ISO 50001 energy management and the data it needs

ISO 50001 is the international standard for energy management systems. It sets out how an organisation runs energy management as a continuous process: an energy policy, a review of where energy goes, objectives and action plans, then measurement to show that energy performance improves. The current edition is ISO 50001:2018, with a 2024 amendment on climate change. ISO 50006:2023 gives the guidance on indicators and baselines, and ISO 50003:2021 sets the rules for the certification audit. The software side, an energy management information system, is covered in the IEMS guide.

This guide is part of the buildings and process performance learning path.

Why organisations adopt it

Article 11 of the recast EU Energy Efficiency Directive requires an enterprise whose average annual energy consumption over the previous three years exceeds 85 TJ (about 23.6 GWh) to have a certified energy management system by 11 October 2027. Between 10 TJ and 85 TJ, an enterprise without an energy management system must have an energy audit every four years, so an energy management system takes the place of the audit cycle. In the UK, an ISO 50001 certificate that covers all of the organisation's energy use is a complete route to ESOS compliance. A certificate that covers only some sites leaves the rest to be assessed in the usual way. The EED guide and the ESOS guide give the detail.

The US Department of Energy reports that organisations which implemented the original ISO 50001 achieved validated energy performance improvements averaging 4.5% a year. The sites chose to join a validation programme, so the figure is subject to selection bias.

In Ireland, SEAI's EXEED grant funds capital projects designed and managed to IS 399, the Irish standard for energy-efficient design. ISO 50001 clause 8.2 requires energy performance to be considered in the design of new and renovated plant. A large SEU project from the action plan can qualify for EXEED if it is designed and managed to IS 399.

The plan-do-check-act cycle and the clauses

ISO 50001:2018 uses the common clause structure of ISO management system standards. Use the table as a map from each stage of the cycle to the clause an auditor will quote.

The energy review and significant energy uses

The energy review (clause 6.3) analyses energy use and consumption from measurement and other data. It identifies the significant energy uses (SEUs): the uses that account for a substantial part of consumption or offer considerable potential for improvement, by criteria the organisation sets. For each SEU the organisation identifies the relevant variables (production, weather, occupancy), the current energy performance and the people whose work affects it.

The standard does not set the criteria. Many organisations rank the uses by consumption and take those that together make up about 80% of the total, which is a Pareto cut-off. They then add any smaller use with a known improvement opportunity. Write the criteria down, because the auditor asks for them.

Without submetering, the ranking is built from nameplate kW multiplied by run hours. That estimate fails in predictable ways:

• A motor nameplate gives shaft output, not electrical input, and most motors run below rated load.
• A load/unload screw compressor runs for all of its hours but is not loaded for all of them. Unloaded, it still draws 15 to 35% of full-load power, according to the Compressed Air Challenge.
• A fan on a variable speed drive at 70% speed draws about 34% of full power in theory, by the fan affinity law (0.7³). Drive losses and fixed system pressure make the real figure higher.
• A chiller's input power changes with load and condenser temperature, so its summer and winter consumption differ.

Each of these can move a use up or down the ranking. A few weeks of metering on the candidate uses is enough to rank them. A baseline needs more: a full production cycle and, for a load that depends on the weather, a full heating and cooling season.

EnPIs and energy baselines

An energy performance indicator (EnPI, clause 6.4) measures energy performance: kWh per tonne, kWh per square metre, compressed air specific power in kW per m³/min. An energy baseline (EnB, clause 6.5) is the reference against which the EnPI is compared. ISO 50006 recommends normalising for the relevant variables, so that a change in production or weather does not look like a change in efficiency.

A simple ratio fails when the SEU has a fixed load. Take a food plant with this monthly electricity model, fitted by regression to twelve baseline months:

E = 180,000 kWh + 310 kWh/t × P + 95 kWh/HDD × HDD

E is the monthly electricity use, P is production in tonnes and HDD is heating degree-days. The 180,000 kWh intercept is the load that runs whatever the output: refrigeration standing losses, lighting, ventilation and air leaks. HDD is in the model because this plant heats its offices and dispatch area with heat pumps. A plant heated by gas would test cooling degree-days instead, because refrigeration work rises in warm weather.

The simple EnPI rose from 634 to 658 kWh/t, which looks 3.8% worse. Two effects cause that. The 180,000 kWh fixed load is spread over 520 tonnes instead of 600, which is 346 kWh/t instead of 300. The 110 extra degree-days add 10,450 kWh. With no change on site, the model therefore expects 704 kWh/t. The normalised EnPI is actual divided by expected: 342,000 ÷ 365,900 = 0.935. The plant used 6.5% less energy than its baseline predicts, a difference of 23,900 kWh.

Checking the model

Check the model before you use it. The FEMP M&V guide quotes the IPMVP tests: each variable needs a t-statistic above 2.0, and the model needs an R² of at least 0.75. Leave out a variable that fails the t-test.

R² does not tell you whether one month's result is real. The root mean squared error (RMSE) of the fit does. Suppose the example model has an RMSE of 9,500 kWh over baseline months that average 380,000 kWh. Its CV(RMSE) is then 2.5%. With twelve months and three coefficients, the model has 9 degrees of freedom, and the 95% t-value is 2.26. The 95% prediction interval for a single month is about ±2.26 × 9,500 × √(1 + 1/12) = ±22,400 kWh. The 23,900 kWh saving is only just outside it. Over three months with independent errors the interval grows by √3, to about ±39,000 kWh, while a steady saving grows by 3. Three months at the same rate give about 71,700 kWh, which is well outside the interval. A reporting month far from the baseline months' average production or weather has a wider interval than this.

Tracking the result with CUSUM

Plot the cumulative sum (CUSUM) of actual minus expected consumption, month by month. While the site runs as it did in the baseline period, the monthly differences fall on both sides of zero and the CUSUM stays near zero. A permanent change shows as a change of slope.

The slope changes from about zero to about −23,400 kWh a month in the month of the project. A slope that flattens again later shows that the saving has stopped, for example because the controls were overridden. The savings guide works through the statistics and the uncertainty of the result.

Revising or adjusting the baseline

Clause 6.5 requires the baseline to be revised when the EnPIs no longer reflect energy performance, when static factors change a lot, or by a method the organisation set in advance. A new production line, a building extension or a change of shift pattern are static-factor changes.

A change that adds a separate, measurable load can be handled by adjusting the baseline instead of fitting it again. The FEMP guide calls this a non-routine adjustment. For example, the food plant adds a 25 kW aeration blower for wastewater treatment that runs continuously. It is not an efficiency measure, so it must not appear as a loss. Meter the blower, and add its metered consumption (about 18,250 kWh in a 730-hour month) to the expected value for each month after it starts. Use the metered value, not the nameplate kW.

Fit the model again when the change alters the relationship with the variables. A new line with a different kWh per tonne, or a new product mix, changes the production coefficient, so an adjustment to the intercept is wrong. Collect new data for the new baseline. In both cases, record the reason, the date, the method and the data used.

The compressed air specific power, chiller COP and heat pump COP calculators give EnPIs for those SEUs.

The data collection plan

Clause 6.6 requires a plan for the data needed to monitor energy performance. It covers at least:

• the relevant variables for each SEU;
• the energy consumption of each SEU and of the organisation;
• the operating criteria for SEUs;
• static factors, where applicable;
• the data specified in the action plans.

Clause 9.1.1 then requires the organisation to monitor its EnPIs, the operation of SEUs and actual against expected energy consumption, and to investigate and respond to significant deviations. The expected consumption comes from the baseline model, so a deviation check needs the energy data and the variable data for the same period.

Meter electricity at 15-minute intervals or finer, so that out-of-hours and base load are visible. Record production in the same periods as energy. A meter day that starts at midnight and a production day that starts at 06:00 put one shift's tonnes against another shift's kWh. Synchronise the clocks, record gaps as gaps, and never fill them with zeros.

Measurement must be accurate and repeatable, and the organisation must keep records of calibration or verification. A Class 0.5S meter under IEC 62053-22 is within ±0.5% of reading from 5% of nominal current upwards, and within ±1% down to 1%. Current transformers fitted on site add their own error, unless the class covers the meter and its current sensors together. In the worked example, a 1% metering error is about 3,700 kWh a month. The model's ±22,400 kWh prediction interval is six times larger. The high-accuracy metering guide explains accuracy classes.

Setting SEU operating criteria

Clause 8.1 requires criteria for the effective operation of each SEU, and clause 9.1.1 requires the organisation to monitor them. Set the criteria from metered data, not from a guess.

Compressed air is the usual first example. Meter the compressor feeder for several weekends when no production air is needed, after a leak survey and repair. Suppose the out-of-production load settles at 24 kW, for air-operated valves and the remaining leaks. Set the criterion at 30 kW, about 25% above that base, for longer than 30 minutes between Friday 22:00 and Monday 06:00. A weekend at 45 kW is 21 kW above the base for 56 hours, which is 1,176 kWh. Over a year of weekends that is about 61,000 kWh. Repeat the base-load measurement after each leak survey, and tighten the limit if the base falls.

Getting ready for certification

1. Collect twelve months of bills for every energy carrier, and reconcile them with any site meters.
2. Meter the likely SEUs and their relevant variables at 15-minute intervals or finer, for at least one production cycle and a heating and cooling season where the weather matters.
3. Define the EnPIs and baselines. Write down the normalisation method, the model statistics, and the rules for revision and adjustment.
4. Set operating criteria for each SEU and alert on them.
5. Complete at least one internal audit and one management review, and keep enough months of EnPI data to show improvement against the baseline.

Certification follows ISO/IEC 17021-1 and ISO 50003. The Stage 1 audit reviews the documented system: the energy review, the SEU criteria, the EnPIs, the baselines and the data collection plan. The Stage 2 audit checks that the system works in practice. ISO 50003:2021 requires the audit team to confirm continual energy performance improvement before the certificate is issued, and again at recertification. A system that is complete on paper but has no data showing improvement does not pass.

ISO 50003 does not set a minimum number of months of reporting data. The evidence must show an improvement larger than the model's uncertainty. In the worked example, one month is marginal and three months are clear. Agree the reporting period with the certification body at Stage 1. The certificate runs for three years, with a surveillance audit every year. At surveillance the organisation must show that improvement actions are being implemented.

ISO 50001 data with EpiSensor

ZEM electricity monitors measure electricity at Class 0.5S for the meter and its current sensors together, and report over Zigbee to the Gateway. A ZEM on the compressor feeder gives the kW for a specific power EnPI and for the out-of-production criterion. ZPC pulse counters read the pulse outputs of gas, water and flow meters. The flow meter on the compressed air header gives the m³/min. ZHM M-Bus interfaces read heat meters. TES temperature sensors measure outside air for degree-days and space or process temperatures. ZIO analogue sensors read 4 to 20 mA transmitters such as a header pressure sensor.

A Gateway running Edge stores the data on site. An Edge calculated device can compute a ratio EnPI, such as specific power or COP, on a clock-aligned interval. An automation rule with a time window and a duration condition can raise an alert or send an email when the compressor exceeds its out-of-production limit. The Data page exports the loaded series as CSV, with a blank cell for a missing reading instead of a zero. Monthly energy and variable data can then go into the regression and the EnPI records without gaps being counted as savings.