Energy codes and building energy simulation

Design loads size equipment, but the owner pays for energy, and the energy code regulates it. This lesson explains how energy codes for commercial buildings work, which code governs the health center, how compliance is shown, and how annual energy is estimated, from the simple degree-day method to hourly simulation. The program runs a small hourly simulation of the building's heating and tests a set of energy conservation measures.

How energy codes work

Commercial energy codes regulate the envelope, the HVAC systems, service water heating, lighting and electrical power. Most offer two paths. The prescriptive path sets requirements component by component: maximum U-values for walls, roofs and windows, minimum equipment efficiencies, fan power limits, required controls such as economizers, setback and demand-controlled ventilation, and limits on the window-to-wall ratio. A building that meets every prescriptive requirement complies. The performance path allows trade-offs: the designer models the proposed building and a reference or baseline building that just meets the prescriptive requirements, and shows that the proposed design uses no more energy, or no more energy cost, than the reference. Trade-offs let a building with, say, more glass compensate with better heat recovery.

For this building the governing code is the Minnesota Energy Code, whose commercial provisions adopt ASHRAE/IES Standard 90.1-2019 with Minnesota amendments and came into force on January 5, 2024. Standard 90.1-2022 is the current national edition and the 2024 International Energy Conservation Code is the current model code, and Minnesota law passed in 2023 directs successive updates of the commercial code toward much larger energy reductions by 2036, so later editions will be stricter. The hospital's own policy requires the building to perform substantially better than the code minimum, and the team uses the performance path to show it.

District energy in energy codes

A building that buys heat and cooling from a district system has no boilers or chillers whose efficiency can be checked, and codes need a rule for comparing it with a reference building that has its own plant. The approaches differ between codes and editions, but the common idea is that the district energy is either converted to an equivalent on-site source at a specified efficiency, or the reference building is given the same district supply. Kofi, who has done this before for Northgate's customers, follows the code's own rules and the authority's published guidance, and documents the method in the compliance report, because it is the first thing a reviewer will question.

Estimating annual energy: degree-days

The simplest annual estimate uses heating degree-days. If a building loses heat at a rate UA per kelvin and has a balance point temperature below which it needs heating, the annual heating energy is UA times the degree-days below the balance point times 24 hours. Published degree-days are usually to base 18 °C, or 65 °F; the true balance point is lower because internal and solar gains supply part of the loss, which is why degree-day estimates are often corrected by a factor. The method is quick and useful for checking orders of magnitude, but it ignores schedules, setback, the timing of gains, ventilation that runs only when the building is occupied, and the way systems behave at part load.

Bin methods and hourly simulation

The bin method improves on degree-days by grouping the hours of a typical year into temperature bins, 5 K wide for example, and calculating the load and system performance for each bin, separately for occupied and unoccupied hours. It captures part-load behavior and schedules reasonably well and can be done in a spreadsheet. Hourly simulation goes further: it calculates the building's loads and systems for each of the 8,760 hours of a typical weather year, with the thermal mass, schedules, controls and equipment curves. Programs such as EnergyPlus, developed for the US Department of Energy and freely available, and commercial tools built on it or on other engines, are used for code compliance and design decisions. Typical-year weather files are constructed from long records to represent average conditions; they are not suitable for checking peak loads.

# Annual heating energy: degree-day estimate versus an hourly simulation on a synthetic typical year
import math, random
SITE = 'us'
d = {'ca': dict(city='Winnipeg', mean=2.8, amp=18.5, hdd_pub=5670), 'us': dict(city='Minneapolis', mean=7.8, amp=16.5, hdd_pub=4250)}[SITE]
random.seed(430)
temps, w = [], 0.0                                   # synthetic hourly outdoor temperatures (a teaching data set)
for hr in range(8760):
    day = hr/24.0; w = 0.97*w + random.gauss(0, 0.95)          # weather swings lasting several days
    t = d['mean'] - d['amp']*math.cos(2*math.pi*(day - 20)/365) + 4.5*math.cos(2*math.pi*((hr % 24) - 15)/24) + w
    temps.append(t)
hdd = sum(max(0, 18 - t) for t in temps)/24
print(f"{d['city']} synthetic year: mean {sum(temps)/8760:.1f} C, minimum {min(temps):.1f} C, HDD18 {hdd:.0f} "
      f"(published normal about {d['hdd_pub']})")
UA = 2431.0                                          # W/K envelope + infiltration (lesson 4)
area, c, oa = 6000.0, 1.21, 4.5
def simulate(UA=UA, eff=0.70, setback=True, usable=0.35):
    # only part of the internal and solar gain offsets perimeter losses; the rest lands in interior zones that need cooling
    q_total = 0.0
    for hr, t in enumerate(temps):
        dow, h = (hr // 24) % 7, hr % 24
        occ = dow < 5 and 7 <= h < 19 or dow == 5 and 8 <= h < 13
        ti = 22.0 if occ or not setback else 18.0
        internal = usable*(400*75*0.7 + 16*area*0.9) if occ else 0.15*16*area
        solar = usable*max(0.0, 25000*math.sin(math.pi*(h - 6)/12)) if 6 < h < 18 else 0.0
        q = UA*(ti - t) - internal - solar
        if occ:
            trec = t + eff*(22 - t)
            q += 1000*c*oa*max(0.0, 18 - trec)
        q_total += max(0.0, q)
    return q_total/1e6                               # MWh
base = simulate()
print(f"Hourly simulation, base design: {base:.0f} MWh/yr = {1000*base/area:.0f} kWh/m2 of heating")
dd_est = UA*24*hdd/1e6*0.8
print(f"Degree-day estimate of envelope heating only, 0.8 correction for gains: {dd_est:.0f} MWh")
print(f"  (the simple estimate leaves out ventilation, setback and the timing of gains)")
for name, kw in (('no night setback', dict(setback=False)), ('energy recovery 0.80', dict(eff=0.80)),
                 ('triple glazing, U 1.0', dict(UA=UA - 941*0.6)), ('no energy recovery', dict(eff=0.0))):
    r = simulate(**kw); print(f"  {name:22s}: {r:5.0f} MWh  ({100*(r - base)/base:+.0f} %)")

Minneapolis synthetic year: mean 8.7 C, minimum -21.2 C, HDD18 4049 (published normal about 4250)
Hourly simulation, base design: 146 MWh/yr = 24 kWh/m2 of heating
Degree-day estimate of envelope heating only, 0.8 correction for gains: 189 MWh
  (the simple estimate leaves out ventilation, setback and the timing of gains)
  no night setback      :   180 MWh  (+23 %)
  energy recovery 0.80  :   130 MWh  (-11 %)
  triple glazing, U 1.0 :    94 MWh  (-35 %)
  no energy recovery    :   285 MWh  (+96 %)

Reading the result

The program builds a synthetic year of hourly temperatures for Minneapolis from the city's mean and seasonal swing with random weather variations, a teaching data set that reproduces the published heating degree-days to within a few percent. It then steps through the year with the building's envelope from Lesson 4, the clinic schedule, night setback, internal and solar gains and the DOAS with energy recovery running only when the building is occupied. Only part of the internal and solar gain is counted as reducing the perimeter heating, because much of it lands in interior zones that need cooling even in winter. The base design uses about 146 MWh of space heating a year, about 24 kWh per square meter. The figure excludes service water heating and the imaging suite, which a full model would include.

The measures show where the energy goes. Removing night setback adds 23 percent. Raising the recovery effectiveness from 0.70 to 0.80 saves 11 percent. Triple glazing with a U-value of 1.0 saves about a third, more than any other measure, which agrees with the design load breakdown in Lesson 4. Removing energy recovery would raise the heating energy by 96 percent: in this building the wheel is not an option but a necessity. The degree-day estimate of envelope heating alone gives a number of the right order, but it says nothing about these measures because it cannot represent ventilation schedules, setback or the timing of gains.

Trade-offs in practice

The measures in the program are typical of the decisions the performance path is used for. The architect would like a little more glass on the community room facade facing the river; the mechanical team would like better heat recovery and lower fan power; the cost consultant would like to drop the triple glazing. Each proposal can be tested with the same model, and the compliance margin, the difference between the proposed and reference results, is the currency in which they are traded. A good team agrees early on how much margin to keep in reserve for changes during construction, because a building that complies by a hair at design development usually fails to comply by the time the last value-engineering change has been made.

Energy measures also interact. Better glazing reduces the heating load, which reduces the value of better heat recovery; better heat recovery reduces the ventilation load, which reduces the value of demand-controlled ventilation. Adding up the savings of measures tested one at a time overstates the combined saving. The model is run with the chosen package as a whole before anyone reports a total, and the hospital's board is shown the package result, not the sum of the parts.

Using simulation well

Energy models are powerful and easily misused. The commonest errors are inputs: schedules that assume the building is empty at night when the urgent care unit is open, equipment densities from nameplates, infiltration rates borrowed from another building, and controls modeled as ideal when the real sequences will be less so. A model calibrated to nothing is a hypothesis. For design decisions, comparisons between options using the same model and inputs are far more reliable than absolute predictions, because many errors cancel. For compliance, the code's modeling rules, which specify how the reference building is built and which inputs must match, take precedence over the modeler's preferences.

A growing number of owners also set targets for energy use intensity, annual energy per square meter, and for carbon. The hospital has set a target for the health center's total site energy use intensity and for its operational carbon, and Kofi's model reports both. After occupancy, the building's actual use will be compared with the model and with Northgate's meter data, and differences will be investigated; the model's assumptions, recorded in the basis of design, make that investigation possible.