AEE CMVP cheat sheet

Adjustments normalize savings for changes in conditions (weather, occupancy, production) independent of the ECM.

Routine adjustments: expected and predictable changes (e.g., weather, production volume) handled via a model.

Non-routine adjustments: unexpected or one-time changes (e.g., building addition, equipment replacement) handled ad hoc.

The adjusted baseline is the baseline energy use modified by both routine and non-routine adjustments.

Savings = (Baseline Energy − Reporting-Period Energy) ± Adjustments.

IPMVP requires that the baseline period be representative of normal operating conditions and that adjustments be based on measured or documented data.

Adjustment methods include: fixed-factor, algebraic, or model-based (e.g., regression) approaches.

Uncertainty of adjustments should be quantified and reported.

Four IPMVP M&V options: A (Retrofit Isolation: Key Parameter Measurement), B (Retrofit Isolation: All Parameter Measurement), C (Whole Facility), D (Calibrated Simulation).

Option A: Some parameters measured, others estimated; estimates must be justified and documented.

Option B: All parameters affecting energy use are measured continuously or periodically.

Option C: Whole-building utility meter data used; best for multi-ECM projects or when savings >10% of total baseline.

Option D: Calibrated simulation used when baseline data are unavailable or for new construction.

Selection depends on: project size, complexity, data availability, and required accuracy.

Each option has different cost and accuracy trade-offs; higher accuracy generally costs more.

M&V costs typically 1–10% of project cost; should be proportional to savings uncertainty.

Options A and B focus on individual ECMs, isolating their energy use from the rest of the facility.

Option A (Key Parameter Measurement): measures parameters with significant impact or high variability; estimates others.

Option B (All Parameter Measurement): measures all parameters affecting the ECM's energy use, often with dedicated meters.

Use when ECM savings are large relative to whole-building use, or when ECMs are independent and can be isolated.

Baseline and reporting-period measurements must be taken at the same boundaries and with same metrics.

Option A estimates must be based on evidence (e.g., manufacturer data, spot measurements) and documented.

Option B is more accurate but costlier; suitable for complex or high-uncertainty ECMs.

Common applications: lighting retrofits, variable frequency drives, HVAC equipment replacement.

Option C uses whole-building utility meters (electric, gas, etc.) to measure total facility energy use.

Best when savings are >10% of baseline whole-facility energy use and multiple ECMs are implemented.

Requires continuous utility data for baseline and reporting periods; often uses regression models with weather.

Less expensive per ECM but cannot isolate individual ECM savings without additional sub-metering.

Adjustments use whole-building regression models (e.g., change-point models) to account for weather and other variables.

Uncertainty analysis is critical; savings must be statistically significant above model uncertainty.

Not suitable for small savings (<5-10%) due to meter and model noise.

Can be combined with Option A/B for verification of individual ECMs.

M&V Plan is a document that outlines the M&V approach, responsibilities, data collection, and reporting.

Key elements: project description, ECM list, M&V option selection, baseline definition, measurement boundaries, adjustment method, data management, reporting schedule, budget.

Develop M&V Plan early, ideally during project design, to ensure data collection and metering are included.

The plan should specify the baseline period, reporting period, and frequency of measurements.

Stakeholders (owner, ESCO, verifier) should agree on the plan before implementation.

The plan should include a description of the uncertainty analysis and how savings will be verified.

Update the plan if significant changes occur during the project.

M&V Plan is often a contractual requirement in performance contracts.

Savings reports must include: baseline energy, reporting-period energy, adjustments, savings, and uncertainty.

Report both energy and cost savings; specify units and currency.

Clearly state the M&V option used and any deviations from the M&V Plan.

Include a discussion of non-routine adjustments and their impact.

Present results in tables and graphs for clarity; include raw data or references.

Report uncertainty at a stated confidence level (e.g., 90%).

Include a comparison with expected savings and explain discrepancies.

Reports should be transparent, reproducible, and follow a consistent format.

Meter accuracy: Class 1 (high accuracy) vs. Class 2 (lower); choose based on required precision.

Metering duration and interval: continuous (15-min intervals) vs. periodic (monthly) based on ECM variability.

Install meters to measure only the ECM's energy use; avoid including unrelated loads.

Calibrate meters regularly; document calibration dates and results.

Consider temporary vs. permanent metering; temporary for short-term, permanent for continuous verification.

Data quality: check for gaps, outliers, and meter drift; use data validation procedures.

Metering cost should be balanced against the value of accuracy; over-metering can be uneconomical.

For Option A, spot measurements or short-term monitoring may suffice for estimated parameters.

Regression models relate energy use to independent variables (e.g., weather, occupancy, production).

Common models: change-point (heating/cooling), polynomial, and time-of-week models.

Model selection based on data patterns and goodness-of-fit (R², CV(RMSE), t-statistics).

CV(RMSE) < 15% (or 10% for Option C) is often used as a calibration criterion.

Use separate models for baseline and reporting periods if conditions changed significantly.

Avoid extrapolation beyond the range of baseline data; use non-routine adjustments if needed.

Model uncertainty should be propagated to savings uncertainty.

Validate models with independent data (e.g., hold-out sample) when possible.