Organizational Energy Use and Performance

Introduction

Energy management begins with a defensible understanding of how an organization uses energy, when it uses energy, and which operational conditions explain changes in consumption. Utility bills alone rarely provide that understanding. They show aggregated energy purchased, but not the equipment, processes, buildings, weather conditions, production volumes, occupancy patterns, or management practices responsible for the result.

Organizational energy performance therefore requires a structured measurement system. The central tasks are to identify significant energy uses, collect reliable data, establish a baseline, define energy performance indicators, normalize consumption for relevant variables, and compare performance with suitable internal or external benchmarks. These tasks support both operational control and strategic decisions such as retrofits, process redesign, procurement, maintenance, and investment appraisal.

The analysis applies across manufacturing, commercial buildings, hospitals, universities, logistics operations, and public-sector organizations. Although the equipment differs, the logic is similar: energy is an input to activities that produce outputs. Performance improves when the organization can distinguish avoidable energy use from energy use that is inherent in its service or production requirements.

Learning objectives

By the end of this chapter, the reader should be able to:

1. Map energy flows and identify significant energy uses.
2. Distinguish energy consumption, demand, load, and performance.
3. Establish a credible energy baseline and reporting boundary.
4. Select appropriate energy performance indicators (EnPIs).
5. Normalize energy data for production, weather, occupancy, operating hours, and other relevant variables.
6. Construct and interpret load profiles and energy-intensity metrics.
7. Apply benchmarking without drawing misleading conclusions.
8. Evaluate data quality, uncertainty, governance, and practical limitations.

1. From energy bills to an organizational energy model

An organization can be represented as a system that receives energy inputs, transforms them through equipment and processes, and produces goods or services. Energy may enter as electricity, natural gas, steam, district heating, fuel oil, biomass, or vehicle fuel. It may be transformed into heat, motion, pressure, lighting, cooling, information services, or transportation.

A useful first distinction is between energy consumption and energy demand:

• Energy consumption is the quantity of energy used over an interval, commonly measured in kWh, MWh, GJ, or litres of fuel.
• Energy demand is the rate at which energy is used at a particular time, commonly measured in kW or MW.
• Load is the time-varying demand profile of a system or facility.
• Energy performance is the measurable result related to energy efficiency, energy use, and energy consumption, considered in relation to relevant variables and conditions.

If a facility uses 120,000 kWh during a month, that is consumption. If its highest fifteen-minute electrical demand during the month is 850 kW, that is peak demand. The two measures are related but not interchangeable. A facility may reduce total kWh while retaining a high peak, or reduce peak demand while using approximately the same monthly energy.

The energy model should define:

• organizational and physical boundaries;
• energy sources and carriers;
• major consuming systems and processes;
• energy flows between systems;
• relevant variables affecting consumption;
• measurement points and data owners; and
• the outputs or services delivered.

A boundary must be explicit. For example, a university may include campus buildings but exclude leased residences; a manufacturer may include production halls but exclude third-party transport. Changing the boundary without adjustment can create an apparent improvement or deterioration that is only an accounting change.

2. Mapping energy use and identifying significant energy uses

2.1 Energy-use mapping

Energy-use mapping is a structured representation of where energy enters, is distributed, transformed, and consumed. At a simple level, an energy balance can be written as:

Here, E_in is energy entering the defined boundary during a period; E_useful is energy delivered to intended services or processes; E_losses is energy dissipated or otherwise lost; E_exported is energy leaving the boundary as a product, exported electricity, steam, or other energy carrier; and E_change represents the change in stored energy within the boundary. For a steady reporting period with negligible storage change, E_change is approximately zero. The equation is an accounting identity, not a claim that every term can be measured with equal precision.

A practical map commonly follows this sequence:

1. List all purchased and generated energy sources.
2. Convert quantities to consistent reporting units where aggregation is required.
3. Locate meters, submeters, fuel records, and process measurements.
4. Allocate energy to buildings, production lines, transport, utilities, and support functions.
5. Identify conversion losses, rejected heat, leaks, and unmetered uses.
6. Link each use to an operational output or service.

Typical energy-use categories include heating, ventilation and air conditioning (HVAC), refrigeration, compressed air, pumping, motors and drives, process heat, steam, drying, lighting, information technology, domestic hot water, fleet transport, and standby equipment.

2.2 Significant energy uses

A significant energy use (SEU) is an energy use that consumes substantial energy or offers substantial potential for improved energy performance. Significance should not be based solely on annual consumption. A small use may be strategically important if it creates safety risks, causes a large demand peak, limits production, or has a highly attractive improvement opportunity.

Organizations may rank uses using a transparent scoring method. For example, each use can be assessed for annual energy share, improvement potential, cost exposure, operational risk, and strategic importance. The resulting score is a management prioritization tool, not a universal scientific threshold. The criteria and weights should be documented and reviewed.

A sound SEU register records:

2.3 Measurement hierarchy

Measurement should be designed around decisions. Whole-site meters establish total consumption; submeters reveal major systems; temporary instruments diagnose particular equipment. Common instruments include electricity meters, thermal energy meters, flow meters, temperature sensors, pressure sensors, occupancy counters, and production counters.

Meter data should be checked for missing intervals, implausible zero values, reversed readings, unit changes, meter replacement effects, and time-zone or daylight-saving inconsistencies. Estimated data should be labelled as estimated rather than presented with false precision.

3. Establishing an energy baseline

An energy baseline is a quantitative reference that permits comparison of energy performance over time or against a defined alternative. It may be a historical period, a model of expected energy use, a production-normalized relationship, or a combination of these.

3.1 Selecting the baseline period

A baseline period should be sufficiently representative of normal operations and should contain reliable data. Selection should consider:

• stable reporting boundaries;
• complete utility and production records;
• normal operating schedules;
• absence of exceptional shutdowns or commissioning periods;
• representative weather and occupancy;
• known process changes; and
• availability of data at the time interval needed for analysis.

A calendar year is convenient, but not automatically valid. A year with an unusually mild winter, an extended shutdown, or abnormal production may be a poor reference. A rolling baseline can be useful where operations change frequently, but it must be governed carefully so that poor performance is not hidden by continual replacement of the reference.

3.2 Baseline types

A static baseline uses a fixed reference period or fixed model. It is easy to communicate but may become less representative after major changes. A dynamic baseline estimates expected consumption under current conditions using relevant variables. It is more informative for changing operations but requires stronger data and modelling controls.

A baseline can be expressed as a simple ratio:

where B is the baseline consumption and E₀ is consumption during the selected reference period. This is appropriate only when comparison conditions are sufficiently similar.

For variable operations, a regression baseline may be used:

Here, Ē is predicted energy consumption for the reporting period; β₀ is the intercept; β_i are fitted coefficients; and X_i are relevant explanatory variables such as production volume, degree-days, occupancy, or operating hours. The model applies only within the range and operating conditions represented by the data. A high statistical fit does not prove causation or guarantee that the model is suitable for control decisions.

The baseline should be frozen, version-controlled, and accompanied by a record of boundary, data sources, assumptions, model form, calibration period, and known limitations. Major structural changes may justify baseline adjustment, but the reason and method must be documented.

4. Energy performance indicators

An energy performance indicator (EnPI) is a quantitative measure used to monitor and evaluate energy performance. A useful EnPI has a clear numerator, a meaningful denominator where appropriate, a defined boundary, a time basis, and a specified data source.

4.1 Common forms

Total consumption is useful for procurement and emissions accounting:

where E_t is total energy in period t, E_i,t is energy from carrier or source i, and m is the number of included sources. Aggregation requires a common energy unit and a stated conversion basis.

Energy intensity relates energy to an output or activity:

where I is energy intensity, E is energy consumed over the specified period, and Q is output or activity, such as tonnes of product, square metres served, patient-days, or passenger-kilometres. Units must be retained, for example kWh/tonne or kWh/m²·year.

A building-use intensity can be written as:

where EUI is energy use intensity, E_annual is annual energy consumption, and A is conditioned or gross floor area as explicitly defined. Comparing EUI requires consistent area definitions, climate conditions, occupancy, operating hours, and included energy carriers.

Percentage change from a reference period is:

where E₀ is reference consumption and E₁ is current consumption. A negative result indicates a reduction in absolute consumption. It does not necessarily indicate improved efficiency if output or operating hours also fell.

4.2 Choosing the denominator

The denominator should represent the service delivered. For a steel plant, tonnes of saleable steel may be more useful than tonnes of raw material. For a hospital, patient-days may be useful, but complexity and clinical activity may require additional variables. For a university, floor area alone may be inadequate if occupancy and laboratory intensity vary greatly.

A denominator can also create perverse incentives. If energy intensity is calculated per unit of output, an organization may appear less efficient during low production because fixed loads are spread over fewer units. Both total consumption and intensity should therefore be monitored, alongside operational context.

5. Normalization of energy consumption

Normalization adjusts energy data to make comparisons fairer by accounting for relevant differences in conditions. Common drivers include weather, production quantity, product mix, occupancy, operating hours, water temperature, and service demand.

5.1 Weather normalization

Heating and cooling loads often depend on outdoor temperature. Degree-days provide a simplified measure. Heating degree-days for a day may be represented as:

Cooling degree-days may be represented as:

where T_b is the selected base temperature and T_mean,d is the mean outdoor temperature on day d. The base temperature is not a universal constant; it depends on building characteristics, internal gains, controls, and climate. Degree-days are screening indicators, not complete thermal models.

A simple weather-adjusted model might be:

where E is energy consumption for the period, HDD and CDD are degree-day totals, Q is production or another activity variable, and the coefficients are estimated from historical data. The model should be tested for coefficient plausibility, residual patterns, multicollinearity, outliers, and stability.

5.2 Production and operational normalization

Suppose a factory uses electricity for both fixed services and variable production. A conceptual model is:

where E_fixed is fixed or standby energy, e_v is variable energy per unit of output, and Q is output. The model assumes a reasonably stable operating regime. If product mix, equipment efficiency, or shift pattern changes substantially, a single linear relationship may be misleading.

Normalization may be physical, statistical, or engineering-based:

• Physical normalization divides by a physical output or area.
• Statistical normalization estimates expected energy from explanatory variables.
• Engineering normalization uses equipment performance relationships, schedules, efficiencies, and process calculations.

5.3 Avoiding over-normalization

Not every available variable should be included. A variable belongs in a model when it is causally relevant, measured reliably, and useful for the decision. Overfitted models can explain historical noise but perform poorly in future periods. Normalization should not remove the effect of a genuine management improvement, nor should it conceal changes in operating practice.

6. Load profiles and demand analysis

A load profile is a time series of energy demand. If interval demand is recorded every Δt hours, energy over a period can be approximated by:

where E is energy in kWh when P_k is average demand in kW during interval k, ∆t is interval duration in hours, and n is the number of intervals. If demand is measured as an interval average, the relationship is direct; if only instantaneous samples are available, the estimate may be inaccurate.

A load profile can reveal:

• morning start-up peaks;
• overnight baseload;
• shift changes;
• coincident operation of large equipment;
• refrigeration or HVAC cycling;
• weekend and holiday patterns;
• abnormal loads caused by faults or controls; and
• opportunities for scheduling, storage, or demand response.

The load factor over a period is commonly defined as:

where LF is load factor, E is energy consumed, P_max is maximum demand during the period, and T is the duration of the period in hours. Load factor is dimensionless and lies between zero and one when the measurements and interval definitions are consistent. It indicates how continuously a system operates relative to its maximum demand; it does not by itself indicate efficiency.

For example, a facility consuming 240,000 kWh in a 30-day month with a peak demand of 500 kW has T = 30 × 24 = 720 hours. Its load factor is 240000 ÷ (500 × 720) = 0.667, or approximately 66.7%. A higher load factor may reduce demand-related costs, but deliberately increasing operation merely to raise load factor may increase total consumption.

7. Benchmarking energy performance

Benchmarking compares performance with a reference. Four useful forms are:

1. Historical benchmarking: current performance against the same organization in an earlier period.
2. Internal benchmarking: comparison among sites, departments, lines, or buildings.
3. Peer benchmarking: comparison with similar external organizations.
4. Best-practice benchmarking: comparison with a technically or operationally leading reference.

A benchmark is meaningful only when boundaries, climate, service level, output quality, occupancy, operating hours, energy carriers, and accounting conventions are sufficiently comparable. A hospital cannot be judged fairly against an office using only kWh/m². A production facility making a complex product should not be compared with a plant producing a standardized product without controlling for product mix.

Benchmarking should use distributions rather than a single supposedly ideal number where possible. Quartiles, control charts, and peer groups can identify unusual performance while acknowledging variation. Benchmarking identifies questions; it does not automatically identify causes or prescribe investments.

8. Worked application: an illustrative manufacturing case

Background

An illustrative food-processing plant operates 5,000 hours per year and produces 12,000 tonnes of product. It purchases electricity and natural gas. Electricity consumption is 6,000 MWh per year, while gas consumption is 18,000 GJ. The site has refrigeration, steam generation, compressed air, motors, lighting, and packaging equipment.

Problem

Management reports that total energy fell after a production decline, but it does not know whether energy performance improved. Demand charges are also high, and the plant experiences a sharp weekday morning peak.

Analysis

Electricity intensity is:

or 500 kWh/t. Gas intensity is:

These indicators provide a more useful starting point than total energy alone. Submetering shows that refrigeration represents 38% of electricity, compressed air 12%, packaging 15%, lighting 8%, and other uses 27%. Refrigeration and compressed air are therefore significant candidates because of their consumption and improvement potential.

The energy team plots fifteen-minute electricity demand. The profile shows refrigeration compressors, ovens, and packaging lines starting within the same thirty-minute period. A control review finds that two compressors operate at full capacity even when product throughput is low. A leak survey finds compressed-air losses during non-production hours.

The team establishes a production-normalized baseline using monthly output and cooling degree-days. The baseline predicts expected electricity for each month. After correcting for actual production and weather, the first six months after intervention show consumption 7.5% below expected baseline. The reduction is not claimed solely from the total bill because production and weather varied.

Approach and solution

The plant implements three measures:

• staggered start-up of packaging and refrigeration equipment;
• compressor sequencing based on suction pressure and actual cooling load; and
• repair of compressed-air leaks combined with an overnight isolation schedule.

The plant maintains a fixed baseline, records the intervention dates, verifies meter continuity, and reports both absolute MWh and normalized MWh per tonne. It also tracks maximum fifteen-minute demand.

Outcome

The interventions reduce normalized electricity intensity from 500 kWh/t to 462.5 kWh/t, a reduction of:

Peak demand falls by 14% because the largest loads no longer start simultaneously. The case illustrates why energy mapping, normalization, and load-profile analysis must be combined. The production-normalized result supports a stronger performance claim than a simple year-to-year comparison.

Learning points

1. A fall in total consumption may reflect reduced output rather than efficiency improvement.
2. Submetering links site results to actionable systems.
3. Demand and energy require separate indicators.
4. Operational controls can produce measurable savings without major capital expenditure.
5. Baseline governance is essential to credible verification.

9. Governance, ethics, and professional practice

Energy data influence capital allocation, public reporting, regulatory compliance, and sometimes employee performance assessment. Data governance should therefore specify ownership, access rights, retention, calculation methods, and correction procedures. Cybersecurity is increasingly important because connected meters and building-management systems may provide routes into operational technology.

Analysts must avoid selective reporting. A manager should not report a favourable energy-intensity metric while omitting reduced production, lower service quality, boundary changes, or increased use of another energy carrier. Estimates and uncertainty should be disclosed. When energy results support financial or environmental claims, the organization should preserve an audit trail from source meter to published indicator.

Artificial intelligence and advanced analytics can detect anomalies, forecast demand, classify load shapes, and optimize equipment schedules. However, models require representative data, explainable outputs, cybersecurity controls, and human oversight. An algorithm that reduces energy by compromising product quality, occupant comfort, safety, or clinical service is not an acceptable energy-management solution.

10. Challenges and limitations

Important limitations include:

• incomplete or poorly calibrated meters;
• inconsistent utility billing periods;
• changing production mix or building use;
• interaction between energy systems;
• rebound effects after efficiency improvements;
• non-linear equipment behaviour;
• small samples and unstable regression coefficients;
• uncertainty in floor area, occupancy, or output data; and
• benchmarks that ignore service quality or local conditions.

Energy analysis should therefore combine quantitative indicators with engineering inspection and operational knowledge. Statistical significance is not the same as operational significance, and a model that predicts well may still provide little insight into the mechanism of consumption.

11. Best-practice implementation sequence

A practical implementation sequence is:

1. Define organizational, physical, and reporting boundaries.
2. Catalogue energy carriers, meters, processes, and outputs.
3. Construct an energy-flow map and rank significant energy uses.
4. Validate meter data and establish data responsibilities.
5. Select a representative baseline period or baseline model.
6. Define a balanced set of EnPIs for consumption, intensity, demand, and service quality.
7. Normalize only for relevant and reliable variables.
8. Analyse interval load profiles and investigate abnormal patterns.
9. Benchmark against comparable internal or external references.
10. Review results with operators and revise controls or investments.
11. Verify outcomes against the baseline and document assumptions.
12. Reassess the system when boundaries, equipment, products, or operating conditions change.

Chapter Summary

Organizational energy performance is established through a chain of evidence rather than a single utility-bill number. Energy-use mapping identifies sources, systems, losses, outputs, and measurement points. Significant energy uses are selected according to consumption, improvement potential, risk, and strategic importance. A baseline provides a controlled reference, while energy performance indicators express consumption, intensity, demand, and change in forms suited to management decisions.

Normalization improves fairness by accounting for weather, production, occupancy, operating hours, and other relevant variables. Load profiles reveal timing, baseload, peaks, and scheduling opportunities. Benchmarking is valuable only when boundaries and service conditions are comparable. Finally, credible performance management depends on data quality, transparent governance, documented assumptions, and protection against misleading claims.

Key Terms

Energy baseline: Quantitative reference used to compare energy performance.

Energy consumption: Quantity of energy used over a specified period.

Energy demand: Rate of energy use at a particular time or interval.

Energy performance: Measurable result related to energy use, consumption, and efficiency.

Energy performance indicator (EnPI): Quantitative measure used to monitor energy performance.

Energy intensity: Energy consumed per unit of output, activity, area, or service.

Energy-use map: Representation of energy inputs, transformations, flows, and uses.

Significant energy use (SEU): Energy use with substantial consumption or improvement potential.

Load profile: Time series showing variation in energy demand.

Peak demand: Highest measured demand during a defined period and interval.

Load factor: Ratio of average demand to maximum demand over a period.

Normalization: Adjustment of energy data for relevant operating conditions.

Heating degree-day: Temperature-based indicator of heating requirement.

Cooling degree-day: Temperature-based indicator of cooling requirement.

Static baseline: Fixed reference period or fixed expected-use model.

Dynamic baseline: Reference model that changes with relevant operating variables.

Benchmarking: Comparison of performance with historical, internal, peer, or best-practice references.

Baseload: Minimum or persistent demand present during a defined operating cycle.

Data boundary: Explicit definition of what energy sources, facilities, activities, and periods are included.

Energy balance: Accounting relationship among energy entering, useful energy, losses, exports, and stored energy.

Review Questions

Short Answer Questions

1. Distinguish energy consumption from energy demand.
2. What characteristics make an energy use significant?
3. Why is a fixed baseline sometimes unsuitable for a changing operation?
4. Define an energy performance indicator and give two examples.
5. What information can a load profile reveal that a monthly utility bill cannot?

Long Answer Questions

1. Explain how an organization can construct an energy-use map from purchased energy to final services.
2. Discuss the principles for selecting and governing an energy baseline.
3. Compare physical, statistical, and engineering approaches to normalization.
4. Explain why energy-intensity benchmarking can produce misleading conclusions when boundaries and service conditions differ.
5. Describe how demand management and energy-efficiency management use different but complementary indicators.

Analytical / Application Questions

1. A facility uses 900,000 kWh annually, has 10,000 m² of floor area, and operates 6,000 hours per year. Calculate its annual EUI in kWh/m²·year and explain why the result alone is insufficient for benchmarking.
2. A plant’s monthly energy is modelled as E = 40,000 + 18Q, where energy is in kWh and Q is tonnes produced. Estimate energy for 2,500 tonnes, identify the fixed and variable components, and state two assumptions of the model.
3. A building shows a falling annual energy total but also a 20% fall in occupancy and a shorter operating schedule. Design a measurement and normalization approach to determine whether energy performance actually improved.

Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes energy demand from energy consumption?

A. Demand is energy used over a period, while consumption is the maximum rate.

B. Demand is the rate of energy use, while consumption is the quantity used over a period.

C. Demand and consumption are identical measures expressed in different currencies.

D. Demand applies only to fuels, while consumption applies only to electricity.

Correct Answer: B. Demand is the rate of energy use, while consumption is the quantity used over a period.

Question 2

Which item is most appropriate for an energy-use map?

A. Only the annual electricity bill

B. Energy inputs, transformations, major uses, losses, and outputs

C. Only equipment purchase prices

D. Employee satisfaction scores without operating data

Correct Answer: B. Energy inputs, transformations, major uses, losses, and outputs

Question 3

A factory uses 6,000 MWh of electricity to produce 12,000 tonnes. What is its electricity intensity?

A. 0.05 MWh/t

B. 0.50 MWh/t

C. 2.00 MWh/t

D. 72,000 MWh/t

Correct Answer: B. 0.50 MWh/t

Question 4

What is the principal purpose of normalization?

A. To remove all variation from measured data

B. To make comparisons fairer by accounting for relevant operating conditions

C. To replace meters with estimates

D. To guarantee that energy use declines every year

Correct Answer: B. To make comparisons fairer by accounting for relevant operating conditions

Question 5

Which variable is most directly associated with weather normalization of heating energy?

A. Heating degree-days

B. Number of suppliers

C. Product brand name

D. Annual maintenance budget

Correct Answer: A. Heating degree-days

Question 6

A site uses 240,000 kWh in 720 hours and has a 500 kW peak demand. What is its load factor?

A. 33.3%

B. 50.0%

C. 66.7%

D. 83.3%

Correct Answer: C. 66.7%

Question 7

Which benchmarking comparison is most likely to be misleading?

A. Two similar buildings in the same climate with equivalent operating hours

B. One hospital compared with a lightly occupied office using only kWh/m²

C. Two production lines making the same product under similar conditions

D. A site compared with its own prior year after adjusting for production

Correct Answer: B. One hospital compared with a lightly occupied office using only kWh/m²

Question 8

In the model E = E_fixed + e_v Q, what does E_fixed represent?

A. Energy that varies directly with output

B. Fixed or standby energy independent of output within the modelled range

C. The maximum possible demand

D. The exported energy only

Correct Answer: B. Fixed or standby energy independent of output within the modelled range

Question 9

Why should a baseline be version-controlled and documented?

A. To allow favourable results to be reported selectively

B. To preserve the reference, assumptions, boundaries, and adjustment history

C. To prevent any future measurement improvements

D. To eliminate the need for operational review

Correct Answer: B. To preserve the reference, assumptions, boundaries, and adjustment history

Question 10

Which statement about artificial intelligence in energy management is most appropriate?

A. An AI model removes the need for measurement validation.

B. AI should optimize energy regardless of safety or service quality.

C. AI can support anomaly detection and forecasting but requires reliable data and human oversight.

D. AI predictions are always causal explanations of energy use.

Correct Answer: C. AI can support anomaly detection and forecasting but requires reliable data and human oversight.