Fault detection and diagnostics: what meter data adds to building FDD

A chiller stages on at 3 a.m. A reheat valve leaks, so an air handler cools air and then warms the same air back up. A rooftop unit runs full tilt on a mild weekend. Nothing has broken, no tenant has complained, and the building keeps running. The only symptom is the meter climbing. Fault detection and diagnostics (FDD) is the practice of catching these operational faults before they turn into a budget overrun, and utility data is one of the most underused inputs to it.

What FDD is, and where meter data fits

FDD is the automated process of spotting equipment and control faults, then pointing to a likely cause. Most FDD today lives inside the building automation system (BAS), reading sensor and control points with rule-based or model-based logic. That is powerful, but it has two limits. Not every building has a rich set of BAS points, and BAS-level FDD is blind to anything it does not meter. Whole-building utility and interval data give an independent second view: the signal that no single controller sees.

This is a different job from two neighbouring practices. Bill-level anomaly detection flags a billing record that looks wrong after the fact. FDD works at the operational level, tied to a specific system and a physical cause, often days or weeks earlier. And an energy dashboard shows the data; FDD interprets that data into a named fault and a corrective action. Meter-based FDD sits between the dashboard you read and the controller you cannot always see.

The faults meter and interval data catch first

Runaway consumption and baseload creep

Interval data at 15-minute or hourly resolution exposes the flat overnight floor of consumption, the baseload. When that floor rises month over month with no change in occupancy or production, something is running that should be cycling off. Baseload creep is invisible on a monthly bill total but obvious in an interval trace, and it is one of the most reliable early warnings of a scheduling or control fault.

Simultaneous heating and cooling

One of the most expensive and least visible faults is heating and cooling the same air at the same time. On the meter it shows up as gas and electricity moving together when the weather does not justify it. Natural Resources Canada lists eliminating simultaneous heating and cooling among the core low and no-cost operational fixes, precisely because it wastes energy without ever tripping an alarm.

Scheduling and off-hours faults

Equipment that starts hours before occupancy, or that never fully drops to an unoccupied state, is a classic fault. The interval trace shows the ramp starting too early or the floor never falling. Because these faults track the clock rather than the weather, meter data catches them without any BAS point at all.

Short cycling and stuck outputs

Equipment that starts and stops too often, or an output stuck wide open, both leave fingerprints in interval data. Short cycling shows up as a rapid sawtooth in demand that wears out compressors and wastes energy on every restart. A stuck valve or damper shows up as consumption that no longer responds to weather or occupancy the way it should. Neither is visible on a monthly total, and both are usually cheap to correct once identified.

How big is the prize

The federal Smart Energy Analytics Campaign tracked buildings using energy management and information systems and found that participants with fault detection and diagnostics tools reached a median 9% energy savings at $0.24 per square foot, with a one to two year simple payback. Faults are not rare edge cases: an official estimate puts the waste from faults in US commercial buildings at roughly 0.7 quads a year, worth nearly $14 billion, and rooftop units alone serve about 90% of buildings. Peer-reviewed work on the prevalence of HVAC faults confirms how common these operating problems are across real building stock.

The data pipeline FDD depends on

FDD is only as good as the data feeding it. That means clean, time-aligned interval streams, consistent units, correct meter-to-space mapping, and gap handling that does not invent readings. This is where messy utility data quietly breaks FDD: a meter relabelled after a tenant change, a bill that spans an odd number of days, or a mix of hourly and monthly granularity will produce false alarms that erode trust in the whole system. Turning bills and interval feeds into standardized, analysis-ready series is the unglamorous work that makes fault detection usable.

Granularity matters as much as cleanliness. Monthly bills can reveal a rising baseline over a year, but only interval data at hourly resolution or finer can separate an off-hours fault from normal daytime use, or catch a fault that appears for only a few hours each day. Where a building has both whole-building interval data and submeters on major loads, FDD can localize a fault to a system rather than just flagging that the building as a whole is drifting.

From a fault list to fixed faults

The failure mode of most FDD rollouts is not missing faults, it is drowning in them. A whole-building analytics layer can generate hundreds of alerts, and an operator who cannot triage them will soon ignore all of them. The discipline that makes FDD pay is ranking each fault by estimated energy cost, comfort impact, and effort to fix, then working the top of the list. Meter data helps here too: the size of a baseload step, or the magnitude of a simultaneous heating and cooling signature, is a direct proxy for the dollars at stake.

False positives deserve the same seriousness as false negatives. A rule that fires every mild night because the baseline never accounted for a new tenant will train the team to dismiss real problems. Commissioning the FDD system itself, tuning thresholds, confirming meter mapping, and validating a handful of alerts against site reality, is part of the work rather than an optional extra. So is ownership: a named person who reviews the queue on a fixed cadence is what turns detection into correction.

Making the savings persist

FDD is not a one-time cleanup. Setpoints drift, overrides pile up, and savings fade unless someone keeps watching. NRCan notes that ongoing commissioning and energy information systems are what keep detected savings from eroding. The natural next step, once meter-based FDD is in place, is to formalize it into a continuous process, which is exactly what monitoring-based commissioning does.