Energy Anomaly Detection

Overview

Energy Anomaly Detection helps you identify unusual energy behavior, estimate its financial impact, and investigate likely causes.

Within ABB Ability™ BuildingPro Suites, this module builds on the platform’s connected building intelligence and proactive AI capabilities to turn operational building data into actionable insights.

If you want the conceptual background first, start here:

Introduction to Anomaly Detection in Smart Buildings

Use Energy Anomaly Detection to:

• detect unusual energy behavior across buildings, sites, and portfolios

• estimate whether anomalies cause overspending or savings

• understand where and when anomalies occur

• investigate likely causes with diagnostics and AI-supported analysis

• validate anomalies and improve future anomaly handling

What is an energy anomaly?

An energy anomaly is a measured behavior that deviates significantly from the expected healthy behavior of an asset or group of assets.

Examples include:

• unusually high energy consumption

• unusually low energy consumption

• abnormal runtime patterns

• deviations that repeat at certain times of day

• deviations that appear only on certain weekdays or timeframes

The module does not only show that something unusual happened. It also helps quantify the impact and guide investigation.

Example

A building usually shows low electricity consumption during the night. If the measured consumption suddenly stays high at night, this can be detected as an anomaly.

How anomaly detection works

At a high level, the module compares measured behavior with expected healthy behavior and identifies significant deviations.

Hierarchical detection with cascade

Energy Anomaly Detection uses a hierarchical approach.

You define the high-level attributes where the module should actively search for anomalies. From there, anomaly evaluation can cascade through the asset hierarchy.

This means:

• enabled attributes define where anomaly detection is actively performed

• if an anomaly is detected higher in the hierarchy, related lower-level assets or attributes can still be involved in the anomaly cascade

• child anomalies help explain how a higher-level anomaly is composed across sub-assets and shorter timeframes

• relevant meters are considered in the cascade if they pass the configured financial impact meter threshold

This hierarchical view helps you move from a portfolio or building-level anomaly down to the likely contributing floors, systems, devices, and time ranges.

Example

An anomaly is detected at building level for daily electricity consumption. The cascade can then show related anomalies on individual floors, and below that on specific devices or subsystems that contributed to the higher-level anomaly.

Context-aware AI analysis

The module uses modern AI-based anomaly detection and analysis to compare actual measured behavior with the expected healthy behavior of the system.

The analysis takes into account contextual factors such as:

• weather

• holidays

• historical operating patterns

This helps distinguish real anomalies from expected changes in consumption caused by external conditions.

In our implementation, the module also combines:

• forecasting models that estimate the expected healthy range

• anomaly scoring based on the deviation between predicted and actual values

• hierarchical grouping of anomalies across assets and timeframes

• rule-based diagnostics

• an AI analysis agent that generates explanations and recommended actions

Future extensions may also consider additional context signals such as:

• people count

• production units

These additional signals can further improve anomaly interpretation in environments where occupancy or production activity strongly influences energy consumption.

Example

Higher heating energy use on a cold winter day may be expected and not anomalous. Higher heating energy use on a mild holiday with low occupancy may be anomalous, depending on the observed pattern and context.

What the module shows

The module helps you work with anomalies across several views:

• Configuration Define how anomaly detection should run and which attributes are actively analyzed.

• Anomaly Center Review anomalies in a hierarchical list, update their status, and open related views.

• Anomaly Detection Statistics Review financial impact, trends, time patterns, anomaly types, sites, and affected assets.

• Anomaly detail Inspect one anomaly in detail, including chart, diagnostics, AI explanation, and validation.

• Anomalies tab Review anomaly information directly from an asset.

• Insight Analytics Visualize anomalies in the Anomaly detection analytic type.

What is configurable?

You can configure the following main settings:

• Anomaly sensitivity Defines how sensitive the module is when classifying deviations as anomalies.

• Anomaly check interval Defines how often anomaly detection is executed.

• Anomaly financial impact alert threshold Defines the minimum estimated financial impact required to trigger an alert.

• Anomaly financial impact meter threshold Excludes meters whose possible daily impact is below the configured threshold.

• Average energy price Used to calculate the estimated financial impact of anomalies.

• Analyzed attributes Defines on which high-level attributes anomaly detection is actively performed.

• Filter options for analyzed attributes Lets you filter by site, tags, asset type, and attribute class before enabling or disabling attributes.

Example

You can configure the module to:

• check anomalies every 15 minutes

• use a medium sensitivity

• alert only if the estimated impact is above a defined financial threshold

• actively search only on selected electricity attributes of selected sites

Financial impact and severity

The module estimates the financial impact of anomalies so that users can focus on what matters most.

Depending on the view, you can review:

• total financial impact

• money saved

• money overspent

• average impact per anomaly

• anomaly severity

• predicted value, actual value, and deviation

This helps prioritize operational action instead of treating every anomaly as equally important.

Example

If an anomaly shows a small deviation but only a very low financial impact, it may be less urgent. If another anomaly shows a similar pattern but a much higher estimated cost impact, it can be prioritized first.

Validation and learning

Users can validate anomalies as:

• Confirmed

• False

They can also add comments.

This feedback supports better future anomaly handling. In your implementation, repeated feedback on false positives helps reduce the probability that similar anomalies will be alerted again.

Example

If a recurring anomaly is repeatedly marked as False because it is caused by a known special operating condition, the system can use that feedback to reduce unnecessary future alerting for similar cases.

Typical use cases

Typical use cases for Energy Anomaly Detection include:

• detecting unexpected increases in lighting, heating, cooling, or plug-load consumption

• identifying buildings, sites, or systems with repeated overspending

• spotting time-based patterns, for example anomalies that occur every night or on certain weekdays

• tracing a building-level anomaly down to contributing floors or devices

• prioritizing investigation based on financial impact

• validating false positives and improving future alerting quality

• supporting operators with AI-generated explanations and recommended actions

Examples

• A lighting system continues running during non-operating hours.

• Cooling consumption rises on weekends although the building is mostly unused.

• A single floor contributes disproportionately to a building-level anomaly.

• A device repeatedly shows abnormal behavior in the same hourly timeframe.

Recommended workflow

A typical workflow is:

1. Configure the module.

2. Select the attributes for active anomaly detection.

3. Review detected anomalies in Anomaly Center.

4. Use Anomaly Detection Statistics to identify patterns and priorities.

5. Open the Anomaly detail page to investigate a specific anomaly.

6. Validate the anomaly and add a comment.

7. Review asset-level anomaly information in the Anomalies tab or in Insight Analytics.

Related pages

Use the following pages for setup and daily work:

Configure Energy Anomaly DetectionAnomaly CenterAnomaly Detection Statistics DashboardAnomaly detail PageAnomalies tabAnomaly detection analytic type

Notes

• ABB Ability™ BuildingPro Suites connects and analyzes building and IoT data to deliver actionable insights and proactive AI capabilities, including anomaly-related intelligence.

• Public product pages confirm the platform’s AI-driven anomaly capability at a high level, while this page describes the implemented module workflow in more detail.