Portfolio-Scale Calibration

Roovie supports calibrating entire building portfolios against real utility data, not just individual buildings.

The workflow combines three systems: the Simulation Studio runs batch simulations across all buildings in a portfolio, the shared meter panel allocates campus-level meter readings back to individual buildings, and the calibration engine computes ASHRAE Guideline 14 metrics for each building and for the campus as a whole.

This is how calibration works at scale — when you have dozens of buildings, shared meters, and the need to validate an entire portfolio's simulation accuracy in a single coordinated workflow.

In One Line

Batch simulate. Allocate shared meters. Calibrate the portfolio.

The End-to-End Workflow

Identify building groups (meter boundaries)
  → Queue all buildings in Simulation Studio
  → Launch batch simulation with consistent profile
  → Enter shared meter actuals per building group
  → Allocate meter readings to buildings using simulation shares
  → Review campus-level calibration metrics
  → Review per-building diagnostic metrics
  → Lock well-calibrated buildings, override anomalies
  → Iterate until portfolio meets thresholds

Step 1: Batch Simulation

Before calibration can happen, every building needs a baseline simulation. The Simulation Studio handles this at scale.

The operator:

1. Creates a simulation profile with standardized settings — date range, physics modes, output options
2. Queues all buildings in the portfolio (individually or by building group)
3. Launches the batch with concurrent execution
4. Monitors progress as buildings complete or fail
5. Retries any failures

The batch runs server-side. One building's failure does not block the others. The operator can close the browser and return later — the batch continues running.

Once complete, every building has a baseline simulation with monthly results that can be used for allocation and comparison.

Step 2: Shared Meter Allocation

Most real portfolios do not have individual utility meters for every building. Campuses, school districts, healthcare systems, and multi-building properties typically share a single meter at the site level.

The shared meter panel takes the aggregate meter reading and splits it across buildings using their simulation results as the proportional driver.

For each month:

Building's allocated energy =
  (Building's simulated kWh / Sum of all group simulated kWh)
  × Actual shared meter reading

The operator enters 12 months of actual meter data — either combined kWh or separate electricity and natural gas — and the system calculates each building's share automatically.

Residual Load

Not all campus energy goes to the buildings. Site lighting, distribution losses, central plants, and EV charging may not be modeled. The residual load setting reserves a percentage of the meter reading for these unmodeled loads before allocating the remainder to buildings.

Applied Locks

When one building's calibration is already good, the operator can lock it. A locked building keeps its current allocation unchanged while the remaining buildings rebalance proportionally from what is left.

Monthly Overrides

For months with known anomalies — a building closure, temporary equipment, unusual occupancy — the operator can manually override the allocation percentage for specific building-month combinations.

Step 3: Calibration Review

After allocation, the system computes calibration metrics at two levels:

Campus-Level Calibration (Primary Target)

The aggregate simulation output for all buildings in the group is compared against the actual shared meter reading. This is the primary calibration target.

• NMBE (Normalized Mean Bias Error) — measures systematic over or under-prediction
• CV-RMSE (Coefficient of Variation of Root Mean Square Error) — measures prediction variability
• Pass/fail against configurable thresholds (default: NMBE within 5 percent, CV-RMSE within 15 percent)

Per-Building Calibration (Diagnostic)

Each building's simulation is compared against its allocated actuals. These metrics are explicitly labeled as diagnostic — they are useful for identifying which buildings are driving the campus-level error, but they are not the same as submetered calibration.

Monthly Breakdown

Both levels include month-by-month comparison:

• Actual versus simulated for each month
• Monthly delta in kWh and percentage
• End-use breakdown comparison (HVAC, heating, cooling, lighting, equipment)
• Visual bar charts and data tables

Step 4: Iteration

Calibration at portfolio scale is iterative:

1. Review the campus-level comparison first — this is the number that matters
2. If the campus comparison is poor, identify which buildings are driving the error using per-building diagnostics
3. Lock buildings that are already well-calibrated
4. Apply monthly overrides for known anomalies
5. Recalculate and review again
6. If model inputs need adjustment (schedules, loads, envelope properties), update the building model and re-run the simulation through the Studio
7. Repeat until the portfolio meets the target thresholds

The system detects stale allocations automatically. If a building is added or removed from the group, or if a designated simulation changes, the interface prompts recalculation before the operator can use outdated results.

What Gets Tracked

Every allocated building carries metadata about where its actuals came from:

• Source type: building (own meter) or shared_meter_group (allocated)
• Basis: measured (real submeter) or allocated (derived from group allocation)
• Group reference: which building group and allocation method
• Stale flag: whether the allocation needs to be refreshed
• Designated simulation: which simulation was used for the allocation ratio

This transparency means allocated actuals are never confused with real submeter readings.

How Batch Simulation and Calibration Connect

The Simulation Studio and shared meter panel are designed to work together:

• Batch simulations create the baseline results that drive allocation ratios
• The shared meter panel uses those results to split actual meter readings
• When models are refined and re-simulated, the allocation automatically updates on recalculation
• The same simulation profiles that ensure consistency in batch runs ensure consistency in calibration comparisons

This connection means calibration is not a separate, disconnected workflow. It is the natural next step after batch simulation.

What Makes This Different

Most calibration tools work one building at a time. When buildings share meters, teams use spreadsheets to manually allocate consumption.

Roovie handles the full portfolio workflow:

• Batch simulation removes the bottleneck of running buildings one at a time
• Shared meter allocation uses model-based ratios instead of manual spreadsheet splits
• Campus-first calibration focuses effort on the right target
• Applied locks protect well-calibrated buildings while others are refined
• Stale detection prevents outdated allocations from persisting
• All allocated actuals are clearly labeled as estimates, not submeter readings

The result is a calibration workflow that scales to portfolios of any size, handles shared meters natively, and maintains transparency about what is measured versus what is allocated.

Bottom Line

Portfolio-scale calibration in Roovie combines batch simulation, shared meter allocation, and iterative refinement into a single coordinated workflow. The Simulation Studio runs consistent baseline simulations across the portfolio. The shared meter panel allocates campus-level meter readings using simulation-based ratios. The calibration engine produces ASHRAE Guideline 14 metrics at both the campus and building level.

The result is a portfolio where every building's simulation accuracy is quantified against real utility data — even when individual submeters do not exist.