loading...
Temperature across the ice around one tube (quasi-steady logarithmic profile at the present heat rate)
Heat flows against time: ice making, ice to the building, direct chiller and the building load
Ice radius against time with the tube surface, the sensor limit and the touching radius
Electric power with the ice bank against a chiller-only plant, and the loop outlet temperature
running...
The science
Why store ice? Melting 1 kg of ice at 0 C absorbs L = 334 kJ, the same as warming 1 kg of water by 80 K, so a tank of ice stores about 93 Wh of cold per kg: a 1 ton-hour of cooling (3.517 kWh) is only 37.9 kg of ice. A chiller can make that ice at night, when the grid is cheap and the outdoor air is cooler, and the building draws it back in the afternoon, so the chiller and its electrical connection can be smaller and run at a steadier load.
Ice on a coil. Cold glycol (about 30 % propylene glycol, freezing point near -13 C) runs through steel tubes immersed in water. Ice grows radially around every tube. Heat has to conduct through the growing annulus, so per metre of tube the resistance is a sum of three terms:
R' = 1/(2 π ri hi) + ln(r0/ri)/(2 π kwall) + ln(R/r0)/(2 π kice) q' = (0 − Tglycol)/R'
Only the last term grows with time, which is why the ice slows itself: the same tube delivers less cold as the shell thickens. The ice mass follows from the latent-heat balance (ice at 0 C; sensible heat of the ice is neglected in the main model):
ρi Lf 2 π R dR/dt = q' M = ρi π (R² − r0²) L N
For a wall held ΔT below freezing and no other resistance this has the classic implicit solution used in the first check: k ΔT t / (ρi Lf) = (R²/2) ln(R/r0) − (R² − r0²)/4.
Glycol side. Each circuit of n runs in series is an effectiveness-NTU heat exchanger against a 0 C bath: NTU = (n Lt / R') / (m cp), ε = 1 − e−NTU, Q = ε m cp (0 − Tin). The film coefficient hi comes from Nu = 4.36 (laminar) or Dittus-Boelter 0.023 Re0.8 Pr0.3 (turbulent) with a blend between Re 2300 and 4000. A glycol circuit warms by a few kelvin, so real ice is thicker near the inlet; the model uses the mean circuit and every run is drawn the same.
Bridging. On a square grid the ice cylinders touch when R = pitch/2. After that the ice merges, water circulation is blocked and the surface area shrinks, so the sensor limit is capped at 99.5 % of this radius. Ice fraction = rhoi π (R² − r0²) L N / (all water in the tank); water in the margin between the bundle and the walls stays liquid.
Melt-out. Warm building return water flows through the tank and melts the ice from the outside. With the bath at 0 C the outlet temperature is Tout = Tin e−NTU, NTU = h A / (m cp) with A = N 2 π R Lt: as the ice shrinks, A falls and the outlet warms. An air bubbler raises h. Once the ice is gone the tank is a mixed bath that warms towards the return temperature.
Chiller and tariff. COP = 0.5 Tevap / (Tcond − Tevap) (half of the Carnot value, kelvin). Ice making evaporates about 3 K below the glycol temperature, direct cooling at 2 C; the condenser is 8 K above a sinusoidal outdoor temperature. The building load, outdoor temperature and the three-step tariff are stylised curves; prices are illustrative, not a market. Pumps are 0.25 kW (glycol) and 0.2 kW (loop) at their reference flows, scaled with flow cubed.
Checks run in the page. (1) The integrator against the closed form; (2) a brute-force radial enthalpy finite-volume solver (explicit, full transient conduction including the ice sensible heat) against the quasi-steady closed form; (3) a Monte Carlo estimate of the ice area of one cell; (4) the glycol-side enthalpy balance against latent heat; (5) a complete charge and melt-out cycle conserving energy; (6) the ton-hour unit conversion. Checks 4 and 5 are bookkeeping audits of the integrator; 1 to 3 are independent of the code that draws the ice.
Honest simplifications. Quasi-steady radial conduction (ice sensible heat neglected: about -0.3 % on the radius after one hour in the finite-volume comparison); uniform mean circuit; no supercooling or nucleation delay; ice density change and tank level ignored; the water bath stays at 0 C while ice exists; tube material treated as steel (k = 45 W/m/K); glycol properties constant at typical values for 30 % propylene glycol near -5 C (cp 3.7 kJ/kg/K, mu 5 mPa s, k 0.43 W/m/K); no stratification inside the tank; no heat gain from the pumps. Material values are typical textbook figures, not measurements of a specific product.