Dual-Method Detection
Rolling Z-score and Isolation Forest detection work together to identify meaningful anomalies.
What Is the Energy Anomaly Detection & Forecasting Accelerator
The Energy Anomaly Detection & Forecasting Accelerator is an AI-powered analysis engine that flags abnormal energy readings for every individually metered piece of equipment in a facility, scores each equipment's health from 0–100, explains what's likely causing each anomaly, and forecasts consumption up to seven days ahead, combining statistical and machine learning detection methods with weather and holiday context.
Built-In Intelligence
Equipment Health Scoring
Every asset receives a 0–100% health score and a facility-wide ranking.
Predictive Maintenance Insights
Every flag includes severity, likely root cause, and a recommended next action.
7-Day Forecasting
Weather- and holiday-aware forecasts with visible accuracy metrics for confidence.
Context-Aware Root Causes
The same spike can mean different things when weather, holidays, and operating context change.
How It Works
Load & prepare
Equipment consumption data is combined with weather and holiday context for the facility's location.
Detect anomalies
Each equipment is analyzed against its own history using Z-score statistics and Isolation Forest machine learning.
Score & explain
Every equipment receives a 0–100 health score, severity level, root cause, and recommended action.
Forecast ahead
Projected consumption extends up to 7 days, incorporating weather and holiday effects.
Report
Equipment-level charts and a two-sheet Excel export make results ready for review and action.
Frequently Asked Questions
01
Is this a live, real-time monitoring dashboard?
The core analysis engine – anomaly detection, health scoring, and forecasting runs today as an automated analysis pass over your energy data. A live dashboard, scheduled runs, and real-time alerting are built around that engine during implementation, tailored to how your facility needs to consume the results.
02
What kind of facility has this actually been tested on?
A big-box retail facility with 558 individually metered circuits, everything from HVAC panels to a single paint shaker, over a six-month analysis window. The same per-equipment approach applies to manufacturing plants, commercial buildings, data centers, and utilities.
03
How does it decide something is an anomaly?
Two independent methods run per equipment: a rolling Z-score compares a reading to that equipment’s own recent average, and an Isolation Forest model flags readings that look statistically unusual in the broader pattern. A reading is flagged if either method catches it.
04
Does every piece of equipment get judged by the same thresholds?
Every equipment is compared only against its own recent history for anomaly detection. Root-cause thresholds, such as a specific temperature or consumption level, can be tuned per equipment type during implementation, since a sign and an HVAC panel don’t share a normal range.
05
What do I get out of a run?
A health score and anomaly count for every piece of equipment, a detailed severity/root-cause/recommendation for every flagged anomaly, a 7-day forecast per equipment, and a two-sheet Excel export available today, with a live dashboard available as part of implementation.
06
Is this a packaged product or a custom build?
It’s an Aquarient accelerator, a proven detection, scoring, and forecasting engine configured to your equipment, your thresholds, and the operational layer your facility needs.
Start Making Your Energy Data Work Smarter
Deploy this accelerator with your energy logs and uncover insights instantly.
Contact our team at contact@aquarient.com
