Most high voltage failures announce themselves months ahead. Acetylene shows up in transformer oil well before the insulation breaks down. A bushing's dissipation factor drifts upward across two test cycles before it flashes over. A switchgear joint runs hot through several load peaks before it arcs. The job of monitoring is to catch those signals while the fix is still a planned, budgeted work order rather than a forced outage on critical load with a year-long replacement wait behind it.

The principle that separates useful programs from dashboard decoration is simple: every monitored parameter needs a threshold and a documented response. This guide covers how to prioritize assets, which parameters carry real diagnostic weight on 35 kV to 220 kV class transformers and switchgear, how to set alarm levels, and how to phase the rollout without drowning your operations team in noise.

Rank Assets by Consequence, Not by Age

Conclusion first: not every asset deserves continuous instrumentation, and spreading sensors thinly across an entire fleet is the fastest way to waste the budget. Rank by the consequence of failure and fund the top of the list. Three questions sort most fleets quickly:

• What does the asset feed, and is there redundancy behind it?
• How long would a replacement take? Large power transformers are built to order, and lead times commonly run a year or more.
• What does an unplanned outage cost in downtime, collateral damage and safety exposure?

The ranking usually puts a small set of units at the top. A 100 MVA 220 kV low-loss power transformer serving a grid substation is the textbook first candidate: high consequence, long replacement lead time, and mature diagnostic methods. At the mid-voltage tier, 35 kV oil-immersed transformers feeding continuous process loads justify monitoring because a single trip can idle a whole production line. Distribution-class pole-mounted units typically warrant scheduled inspection and oil sampling rather than continuous monitors, since unit costs are lower and spare availability is better.

The Parameters That Carry Real Diagnostic Weight

Oil-filled power transformers

Four parameter groups cover most of the diagnostic value:

• Dissolved gas analysis (DGA). Faults inside the tank, whether overheated conductors, partial discharge or arcing, generate characteristic gases that dissolve in the oil. Hydrogen and acetylene are the sharpest early indicators, and the ratios between gases point to the fault type. Laboratory sampling every 3 to 6 months suits most units; continuous multi-gas monitors earn their cost on the top criticality tier.
• Thermal data. Top-oil and winding hot-spot temperatures read against load. Sustained hot-spot temperatures above design limits accelerate paper aging sharply, which is why loading guides such as IEC 60076-7 treat thermal duty as the primary life-consumption driver.
• Oil quality. Dielectric strength, moisture content and power factor show whether the insulation system is drying out or contaminated.
• Bushings and the tap changer. Capacitance and dissipation factor trends from bushing test taps, plus operation counts and contact condition on the load tap changer, which is consistently one of the most failure-prone subassemblies in industry failure surveys.

Metal-enclosed HV switchgear

Switchgear failures concentrate in connections and insulation surfaces, so the monitoring mix shifts accordingly:

• Partial discharge surveys using transient earth voltage and ultrasonic methods, quarterly at minimum on critical lineups. PD activity usually precedes insulation breakdown by months.
• Infrared thermography of busbar joints and cable terminations, taken under meaningful load. Infrared inspection windows in metal-clad panels let technicians scan energized equipment without opening enclosures, which removes the main safety reason for skipping rounds.
• Breaker records: operation counters, operating times and mechanism condition, because a breaker that fails to clear a fault turns a contained event into a bus-level one.

One procurement note: monitoring provisions are cheapest to specify while the lineup is still on the drawing board. When tendering for a metal-enclosed high voltage switch cabinet, ask the manufacturer for panel layouts that leave room for PD sensors and infrared windows, and for wiring that brings sensor signals to an accessible terminal strip.

Condition-Based Beats Calendar-Based for High Voltage Fleets

Three maintenance strategies are on the table: reactive (run to failure), preventive (fixed calendar or operation intervals) and condition-based (act when measured condition crosses a threshold). For high voltage assets, condition-based monitoring wins, and the reasons are structural rather than fashionable.

Run-to-failure is only rational where failure is cheap and harmless, which describes almost nothing at 35 kV and above. Fixed-interval maintenance has the opposite flaw: it ignores actual condition. A fault that develops the week after a scheduled oil sample keeps growing until the next one, and a healthy unit still gets opened, inspected and resealed because the calendar says so, even though every intrusive intervention carries its own risk of human error.

Condition-based monitoring matches the physics. The dominant failure modes of transformers and switchgear develop progressively and leave measurable signatures: gas generation, PD activity, temperature rise. Frameworks such as the IEC 63097 smart energy roadmap formalize exactly this model, with continuous condition data feeding asset management decisions. The discipline to plan for is procedural rather than technical: a condition-based program without written responses drifts back into reactive maintenance with extra sensors attached. If you start with one technique, start with DGA; our dissolved gas analysis testing guide walks through sampling practice, fault-gas patterns and interpretation.

Set Warning and Action Thresholds Before the First Reading

Decide what each parameter means before data starts arriving, or the program will generate arguments instead of decisions. Two tiers are usually enough: a warning level that triggers increased sampling and investigation, and an action level that triggers planned de-energization or load transfer.

Two habits keep thresholds workable. First, rate of change often matters more than the absolute value: a steadily rising hydrogen trend at a moderate concentration can be more urgent than a stable reading slightly above a generic limit. Second, tailor limits to the asset instead of copying a generic table. For a large power transformer, even a few ppm of acetylene in oil is a genuine finding, because acetylene points to arcing or very localized high temperatures. For top-oil temperature, alarm settings in the 85 to 95 °C range are common practice, adjusted to the unit's design and loading pattern.

Finally, give every alarm an owner and a written response: who gets notified, what gets checked within 24 hours, and what escalates to an outage decision. Programs that send alarms into an unattended shared inbox train everyone to ignore alarms.

A Five-Step Rollout

Build the program as a sequence rather than a big-bang project:

1. Inventory and rank. List every high voltage asset with its voltage class, load served, redundancy and replacement lead time, then score the consequence of failure.
2. Capture baselines. Factory acceptance tests and commissioning measurements, including SFRA signatures, oil screens, dissipation factor and PD levels, are the reference points every later trend depends on. Insist on complete, transferable test records at purchase.
3. Match method to criticality. Continuous online monitoring for the top tier; routine offline testing (DGA, thermography, PD surveys) for the middle; visual inspection and counter reviews for the rest.
4. Define alarms and responses. Two-tier limits per parameter, with documented actions and named owners.
5. Review annually. Retune thresholds after every intervention and after any change in loading pattern, since a transformer moved to cyclic peak duty ages differently than one on steady base load.

One more point belongs in the procurement phase, because retrofitting is where budgets bleed. Monitoring readiness is cheap to specify at purchase and expensive to add later: accessible oil sampling valves, bushing test taps, winding temperature indicators, mounting space and wiring for sensors, and open communication protocols such as IEC 61850 or Modbus so the data reaches your existing SCADA instead of a proprietary silo. Ask bidders to price these as line items. A supplier that builds the transformer, the switchgear and the prefabricated substation as a matched set, as Jiangsu Dingxin Electric does across its 35 kV to 220 kV ranges, can integrate those provisions at the factory rather than leave them to site improvisation.

The economics close the argument. Monitoring hardware and routine testing on a critical transformer or switchgear lineup cost a small fraction of a single forced outage: one emergency repair, one oil processing campaign, or one year-long wait for a replacement power transformer. Start with the consequence ranking, put DGA and thermography on the top tier, wire every alarm to a named response, and expand from the units that have earned the sensors. That is the whole playbook, and it works at any fleet size.