What actually stops a high-utilisation gas-engine power plant? The answer is not available from a count of engine failures alone.
This anonymized case uses K4 field operating data from a three-engine distributed power plant. Its daily log records fuel-supply interruptions, heat limitations, weather and grid events, scheduled work, repairs and records that cannot be assigned to a specific cause.
Operating context
| Observation period | 2 November 2025 – 31 August 2026 |
|---|---|
| Plant configuration | 3 × Jenbacher J320 |
| Reporting basis | 303 calendar reporting dates / 302 complete 24-hour intervals |
| Total possible unit-hours | 21,744 h |
| Recorded engine-hours | 20,985 h |
| Electricity generation | 18,893 MWh (18.893 GWh) |
| Average output while running | Approximately 900 kW per engine |
Dataset and methodology
K4 reviewed the daily operating log for each unit. The log spans 303 calendar reporting dates. The utilisation denominator uses 302 complete 24-hour intervals across three units: 21,744 possible unit-hours. Recorded engine-hours are the logged running hours; the difference is 759 lost unit-hours. No additional 72 unit-hours have been reconstructed or inferred.
The resulting 96.5% is engine-hour utilisation calculated from the operating log. It is not contractual availability, technical availability, guaranteed availability, OEM reliability or a general statement of fleet reliability.
Downtime labels are K4’s analysis of the daily log and include external causes. Where the record supported only a partial-day difference, but not a defensible cause, K4 left it unclassified rather than assigning it to equipment failure.
Headline operating results
Across the reporting record, the three units accumulated 20,985 engine-hours and generated 18,893 MWh. Dividing production by running hours gives an average electrical output of approximately 900 kW per running engine.
The operating log shows high utilisation. It does not show that every lost hour was an engine event.
What stopped the plant?
The 759 lost unit-hours were classified as follows. Percentages are shares of lost unit-hours and may not total exactly 100% because of rounding.
Lost unit-hours by log category
759 h total · bar length represents share of lost unit-hours
Interpretation
The largest category is deliberately neutral: 389 hours were unclassified or arose from partial-day records. The log does not support presenting those hours as equipment failure.
Among classified causes, gas supply was largest at 166 hours. Ambient heat, thunderstorms and grid or consumer events together accounted for another 134 hours. Planned maintenance represented 41 hours. Repairs and other causes represented 29 hours.
Plant output depends on an operating system beyond the engines. Fuel infrastructure, cooling and ambient design, weather resilience, grid interface, consumer demand, maintenance planning and data quality all affect realised utilisation.
Normalized fuel benchmark
This is an engineering-normalized reference benchmark, not measured site gas consumption.
The site’s recorded gas-meter totals are excluded because they are inconsistent with the electrical production data. No actual site electrical efficiency is calculated from those readings.
The manufacturer/project performance curve gives 274 Nm³/h at 1,063 kW, 211 Nm³/h at 796 kW and 148 Nm³/h at 529 kW, using a reference gas LHV of 9.5 kWh/Nm³. Interpolation at approximately 900 kW gives approximately 236 Nm³/h.
Applying the normalized figure to 18.893 GWh gives approximately 4.95 million Nm³. This curve-based comparison value is not measured consumption at the plant.
Engineering lessons
- Define the metric. Running hours divided by possible unit-hours is utilisation, not a contractual or OEM reliability measure.
- Separate engine and system causes. Fuel, ambient, weather and grid constraints can stop production without being engine failures.
- Keep unknowns visible. Partial-day and incomplete records should remain neutral until evidence supports a cause.
- Treat maintenance separately. Scheduled intervention and corrective repair answer different operating questions.
- Normalize cautiously. A performance-curve fuel benchmark supports comparison, but cannot replace validated metering.
The operating question
A headline utilisation figure is useful only when its basis is explicit and the lost hours are investigated. This record does not support reducing downtime to “engine failures”. It supports protecting fuel supply, designing for ambient conditions, planning for weather and grid events, scheduling maintenance, investigating repairs and improving log completeness.
Turn operating records into engineering decisions
K4 reviews gas-engine operating data, maintenance evidence and technical risk to identify what limits production and where further evidence is needed.