Thermal demand forecasting is ultimately a physics problem: predict how much cooling or heating the building will need at each hour of the day, given the building's physical properties and the forecast boundary conditions. The accuracy of that prediction depends heavily on how well the model captures the building's envelope — the thermal interface between the conditioned interior and the outdoor environment.
The envelope variable that has the most disproportionate impact on thermal demand modeling is the window-to-wall ratio (WWR): the fraction of the building facade that is glazed versus opaque. A building with 20% WWR and a building with 60% WWR at the same square footage, same location, and same occupancy have fundamentally different thermal demand profiles — different peak timing, different sensitivity to solar orientation, and different forecast model structure requirements. Getting this wrong in the model produces systematic errors in exactly the high-demand conditions that demand charge avoidance depends on.
The Physics of Solar Heat Gain Through Glazing
Opaque walls transfer heat primarily through conduction — the temperature differential across the wall, multiplied by the assembly's U-value (thermal transmittance). This is a relatively straightforward relationship: double the indoor-outdoor temperature differential and you roughly double the conduction heat gain. The dynamics are governed by the assembly's thermal resistance and the exterior surface's solar absorptance, but the heat gain is generally proportional to the steady-state temperature difference.
Glazing transmits heat through a different mechanism: solar radiation. The solar heat gain coefficient (SHGC) of a window assembly determines what fraction of incident solar radiation passes through the glazing into the interior. A window with SHGC of 0.40 transmits 40% of incident solar radiation as heat gain; one with SHGC of 0.25 transmits 25%. This solar heat gain is additive to the conduction heat transfer through the glass, which is driven by the indoor-outdoor temperature differential.
The solar radiation component is what makes high-WWR buildings dramatically more sensitive to time-of-day and solar orientation than low-WWR buildings. At 2 PM on a clear summer day with direct west-facing sun, a west-facing curtain wall with 60% WWR and SHGC 0.35 can generate solar heat gains of 15–25 W/sq ft of glazed area — equivalent to about 50,000–80,000 Btu/hr for a typical floor of a mid-size office building. That solar pulse arrives in the afternoon when outdoor temperatures are also peaking, creating a double load that the forecast model must capture accurately to predict afternoon cooling demand correctly.
Orientation as a Demand Pattern Determinant
Building orientation — which facade faces south, east, west, and north — has a first-order effect on when peak solar loads occur and how they interact with occupancy patterns.
South-facing facades receive maximum solar exposure in winter (when the sun angle is low) and minimum in summer (when the sun angle is high, making south-facing glass less exposed than east or west). This makes south-dominated buildings heating-dominant in winter and relatively moderate in summer cooling compared to their WWR would suggest.
East-facing facades receive maximum solar exposure in the morning, with peak solar gain 8–10 AM in summer. This creates a morning cooling challenge that compounds with the morning HVAC startup demand peak — both are happening simultaneously, and the forecast model needs to account for the solar contribution to the morning load for east-facing buildings, not just the conduction load that would drive the schedule in a low-WWR building.
West-facing facades are the highest cooling demand risk: maximum solar exposure in the afternoon when outdoor temperatures are also highest, and occupancy is still at near-peak levels. A high-WWR building with significant west exposure may have its peak cooling demand not at 2 PM but at 4–5 PM, after peak occupancy but when solar load is still maximum. BMS schedules designed for a 2 PM afternoon peak cooling window will under-provision for this building type's actual demand timing.
This orientation-dependent demand timing is something a data-driven forecast model can learn from historical BMS data — the model will observe that cooling demand in this building peaks later in the afternoon and adjust its predictions accordingly. But a model that doesn't have enough training data (a new building or a recently renovated one) needs the building geometry and orientation information explicitly to initialize the model correctly before the training data provides enough signal.
Thermal Mass vs. Thermal Conductance in High-WWR Buildings
Glass has very low thermal mass compared to concrete or masonry. A building with 60% WWR has a much higher ratio of low-mass glazing to high-mass concrete than a building with 20% WWR. The practical consequence for demand management is that the thermal mass pre-cooling strategies that work well in heavy-mass buildings are less effective in high-WWR buildings — there's less thermal flywheel to charge overnight because a larger fraction of the envelope is glass.
High-WWR buildings also have shorter thermal time constants because the glass envelope conducts heat much faster than an insulated masonry wall. The building heats up faster when outdoor temperature rises and cools down faster when it drops. Pre-cooling a high-WWR building 6 hours before occupancy provides less residual benefit than pre-cooling a low-WWR masonry building 6 hours before occupancy, because the glass envelope allows the stored cooling to dissipate faster.
For demand management in high-WWR buildings, this means the pre-conditioning window needs to be closer to occupancy start — 2–3 hours rather than 4–6 hours — and the staging ramp needs to be spread across a narrower window. The demand event risk is harder to shift far off-peak because there's less thermal buffer to maintain. High-WWR buildings typically benefit more from demand smoothing (spreading the cooling ramp over time to reduce peak kW) than from temporal load shifting (moving the load earlier in the day), because temporal shifting doesn't work as well when the envelope's thermal time constant is short.
We're not saying high-WWR curtain wall buildings can't benefit from forecast-based demand management — they can, and the economic opportunity is significant in buildings with large west-facing glass. We're saying the strategy needs to be calibrated to the envelope's actual time constant, not borrowed from models built for heavy-mass buildings with different thermal dynamics.
Getting Envelope Parameters Into the Model
The ideal data source for envelope parameters is the building's energy model documentation — the EnergyPlus or DOE-2 simulation file created during design or commissioning, which contains precise U-values, SHGC values, WWR by facade, and orientation data. These files exist for most buildings that received an energy code compliance analysis at permitting, though tracking them down years after construction can require working through the architect's project records.
When design documentation isn't available, envelope properties can be estimated from field inspection: facade photographs with measured grid analysis for WWR, window specification labels or building product submittals for SHGC and U-value, and IR thermography during moderate weather for effective U-value verification. This field-derived data is less precise than design documentation but sufficient for forecast model initialization.
One practical note: the effective SHGC of an installed window assembly changes over time as solar films degrade, glazing seals fail (creating condensation that reduces solar transmittance), and interior shading devices (blinds, shades, solar film added after occupancy) modify what the original spec shows. The operating building's effective SHGC may differ meaningfully from the as-specified value. A forecast model calibrated to actual energy data — which reflects the real operating envelope — will self-correct for these deviations over time, whereas a physics-only model initialized from design specs will carry the error forward until it's explicitly corrected.
In our experience, the buildings where envelope modeling errors most significantly affect forecast accuracy are those where substantial building modifications occurred after original construction — window replacement programs that changed SHGC, added exterior shading structures, or envelope repairs that changed effective infiltration rates. When scoping a new building for pilot, we ask specifically about any post-original-construction envelope modifications in the past 5 years, because these are the cases where design documentation is systematically unreliable.