9. The water loop of the boiler house: why the gas looked 1 % off, and why it is not

Question. In the house before renovation, the study first reported FastSmoothBuildings's gas 3.3 % above Buildings, then 2.6 %, for 1.7 % more heat in the rooms. Does FastSmoothBuildings lose or misplace heat in its water loop (boiler, pump, three-way valve, six radiators with thermostatic valves), or burn more gas per unit of heat?

Answer. No. Both water loops close their energy balance from the energy meters within 0.15 % over the month, the jacket losses of the boilers are identical (127.3 kWh), and the two libraries burn the same gas per unit of heat within 0.2 % (1.230 and 1.233 kWh of gas per kWh delivered by the radiators). FastSmoothBuildings burns 1.9 % more gas because it delivers 1.7 % more heat. The 1 % in excess came from the study itself, which read the gas from samples of the fuel power five minutes apart. Those samples catch or miss the two-minute bursts of the boiler when the valves open in the morning: −0.75 % for Buildings, +0.6 % for FastSmoothBuildings. The study now reads both meters (study/buildings/harness.py).

Data: results/water_loop_month.json (28 days, by window), results/water_loop_restart.json (days 8 to 10), results/water_loop_restart_table.csv (the morning restart), results/water_loop_components_restart.csv, results/water_loop_restart_tau0.5.json; script scripts/water_loop.py.

Method

The base case is simulated with both libraries in OpenModelica, keeping the variables of the loop. The energy balance of each loop, in kWh over a window:

• gas: the increments of the fuel meter (EFue, an integrator in Buildings, a state in FastSmoothBuildings), exact whatever the output interval; next to it, the integral of the sampled fuel power, as the study computed it before;
• into the water: efficiency (0.85) times the gas, minus the jacket loss of the boiler to its room. Buildings 13's BoilerPolynomial injects QWat_flow = eta*QFue_flow + UAOve.Q_flow into the water and then conducts UAOve out of it again, so its loss counts twice; FastSmoothBuildings's port of it does the same with a doubled UA. Plus the pump power, which Buildings adds to the water;
• out of the water: the heat of the six radiators to their rooms (convective and radiant ports);
• stored: the change of the heat in the water and metal of FastSmoothBuildings's radiators and boiler;
• residual: in − out − stored, the energy the loop would create or lose.

The month (28 days of January)

	Buildings	FastSmoothBuildings
Gas from the meter	3390.6 kWh	3454.2 kWh (+1.9 %)
Gas from the 5-minute samples (what the study read)	3365.3 kWh (−0.75 %)	3475.6 kWh (+0.62 %)
Jacket loss of the boiler	127.3 kWh	127.3 kWh
Pump heat	4.0 kWh	(not modelled)
Into the water	2758.7 kWh	2808.8 kWh
Radiators to the rooms	2755.5 kWh	2801.9 kWh (+1.7 %)
Stored in water and metal	(not computed)	2.6 kWh
Residual	3.2 kWh	4.3 kWh
Gas per unit of heat	1.230	1.233

The first day is the warm-up from the initial temperatures (20 degC everywhere): both libraries burn 71 to 72 kWh of gas and leave 2.2 kWh in the water and metal of the loop (FastSmoothBuildings's storage is computed, Buildings' is its residual of that day). After it, the residuals are 1.0 kWh (Buildings) and 3.9 kWh (FastSmoothBuildings) on 2700 and 2750 kWh: 0.04 and 0.14 %.

The morning restart

At 06:00 the bathroom's set point steps up, at 07:00 those of the living spaces. Day 8, every 15 minutes:

Time	Buildings: firing, water, gas	FastSmoothBuildings: firing, water, gas
05:45	0.13, 74.87 degC, 2.4 kW	0.13, 74.87 degC, 2.4 kW
06:00	0.13, 74.87 degC, 2.5 kW	1.00, 73.97 degC, 18.8 kW
06:15	0.15, 74.85 degC, 2.8 kW	0.16, 74.84 degC, 3.1 kW
06:30	0.33, 74.67 degC, 6.3 kW	0.27, 74.73 degC, 5.0 kW
07:00	0.24, 74.76 degC, 4.5 kW	0.81, 74.19 degC, 15.2 kW
07:15	0.62, 74.38 degC, 11.7 kW	0.61, 74.39 degC, 11.5 kW

When a thermostatic valve opens, cold water from its radiator returns to the boiler; the water dips and the proportional controller on the boiler temperature (gain 1/K, set point 75 degC, the same in both libraries) fires up to full power for a minute or two. Both libraries do it: the boiler water reaches its lowest at 72.0 degC in Buildings and 72.2 degC in FastSmoothBuildings over these days. FastSmoothBuildings's set points change as a smooth step of 60 s centred on the switch time, so its burst starts 30 s before the hour, where Buildings' starts at the event. That is why FastSmoothBuildings's extra gas shows by hour of the day at 05:00 to 07:00 (results/study_hour_of_day.csv), and why samples five minutes apart misread it: a sample that falls in a two-minute burst at 18.8 kW counts it for five minutes, a sample that falls next to it counts none.

Two leads that were not the cause

• The compliance of the pressure nodes. FastSmoothBuildings's pressure-driven circuits give the inlet of each component a small compliance (tauHyd = 5 s) whose stored mass the energy balance does not follow. With tauHyd = 0.5 s the residual over days 8 to 10 is 0.39 kWh against 0.26 kWh at 5 s, on 192 kWh: no difference (water_loop_restart_tau0.5.json).
• A component that loses heat. The balance of each FastSmoothBuildings component from its port flows and temperatures, sampled every 5 minutes (water_loop_components_restart.csv), does not close: it sums to 42 W where the meters give 5 W, and its terms depend on the reference temperature of the enthalpy, because the samples do not resolve the flow transients of the valves. That per-component audit first suggested a leak in the radiators; the exact balance from the meters shows there is none.

What it changed

• The study keeps the meters of Buildings (EFue.y, EEle.y, EFan.y) and reads every production from its meter first (study/buildings/harness.py): the gas of the house before renovation is +1.9 % for +1.7 % heat.
• Nothing in the library: FastSmoothBuildings's boiler, pump, valves and radiators conserve energy and burn gas per unit of heat as Buildings does.