Weather normalization in measurement and verification (M&V) is the process of adjusting energy consumption data to account for variations in weather. Because weather is one of the most significant drivers of facility energy use, normalizing this data ensures that your energy savings calculations are accurate and verifiable. This normalization is done through a multivariate regression model.

How Weather Normalization Works in M&V

In the regression model, you are able to account for specific weather metrics to quantify a direct relationship between your facility’s energy consumption (kWh, KW, therms, etc.) and the outside temperature.

Heating and Cooling Degree Days (HDDs and CDDs)

The most common metrics used are Heating Degree Days (HDDs) and Cooling Degree Days (CDDs).

For example, consider a chiller project that primarily cools a building. This system would naturally have much higher energy consumption in the summer months compared to the winter months. By factoring in CDDs, the model accounts for this expected seasonal spike so it isn’t mistakenly calculated as a loss of energy efficiency.

Two Approaches to Normalizing Consumption Data

When normalizing consumption data for weather, there are two primary modeling approaches you can take depending on your project requirements and utility program standards.

Approach 1: The Forecast Model

In this approach, you first build a baseline model. Then, you predict what the facility’s consumption would have been without the energy conservation measure (ECM) by applying that baseline model to the reporting period conditions (the period after the ECM is installed).

Essentially, you are using actual weather data from the reporting period to calculate: “If we did not install this ECM, here is what the consumption would have been.” This is known as a forecast model.

Approach 2: The Normalized Operating Conditions Model

The second option is to use a normalized operating conditions model. This approach is commonly used in Normalized Metered Energy Consumption (NMEC) utility programs.

Instead of just one model, this approach develops two distinct models:

• One for the baseline period.
• One for the reporting period.

You then take the results of each model and simulate them under a standardized, average weather condition (for example, TMY3 or TMYx weather data files). Essentially, you are using a standardized average temperature to evaluate both models on a completely level playing field.