Boiler energy audit: measure the losses, then rank what is recoverable

The direct method — steam energy out divided by fuel energy in — requires two numbers a bagasse or biomass plant does not reliably possess: fuel mass flow and as-fired heating value. Bagasse arrives at ~50 % moisture and LHV 7.2–7.5 MJ/kg, and both figures move with the cane, the mill and the weather; belt scales drift and moisture is sampled, not measured continuously. Divide two uncertain numbers and the "efficiency" that results carries an error band wider than any improvement worth making.

The indirect method of BS 845 / ASME PTC 4 works the other way: measure each loss leaving the boiler, subtract from 100 %. Flue-gas temperature and O₂ fix the dry-gas loss; fuel analysis fixes the moisture losses; ash sampling fixes the unburnt-combustible loss; a radiation allowance covers the casing. Each loss is measured with instruments that can be checked, and — the real point for an operating plant — each loss is individually attackable. The audit output is not one flattering number; it is a map of where the fuel goes. This service sits inside Arrow Energy Co., Ltd.'s boiler and emission-control services and shares its measurement discipline: no basis, no number.

Three rules protect every number the audit produces. First: record feedwater and condensate-return temperature at every load point, every time. Steam-per-tonne-of-fuel comparisons finer than about ±2.5 percentage points are impossible without them — a 10 °C drift in feedwater temperature shifts the heat required per tonne of steam by more than a real economizer's benefit, and it drifts for free with deaerator pressure and return fraction. Second: zero the CO cell before touching the burner. Combustion trim navigates by the CO breakpoint; a cell with 80 ppm of zero offset moves the apparent breakpoint and the "optimised" setting with it. Zero and span first, tune second. Third: one basis throughout. Every figure carries its basis — GCV or NCV, load point, fuel moisture at sampling time — and before/after comparisons are made only within one basis. These are the same habits our ESP and boiler commissioning teams apply to guarantee tests, applied to heat instead of dust.

What savings can a boiler energy audit typically identify?

On units running high excess air and a hot stack, the recurring finds are: excess-air trim to λ ≈ 1.15, worth +0.7–0.8 percentage points for tuning cost; an economizer retrofit, typically +3.1–4.0 pp via the ~20 °C ≈ 1 pp stack-temperature mechanism; air preheating where the economizer approach is exhausted; blowdown heat recovery and condensate return, each typically fractions of a point to ~1 pp depending on current practice. All figures are indicative until measured on the unit.

The recoverable-loss map is walked in payback order. Trim first — it needs instruments and discipline, not steel. Then heat recovery: a boiler economizer converts stack heat to feedwater heat; where feedwater temperature is already high, a boiler air preheater takes the same stack heat into combustion air instead — the audit states which surface fits the temperature profile and the acid-dewpoint margin, since driving a 203 °C stack toward the ~55 °C-above-dewpoint floor is the whole game. Then char: unburnt-in-ash sampling separates grate, air-distribution and fineness causes. Then water-side housekeeping: blowdown rate against conductivity, flash-heat recovery, condensate return fraction.

How does the A/B/A test protocol verify a saving claim?

Three periods on matched conditions: A — baseline without the measure; B — with it; A again — measure removed or off. Loads matched, fuel lot held where possible, feedwater and return temperatures logged, CO cell zeroed before any tuning, one efficiency basis throughout. A real saving appears in B and disappears in the second A. If the second A does not return to baseline, something else changed and the test is void.

The protocol exists because boiler plants are besieged by devices claiming 5–10 % fuel savings on physics that cannot survive a loss inventory. Arrow's position is neutral and empirical: we will adjudicate any vendor's claim with the customer's own data, under the A/B/A protocol, and sign the result either way. The same protocol verifies our own recommendations after execution — the audit that proposed an economizer is re-run on the same basis to bank the measured percentage points.

Typical duration (indicative). Preparation and instrument mobilisation 1–2 weeks; on-load measurement campaign 3–5 days across the plant's real load range; analysis and reporting 2–3 weeks. An A/B/A verification adds its own three test windows, sized to the plant's load stability — commonly 1–2 weeks total.

Deliverables.

• Measured loss inventory per BS 845 / ASME PTC 4, per load point, basis stated (GCV/NCV, fuel moisture at test).
• Instrumentation record: instrument list, calibration status, zero/span log — so the numbers can be defended later.
• Ranked measure list: each measure with its mechanism, its quantified gain in percentage points, its cost class, and its payback arithmetic on stated fuel value and operating hours — not a brochure of options.
• Budget-level specifications for recommended heat-recovery hardware, sized from the measured temperature profile.
• A/B/A protocol document, ready to execute for any future claim, plus the adjudication report where a test was run.
• Re-measurement offer: the same audit basis re-run after execution to verify the banked gain.

Guaranteed figures — for efficiency as for emissions — are stated per project after the technical assessment, on a stated basis. What the audit publishes beforehand is the measured map, and a ranked route through it.