Water OS

The operating system for a water system: monitor, diagnose, predict, optimise, and, within hard limits, control.

What it does

A Treatment Compiler: given source water, required output, volume, energy and cost, choose the treatment train and its setpoints, and re-plan as the source changes.
Water-energy co-optimisation: when water is scarce spend energy, when energy is scarce save it, inside the envelope.
Every plan is checked by a deterministic constraint layer before anything moves. The AI proposes; the constraint layer and the PLC decide.

What it is not

No language model is ever in the control loop of a safety-critical actuator. Water OS starts at L0 (observe) and earns each level.

Status discipline

Four labels only: Shipping, Research track, Designed, Planned. A label changes when the record supports it, not before. Current label: Research track.

See the roadmap

Treatment Compiler · specimen

Nodes/PLN-001/Plan

Live · simulated feedPlan #412Replan on ±15% conductivitySpecimen

Inputs · measured

sourcereclaimed effluent

turbidity3.1 NTU

conductivity640 µS/cm

TOC6.0 mg/L

microbialhigh risk

targetcooling make-up · CM-2

volume4.8 m³/day

energyPV 1.9 kW · batt 38%

envelope61% · Aug

Candidate trains · predicted

trainin-speckWh/m³$/m³

AUF · GAC · CF · UV · Cl0.970.90.62

BUF · GAC · CF · RO · RM · UV · Cl0.992.41.14

CUF · IX · CF · UV · Cl0.911.10.71

Every plan is written to the ledger before it is executed, and the constraint layer must clear all twelve rules first.

Constraint layer · deterministic

UF in pathPASS

UV in pathPASS

UV dose ≥ 40PASS

Cl residual 0.5 to 1.5PASS

recovery ≤ 85%PASS

TMP ≤ 0.8 barPASS

predicted spec ≥ 0.95PASS

energy within planPASS

envelope not exceededPASS

sensors trusted ≥ 0.9PASS

no pinned module overriddenPASS

divert valve healthyPASS

Why this plan

A meets the spec at a third of B’s energy. B is held in reserve and selects automatically if conductivity passes 780 µS/cm. C fails the microbial margin.

PLAN A · RELEASED TO PLC · 14:02:11Z · SIGNED

Next: Node