Energy data
Article 22 of 25 · Energy insight and dataAnomaly detection: recognising energy consumption deviations: a guide for businesses
Direct answer
Anomaly detection in energy consumption means recognising deviations between what you actually use and what a normal pattern leads you to expect. In practice you compare the measurement with a baseline and watch for night consumption that fails to fall, sudden peaks, a gradual rise or unexpected weekend use. Each deviation is a prompt to look for the cause.
- Clear definition
- Data-driven assessment
- Risks and opportunities visible
- Practical next steps
Anomaly detection: recognising energy consumption deviations: scattered information versus Energy Intelligence
Anyone buying energy without looking at the consumption curve misses the early signs that something is wrong. A valve that sticks, a setting that never returns after maintenance, a device left running at night: it costs money before anyone notices. Anomaly detection is about those early signs. It matters to everyone with a large consumer connection or smart meters that record consumption per quarter-hour or hour, from office to production site.
- You recognise a deviation by comparing the current measurement with a baseline: the pattern that fits your normal operation.
- Common signals are night consumption that does not drop to a low, a sudden peak, a gradual rise over weeks, and consumption on days you are closed.
- After a signal you look for the cause in three directions: a fault, a wrong setting or changed behaviour.
Insight
Traditional approach
Information is scattered across portals, documents, invoices or separate spreadsheets.
Modern approach
Data, context and interpretation are brought together into a clear decision picture.
Decision-making
Traditional approach
Choices are made based on averages, assumptions or occasional analyses.
Modern approach
Scenarios, KPIs and current measurement data make the trade-off more concrete and repeatable.
Follow-up
Traditional approach
Actions often stay non-committal or disappear into separate reports.
Modern approach
Follow-up actions, monitoring and reporting are linked to the same energy data.
Which deviations do you meet?
Deviations have recognisable shapes. Night consumption that fails to fall is the best known: when a building or process stands still, the curve should drop to a low. If that line stays high, something keeps running that could be off. A sudden peak points to something that switches on abruptly, such as a faulty thermostat or a machine that jams. A gradual rise over weeks is deceptive, because on any single day it barely shows; think of clogged filters or wear. Unexpected weekend consumption means installations run while no one is there. Each shape tells you something different about the likely cause.
- Night consumption that does not drop to a low: something keeps running that could be off.
- A sudden peak: a device or setting that changes abruptly.
- A gradual rise over weeks: fouling, wear or creeping standby use.
- Consumption on days you are closed: installations running needlessly.
How do you spot them?
You only recognise a deviation once you know what is normal. So you work with a baseline: the consumption pattern that fits your normal operation, corrected for factors that logically drive consumption, such as production, occupancy and outside temperature. This approach is the heart of energy monitoring within an energy management system under ISO 50001. Place the measurement next to that baseline and deviations stand out. Alarms that trigger on an over- or under-run do this automatically. Comparison also helps: the same week against last year, or site against site. If one location deviates consistently, something is going on there. A deviation is never a conclusion, only a reason to look.
- Build a baseline from a representative period of normal consumption.
- Correct for driving factors: production volume, occupancy and weather.
- Set alarms on an upper and lower bound around expected consumption.
- Compare between periods and between comparable locations to find outliers.
What do you do after a signal?
A signal is the start of a search, not a final verdict. First look at when the deviation began and whether that coincides with an event: maintenance, a fault, a new machine or a schedule change. That narrows down the location. Then look for the cause in three directions. A fault: a sensor, valve or pump that no longer works well. A setting: a timer, thermostat or control that is wrong, often after maintenance or a power interruption. Or behaviour: equipment left on, doors left open, a process used differently. Fix the cause and then keep watching the curve. If the deviation stays away, the intervention was on target.
- Establish the start moment and link it to an event to narrow the location.
- Test for a fault: sensors, valves, pumps or other failing hardware.
- Test for a setting: timers, thermostats and controls after maintenance or a fault.
- Test for behaviour: equipment left on or processes used differently.
- After the fix, check whether the deviation disappears from the curve.
Curious what this looks like with your own data?
In a no-obligation call, a specialist looks at your meters, sites and energy questions with you. Response within one business day.
Search the knowledge base
Find the answer to your question.
Search using your own words. Abbreviations and spelling variants are recognised, so EMS also finds the articles on energy management systems.
Topic
