Lakeshore is a water-cooled plant — four ZHT300KCE scrolls from the catalog on R513A, a coaxial evaporator and condenser, a condenser-water pump and a cooling tower with a Eurovent 35/30/25 rated point on Braun's enthalpy-basis model — holding 6.5 °C for three chilled-water handlers, each with a gas-boiler stand-in for heat and ten cubic metres of loop water declared as a buffer on its intake.

On New York's year the plant delivers 623 MWh on 138 MWh of electricity — a seasonal COP of 4.5 with its tower fans and pump. Until 4 September the stand-in cycled four fixed-speed scrolls where a centrifugal unloads on a drive, and the case said so; drawn that way the same plant reads 209 MWh, a COP of 3.0. It now carries a variable-speed centrifugal's part-load curve — a Carrier 19XR's, from the public EnergyPlus reference dataset, second-hand and a generation old, so it under-claims rather than over-claims — and draws three-eighths of its full-load power at half load. Every room is held to within seven hours. The plant carries a waterside economiser: a plate exchanger between the tower loop and the chilled-water return, ahead of the evaporator, with a free-cooling stage the controller tries before the chiller. On this campus it is worth 5.5 MWh a year, all of it the plate pre-cooling the return ahead of a running chiller: the free-cooling stage on its own never carries an hour of New York's year, because the handlers' airside economisers already take the load whenever the tower is that cold. A waterside economiser earns more on loads that persist in cold weather, a data hall or a laboratory without outside air. The boilers burn 477 MWh of gas for 406 MWh of heat, 132 MWh of it the reheat the humidity ceilings ask for.

The rooms read 588 / 1 683 / 865 humid hours, and they are hours on the bound, not over it: no room ends an hour more than 1 % above its 0.012 kg/kg ceiling, inside the controller's own tolerance. This page used to say that a boiler-fired reheat on an 8 m³/s handler does not hold its ceilings on New York summers; on the model as it stands it does, the humid-hour count reading a room that sits on its bound as a humid hour, and the price is the 132 MWh of reheat. Whether the cleanroom's face-and-bypass, which reheats with room air for nothing, takes that heat off is the next thing to measure here.

What it reads on New York

Cooling / heating	623 MWh / 406 MWh (132 of it ceiling reheat)
Electricity / gas	269 MWh / 477 MWh
Plant	623 MWh delivered on 138 MWh, seasonal COP 4.5 on a centrifugal's part-load curve (209 MWh as cycling scrolls; 144 without the waterside economiser)
Rooms	every one held to within 7 hours; humid hours 588 / 1 683 / 865, none of them more than 1 % over the 0.012 kg/kg ceiling

What it taught

• A plant this size is a centrifugal, and a centrifugal's part-load curve is a third of its annual electricity: 209 MWh as cycling scrolls, 138 on a variable-speed 19XR's published shape — the same rooms, the same hours. The curve is second-hand (EnergyPlus's dataset) and a generation old, and the validation page says so.
• A loop's thermal mass is a configuration, not a default: 10 m³ a receiver costs comfort hours because the plant answers a load step later.
• A humidity ceiling with reheat is a finding to measure, not to hide.

The machines

• Water-cooled plant, 600 kW class — open in the builder
• Campus AHU, 8 m³/s, boiler heat — open in the builder

Sources

• Copeland PSS rating database (ZHT300KCE, R513A)
• EnergyPlus reference dataset Chillers.idf — Carrier 19XR 742 kW VSD part-load curve (CoolTools)
• EVAPCO LPT bulletin and Marley Aquatower tables (tower model anchors)
• Braun, Klein & Mitchell 1989 (tower ε-NTU)
• PVGIS TMY, New York

Readings of the shipped model on the building's own PVGIS typical meteorological year; engineering estimates from catalog data and published correlations. The building runs on every plan, including Free, a few times a day.