> ## Documentation Index
> Fetch the complete documentation index at: https://documentation.kodelabs.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Energy modeling and M&V
> How EnerG uses IPMVP options, adjusted baselines, and regression models to verify savings and forecast usage
Energy modeling links meter data to savings verification and forecasts. EnerG follows the International Performance Measurement and Verification Protocol so savings stay comparable across projects. This page explains the methods; the how-tos train the models.
To build models, see [Create an M\&V model](/products/energ/energy-modeling/mv-models) and [Forecasting](/products/energ/energy-modeling/forecasting).
## What M\&V confirms
Measurement and Verification (M\&V) confirms that reported savings reflect real performance. It documents savings for audits, incentives, and internal governance.
EnerG trains a model on a baseline period, then compares later usage against the model's **adjusted baseline**: what the building would have used without the measure, under the same conditions.
## IPMVP options
IPMVP offers several options. EnerG supports two.
| Option | Boundary | When to use it |
| -------- | --------------------- | ------------------------------------------------------------ |
| Option B | Meter or system level | A measure targets specific equipment with dedicated metering |
| Option C | Whole building | A retrofit or program is best measured at billing boundaries |
Option B isolates a single system when metering separation is strong. Option C reads the whole-building signal when boundary-level bills represent performance.
## Meter types for modeling
You build models from two meter types.
* **Utility bill meters** provide monthly granularity from billing periods.
* **Interval meters** provide hourly or 15-minute granularity for finer load shapes.
Pick the meter type that matches your data and the savings story you need to tell. Interval models capture time-of-day patterns that monthly bills cannot show.
## Drivers and model fit
Models regress energy against drivers so you separate real savings from noise. Common drivers include weather (degree days), occupancy, and interval-of-week patterns.
EnerG reports fit with three accuracy metrics.
| Metric | What it measures |
| --------- | ------------------------------------- |
| R-squared | Share of variation the model explains |
| CV-RMSE | Scatter of predictions around actuals |
| MAPE | Average percent error |
A model that falls below your thresholds flags as not acceptable. Widen the baseline, adjust variables, or improve meter completeness before you rely on it. See [Calculations and parameters](/products/energ/reference/calculations-and-parameters#model-accuracy) for typical thresholds.
## Forecasting
Forecasting projects future consumption from trained models. It extends past performance forward using the drivers you trained, such as weather.
You link forecasts to capital planning measures so projected use supports funding decisions and scenario work.
## Next steps
Configure meters, baselines, and drivers to train a savings model.
Read savings, charts, and accuracy metrics after a run completes.
Project one to 12 months of consumption from trained models.
Review how baseline windows and weather models set the reference.