Monitoring performance degradation on a heat recovery steam generator

A heat recovery steam generator takes exhaust energy from a gas turbine and makes steam. Its performance is a heat transfer problem across several pressure levels, and degradation shows up as less steam, at lower conditions, from the same exhaust energy.

Three things degrade. The gas-side surfaces foul, adding resistance to heat transfer and pressure drop. The water side scales or corrodes, adding its own resistance and eventually threatening tube integrity. And the mechanical structure fatigues under thermal cycling, which is a life problem rather than a performance one.

The commercial consequence is combined-cycle efficiency, and it compounds: gas-side pressure drop costs turbine output directly, while lost steam production costs steam turbine output. A degradation that looks modest in the HRSG data can be worth a great deal at the plant boundary.

The primary indicators are pinch and approach temperatures. Pinch is the difference between gas temperature leaving the evaporator and saturation temperature at that pressure. Approach relates feedwater temperature entering the evaporator to the same saturation point. Both widen as heat transfer degrades.

Neither means anything unnormalised. Both depend on gas turbine load, ambient conditions, exhaust flow and steam pressure, all of which vary constantly on a cycling plant. A raw pinch trend is a trend in duty.

The workable approach is to compare like with like: select periods at comparable load, ambient and steam conditions, and trend the corrected values across those windows only. On a cycling plant those windows are short and scattered, which is exactly why manual analysis rarely gets done and why this is worth automating.

Both produce similar output symptoms, and the separation comes from looking at pressure drop and the temperature profile together rather than at either alone.

Gas-side fouling adds resistance to heat transfer and adds gas-side pressure drop. Rising differential pressure across the gas path alongside widening pinch is the characteristic pair.

Water-side scaling adds resistance to heat transfer without touching gas-side pressure drop. Widening approach with a flat gas-side differential pressure points inside the tubes rather than outside them.

The distinction matters because the remedies are entirely different: a gas-side clean is a mechanical exercise, and a water-side problem is a chemistry problem that has already been happening for a while by the time it is visible in performance.

An HRSG on a plant that starts and stops daily is doing something its design case may not have assumed. Thermal transients during startup and shutdown load the thick-walled components, the headers and the tube-to-header connections, and the damage accumulates with cycles rather than with hours.

This is a remaining-life question rather than a performance one, and it needs different instrumentation: metal temperatures at the components that matter, ramp rates, and a cycle count with the depth of each transient rather than a simple total.

It is worth separating explicitly in any monitoring specification, because a performance monitoring system that watches efficiency will not see it at all. The plant can be performing well and consuming life quickly at the same time.

The mitigations are operational as much as mechanical: managed ramp rates, attention to drainage and to warm-keeping between starts, and holding startup within the ramp limits rather than beating them for a faster synchronisation.

Water-side problems are commonly visible in chemistry before they are visible in performance. Conductivity, pH, dissolved oxygen, iron and silica carry the information, and departures precede scaling and corrosion by a long margin.

This makes chemistry the cheapest early indicator available on an HRSG, and it is frequently recorded manually, in a logbook, at a frequency chosen decades ago. Bringing it into the same record as the performance data is one of the higher-return pieces of work available on this asset.

Cycling makes it harder, because a plant that stops daily has repeated periods of layup, and layup practice determines how much oxygen the system sees. A well-run performance monitoring programme that ignores layup chemistry is watching the symptom and missing the cause.

Pinch and approach at each pressure level, normalised, over comparable operating windows. Gas-side differential pressure, normalised for flow. Steam production against exhaust energy. Attemperator flows, which move as heat transfer distribution changes and are frequently the earliest performance signal available. Metal temperatures and ramp rates for the fatigue question. And the chemistry set alongside all of it.

The reference is the difficulty. A commissioning performance test under controlled conditions is what the trend should be measured against, and design data is a poor substitute because it describes an idealised unit rather than the one installed.

Where no commissioning reference exists, the honest position is that absolute degradation cannot be recovered and the plant should establish a reference now, accept that it describes the current condition, and trend from there. That is the same conclusion as on every other asset where the baseline was not captured, and the same lesson: it is cheap at commissioning and impossible afterwards.