Inside the high‑purity water play powering Indonesia’s CCGTs

In combined‑cycle gas turbine plants (CCGTs; gas turbine plus steam cycle), water looks deceptively simple. The machines are relatively water‑efficient at ~1–2 L/kWh (thundersaidenergy.com), but the [steam side is unforgiving](https://beta.co.id/en/blog/the-ppb-problem-why-steam-purity-makes-or-breaks-turbines-in-ammonia-plants): [any impurity in makeup water](https://beta.co.id/en/blog/ccgts-ppb-problem-why-ultrapure-steam-makes-or-breaks-turbine-output) can trigger scale and corrosion across high‑pressure boilers and turbogenerators.

That purity mandate collides with scarcity. In Indonesia, agriculture consumes >70% of available supply (researchgate.net), pushing utilities to prioritize reuse and efficient treatment. A case in point: the Grati CCGT in East Java retrofitted a brackish‑water RO on its WWTP effluent, cutting effluent conductivity from ~500 μS/cm to <20 μS/cm, reusing 50% of wastewater (~1,600 m³/month) and saving IDR 320 million/year with a 5‑year payback (papers.iafor.org; papers.iafor.org). A move like that aligns with boiler‑feed targets typically set at <0.1 μS/cm conductivity and <20 ppb silica (ebrary.net; ebrary.net).

The following design guide is for water treatment specialists. It draws directly on industry sources to outline pretreatment, core demineralization, and final polishing with the parameters and trade‑offs CCGTs need.

Makeup purity targets and local constraints

CCGT makeup systems are engineered to deliver ultra‑pure water to the HRSG (heat recovery steam generator; the steam generator that recovers heat from gas turbine exhaust). OEM guidelines typically require <0.1 μS/cm conductivity and <10–20 ppb silica, with sodium often <5 ppb (ebrary.net; ebrary.net). In water‑stressed environments such as Indonesia (agriculture >70% of supply, researchgate.net), reuse can materially trim raw water intake — as shown at Grati’s brackish‑water RO retrofit (papers.iafor.org; see also papers.iafor.org). Linking such retrofits to brackish service is a natural fit for brackish-water RO in power settings.

Although CCGTs consume ~1–2 L/kWh (thundersaidenergy.com), seasonality can bite: river TDS (total dissolved solids; the sum of dissolved ions) can spike ~4× during drought (power-eng.com). Treatment trains must be sized for both the averages and the outliers.

Raw water characterization and pretreatment train

Start with a full analysis: suspended solids, hardness, silica, chloride, organics — and how they swing over seasons. Pretreatment’s job is to strip particulates and fouling agents to protect downstream membranes and resins (power-eng.com; power-eng.com).

For high‑turbidity sources, coagulation/flocculation is followed by a clarifier designed at a surface overflow rate of ~5–10 m³/m²·h. Expect turbidity to fall to ~1–5 NTU (Nephelometric Turbidity Units; a measure of cloudiness). Without robust clarification, particulate loads can blind filters or RO membranes.

Clarifier effluent typically goes to multimedia filtration. A dual‑media bed using sand media beneath anthracite captures residual turbidity. Plants report clarifier+filter effluent at ~0.3–1.0 NTU, with well‑designed filters reliably delivering <0.5 NTU (power-eng.com). Multi‑layer designs using anthracite media extend run times and help protect downstream elements.

Many modern lines add microfiltration or ultrafiltration (UF; pressure‑driven membranes that remove fine particles and colloids) ahead of RO (reverse osmosis; a high‑pressure membrane that rejects dissolved ions). One case showed swapping a sand filter for UF dropped turbidity to ~0.03 NTU and stretched RO cartridge changeouts from ~3 weeks to ~3 months (power-eng.com; power-eng.com).

Activated carbon is a staple to remove chlorine and organics upstream of membranes and ion exchange; in practice, plants use activated carbon before final fine filtration. A 5–20 µm cartridge filter then captures remaining fines (power-eng.com). Antiscalant dosing and pH adjustment may also be used prior to RO to inhibit scale; tight control is aided by a dosing pump and, where applicable, membrane antiscalants.

Each pretreatment step should be sized for peak flow with redundancy. Typical loss‑of‑charge and backwash requirements must be balanced against uptime. Coagulant dose and filter area swing with turbidity, hardness, TOC (total organic carbon), iron, and manganese in the raw source. As Fig. 3 below illustrates, the contemporary design uses micro/UF upstream of RO to ensure very low suspended solids, enabling cartridge filters to last months (power-eng.com; power-eng.com).

Core demineralization: IX versus RO (and CEDI)

Bulk demineralization typically follows one of two paradigms: ion exchange (IX; resins that swap ions in water with ions on a solid) or membranes. Traditional two‑bed IX uses strong‑acid cation in Na⁺ form (softens and acidifies) followed by strong‑base anion in OH⁻ form (power-eng.com). Typical mixed‑bed goals are >20 kgr·eq capacity per m³ of resin; run lengths depend on raw TDS and flow. Regeneration uses HCl or H₂SO₄ for SAC and NaOH for SBA, consuming chemicals and producing dilute brine waste. Conventional two‑bed systems can deliver outlet conductivity ~0.5–2 µS/cm (power-eng.com), but require frequent regeneration when treating tens–hundreds of mg/L TDS. Plants opting for IX can tap complete systems via ion exchange packages, and resin selection via ion-exchange resin portfolios.

Mixed‑bed polishing (a single vessel mixing cation and anion resins) achieves near‑equilibrium purity — ~18 MΩ·cm (~0.055 µS/cm) — removing residual ions and meeting ASME‑grade feed (silica <10–20 ppb, pH ~9) (ebrary.net; ebrary.net). Portable resin “bottles” are often swapped by vendors, avoiding on‑site regeneration (power-eng.com). Note: mixed beds regenerate even more frequently than two‑bed systems (due to higher resin utilization) and their brine waste may require neutralization if Title V regulations apply. For package solutions, final polishers are available as mixed-bed units; integrated deionizers are offered as demineralizers.

Membranes have surged in power makeup trains. Reverse osmosis rejects ~95–99% of dissolved salts, producing 1–2 µS/cm permeate that slashes loading on any downstream IX resins (power-eng.com). One study noted RO ahead of IX “greatly reduced loading on the IX resins,” cutting regeneration frequency and chemical demand (power-eng.com). RO also removes >90% of silica and TOC — persistent IX troublemakers — with RO+CEDI (continuous electrodeionization) outperforming IX alone on silica/organics (power-eng.com; power-eng.com). RO does carry a concentrate stream (~20–30% of feed) and pump energy of ~0.5–5 kWh/m³, while CEDI consumes ~1 MJ/m³ compared with a high‑pressure RO pump at 20–50 MJ/m³ (power-eng.com). Practical RO recovery sits ~80–90% for typical river water to balance intake versus scaling risk and concentrate disposal.

Modern plants often combine RO with either a final mixed bed or CEDI. CEDI offers “chemical‑free” continuous polishing to mixed‑bed purity, but it needs very low hardness (often achieved only after RO) and is sensitive to organics and chlorine (power-eng.com; power-eng.com). Packaged EDI units and membrane lines from membrane systems portfolios are standard in new builds.

Selection ultimately hinges on raw TDS and economics (power-eng.com). Low‑TDS feeds (<~200–300 mg/L) often favor conventional IX; where solids are higher or pressure to curb chemical waste is stronger, RO pre‑treatment with IX or CEDI polish is preferred. Many new CCGTs follow the “post‑RO + portable mixed‑bed” model (power-eng.com), reflecting the shift toward membranes (power-eng.com; power-eng.com).

Final polishing and conditioning steps

If not using CEDI, a mixed‑bed ion exchanger on RO permeate drives conductivity well below 0.1 µS/cm and throttles silica into single‑digit ppb (ebrary.net; ebrary.net). These “bottles” are typically swapped monthly by service companies. Plants may also deploy a condensate polisher to guard quality during upsets.

Dissolved gases are managed via deaeration. Pre‑heating and feeding the demineralized stream to a liquid deaerator (spray or steam‑heated) takes O₂ to <20 ppb; an oxygen scavenger (hydrazine or organic sulfite) is dosed afterward (ebrary.net). Some high‑pressure HRSG designs use vacuum or membrane degasifiers, but liquid deaeration is standard.

Demineralized feedwater is typically volatiles‑treated (AVT‑Na; volatile chemicals are added for pH control) in CCGTs. Ammonia (and sometimes hiding O₂ scavengers) is added to reach ~9.4 pH, ensuring ~9.6 in the LP drum (ebrary.net). Cation conductivity (<0.2 µS/cm after degassing) is trended to detect ionic leaks.

Final units are sized for boiler‑feed flow, with resin holds and cartridges specified to ride through upsets (e.g., membrane flushes). Continuous analyzers for conductivity, silica, and pH with alarm setpoints, plus automated blowdowns based on these signals, keep quality on spec. As a rule of thumb, ultimate maxima follow OEM boiler/HRSG guidance — <0.1 µS/cm conductivity; silica <10–20 ppb; Na <5 ppb (ebrary.net; ebrary.net).

Waste, energy, and regulatory trade‑offs

IX regeneration produces spent acids/bases — roughly ~1–2 m³ regenerant per 100 m³ treated — which are typically neutralized and discharged under NPDES permits. RO creates a high‑TDS concentrate (~10–30% of feed) that needs disposal (e.g., evaporation ponds or sewer, if permitted). In Indonesia, adhering to effluent limits on flow and salt load matters; many plants recycle regenerant between beds or reuse RO concentrate for non‑critical makeup (e.g., cooling towers) to reduce discharges.

Energy matters too. Pumping and treatment usually consume <2% of plant output. RO pumps are one of the larger auxiliary draws; options such as high‑efficiency pumps and VSDs can move the needle. Notably, CEDI units use ~1 MJ/m³, whereas a high‑pressure RO pump can consume 20–50 MJ/m³ (power-eng.com). If local electricity tariffs are high, these figures influence the IX‑versus‑RO trade‑off.

Water reuse is a strategic lever. The Grati BWRO project reused 50% of wastewater (~1,600 m³/month), dropped conductivity from ~500 μS/cm to <20 μS/cm, and saved IDR 320 million/year with a 5‑year payback (papers.iafor.org; papers.iafor.org), underscoring the business case for retrofits.

On standards, Indonesia does not impose a unique boiler‑feedwater spec beyond international codes, but environmental permits emphasize water‑use efficiency and discharge limits. The government’s PROPER program rewards pollution reduction (a gold rating requires minimal effluent). Designs that drive toward near‑zero liquid discharge — e.g., reusing brine — can yield reputational and regulatory benefits. Teams should consult ESDM (Energy Ministry) and KLHK (Environment) on effluent quality and handle any B3 (hazardous) waste per government standards.

Reliability is operational. As [40] notes, operator skill and monitoring are crucial (power-eng.com). Build in redundancy — parallel filters, IX vessels, RO skids — so one unit can go offline without starving the HRSG. Automate IX regenerations and RO cleanings, include cross‑connects for bypass or interstage feed, and layer in digital control for predictive maintenance (e.g., tracking resin capacity exhaustion or RO flux decline).

Best‑practice configuration and metrics

Robust pretreatment pairs coag/floc with a clarifier and multimedia filters; many sites supplement with UF to deliver <0.5 NTU feed, with <0.1 NTU ideal to protect membranes and resins (power-eng.com). The core demineralization often runs on RO as a first stage (single or two‑pass depending on raw TDS), followed by IX polishing. RO+CEDI or RO+mixed‑bed has increasingly displaced conventional two‑bed systems (power-eng.com; power-eng.com), minimizing chemical use while producing 10–15 MΩ·cm‑quality water. Membrane‑first lines are well served by modular membrane systems.

Final polish drives feed conductivity ≪0.1 μS/cm and silica to <10–20 ppb (ebrary.net; ebrary.net), followed by deaeration to <20 ppb O₂ and pH adjustment to ~9.4. Continuous analyzers and automated blowdowns close the loop. Supporting equipment and spares from water treatment ancillaries help maintain uptime.

Sustainability measures, including recycling backwash or regenerant streams and tapping non‑traditional sources (e.g., treated municipal/industrial effluent), reduce freshwater intake — as demonstrated at Grati (papers.iafor.org; papers.iafor.org). If discharge limits tighten, designing for ZLD (zero liquid discharge) by adding evaporators or crystallizers on RO brine can be considered.

By matching pretreatment to raw water swings, right‑sizing IX/RO/CEDI, and tightening final polish and controls, plants can hit stringent chemistry specs (e.g., ASTM/ASME‑aligned targets at ebrary.net; ebrary.net) while cutting chemical and energy costs. As industry reports note, leveraging modern membranes and continuous deionization improves both quality and environmental performance (power-eng.com; power-eng.com).

Sources: industry and regulatory references as cited above — including papers.iafor.org; ebrary.net; power-eng.com; power-eng.com; power-eng.com; and power-eng.com.