Refinery wastewater’s last mile: why ultrafiltration is beating bed filters

After an API separator (gravity oil–water separation) and a DAF unit (dissolved air flotation), many refineries still need a final “polish” to push oil and suspended solids to very low levels. The choice typically comes down to advanced membranes—most often ultrafiltration (UF)—or conventional media beds.

New field data put a sharper point on the tradeoff. A Tehran refinery study found a 30‑kDa UF module cut oil & grease by 99.2% and TSS (total suspended solids) by 100%, with permeate clean enough for the strictest “special region” discharge standards (pr.ripi.ir) (pr.ripi.ir).

In parallel, a PolyCera UF pilot treating API effluent around 100 mg/L oil and 1000 mg/L solids achieved effluent of <10 mg/L oil and <30 mg/L solids (www.polycera.com), and in practice UF permeate often meets reuse criteria directly (e.g., <10 mg/L oil) (www.polycera.com) (pr.ripi.ir).

Polishing after API and DAF

As a finishing step downstream of the primary train, often including a dissolved air flotation unit, refineries compare membrane performance to granular media. A DAF stage can be implemented with packaged units such as DAF systems to knock down bulk oil and solids before the polish.

Membrane filtration covers MF (microfiltration) and UF (ultrafiltration, a pressure-driven barrier that rejects emulsified oil and fine particles). Plants commonly use ultrafiltration here for tighter effluent control.

Membrane filtration results (UF/MF)

Industrial trials show polymeric UF can remove virtually all remaining oil and solids. In the Tehran case, the 30‑kDa module reached 99.2% oil & grease removal and 100% TSS removal, meeting “special region” discharge (pr.ripi.ir) (pr.ripi.ir).

A PolyCera pilot treating API‑effluent (≈100 mg/L oil, 1000 mg/L solids) delivered <10 mg/L oil and <30 mg/L solids, with the vendor noting UF permeate frequently meets reuse thresholds like <10 mg/L oil, and that it can “reduce the burden of subsequent treatment” (www.polycera.com) (www.polycera.com).

Media filtration benchmarks

Conventional deep‑bed media filters—sand/anthracite or oil‑adsorbing beds—remove much of the residual oil but are less thorough. Benchmark data place media at roughly ~60–85% oil removal; an EPA dual‑media study reported ~68–86% oil removal under optimized backwash (nepis.epa.gov). Effluents often remain 10–30 mg/L oil.

Within this category, specialized adsorptive beds perform better. Siemens reports walnut‑shell filters taking 100 ppm feed oil to <10 ppm in discharge, and PerforMedia media hitting the same <10 ppm even with 1000 ppm feed surges (magazine.watertoday.org). These media also trap fine solids and release them during backwash (magazine.watertoday.org).

Typical dual‑media beds use sand and anthracite; many operators select sand/silica media for 5–10 micron capture and pair it with anthracite for multi‑layer filtration.

Effluent quality comparison

UF polishing readily meets “very low” oil and TSS targets: refinery effluent after UF has been reported at <10 mg/L oil and <30 mg/L TSS—typically well below regulatory limits (www.polycera.com). By contrast, media alone often leaves ≈5–20 mg/L oil and several tens of mg/L TSS unless followed by other steps.

The qualitative takeaway from case studies is consistent: membranes produce superior water quality that “reduces the burden of subsequent treatment” (www.polycera.com), whereas media rely on mechanical capture and, depending on media, adsorption.

Capital and operating costs

CAPEX (capital expenditure) for UF is higher. A UF plant includes high‑cost membranes (tubular or hollow‑fiber), high‑pressure pump skids, and CIP (clean‑in‑place) tanks. Roughly, a UF skid can cost 2–5× more per m³/day capacity than a comparable deep‑bed filter, though design specifics drive variation.

Multimedia filters require only gravity or low‑pressure vessels and media—a much lower‑cost setup per unit flow. OPEX (operating expenditure) is usually lower too: media filters mainly run low‑power backwash pumps and may lose a small amount of media over time.

UF systems incur continuous energy draw and maintenance for pumping, fouling control, and periodic chemical cleanings. In the PolyCera pilot, the system operated without adding treatment chemicals during normal operation, but it required regular backwashing and in‑place chemical cleaning to restore flux (www.polycera.com).

Advanced adsorptive beds (e.g., walnut‑shell, PerforMedia) are backwashed in situ and not consumed—costs center on pumps and backwash water (magazine.watertoday.org). By contrast, GAC (granular activated carbon) filters incur higher chemical cost because the carbon is consumed and must be replaced; walnut/PerforMedia avoid that (magazine.watertoday.org). Where carbon is appropriate, operators may use activated carbon media in separate adsorption trains.

Sludge, maintenance, and footprint

In operation, UF can reduce sludge volume and hazardous waste: its concentrate stream can be routed back to an oil‑skimmer rather than creating a new waste stream (www.polycera.com). Media filters generate particulate backwash that must be handled.

Membrane modules have finite life—often 5–10 years—and need routine CIP with acids or oxidants to manage fouling. Media filters generally need only top‑up or periodic replacement of sand/coal and occasional manual bed cleaning. From a staffing perspective, high‑pressure UF requires more sophisticated oversight than gravity filters.

Footprint and complexity diverge: membrane systems are compact but complex (controls and chemical systems), while media filters are larger for the same flow but simpler. If space is tight or reuse is targeted, UF’s higher treatment efficiency can be worth the complexity; otherwise, media offers a simpler polish.

Compliance and reuse economics

The benefit of better polishing sharpens when discharge limits are strict. One cited benchmark: Indonesian upstream petrochemical standards allow ~15 mg/L oil‑and‑grease (it.scribd.com). A media filter may marginally achieve this, but UF routinely drives levels far lower (often <5 mg/L). If effluent surcharges or compliance fees hinge on very low oil content, the higher CAPEX of membranes can be justified; if limits sit around 10–15 mg/L and only moderate polishing is needed, a low‑cost media filter may suffice.

Water reuse changes the calculus. High‑quality UF permeate can “be reused within the industrial process, obtaining significant savings” (www.berghofmembranes.com). The PolyCera pilot also reused the UF concentrate by returning it to an oil‑skimmer, avoiding a new waste stream (www.polycera.com). No additional chemicals were needed for UF during normal operation, lowering ongoing costs (www.polycera.com). Multifiltration can also cut costs by reducing DAF load, saving coagulant dosing, but still leaves sludge to handle.

Quantification and decision points

Quantitatively, suppose multimedia polishing reduces oil from 15 to 5 mg/L versus UF reducing to 1 mg/L. If effluent surcharges or water reuse revenues are sensitive to these levels, one can compute the net present value of adding UF (see, for example, refinery MBR pretreatment discussions: www.mdpi.com). In the absence of site‑specific flow and cost data, the decision centers on: (a) volumes treated, (b) regulatory stringency or reuse value, and (c) relative capital budgets.

Lifecycle and risk management

Over decades, a membrane system’s higher O&M may be offset by avoided fines or deferring other upgrades. If adding UF allows a plant to meet the tightest “special region” standards (pr.ripi.ir), it may avert multi‑year compliance projects or penalties. Some published cases suggest adding UF/MBR can be net‑positive when factoring in high‑quality discharge benefits (pr.ripi.ir) (www.polycera.com).

Bottom line on polishing choices

Advanced polishing can push refinery effluent to very low oil/solids. UF consistently outperforms media—often producing <10 mg/L oil and essentially zero suspended solids (www.polycera.com) (pr.ripi.ir)—but at higher capital and maintenance cost. Media filters (sand/anthracite or oil‑adsorbing beds) are cheaper and simpler, removing a large fraction of residual oil (down to ~10 mg/L) (nepis.epa.gov) (magazine.watertoday.org) but generally cannot match UF purity. Where sub‑10 mg/L targets or reuse drive value, UF’s superior removal often justifies its costs (www.polycera.com) (pr.ripi.ir); if only moderate polishing is required, a well‑designed multimedia bed can be a cost‑effective alternative (magazine.watertoday.org) (nepis.epa.gov).

Sources: Industrial case studies and reviews (www.polycera.com) (pr.ripi.ir) (magazine.watertoday.org) (nepis.epa.gov) (it.scribd.com); engineering reports and vendor data (www.polycera.com) (magazine.watertoday.org).