Palm oil mills move eye-watering volumes of water and wastewater — and their boilers only run as clean as the feedwater. Here’s the full, engineer‑level design that turns murky rivers and POME into reliable steam, with the numbers, specs, and sources to match.
Industry: Palm_Oil | Process: Boiler_&_Power_Generation
Start with scale: even a thin layer in a boiler “increases fuel consumption…leading to tube failures,” as palm-mill studies put it (researchgate.net). And “small improvements” in boiler water quality “save a large amount of fuel and reduce CO₂ emissions” (researchgate.net) (researchgate.net).
The volumes are massive. Palm oil mills typically draw 1–1.2 m³ of water per tonne of fresh fruit bunch (FFB) (researchgate.net). Global crude palm oil (CPO) output is ~79 Mt per year, and producing each tonne of CPO yields roughly 2.5 t of palm oil mill effluent (POME) that’s 94–96% water by weight (mdpi.com) (mdpi.com). That math implies ~200 Mt of POME (≈200 million m³) annually (mdpi.com).
Makeup water needs are just as large: producing 1 t of CPO typically requires ~5–7.5 t of water (mdpi.com). In a representative 30 t/h FFB mill (20% oil yield), steam demand is ~18 t/h — roughly 18 m³/h of water (researchgate.net). Robust pretreatment followed by demineralization — “external” plus “high‑grade” polishing — is therefore non‑negotiable in mills that often struggle to run even moderate boiler cycles (<5) without fouling.
Raw water quality and pretreatment
Mill feedwater often comes from rivers, lakes, or wells with high turbidity, organics, and variable minerals. In Johor, Malaysia, estate wells logged turbidity ~14 NTU and aluminium at 0.99 mg/L — above typical drinking-water norms (pmc.ncbi.nlm.nih.gov). Many palm boilers draw virtually no condensate makeup (“often none from sterilizer or live-steam heat”), forcing treatment from raw each shift (thesawit.blogspot.com).
The conventional train starts with coagulation/flocculation and clarification. pH is adjusted (often 6–8 with lime or soda ash), coagulants are dosed, and flocculants grow settleable flocs. Clarifiers typically run 1–4 hours detention (slideshare.net). Palm-mill practice emphasizes removal of turbidity, colloids, iron/manganese, and suspended oils (thesawit.blogspot.com). A properly sized clarifier at this stage unloads the rest of the system.
Jar tests determine the coagulant dose; in Indonesian mill conditions, 20–100 mg/L is common, based on similar cases (pmc.ncbi.nlm.nih.gov) (thesawit.blogspot.com). When footprint is tight, a lamella settler can shrink clarification area while preserving 0.5–4 hour detention guidance.
Post-clarifier, multimedia filtration polishes out carryover and fines. Filters typically run 5–15 m/h and are backwashed until filtrate is <1 NTU; anthracite-sand stacks are common. A dosing pump keeps pH and coagulants on spec through turbidity swings.
Filter media choices matter. Dual media with sand/silica removes 5–10 µm particles, while anthracite extends run length in a multi-layer bed. Many mills add activated carbon to reduce organics, color, and taste/odor before storage.
Disinfection is applied when surface water risks biofouling. UV provides 99.99% pathogen kill without residuals — a fit for ultraviolet systems with low operating cost. If chlorination is used, residual chlorine must be removed to protect membranes and ion exchange resins; keep residual <0.1 mg/L (watertechnologies.com) and neutralize with a dechlorination agent when membranes are downstream.
For tighter pretreatment, ultrafiltration (UF) can precede reverse osmosis (RO). One case logged ~89% turbidity reduction and SDI<3 ahead of RO (researchgate.net), aligning with UF pretreatment used for surface waters. Targets after pretreatment are <1 NTU turbidity, TSS <10 mg/L, and iron/manganese <0.1 mg/L.
Softening and hardness removal
Untreated feed still carries calcium and magnesium that will precipitate as boiler scale. The standard fix is sodium-cycle ion‑exchange softening (replacing Ca²⁺/Mg²⁺ with Na⁺). A softener at this point typically cuts hardness to ≈0.1–0.3 mg/L as CaCO₃ (lenntech.com), with 95–99% removal.
As a rule of thumb, 150 mg/L hardness can be reduced to <5 mg/L (often ~0) after softening. Resin capacities of 15–25 kgr/ft³ set the service run; regeneration with 6–10% NaCl typically consumes ~50–100 kg salt per m³ of resin per cycle and yields ~10% brine waste. A dealkalizer (weak‑acid cation) may be used to reduce transient alkalinity and CO₂. When TDS reduction is needed, engineers add full ion exchange demineralization steps (SAC in H‑form).
Upstream media must be compatible and maintainable. Robust steel filter housings tolerate high backwash pressures in industrial service, guarding resin beds from solids surges.
Demineralization: ion exchange vs. RO
Two main approaches deliver demineralized water after softening: classical ion exchange in acid/caustic cycles, and RO-based trains. In IX, water passes a hydrogen-cycle strong acid cation (SAC) resin, then a strong base anion (SBA) resin — often followed by a mixed‑bed polisher. A two‑bed demineralizer routinely produces <1 µS/cm, and a final mixed‑bed pushes to ~0.1–0.5 µS/cm with silica often <0.1 mg/L; “nearly complete removal of all ions, including carbon dioxide and silica, is required” for high‑pressure service (watertechnologies.com), and designs target silica <0.02 ppm per ASME‑type expectations (watertechnologies.com).
Performance is chemistry‑limited. A well‑regenerated H/Cl system produces ≈0 mg/L residual hardness and typical conductivity <0.5 µS/cm. Weak ions such as CO₂ and silica are partly removed by SBA, so weak-base anion stages or the final mixed‑bed often “mop up” to meet silica specs that can be 2–10 mg/L depending on boiler pressure (lenntech.com). Regeneration uses roughly 0.5–1.0 m³ of 4–10% HCl per m³ H‑cycle run and similar 2–5% NaOH for SBA; the waste volume is small but highly concentrated.
RO strips salts physically, delivering ~90–99% rejection. In boiler duty, UF or fine filtration precedes a first‑pass RO, with optional second‑pass RO or EDI. Typical brackish RO recovers 75–85%; higher (≥90%) is possible with staging and antiscalant dosing. A brackish-water RO treating clarified, softened water yields ~5–30 mg/L TDS (≈10–50 µS/cm). UF+RO on POME effluent has produced boiler-quality permeate at ~21.5 mg/L TDS with pH ~6–7 and COD ~5 mg/L (mdpi.com).
Silica is the swing factor: single-pass RO often leaves several mg/L (≈90% rejection), so a second pass or IX polish is common to meet ASME silica lines that can be ~8 mg/L at ~60–70 bar (lenntech.com). Many hybrids run softening → RO → strong polisher. Membrane selection matters: engineers routinely specify Filmtec RO membranes for stable brackish service.
Economics diverge. Switching from IX to RO in one study increased process flow ~17.5× and blowdown volume ~150%, while making brine 20–30× less concentrated (researchgate.net). IX concentrates chemical use and minimizes wastewater volume; RO uses more electricity and produces larger, lower‑TDS streams. A membrane system often prevails where reducing chemical handling is a priority and raw hardness is modest.
Key metrics and boiler outcomes
Purity expectations are boiler‑driven: hardness ≈0–0.1 mg/L, alkalinity ≈0, and silica as low as <0.5 mg/L for higher pressures (watertechnologies.com) (lenntech.com). Softening plus H/Cl exchange hits these targets; RO plus polishing can as well and has been shown to return COD ~5 mg/L and TDS ~21 mg/L from POME (mdpi.com).
Recovery differs: ion exchange “recovers” nearly all water except small regeneration losses, while RO recovery depends on salinity and staging. With better feed purity, cycles of concentration can be raised (often mills run 3–10), trimming blowdown losses. One calculation for a 27 t/h steam boiler: raising boiler water from 100 ppm feed TDS to a 1900 ppm limit requires ~1.5 t/h blowdown (~5.5–5.6%) — formula Blowdown = Steam Flow/(TDS_max/TDS_feed – 1) (renews.my.id).
Deaeration, conditioning, and blowdown control
After demineralization, dissolved gases must go. A steam‑heated deaerator typically raises feed to ~105°C and vents O₂ and CO₂, achieving <0.02 mg/L O₂ (≈7 ppb) per ASME‑OT practice (watertechnologies.com). Residual oxygen is scavenged chemically (e.g., sodium sulfite or hydrazine) using an oxygen scavenger program.
Feed pH is held around ~7.5–9 and boiler water kept alkaline (pH 9–9.5) with phosphate for any residual hardness. Neutralizing amines minimize carbonic acid in returns; a neutralizing amine helps stabilize pH and reduce corrosion at low dosage.
Internal treatments keep scale and deposits at bay. A scale control program complements phosphate/polymer practices, while alkalinity control prevents swings. Operators monitor pH, conductivity, hardness, and silica daily; conductivity meters and flow indicators must be calibrated to µS/cm accuracy for spec compliance.
Blowdown is set to maintain boiler-water TDS within design — often 1500–3000 ppm for low-hp units and up to 6000–8000 ppm for high‑pressure units. Higher feed purity allows higher cycles and lower continuous blowdown, saving makeup water and heat.
Performance and business outcomes
With this multistage design, mills eliminate most scale/corrosion problems (>90%), cutting forced outages and tube failures. One mill that upgraded to RO+IX cut blowdown by over 50%, saving ~5000 m³ of freshwater annually and reducing outages. Removing 1 mm of scale improves heat transfer by ~2–4%, directly reducing fuel burn; in biomass boilers firing palm fiber/shell, that margin becomes exported power or certificate revenue. The studies cited emphasize that even small quality improvements “save a large amount of fuel and reduce CO₂ emissions” (researchgate.net) (researchgate.net).
There’s a regulatory tailwind: reusing clarified and purified water reduces raw intake and blowdown discharge, aiding compliance with Indonesian Government Reg. No. 82/2001 and 68/2016 (as framed in mill practice). Advanced examples show >80% process water recycling. A 2025 review reported UF/RO routinely returning boiler‑quality water from POME (permeate ~COD 5 mg/L, TDS 21 mg/L) (mdpi.com), aligning with ZLD‑leaning strategies. A set of ancillaries — from tanks to instruments — underpins stable operation.
Design recommendations and sizing
Flow-balance to steam. For a 30 t/h FFB mill (~18 t/h steam), feedwater demand is ~18 m³/h. Size clarifiers and filters at 1.5–2× for redundancy and backwash. Resin volumes should be engineered to hardness/TDS load; for example, 20 kgr/ft³ resin treats ~200 mg/L hardness for ~20 bed volumes per run. A high‑quality ion‑exchange resin selection is key to predictable run lengths.
RO skids are sized to meet daily steam needs (e.g., a 50 m³/h plant yields ~1200 m³/d at typical recoveries). Where chlorine is present, pre‑RO filtration down to a cartridge filter helps protect membranes from floc carryover and filter fines.
Automation reduces variability. Coagulant dosing can be driven by online turbidity; backwash triggers by differential pressure; resin regenerations by conductivity breakthrough or schedule. Chemical prep and dosing benefit from reliable parts and consumables inventories to avoid unplanned outages.
Regeneration needs discipline. Separate softeners plus a mixed‑bed polisher enable ultra‑purity; polisher regeneration might be monthly, though daily is feasible with countercurrent designs. Spent acid/caustic and brines must be neutralized and managed appropriately.
Membrane care is continuous. Include antiscalant for silica/heavy metals, design for moderate recovery (70–80%), and schedule frequent CIPs. A membrane antiscalant limits fouling between cleans, and membrane cleaners help restore performance toward 95%+ of nameplate.
Operation ties it together. Train operators to spot carryover symptoms (foaming, conductivity jumps). Oxygen scavenging (e.g., sulfite ~50 ppm) plus hydrazine/amine conditioning supports corrosion control. Blowdown can be routed to RO reject recycle where practical. A mixed‑bed IX polisher downstream of RO simplifies silica compliance when running higher‑pressure boilers.
Summary of expected outcomes
The complete train — clarification, filtration, softening, demineralization, and deaeration — yields virtually 100% removal of hardness and iron, silica down to <0.5 mg/L, and treated‑water conductivity ~0.1–1 µS/cm (watertechnologies.com) (lenntech.com). Well‑run systems keep silica in steam condensate <0.02 ppm, hardness <0.01 mg/L, and oxygen <7 ppb — bounds beyond which damage‑free operation is expected (scribd.com) (watertechnologies.com).
The payoff: higher boiler cycles, lower blowdown, and fewer tube failures — with quantified examples including 50%+ blowdown cuts (~5000 m³/y freshwater saved) and 2–4% fuel savings from removing 1 mm scale. These are the practical gains behind why palm mills invest in polishing trains and in consumable‑heavy assets like demineralizers and mixed‑beds rather than run at chronic <5 cycles and inflated fuel burn.
Underpinning these outcomes are the same field notes mills have traded for years: palm raw water needs rigorous clarification and filtration (thesawit.blogspot.com), clarifiers hold 1–4 hours (slideshare.net), UF+RO can lift POME‑derived water to boiler quality (mdpi.com), and scale avoidance is the cheapest fuel on site (researchgate.net).
