A physics-based, browser-based simulator of a commercial building's condenser-water cooling loop — two evaporative cooling-tower cells and lead/standby condenser-water pumps. One self-contained HTML file: no build, no dependencies, no backend.
It's a control and efficiency test bench, not a data historian — a place to rehearse the cooling-tower sequence of operations, try setpoints, compare pump/fan staging, and see the energy cost, against a model calibrated to real plant telemetry. Open index.html and it runs.
- Live animated schematic — pipe color tracks water temperature, dash speed tracks flow, fans and pumps spin at their modeled speed, with live values at each sensor and valve.
- The sequence runs the plant — tower staging, CWS low-limit protection, freeze logic, and economizer gating operate automatically; or take any actuator to HAND to override the sequence (including its safeties) and drive the plant into failure modes on purpose.
- KPI faceplate — supply / return temperature, approach to wet-bulb, loop flow, heat rejection, and live electrical power draw (fans + pump).
- Everything adjustable, live — setpoints, sequencing thresholds, tower and hydraulic constants, PID gains, ambient (OAT / wet-bulb), and a weather-driven building-load model.
- Optional live weather — pick any US METAR/ICAO station (e.g.
KJFK) or a globallat,lonpair, and the sim drives OAT + wet-bulb from real observations (NWS, with an Open-Meteo fallback). It's the only network call; everything else runs locally. - Alarm flags — cavitation, low flow, low / freeze CWS, freeze risk, OAT lockout, over-HP, air ingestion, basin overflow, pump deadhead, basin low, and ΔT clamp.
- Pumps — affinity laws on a manufacturer head curve with a part-load efficiency curve; series / parallel hydraulic resistances are solved against the curve to find the operating flow.
- Cooling tower — an open evaporative tower in effectiveness (ε-NTU) / air-side Merkel form on the wet-bulb enthalpy driving force; airflow follows fan affinity, with a natural-draft floor when the fan is off.
- Thermal transport — first-order lags on the hot and cold legs, pinned to the measured loop transients.
- Building load — economizer-gated: a 24/7 IT/CRAC floor, an occupancy-scheduled AHU load, and bidirectional water-source heat pumps that add to or pull heat from the loop.
The hydraulic resistances, pump differential-pressure curve, transport volumes, and tower air-side cap are fit to building-automation (BAS) field trends; the equipment skeleton — tonnage, pump horsepower, pipe sizes, basin volume — is the real installed plant.
It's a model, best used for relative comparisons (this staging vs. that one, this setpoint vs. that one) and for rehearsing sequences. In summer, fed true wet-bulb, steady-state supply temperature tracks the field to about 0.94 °F RMS across the load range. Known limits:
- Cold-weather behavior is unvalidated — there's no sub-freezing field data yet, so the winter freeze / bypass physics and fan-speed sensitivity are provisional.
- Loop load is treated as equal to tower heat rejection — the compressor heat-of-rejection uplift (~×1.2) isn't applied yet, so duty reads slightly conservative.
- Differential pressure is modeled as a function of flow — correct for this constant-flow building, but it can't reproduce a measured DP-reset schedule.
- The building-load model is a warm-season fit — winter / heating-season loop load is modeled, not measured.
- Quasi-steady thermal response — steady-state behavior is the trustworthy part; transient accuracy isn't characterized.
- Run it — open
index.htmlin any modern browser. To serve it live, use GitHub Pages (Settings → Pages → deploy from branch, root folder). - Drive it — press Pause/Run to let the sequence operate the plant, or take an actuator to HAND to override it. Watch the faceplate and the power draw figure to weigh the energy cost of staging and setpoint choices.
- Adapt it to another plant — the model is generic in form and specific in numbers. Most site constants live in the
CFGobject; a few engine constants (pump curve, basin volume, fan kW, alarm thresholds) sit nearby — all near the top of the<script>block, with inline notes on what each one is. Re-fit the hydraulics from a couple of pump-speed points and the tower from a fan / wet-bulb sweep. The SVG schematic is hand-laid for a two-cell layout, so a different plant topology means editing the SVG and its render bindings, not just the numbers.
Vanilla HTML / CSS / JavaScript — no frameworks, no build, no dependencies. A single file you can read top to bottom.
MIT — see LICENSE.
An engineering and teaching tool. All outputs are model estimates and carry the limitations above; verify independently before acting on a real plant. The build is de-identified — no site, owner, tenant, or equipment-make information — while keeping the real equipment profile for realism.