Webinar Recap: Online DGA Monitoring with Monitor Watch — January 2025

On January 15, 2025, Delta-X Research hosted a webinar focused on online dissolved gas analysis monitoring with Monitor Watch. The session drew utility engineers, asset managers, and testing professionals from across North America, and the questions submitted before and during the session reflected where organisations currently stand in their thinking about continuous transformer monitoring.

This post covers the core technical content from the webinar and addresses the questions that came up most consistently.

The Case for Continuous Monitoring

The webinar opened with the fundamental reason continuous gas-in-oil monitoring adds diagnostic value over periodic laboratory sampling: a question that needs a direct technical answer rather than a general appeal to "more data."

The answer is lead time. Periodic sampling programmes, typically quarterly for high-priority transformers, create observation windows that are 90 days wide. Many serious transformer fault mechanisms develop on shorter timescales than this. Thermal faults driven by contact deterioration or insulation breakdown can initiate and accelerate to advanced stages within weeks. Electrical fault development, partial discharge progressing to arcing, can occur in days. A transformer that receives a clean DGA report in January may be approaching failure in February.

For a transmission transformer with no nearby backup, a critical station role, or an 18–24 month replacement lead time, the 90-day blind window in a quarterly sampling programme represents a risk that is disproportionate to the cost of continuous monitoring. CIGRE TB 630 [1], produced by Working Group A2.44 specifically on online monitoring for transformers, frames this lead-time argument as the primary justification for online monitoring deployment on high-consequence assets.

Online DGA monitors, permanently installed sensors measuring dissolved gas concentrations continuously or at intervals of hours, close this window. The question is not whether online monitoring provides more timely data; it does. The question is whether that additional data can be analysed reliably enough to be useful.

How Monitor Watch Applies R-DGA to Online Data

The core analytical challenge of online monitoring is signal noise. Laboratory DGA extractions follow a controlled procedure that isolates the dissolved gas measurement from short-term oil disturbances. Online sensors measuring continuously do not have this luxury. Their readings reflect temperature cycling, load variation, dissolved air fluctuation, oil flow patterns, and sensor drift, in addition to any gas actually generated by fault activity.

Dukarm [2] established that analytical uncertainty in laboratory DGA is already substantial: coefficients of variation in the 10–20% range for most gases under normal laboratory conditions. Online sensor data can exhibit far larger short-term variation. Applied directly to R-DGA calculations without conditioning, raw online sensor data generates spurious movements in CSEV (Cumulative Severity) and HF (Hazard Factor) metrics [3] that would undermine confidence in the monitoring programme.

Monitor Watch addresses this with signal processing calibrated for online DGA sensor data characteristics before the R-DGA calculations are applied. The conditioned data stream is then processed using the same CSEV and HF framework that Transformer Oil Analyst™ (TOA) applies to laboratory samples [3]. The result is a single, consistent risk picture for each transformer, regardless of whether its DGA record is built from laboratory samples, online monitor readings, or a combination of both.

This analytical consistency is important for utilities managing mixed fleets: some critical units monitored continuously, others sampled periodically. Without it, two separate and potentially inconsistent risk assessments, one from online data and one from lab data, must be reconciled manually for every transformer where both data types exist. Monitor Watch eliminates that problem by applying one method to both.

Q&A: Common Questions from Attendees

The following questions recurred consistently across attendees. They are worth addressing directly.

How does Monitor Watch handle sensors from different manufacturers?

Monitor Watch is designed to accept data from a range of commercially deployed gas-in-oil sensors, including multi-gas monitors from the major instrument suppliers in the market. Data ingestion, unit conversion, and normalisation are handled within the platform. The R-DGA analytical output is consistent regardless of sensor hardware, which allows utilities to make hardware decisions based on field performance requirements without analytical workflow implications.

What happens when sensor data and periodic laboratory samples disagree?

Discrepancies between online sensor readings and concurrent laboratory extractions occur and usually have identifiable causes: sampling location relative to sensor position, oil circulation patterns, sensor calibration drift, or sample handling. The system flags significant divergences for review. Persistent divergence is worth investigating. It typically resolves to a sensor calibration issue or a sampling technique question, and the investigation itself often produces useful information about the transformer's oil circulation behaviour.

Is the alert logic threshold-based, the same as IEEE C57.104?

Monitor Watch alert logic goes beyond fixed concentration thresholds. While threshold-based alerts consistent with IEEE C57.104-2019 [4] and IEC 60599 [5] are available and configurable, the system's most valuable alert capability is rate-of-change based: flagging when gas generation appears to be accelerating, even before concentrations reach standard alert levels. This rate-of-change sensitivity is the direct analogue of HF's function in TOA [3] and represents the primary detection advantage of continuous monitoring over periodic sampling. A transformer whose gas concentrations are below all thresholds but rising at 15% per week is a different risk than one with the same current concentrations that have been stable for three years.

What does implementation typically look like for a utility new to online monitoring?

Most utilities begin by deploying online monitoring on a small number of high-consequence transformers. These are typically units identified through fleet screening as combining elevated CSEV or HF with high criticality scores: no backup, long replacement lead time, or critical load. This approach allows the asset management team to build familiarity with the data management workflow and Monitor Watch alert handling before scaling deployment. Delta-X Research provides implementation support through the integration, data connection, and commissioning phases.

How does continuous monitoring data interact with existing outage scheduling?

This is one of the most practically important questions. The value of early fault detection is only realised if the detection triggers an appropriate maintenance response within the available outage windows. Monitor Watch rate-of-change trends provide lead time that can be counted in weeks rather than quarters, sufficient in many cases to plan a targeted inspection or maintenance intervention at the next scheduled outage opportunity rather than requiring an emergency response. The integration of monitoring outputs with outage planning processes is a workflow question that the Delta-X Research team is glad to work through with utilities at any stage of programme development.

Accessing the Webinar Recording

If you were unable to attend the January 15 webinar or would like to review the content, contact us to request access to the recording and slide deck.

For detailed information on Monitor Watch capabilities, visit the Monitor Watch page. For the technical foundations of R-DGA methodology underlying the analytical framework, visit the Science page.