Water Units at a Glance

Water Units at a Glance#

A one-page map of the difflow_water plugin: every unit, what it takes and makes, what model it is, and how far it has been validated. The full treatment (equations, parameters, references) is in Water Treatment and Desalination Unit Operations.

How the units connect (phases 1 and 1b)#

 intake seawater (H2O + TDS)
        │
        ▼
 PressureExchanger ──bypass──► WaterPump (HP) ──┐
   ▲      │                                     ├─► ROStage / ROTrain ──► permeate (product)
   │      └──hp_feed──► WaterPump (booster) ────┘       │
   │                                                    │
   └──────────────── concentrate ◄──────────────────────┘
          │
          └──brine──► disposal

 economics.lcow(...) ◄── RO area, pump works, PX flow

Units#

Unit	Kind	In → out	Model	Validated how
`seawater` properties	library	(T, S) → ρ, μ, φ, π	Sharqawy et al. 2010 Eqs. 8, 22, 23, 49	Coefficients read against the paper (content verified); within Sharqawy’s stated accuracy of TEOS-10 (density 0.07 %, cp 0.15 %, osmotic pressure 0.6 %; notebook 77)
Pitzer chemistry, `speciate`	library	element totals, T → speciation, pH, saturation indices	Harvie-Møller-Weare Pitzer model as in PHRQPITZ; PHREEQC pitzer.dat parameters	Equations read against PHRQPITZ (content verified); SI within 0.0004 and pH within 3e-5 of PHREEQC, seawater to 3x concentrate, 15-35 C
`ROStage`	palette	feed → concentrate, permeate	Solution-diffusion, film theory, plug flow; WaterTAP channel correlations; optional saturation indices	Within 0.3 % of WaterTAP RO1D (160 elements) on four cases
`ROTrain`	palette	feed → concentrate, permeate	ROStages in series with interstage boosters	No physics of its own
`PressureExchanger`	palette	brine, feed → hp_feed, brine, bypass	Efficiency form, optional leakage and mixing	Within 0.02 % (outlet pressure) of WaterTAP in the plant, with and without leakage/mixing
`WaterPump`	palette	inlet → outlet	Isothermal, `V dP / eta`	Within 0.24 % of WaterTAP in the plant
`economics`	library	sizes, power, product flow → LCOW, SEC	WaterTAP 1.8.0 costing structure and defaults	LCOW within 0.16 % of WaterTAP
`network`: `WaterNetwork`, `NetworkModel`	library	network, demands, leaks, settings → heads, pressures, flows	EPANET 2.2 equations, equation-oriented Newton; valves as complementarity (sharp or smooth); INP reader; extended-period simulation with controls	Within 2.7e-5 m (Net1, Net3) and 1.6e-3 m (ky10, 920 junctions) of EPANET 2.2; over 24 h / 7 days, tank levels within 1e-4 m (Net1) / 3.4e-2 m (Net3), statuses identical; chlorine within 0.01 mg/L, age 0.16 h (Net1); pump energy and differentiable speed schedules
`MVCDesalinator`, `mvc_performance`	palette	feed → distillate, brine	WaterTAP’s single-stage MVC plant: evaporator, compressor, condenser, preheaters, pumps, costing; one Newton solve	Within 4e-7 of WaterTAP 1.8.0 on every output; its LCOW optimum reproduced with `jax.grad` + SLSQP (1e-7)
`MEDDesalinator`, `med_performance`	palette	feed → distillate, brine	Parallel-feed MED: per-effect mass, salt and energy balances on Sharqawy properties	No WaterTAP counterpart; balances only (no fitted coefficients)
`ActivatedSludgePlant`, `activated_sludge_plant`	palette	influent → effluent, wastage	BSM1: ASM1 reactors + Takács settler, steady state, one equation set	BSM1 report Tables 6-7 within 3.2e-3 (printed digits); WaterTAP’s BSM1 example within 1.3e-7
`AnaerobicDigester`, `anaerobic_digester`	palette	influent → effluent, biogas	ADM1 as implemented for BSM2 (Rosen & Jeppsson 2006), algebraic pH, steady state	Their 13-digit verification case within 1.7e-7; WaterTAP’s ADM1 example within 2.5e-4
`ASM1Reactor`	palette	inlet → outlet	ASM1 CSTR at steady state; KLa or DO setpoint	BSM1 Table 6 reproduced reactor by reactor within 0.4 % (notebook 79)
`TakacsSettler`	palette	feed → effluent, underflow	10-layer Takács settler, steady state	BSM1 Table 7 layers within 0.13 % from the printed feed (notebook 79)
`Sedimentation`	palette	feed → clarified, sludge	Ideal basin, overflow-rate removal, Stokes settling	Derived from the ideal-basin assumptions; Stokes as stated in Messina et al. 2015 (read)
`Filtration`	palette	feed → filtrate, retained	Yao colloid filtration (clean bed), Carman-Kozeny head loss	Yao’s efficiency cross-checked against Messina et al. 2015; no independent numerical check
`Disinfection`	palette	feed → treated	Chick-Watson in CSTRs in series; CT, log inactivation	Fitted to the EPA SWTR CT table for Giardia (within 3 %); reproduces 3 logs at the table’s CT (notebook 78)
`GACAdsorber`	palette	feed → treated, removed	WaterTAP’s CPHSDM (Hand et al. 1984)	Within 2e-14 of WaterTAP 1.8.0 on four cases
thermal properties	library	(T, S, P) → c_p, h, h_fg, p_v, BPE	Sharqawy 2010 Eqs. 9, 36, 37, 43, 53-55; Nayar 2016 digits via WaterTAP	Content verified to the printed digits; brine vapor pressure equal to WaterTAP’s to 1e-9
`swro_performance` / `build_swro_flowsheet`	library	intake → LCOW, SEC, recovery, …	The plant above	Against WaterTAP’s `RO_with_energy_recovery` layout

Palette: registered with the editor through the plugin’s register() and usable as a Flowsheet operation. Library: called from Python; still differentiable.

What to trust#

Seawater properties: published correlations, coefficients checked against the source, inside their stated ranges.
RO and plant numbers: they reproduce WaterTAP to 0.3 %. That checks the implementation against an independent code with the same physics, not the physics against a plant. The channel correlations behind the mass-transfer coefficient and pressure drop are WaterTAP’s and are unverified against their papers.
Scaling: the Pitzer equations are checked against their source and the numbers against PHREEQC; the parameters themselves are PHREEQC’s.
MVC: equal to WaterTAP’s model to 4e-7: an implementation check (same equations, same properties), not a plant validation. MED rests on mass and energy balances only and has not been cross-checked.
Costs: WaterTAP’s default parameters, a benchmark convention, not a cost estimate for a real plant.

Known gaps#

Every unit is taken on its own, with its check, in notebooks 77-79 (per-unit examples and checks). Against WaterTAP 1.8.0 the plugin covers its seawater RO, MVC, GAC, ASM1 (BSM1) and ADM1 flowsheets, about 15-20 % of WaterTAP’s unit models. The full list of what is missing (nanofiltration, electrodialysis, ion exchange, the zero-order library, BSM2, …) is in Coverage relative to WaterTAP.

RO transports sea salt as one solute; ions are resolved only for scaling (equal rejection assumed). Saturation indices are at 1 atm, without kinetics; no Sr, Ba, silica or boron.
No nanofiltration yet.
Networks: water quality is not differentiable; no rule-based controls, volume curves or quality sources.
Thermal: no MSF; MED without inter-effect preheaters or flash boxes and without an independent cross-check.
Treatment: no coagulation, softening, ion exchange, UV or UF/MF; filters are clean-bed; no treatment costing.
Wastewater: steady state only, ASM1 at 15 C; no ASM2d or ASM3 yet; no ASM1-ADM1 interfaces; no costing.