Introduction
Produced water is the largest single stream on most mature oil facilities, and on an early production facility it grows faster than anyone plans for. The reservoir starts dry-ish and ends up making far more water than oil; a field at 80% water cut is making four barrels of water for every barrel of oil. All of that water leaves the three-phase separator carrying dispersed oil, dissolved hydrocarbons, solids, and chemicals — and it has to be cleaned to a number, monitored against that number, and disposed of in a way a regulator will accept.
The produced water treatment system is the part of the facility that does this. It is rarely glamorous and almost always the first thing squeezed when topsides weight gets tight, which is exactly why it is the system that ends up constraining production a few years in. When the water plant cannot keep up, you cannot process the oil — the facility chokes back to whatever the water system can clean.
This post covers what the water leaves the separator carrying, the treatment train that cleans it, the discharge specification that governs the design, and the very different demands of overboard discharge versus reinjection.
Why Produced Water Is the Hidden Bottleneck
Oil-in-water (OiW) leaving a three-phase separator is typically a few hundred mg/l of dispersed oil — droplets fine enough that they did not separate by gravity in the residence time available. The separator is sized for the oil and gas; the water leg gets whatever cross-section is left. So the water arrives at the treatment plant already needing real work, and the volume only ever increases over field life.
The trap is designing the water system for first-oil conditions. Water cut climbs, total liquid handling rises, and the chemistry changes — heavier emulsions, more solids, scale, sometimes souring. A water plant sized for day-one rates is a facility-wide rate limit by year three. On a fixed-weight MOPU or EPF deck there is rarely room to retrofit, so the water system has to be sized for the late-life water rate, not the first-oil one. This is the single most common produced-water design error, and it is expensive precisely because it cannot be fixed cheaply later.
The Treatment Train — Bulk to Polishing
Produced water is cleaned in stages, each removing a finer fraction of oil than the last. The physics is Stokes' law: a dispersed oil droplet rises through water at a velocity proportional to the square of its diameter and the density difference, and inversely to viscosity. Big droplets separate easily by gravity; the fine droplets that survive the separator need centrifugal force or flotation to bring them out in a sensible footprint.
| Stage | Duty | Typical OiW out |
|---|---|---|
| Primary / bulk separation (skim tank, CPI, or the separator water leg itself) | Knock out the easy, large droplets and free oil | a few hundred mg/l |
| Deoiling hydrocyclone | Centrifugal separation of finer droplets — ~1000 g in a static conical liner, no moving parts | ~20–40 mg/l |
| Degasser / flash drum | Recover gas broken out across the cyclone; protect downstream | — |
| Flotation (CFU / IGF / DGF) | Gas bubbles attach to fine oil droplets and float them off — the polishing step | <20 mg/l |
| Media / nutshell filter (reinjection only) | Remove residual oil and solids to protect injectivity | single-digit mg/l + low TSS |
The deoiling hydrocyclone is the heart of most offshore water plants: compact, passive, and tolerant of motion, which matters on a floating or jack-up facility. It works on the pressure differential ratio (PDR) across the liner and rejects a small oily underflow (the reject ratio, typically 1–3% of feed) back to the slops or closed-drain system. It needs feed pressure to develop the spin — lose pressure and it stops working, so the upstream hydraulics matter as much as the cyclone itself.
The compact flotation unit (CFU) then polishes the cyclone effluent. Fine gas bubbles attach to the remaining oil droplets and carry them to a froth layer that is skimmed off. CFUs combine cyclonic and flotation action in a short-residence vessel, which is why they have largely displaced large induced-gas-flotation (IGF) tanks on space- and weight-limited decks.
Setting the Discharge Specification
The number the plant is designed to hit comes from the discharge route and the regulator, not from the process engineer's preference:
- Overboard discharge to sea is governed by regional regimes. The OSPAR convention sets a 30 mg/l monthly-average dispersed-oil limit for the North-East Atlantic; the US NPDES regime uses 29 mg/l monthly / 42 mg/l daily-maximum; many other jurisdictions have adopted the 30 mg/l benchmark. The limit is on dispersed oil, measured by a defined method, sampled and reported — so the design target is comfortably below the consent figure to leave margin for upsets.
- Reinjection (produced water reinjection, PWRI, or disposal to a dedicated well) has no overboard limit but a much tougher injectivity spec: tight oil-in-water and total suspended solids, because oil and solids plug the formation face and kill the injection well over time. The water quality required to keep an injector alive is often far cleaner than any overboard limit.
The discharge spec also drives the monitoring. Overboard streams need an OiW analyser and a sampling regime that matches the regulator's defined method; you cannot demonstrate compliance against a method you did not design the sampling for.
Reinjection vs Overboard — Different Problems
These two routes look similar on a block diagram and are completely different to engineer:
- Overboard is about hitting a dispersed-oil number reliably through upsets. The risk is a slug of oil or a chemical that tightens the emulsion breaking through to sea and tripping a consent breach. The design margin is in the flotation polishing and in surge/skim handling.
- Reinjection is about protecting a well for years. The risks are solids and oil plugging the formation (injectivity decline), scale where incompatible waters mix, souring from sulphate-reducing bacteria, and the integrity of the injection string. The design margin is in solids removal, filtration, and water-compatibility chemistry — and the consequence of getting it wrong is a dead injector and nowhere to put the water.
Many facilities do both: reinject when the well will take it, discharge overboard when it will not, with the plant sized for the worse case. The switchover logic and the surge capacity to ride through it are part of the design, not an operations afterthought.
Worked Example — Sizing the Polishing Duty
Scenario: an EPF at late-life is making 50,000 bwpd (≈ 331 m³/h) of produced water. The separator water leg delivers it at ~400 mg/l dispersed OiW. The consent for overboard discharge is 30 mg/l monthly average, so the design target is ≤ 20 mg/l to hold margin.
The cyclone stage: a bank of deoiling hydrocyclone liners takes the 400 mg/l feed down to ~30–40 mg/l. With a reject ratio of 2%, the oily underflow is ≈ 6.6 m³/h routed to the closed-drain/slops system — small, but it has to have somewhere to go, and that recycle loads the upstream separator. Forget to account for the reject and the slops system overflows.
The flotation stage: the CFU polishes 30–40 mg/l down to the ≤ 20 mg/l target. Crucially it is sized for the full 331 m³/h at late-life water cut, not the first-oil rate — the unit installed on day one has to still make the number when the field is mostly water.
The droplet check: by Stokes' law a 30 µm oil droplet (Δρ ≈ 150 kg/m³, water viscosity ≈ 0.7 cP at temperature) rises at only ~0.1 mm/s — far too slow to remove by gravity in any sensible vessel. That is the whole reason the cyclone (centrifugal acceleration) and the CFU (bubble attachment) exist: they remove the fine droplets that gravity never will.
Common Pitfalls
- Sizing for first-oil water rates. The single most expensive produced-water mistake. Water cut climbs over field life; size the plant for the late-life rate or it becomes the facility bottleneck.
- Ignoring the cyclone reject. The oily underflow recycles to slops and reloads the upstream separator. Account for it in the water balance and give it somewhere to go.
- Starving the hydrocyclone of pressure. A deoiling cyclone needs the feed pressure differential to develop the spin. Poor upstream hydraulics or an undersized boost pump and it simply stops separating.
- Chemical carryover that tightens the emulsion. Overdosing demulsifier, or the wrong scale/corrosion inhibitor, can stabilise the very oil droplets the plant is trying to remove. Treat chemistry as part of the water design, not just the oil train.
- Designing overboard treatment but specifying reinjection later. Reinjection needs solids and filtration the overboard plant never had. If reinjection is even possible, design the solids-removal capability in from the start.
- No surge / skim capacity for upsets. A slug of free oil from an upstream upset will blow straight through a plant with no buffer and trip the discharge consent. Build in surge handling.
Conclusion
Produced water treatment is the system that quietly decides how long a facility can keep producing. The oil and gas trains get the attention; the water plant gets the leftover weight — and then constrains the whole facility when the water cut climbs and it cannot keep up.
Clean the water in stages — bulk separation, deoiling hydrocyclones for the fine droplets, flotation to polish — and size every stage for the late-life water rate, not first oil. Set the target below the discharge consent so upsets do not become breaches, and if reinjection is on the table, design the solids removal and filtration in from day one. Get the water system right and the facility produces to its real limit; get it wrong and you choke back the oil to whatever the water plant can clean.
