What Is an EnPI? How to Set Energy Performance Indicators

An energy performance indicator (EnPI) is a value or model that expresses the energy performance of a facility, system or piece of equipment in measurable terms. Total kWh alone does not describe performance: when production rises, the weather turns cold or occupancy increases, higher consumption does not mean lower efficiency. A well-chosen EnPI separates these effects and answers the question “are we using less energy under the same conditions?” Because the general framework of ISO 50001 is covered in a separate article, this one focuses on how to choose an EnPI, how to build its baseline and how to test the model. The main guidance document on the subject is ISO 50006.

EnPI and Energy Baseline (EnB) Work Together

An EnPI on its own is just a number; it needs a point of comparison to become meaningful. The energy baseline (EnB) quantifies energy performance during a selected reference period. The EnPI value in the reporting period is compared with the EnB, and the difference is interpreted as a change in energy performance. The baseline period should represent normal operating conditions and be long enough to capture seasonal effects; for most facilities this means at least one full annual operating cycle. Months that include breakdowns, long shutdowns or unusual production should be assessed with the reasoning documented.

Types of EnPI

ISO 50006 groups EnPIs into four types by level of complexity. Which one to choose depends on the number of variables that affect consumption and the quality of the available data.

Ratio indicators are easy to understand but carry an important trap: if consumption has a fixed base load component, kWh/ton rises when production falls, and performance appears to worsen even though efficiency has not changed. A regression model avoids this error because it separates base load as the constant term.

Setting an EnPI Step by Step

1. Identify the user and the purpose: Senior management needs a site-wide indicator, while the operations team needs indicators at system or equipment level.
2. Define the EnPI boundary: The site, building, process or equipment boundary must match the existing metering structure; no indicator can be built for a boundary that cannot be measured.
3. Identify significant energy uses: Consider separate indicators for chillers, compressed air, boilers and furnaces that stand out in the energy review.
4. List relevant variables: Identify factors that change routinely, such as production volume, degree days, occupancy and operating hours.
5. Separate static factors: Record conditions that normally do not change but affect the baseline when they do, such as floor area, shift pattern or product type.
6. Plan data collection: Define the source, frequency and responsible person for each variable; energy and variable data must cover the same time intervals.
7. Select the baseline period and calculate the EnB.
8. Test and document the model: Record goodness-of-fit statistics, coefficient significance and acceptance criteria.

Weather Normalization with Degree Days

In buildings with heating and cooling loads, outdoor temperature is one of the strongest variables. Using degree days instead of the monthly average temperature captures the non-linear relationship between temperature and energy demand more accurately:

• Heating degree days (HDD): HDD = Σ max(0, T_base,heating − T_avg,day)
• Cooling degree days (CDD): CDD = Σ max(0, T_avg,day − T_base,cooling)
• Multivariable model: E = b0 + b1·HDD + b2·CDD + b3·x3 (b0 is base load, x3 could be production or occupancy)

The base temperature depends on internal gains and operating setpoints of the building; the chosen value should be documented with its rationale and kept the same in the baseline and reporting periods. In buildings cooled with electricity, HDD and CDD together usually give meaningful results, while in buildings heated with natural gas HDD alone is often sufficient.

Model Adequacy: R² and Beyond

The coefficient of determination, R², shows how much of the variation in consumption is explained by the variables in the model and takes a value between zero and one. A high R² alone does not make a good model; the following checks should be done together:

• Is the number of observations sufficient for the number of variables? Building a model with many variables on few months produces a result that fits the past very closely but cannot explain the future.
• Are the coefficient signs physically meaningful? A negative HDD coefficient in a heated building points to a data or boundary error.
• Are the residuals (actual minus model) randomly distributed, or is there a systematic deviation in certain months?
• Was the acceptance threshold set in advance and written into the energy management system documentation?

Once the model is accepted, actual variable values from the reporting period are fed into it. The difference Expected consumption − Actual consumption gives the change in performance with conditions accounted for. A CUSUM chart, which plots the cumulative sum of these differences, is a common method that makes it easier to see from which month performance started to change. When static factors change, for example a new building wing, a product change or a new measurement boundary, the EnB should be adjusted with the reasoning documented.

Calculating and Tracking EnPIs with 6S Teknoloji

The ISO 50001 module in 6S Teknoloji supports the calculation side of the steps above. Monthly values are entered for variables such as number of employees, floor area, HDD, CDD, occupancy rate, production volume, patient days and number of students; monthly HDD and CDD values can be filled in automatically from weather archive data, using a base temperature of 18 °C for heating and 22 °C for cooling. The user selects which variables enter the model; the platform fits an ordinary least squares regression to monthly electricity consumption from the main energy source and shows the equation, the R² value, the number of observations and a chart comparing actual and modeled monthly consumption. Models with an R² of 0.75 or higher are marked as adequate; building a model requires at least two more monthly observations than the number of selected variables. The Excel report lists actual and reference monthly consumption together with the deviation between them.

In addition, electricity consumption is normalized as kWh/(HDD+CDD) and natural gas consumption entered from invoices as m³/HDD, and facilities in a portfolio can be ranked by normalized consumption or by kWh/m². A weather sensitivity analysis that links daily consumption to HDD and CDD shows base load and degree-day slopes separately. Investment and monthly return records for efficiency projects are entered by the user; the platform does not calculate those amounts itself. Choosing the EnPI boundary, the baseline period and the acceptance criteria is the decision of the energy manager; 6S Teknoloji provides the data and calculations needed to put those decisions into practice.