Detection
A virtual sensor is a calculated metric derived from existing physical sensors — filling data gaps without installing new hardware.
A virtual sensor is a derived measurement. Instead of reading a physical probe, you compute the value from the readings of one or more other sensors. Supply-air enthalpy from temperature and humidity. Occupancy from CO₂ trend. Equipment runtime from power draw. The virtual sensor is the answer to "we don't have a probe for that — but we have enough to compute it."
Virtual sensors solve two problems. First, they fill data gaps without procurement, installation, and BMS re-commissioning. Second, they act as cross-checks against the physical probes: if the virtual occupancy and the physical occupancy disagree, one of them is wrong, and you can use the historical accuracy of each to decide which.
In Explore, virtual sensors are a first-class object. They're versioned, watched for drift like physical sensors, and used by the Frostdynamics engine to validate inputs before they affect a forecast or fire an alert.
Related terms
See it in product
This is the engine that ships sensor intelligence as a product. Anomaly detection across six methods, forecasting with explicit confidence bounds, continuous compliance, and what-if simulation — all grounded in your own telemetry, all explainable, all auditable.
See FrostLogic Explore in action