Data infrastructure
The data pipeline that makes forecasting possible — and auditable.
A production forecasting system is only as reliable as its data supply chain. Gridvynt’s infrastructure layer handles the ingestion, normalization, gap-filling, and time-series storage of SCADA actuals, historian feeds, and NWP grid data — with full lineage from raw SCADA measurement to published forecast output. Every point in the forecast has a traceable data provenance.
Infrastructure specs
What the data pipeline handles.
Sub-5-Minute SCADA Latency
SCADA actuals and meter production data are ingested and staged for the ML correction layer within 5 minutes of measurement timestamp. This latency target is required for meaningful real-time recalibration — a 30-minute lag would mean the ML model is training on data that no longer reflects current site conditions during ramp events.
99.7% Uptime SLA
Forecast API and data ingestion pipeline availability is contractually backed at 99.7% for Portfolio and Enterprise customers — because a missed 06Z model run has direct consequences for your day-ahead scheduling. SLA tracked and published on the status page; incidents logged with resolution timeline.
Native SCADA and Historian Connectors
Pre-built connectors for OSIsoft PI System (read actuals + write forecast P-tags), OATI WebSmarts (push and pull), and generic Modbus TCP / DNP3 SCADA endpoints. Gridvynt reads from your production historian — no separate data extraction process required from your operations team.
Audit-Grade Data Lineage
Every forecast point traces to: the NWP model run that produced it (model ID, run timestamp, grid resolution), the SCADA actuals window used for ML training, and the ML model version in production at time of publish. Lineage records are immutable and available via API for dispute resolution and regulatory inquiry.
Questions about data integration?
Our integration team has connected Gridvynt to Pi System, OATI, and generic SCADA endpoints. Tell us your stack — we’ll tell you the integration path.