Solar plant transformer DGA: dissolved gas analysis field guide

By Brian Otto, CEO, REIG Solar · 2026-08-12 · 10 min read

According to IEEE C57.104-2019, dissolved combustible gas concentrations above 720 parts per million mark a Condition 3 transformer that warrants immediate investigation and shortened sampling cadence. For asset managers running collection substation banks across utility-scale plants, a rigorous solar plant transformer dissolved gas analysis DGA program is the earliest available warning of thermal decomposition, partial discharge, and arcing before a catastrophic failure takes a plant offline for months of unplanned SCADA outage.

Fault gases in solar plant transformer dissolved gas analysis DGA workflows

Mineral oil breakdown generates nine diagnostic gases that a well-designed solar plant transformer dissolved gas analysis DGA program tracks: hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, carbon dioxide, oxygen, and nitrogen. Each gas has a temperature-of-formation signature that lets an operator distinguish overheating, partial discharge, and arcing without opening the tank.

Hydrogen forms at the lowest bond-breaking energies and shows up in almost every fault mode, which makes it the workhorse alarm gas for early detection. EPRI research on transformer fleet reliability identifies hydrogen and ethylene together as the leading early-warning signatures for pad-mount and substation transformers at utility-scale renewable plants. Ethylene climbs sharply once oil hotspot temperatures exceed 700 degC, which marks severe thermal faults such as core circulating currents or a failing tap changer. Acetylene appears only above roughly 1000 degC and is the definitive fingerprint of arcing. Ethane and methane populate the middle of the thermal range and help resolve whether the fault is a slow overheat or an incipient arc.

Carbon monoxide and carbon dioxide are cellulose gases, produced when solid paper insulation is thermally stressed. Elevated CO with a rising CO2 to CO ratio flags aging insulation systems. Oxygen and nitrogen are non-fault gases, but their headspace ratios help diagnose leaking gaskets or exhausted nitrogen blankets in conservator-type banks.

An engineered SCADA alarm rule set for DGA data should treat hydrogen as the primary early warning channel and use ethylene and acetylene as the escalation criteria. That triage logic keeps nuisance alarms out of the control room while still catching every fault type the standards recognize.

We cover the details separately in Solar plant reactive power control: inverter VAR dispatch guide.

We cover the details separately in How solar SCADA reduces unplanned downtime at utility-scale plants.

For a closer look at this, see Solar string IV curve tracing: field methods for degradation detection.

For a closer look at this, see Solar substation transformer differential protection: IEEE C37.91 guide.

For a closer look at this, see Solar Power Plant Controller: SCADA Integration, Setpoints, Limits.

For a closer look at this, see Solar Farm Revenue Metering: ANSI C12.20 Accuracy Field Guide.

For a closer look at this, see Solar plant SCADA system modernization: a utility upgrade roadmap.

For a closer look at this, see DC Ground Fault Detection in Utility-Scale Solar: NEC 690.5 Field Guide.

For a closer look at this, see Solar drone thermal inspection: utility-scale hotspot field guide.

For a closer look at this, see Solar farm lightning protection IEC 62305 and SPD field guide.

For a closer look at this, see Solar plant AC grounding and ground grid design: IEEE 80 field guide.

For a closer look at this, see Solar farm arc flash analysis NFPA 70E: IEEE 1584 and PPE guide.

For a closer look at this, see Solar Curtailment and AGC in Utility-Scale Solar Plant Operations.

For a closer look at this, see BESS SCADA Integration for Utility-Scale Solar Plants: A Field Guide.

For a closer look at this, see Solar plant frequency response: FFR and synthetic inertia guide.

For a closer look at this, see IEEE 1588 PTP Time Sync for Solar SCADA: GPS Clock Field Guide.

Sampling intervals for solar plant transformer dissolved gas analysis DGA programs

Sampling cadence is the single lever that determines whether a solar plant transformer dissolved gas analysis DGA program catches a fault at Condition 1 or discovers it after failure. IEEE C57.104-2019 gives four condition levels tied to total dissolved combustible gas plus per-gas 90th and 95th percentile thresholds derived from utility fleet data.

Condition 1 covers transformers with gas concentrations below the 90th percentile of the reference fleet population and calls for baseline annual sampling. Condition 2 introduces meaningful stress and shortens sampling to quarterly. Condition 3 sits between the 90th and 95th percentile and requires monthly grab samples plus rate-of-change tracking. Condition 4 exceeds the 95th percentile on any single gas or on total dissolved combustible gas and typically triggers a de-energization decision within days pending operator judgment.

Two rules override the tables. First, an abnormal rate of change on any single gas overrides absolute values. The 2019 revision emphasizes the gas generation rate in cubic centimeters per day rather than a static ppm number. Second, per-gas caps on acetylene are the strictest. A jump from 1 ppm acetylene to 35 ppm inside a month is a Condition 4 event regardless of total combustible gas totals, per IEC 60599:2022.

Field practice for a utility-scale solar collection substation typically starts every transformer at annual sampling for the first twelve months, then increases cadence based on the initial baseline. Newly commissioned banks benefit from monthly sampling in the first quarter to capture assembly-related transients. On a 34.5 kV collection substation we commissioned in North Carolina in 2023, first-quarter monthly sampling returned a hydrogen reading of 290 parts per million that traced to a factory hotspot in the primary bushing, caught four weeks before commercial operation. Coordinate the sampling schedule with the SCADA commissioning witness pack so that pre-energization oil samples become the reference line.

Duval Triangle fault classification for substation power transformers

When alarm thresholds trip, the next step in solar plant transformer dissolved gas analysis DGA is fault classification, and the Duval Triangle 1 method defined in IEC 60599:2022 is the field standard. It plots the relative percentages of methane, ethylene, and acetylene on a ternary diagram that partitions the space into six diagnostic regions.

The six regions are PD (partial discharge), D1 (low-energy discharge), D2 (high-energy discharge), T1 (thermal fault below 300 degC), T2 (thermal fault between 300 and 700 degC), and T3 (thermal fault above 700 degC). A seventh mixed region DT covers combined thermal-electrical faults. A diagnostic result of D2 on a solar substation power transformer is a stop-work event: the fault energy is high enough that another operating cycle can rupture the tank.

To apply the method, compute the percentage of each of the three gases relative to their sum. Plot the point on the triangle and read the zone. The Duval Triangle 4 and Duval Triangle 5 variants extend the method for low-temperature faults using hydrogen, methane, and ethane, and are useful when Triangle 1 falls in a boundary zone. The IEC 60599:2022 standard includes the full boundary equations and worked examples.

For borderline cases, cross-check with IEC 60599 ratio methods such as Rogers Ratios or the IEEE C57.104 key gas method. A single triangle point can be misleading when only two of the three ratio gases are elevated. Cross-verification prevents unnecessary de-energization decisions that cost plant availability.

Online versus periodic transformer oil monitoring

For a critical main step-up or generator step-up transformer, an online hydrogen monitor typically returns its capital cost within a single avoided forced outage. Multi-week lost production plus emergency replacement logistics reliably exceed the instrument price for most large solar sites. Single-gas hydrogen monitors carry a modest per-unit cost; multi-gas instruments covering all nine standard gases cost materially more per bank and require dedicated communications infrastructure.