Introduction
Water injection is the least glamorous system on most facilities and the one with the longest memory. A separator that underperforms announces itself this week. An injection system that delivers slightly wrong water announces itself in three years — as hydrogen sulphide appearing in a previously sweet field, as barium sulphate scale in the near-wellbore region that no acid will dissolve, or as an injectivity decline that steadily raises pump discharge pressure until the wells will not take the design rate.
What makes the system unusual is that its acceptance criteria are not process numbers at all. They come from the subsurface: dissolved oxygen in parts per billion, sulphate in milligrams per litre, solids by size distribution rather than concentration. The facilities engineer inherits these as a specification and is expected to build a train that meets them continuously, offshore, on a weight and space budget.
This post covers where those numbers come from, what each treatment stage can actually deliver, and the decisions that most often get made too late.
The Specification Comes From the Reservoir, Not the Process
Before sizing anything, three subsurface questions need answers, and all three belong to reservoir engineering rather than facilities:
- Is the formation water high in barium or strontium? If it is, mixing it with sulphate-rich seawater precipitates barium or strontium sulphate. Unlike carbonate scale, sulphate scale is effectively acid-insoluble — prevention is the only economic strategy.
- Is the reservoir souring-prone? Sulphate-reducing bacteria convert sulphate into hydrogen sulphide in the formation. A sweet field injected with untreated seawater can turn sour over years, which changes the facility's entire materials and sour-service basis long after the steel is in place.
- What will the formation actually accept? Pore-throat geometry sets the tolerable particle size, and the classic bridging criteria describe three distinct outcomes rather than one threshold. Particles larger than about one-third of the pore-throat diameter bridge at the rock face and build an external cake — that lifts injection pressure, but it is at least a surface phenomenon. Particles between roughly one-third and one-seventh invade and bridge inside the pore network, and that internal plugging is the damage that cannot be reversed. Below about one-seventh they pass through. The design target is the one-seventh figure, not the one-third.
If these answers do not exist yet, the honest engineering position is to design the train with space and tie-in provision for the stages that may later prove necessary — particularly sulphate removal, which is the one nobody wants to retrofit onto a weight-limited deck.
Oxygen: The First Number, and the Hardest
Seawater at surface conditions is saturated with dissolved oxygen — roughly 8 mg/l (8,000 ppb) at 15 °C, falling to nearer 7.3 mg/l at 20 °C. Injection specifications typically sit at 10 to 20 ppb, which means removing better than 99.7% of what arrives, and holding that around the clock.
The reason is corrosion. Oxygen is far more aggressive to carbon steel than the carbon dioxide most upstream systems are designed around, and at injection pressures the corrosion products themselves become solids that plug the formation. Oxygen also supports the aerobic bacteria that foul filters and pipe walls.
Two mechanical routes dominate:
- Vacuum deaeration — a packed tower held under vacuum by ejectors or liquid-ring pumps, usually in two or three stages. This is the workhorse; both of FEEC's seawater injection packages — the 6,000 BPD SWIM and the 100,000 BPD Gulf of Mexico topside — use it.
- Gas stripping — counter-current contact with fuel or inert gas, which drives the partial pressure of oxygen down further than vacuum alone. Often combined with vacuum in the same tower.
Either route is then polished chemically with an oxygen scavenger, typically sodium or ammonium bisulphite. The stoichiometry is modest: the reaction consumes about 6.5 kg of sodium bisulphite per kg of oxygen, and practice dictates dosing above that to maintain a measurable residual.
A worked example, at SWIM scale. 6,000 BPD is 954 m³/day, or 39.8 m³/h. The deaerator outlet and residual below are illustrative design figures, not project data:
Raw seawater at 8 mg/l O2 8 g/m³ × 39.8 m³/h = 318 g/h oxygen
After vacuum deaeration, 30 ppb 0.03 g/m³ × 39.8 = 1.2 g/h (99.6% removed)
Scavenger polish, 30 to <10 ppb 0.02 g/m³ × 39.8 = 0.8 g/h oxygen
bisulphite, stoichiometric 0.8 × 6.5 = 5.2 g/h
bisulphite to hold 1 mg/l residual = 40 g/h
Two things fall out of this arithmetic. The mechanical stage does virtually all the work — 99.6% of the oxygen — and the chemical stage exists only to catch the last fraction and absorb upsets. And the scavenger consumption is dominated not by the oxygen it reacts with but by the residual being maintained, roughly eight times the stoichiometric demand. Note also that both figures are active sodium bisulphite; commercial product is typically a 35–40% solution, so the dosing pump behind them is passing nearer 100–115 g/h. Sizing the chemical package on stoichiometry alone leaves it undersized twice over.
The trap worth naming: water leaving a vacuum deaerator sits, for practical purposes, at its bubble point. The booster pump taking suction from that vessel has essentially no vapour-pressure margin, and its NPSH available is whatever the liquid level and the suction line give it. This is precisely the bubble-point NPSH case where using a laboratory vapour pressure instead of the vessel's own conditions produces a margin that exists only on paper. Deaerator bottoms pumps are elevation-driven machines, and the deck layout has to reflect that before the structure is fixed.
One caution on borrowing margin rules: the reduced NPSH margin sometimes accepted for hydrocarbons near their bubble point does not transfer to this duty. At 20 to 30 mbara, cold water has a vapour-to-liquid specific volume ratio of the order of 10⁴ to 10⁵, so a trace of flash produces a large vapour volume and the thermodynamic suppression credit is simply not available. A deaerator bottoms pump needs the full margin policy.
Oxygen ingress after treatment deserves equal attention: vent gooseneck design, seal flushes, sample points, and the chemical injection package itself are all routes for atmospheric oxygen to re-enter water that has just been expensively deoxygenated.
Sulphate: The Weight-and-Space Decision
Seawater carries roughly 2,700 to 2,900 mg/l of sulphate. If the reservoir engineers have flagged either scale or souring risk, the mitigation is a sulphate removal unit — nanofiltration membranes that reject sulphate while passing most of the monovalent ions, typically bringing it down to the region of 40 to 100 mg/l.
This is a genuine facilities decision rather than a detail:
- It is heavy and large — pre-filtration, membrane trains, cleaning-in-place systems, and the associated chemicals.
- Membranes are chlorine-sensitive. If hypochlorite is dosed at the lift caisson for biofouling control, it must be removed before the membranes, usually with bisulphite, which couples the biocide and scavenger strategies together.
- It is the stage most likely to be deferred at concept stage and the most expensive to add later.
The alternative to removal is management: scale inhibitor squeeze treatments into the formation, and a souring-monitoring regime. That trade — capital and topsides weight now against intervention cost and souring risk later — is one the facilities team should force into the open early, because it drives layout, weight, and the material selection that follows from whether the field is expected to remain sweet.
Solids: The Damage You Cannot Undo
Filtration is usually specified in two stages: coarse removal at the intake — band screens or strainers handling marine debris — followed by fine filtration, commonly to around 2 µm absolute, sized against the pore-throat criterion above.
Two points are worth making. The first is that concentration alone is the wrong specification. Total suspended solids in mg/l says nothing about whether the particles will bridge at the rock face; particle size distribution is what matters, measured by particle counting, with the membrane filter test read alongside it to capture plugging tendency and solids loading. The second is that formation damage is largely irreversible. Near-wellbore plugging can sometimes be partially recovered by back-flowing or acidising, but the economics of a workover on an injector are rarely attractive, so the filtration train is protecting an asset that cannot easily be repaired.
Corrosion products count as solids too, which is why oxygen control and solids control are the same problem viewed from different ends.
Bacteria: The Slow Failure
Sulphate-reducing bacteria are the mechanism behind reservoir souring and a significant contributor to microbiologically influenced corrosion. Control is a regime rather than a unit operation:
- Continuous low-dose biocide with periodic shock dosing, since bacterial populations adapt to a single continuously-dosed chemistry.
- Chlorination at the intake — commonly electrochlorination at the lift pump caisson — to control marine growth and biofouling before it reaches the filters.
- Chemistry sequencing that acknowledges itself. Hypochlorite upstream, dechlorination before membranes, scavenger after deaeration, scale inhibitor downstream. Each chemical has somewhere it must not be, and packages designed in isolation produce conflicts that only surface during commissioning.
What Water Quality Does to Materials
The water quality specification and the materials selection are the same decision. With dissolved oxygen genuinely held below 10 ppb, deaerated water systems can use carbon steel for much of the downstream train. Upstream of the deaerator — raw, aerated seawater — carbon steel is not viable, and the selection runs to super duplex, 6Mo austenitics, titanium, copper-nickel, or non-metallics such as GRE depending on pressure and duty.
This is why an oxygen excursion is more serious than its duration suggests. The general corrosion rate at 100 ppb is low enough that it will not eat a corrosion allowance in months — the damage is localised. Pitting and under-deposit attack proceed at penetration rates far above the general rate, and the iron oxide they generate travels straight to the formation as the solids the filtration train was installed to remove. The specification is not a target to be approached; it is the premise the rest of the design rests on.
Common Mistakes
- Accepting the water quality spec without its basis. Ask which risk each number addresses. A 20 ppb oxygen limit and a 40 mg/l sulphate limit imply particular subsurface concerns, and knowing them tells you which numbers are hard and which have room.
- Sizing the scavenger package on stoichiometry. As above — residual control dominates consumption by roughly eight times, before the active-to-product conversion is applied.
- Treating the deaerator bottoms pump as an ordinary transfer pump. It is a bubble-point NPSH case, and it drives deck elevation.
- Leaving no space for a sulphate removal unit on a field where souring has not yet been ruled out.
- Designing the chemical packages independently of one another, then discovering during commissioning that the biocide attacks the membranes.
- Specifying solids by concentration alone, with no particle size distribution and no link to pore-throat geometry.
- Ignoring oxygen ingress paths downstream of the deaerator, which can quietly return the water to a condition the materials were never selected for.
Conclusion
Injection water quality is a subsurface specification delivered by a facilities system, and most of its failures come from that handover being incomplete. The oxygen number sets the corrosion regime and the materials; the sulphate number decides whether a membrane plant occupies the deck; the solids number protects an asset that cannot be repaired economically; and the bacterial regime determines whether the field the operator sanctioned as sweet is still sweet in a decade.
None of these are difficult to meet with a well-designed train. They are difficult to meet continuously, which is the actual engineering problem — and the reason the injection system deserves the same design attention as the production train it so often sits behind.
