How To Validate Energy Efficiency Improvements With Regression Analysis

Published: September 14, 2022.

Northmore Gordon

Regression analysis is essential because it decouples the impacts of site changes, such as seasonal or production changes, from the energy savings resulting from an Energy Conservation Measure (ECM). Without regression analysis, there is no way to independently validate claimed savings and ensure business outcomes were achieved. Energy consumption is often impacted by a complex interaction of multiple factors, so solely examining energy consumption over time will paint a very narrow picture. Regression analysis provides a scientific and thorough validation method that establishes what energy use would have been without the intervention under comparable conditions - the central question in measurement and verification.

Energy efficiency improvements offer businesses opportunities to reduce electricity use, reduce fuel demand, and save on operational expenses. Standardized best practices are available to effectively implement Energy Conservation Measures (ECMs) and provide stakeholders with confidence that the ECM is delivering verifiable energy and cost savings.

Having a sound strategy to verify ECM results is paramount to the overall success of the project. The International Performance Measurement and Verification Protocol (IPMVP) and other related frameworks stress the use of regression analysis as the preferred option. In fact, this approach is essential in order to decouple the impacts of changes at the site, including seasonable or production changes, from the energy savings from resulting from the ECM. Without regression analysis there is no way to independently validate the claimed savings and to ensure the business outcomes were achieved.

Regression analysis is a powerful and flexible tool to gauge the impact of various operational and environmental factors (independent variables) on a facility's energy consumption. It is a valuable exercise to reliably communicate the effectiveness of an ECM. Doing so can help build momentum and support for future energy efficiency projects.

As with any useful analysis, it is critical to start with reliable data and to understand the level of accuracy required. The process for regression analysis can be laid out in six steps.

The Steps To Developing Regression Model

First, all independent variables must be identified. With regard to verifying ECMs, these are parameters that are understood to influence the facility's energy consumption. All the variables can be included into a single model, so that the pre- and post-performance can be compared. Common independent variables included in energy regression models include:

  • ambient temperature
  • plant production
  • facility occupants
  • heating degree days
  • cooling degree days

The next step is to collect data. It is impossible to determine the extent to which something is absent without having a solid reference to how much was present before. For energy conservation, this means establishing a baseline period for a complete operating cycle, sometimes a full calendar year, immediately before ECM implementation. As the reference point to gauge ECM effectiveness, the baseline period must be representative. Anomalies and apparent outliers in the data should be addressed to ensure only valid data points are used.

Graphing the data to get a visual representation of the independent variable impact on energy consumption can be useful if a single variable is responsible for the majority of the variation in energy use, however, if there are multiple variables a graph is less useful.

Performing a regression analysis establishes a model, or algorithm for the relationship between energy consumption and the variables. In some cases, a simple regression will suffice. In many cases, however, there are multiple independent variables that influence energy consumption. In these situations, more complex multiple regression models will need to be developed to accurately depict the dynamic effect that the variables have on energy usage.

The final step is to ensure the effectiveness of the regression model. Several statistical tests can help determine if the model is a fair and accurate representation of the system. While not an exhaustive list, the fundamental tests to validate a regression model are R-squared t-statistics and residuals.

Planning Out Your Next Efficiency Project

Energy consumption is often impacted by a complex interaction of multiple factors. Therefore, solely examining energy consumption over time will paint a very narrow picture. It is important that the effect of an ECM is validated in a scientific and thorough manner.

Northmore Gordon uses a regression analysis approach based on the IPMVP to provide a statistically valid measure of the savings from an energy efficiency project or group of projects. Using this approach, we can also give the statistically verifiable likelihood of the savings results being in between upper and lower bounds. Additionally, we use regression analysis to create energy efficiency and emission reduction certificates that can generate additional revenue for businesses.

From inception to execution, Northmore Gordon supports commercial and industrial businesses to implement ECMs and realize savings through proven decarbonisation strategies. Energy efficiency improvements are cornerstone measures in accelerating toward net zero emissions - and every ECM project should include a plan to validate the project's effectiveness.

Using this guidance

Check the site's operating conditions, data quality, project stage and current program settings before turning the dated discussion into a scope or investment decision.

Why is regression analysis important for validating energy efficiency improvements (ECMs)?

Regression analysis is essential because it decouples the impacts of site changes, such as seasonal or production changes, from the energy savings resulting from an Energy Conservation Measure (ECM). Without regression analysis, there is no way to independently validate claimed savings and ensure business outcomes were achieved. Energy consumption is often impacted by a complex interaction of multiple factors, so solely examining energy consumption over time will paint a very narrow picture. Regression analysis provides a scientific and thorough validation method that establishes what energy use would have been without the intervention under comparable conditions—the central question in measurement and verification.

Which framework recommends regression analysis as the preferred approach for verifying ECM results?

The International Performance Measurement and Verification Protocol (IPMVP) and related frameworks stress the use of regression analysis as the preferred option for verifying Energy Conservation Measure (ECM) results. We use a regression analysis approach based on the IPMVP to provide a statistically valid measure of the savings from an energy efficiency project or group of projects, including giving the statistically verifiable likelihood of savings results falling between upper and lower bounds. This approach aligns with the practical M&V brief structure that separates engineering estimates from verified energy results and ensures accountability in energy management and sustainability initiatives.

What are the six steps for developing a regression model to validate energy efficiency improvements?

The process involves: (1) identifying all independent variables that influence the facility's energy consumption; (2) collecting data, including establishing a representative baseline period for a complete operating cycle (sometimes a full calendar year) before ECM implementation; (3) addressing anomalies and outliers to ensure valid data points; (4) graphing the data to visualize the impact of independent variables; (5) performing a regression analysis to establish a model or algorithm for the relationship between energy consumption and the variables; and (6) ensuring the effectiveness of the regression model using statistical tests such as R-squared, t-statistics, and residuals analysis.

What common independent variables are included in energy regression models, and why do multiple variables require different modeling approaches?

Common independent variables in energy regression models include ambient temperature, plant production, facility occupants, heating degree days, and cooling degree days. A simple regression will suffice when a single variable is responsible for the majority of the variation in energy use. However, in many cases there are multiple independent variables that influence energy consumption, requiring more complex multiple regression models to accurately depict the dynamic effect that the variables have on energy usage. Graphing data alone is useful only when one variable dominates; with multiple variables, regression analysis becomes necessary to separate their individual contributions to total energy consumption.

Why is establishing a representative baseline period before ECM implementation critical for regression analysis?

A representative baseline period for a complete operating cycle—sometimes a full calendar year—is essential because it captures normal variations in production, weather and occupancy before the ECM is implemented. Without this complete cycle, the baseline may not reflect actual operating conditions, leading to inaccurate comparison and verification of savings. The baseline establishes the reference standard against which the ECM's impact is measured, and this foundational accuracy directly affects whether the regression model can reliably isolate the energy savings from the ECM itself rather than from changes in production volume, seasonal conditions or other operating variables.

What statistical tests validate a regression model's effectiveness, and what do they indicate?

Fundamental tests to validate a regression model are R-squared, t-statistics, and residuals. R-squared measures how well the model explains the variation in energy consumption; t-statistics assess the significance of individual independent variables; and residuals analysis identifies whether the model's predictions deviate from actual observations in a systematic or random way. These tests together establish whether the regression model reliably captures the relationship between energy consumption and the influencing variables, and whether the predicted energy baseline—and therefore the calculated savings—can be trusted as statistically valid. Strong statistical validation is necessary before accepting savings for reporting, certificates, or investment decisions.

How does Northmore Gordon integrate regression analysis into its measurement and verification services?

We use regression analysis based on the IPMVP to provide a statistically valid measure of the savings from an energy efficiency project or group of projects. Our approach includes project-boundary selection, baseline modelling, savings calculation and reporting that establish credible, defensible outcomes. We also use regression analysis to create energy efficiency and emission reduction certificates that can generate additional revenue for businesses. The regression model produces not only a point estimate of savings but also the statistically verifiable likelihood of results falling between upper and lower bounds, supporting both internal performance tracking and formal certificate claims.

What is the practical relationship between measurement and verification planning and the success of regression analysis?

Early measurement planning is essential because regression analysis requires quality data collected systematically before and after ECM implementation. If metering and data collection are not planned before installation begins, baseline establishment and post-project monitoring may become difficult or impossible. The M&V brief specifies instruments, intervals, quality checks and responsibilities for data access to ensure the data exists and is structured appropriately for regression modeling. We recommend that measurement and verification approaches be considered during project scoping—before commitments are finalized—to preserve evidence and enable statistically robust validation of actual outcomes against baseline conditions.

Images on This Page