entirely. Fleet DGA is less about any single interpretation and more about prioritization and data: getting the right coverage on the right assets, making heterogeneous data comparable, and surfacing the handful of transformers that actually warrant attention.

This matters more now than it did a decade ago. The U.S. Department of Energy reports that the average large power transformer is around 40 years old, with more than 70 percent over 25 years old, and lead times for replacements now run to 36–60 months. An aging fleet facing multi-year replacement times is exactly the fleet you want to watch closely — and can least afford to watch blindly.

The three hard problems of fleet DGA

1. Coverage: which assets earn an online monitor?

You cannot instrument every transformer, and you should not try. Online DGA monitoring is powerful but finite, so it belongs on the assets where an early warning is worth the most: the highest-criticality units, those already showing diagnostic movement, and those hardest or slowest to replace. The rest of the fleet stays on periodic laboratory sampling. IEEE C57.143 frames how monitoring equipment is applied; the fleet-level decision is a risk-based allocation, not a uniform rollout.

2. Data: making heterogeneous samples comparable

Fleet DGA data is messy by nature. Samples arrive from different laboratories and from online monitors that differ in gas coverage, calibration approach, and communications reliability. Dates, units, and asset identities are inconsistent. Operating context — load, cooling, oil processing, maintenance — lives in other systems. Before any fleet ranking is trustworthy, all of that must be normalized into comparable trends. CIGRE TB 783 addresses DGA monitoring systems, and TB 630 frames the intelligent condition-monitoring systems that fleet-scale work depends on.

3. Signal: finding the few assets that matter

The value of a fleet view is not “more data” — it is finding the few assets whose gas behavior deserves a human’s time. That means comparing each asset against its own history and against fleet norms, distinguishing normal seasonal movement from genuine acceleration, and surfacing candidates for repeat sampling, inspection, or escalation. The interpretation methods still apply per asset — the Duval triangle, gas ratios, and key-gas method, which you can run on any sample with the interactive DGA Analyzer — but at fleet scale the first job is triage.

From fleet data to fleet decisions

CIGRE’s international reliability survey (TB 642) puts substation-transformer major-failure rates at roughly 0.53 percent per year — low enough that most of a fleet is healthy at any moment, which is precisely why triage matters: the point of fleet monitoring is to spend engineering attention where the evidence justifies it. A workable fleet DGA operating model looks like this:

• Baseline the whole fleet with periodic sampling and consistent data capture.
• Allocate online monitors to the highest-consequence and most-degraded assets.
• Normalize lab and monitor data into comparable, quality-flagged trends.
• Rank assets by trend movement and criticality, not gas value alone.
• Route the short list into a human-reviewed evidence package for a decision.
• Feed back the outcomes into spare and replacement strategy.

That last step is where fleet DGA earns its keep in today’s market. Knowing which assets are degrading changes where you position scarce spares and how you sequence replacements — the logic behind a large power transformer spare and resilience planner and the AI-data-center transformer bottleneck that makes spare positioning a board-level question. For the full fleet picture, the 2026 Power Transformer Fleet Intelligence Report sets out the aging, lead-time, and demand pressures in one place.

How GridAPM approaches fleet DGA

GridAPM does not replace laboratories, monitors, or engineers. It normalizes fleet DGA into comparable trends, flags data-quality gaps and contradictions, ranks assets by movement and consequence, and drafts source-linked evidence packages for the assets that warrant review — so a qualified team decides what to monitor, retest, inspect, or escalate, with every recommendation traceable to the gas data and standard behind it. It is condition-based maintenance (why it beats a pure calendar) applied at fleet scale, with a human in the loop.

A fleet is only as safe as its weakest, least-watched asset. The goal of DGA monitoring for transformer fleets is to make sure that asset is never a surprise. Request a GridAPM pilot to scope a fleet DGA workflow, or try the DGA Analyzer on a single sample first.