We speak the protocols your plant already speaks.
Modbus, OPC-UA, MQTT, inverter clouds, time-series databases, CSV — whatever your fleet looks like today, we integrate in 2–3 weeks without hardware changes.
If your stack speaks it, we ingest it.
Inverter cloud APIs, industrial protocols, time-series databases, and flat files — the four shapes utility-scale PV and BESS data actually arrives in.
For portfolios without full SCADA — we pull from the vendor portal.
Direct plant-side connections over the control network.
Read directly from your existing telemetry store.
For bootstrapping, archive backfills, or air-gapped sites.
Brand names and logos are the property of their respective owners and are shown here to indicate supported data sources. No endorsement or partnership is implied unless explicitly stated.
Modbus — the backbone of utility-scale solar
Nearly every string inverter, transformer, weather station, and power meter in a utility-scale PV or BESS plant exposes its registers over Modbus. It's simple, battle-tested, and unglamorous — which is exactly why it runs the fleet.
The platform reads Modbus through two transports:
Modbus TCP
IP-based. The modern default. Runs over the control-network VLAN or a cellular tunnel to a plant-side gateway.
- Port 502 (standard) or custom
- Slave ID per device; unit IDs discoverable
- Works behind NAT with a secure outbound tunnel
Modbus RTU
Serial RS-485 daisy-chained between devices. Still the reality for many legacy plants. Bridged via a Modbus-RTU-to-TCP gateway or an edge collector.
- 9600, 19200, 38400, or 115200 baud
- 8N1 / 8E1 / 8O1 frame configurations
- Up to 247 devices per chain
How our Modbus adapter works
Auto-probe unit IDs in your address range. Match device signatures to our library (Huawei SUN2000, SMA Sunny Tripower, Sungrow SG, Fronius Tauro, ABB PVS, Power Electronics, etc.).
Apply a pre-built register map for known devices, or use our mapping wizard for unknown ones. Every register gets a normalised name (e.g. active_power_kw, dc_voltage_v).
Apply scaling factor, unit conversion, and bit-pack decoding (16/32/64-bit, big/little-endian, signed/unsigned). One wrong gain factor kills a whole analysis — we catch those early.
Configurable intervals per register group. Hot metrics every 5 s, cold metrics every 5 min. Back-pressure and retries. Compresses on the wire.
Why this matters for the ML
A physics-ML digital twin is only as good as the data feeding it. If register scaling is wrong, or one inverter reports in W while another reports in kW, the residuals are junk and every alert is a false positive. Our adapter catches scaling and unit mismatches during discovery — before a single prediction runs.
Every way your data shows up
From legacy SCADA to inverter-vendor clouds to a folder of CSV files.
Industrial Protocols
Direct plant-side connections, usually over the control network.
- Modbus TCPIP-based over Ethernet. Our default for modern plants.
- Modbus RTUSerial (RS-485). Bridged via gateway to the platform.
- OPC-UAStandard for SCADA systems. Certificate-based auth.
- MQTTPublish/subscribe. Good for rooftop and hybrid fleets.
- IEC 61850Substation protocol. Available on request.
Inverter Cloud APIs
For portfolios without full SCADA, we pull from the vendor portal.
- Huawei FusionSolarOAuth2. Plant, device, and alarm endpoints.
- SMA Sunny PortalPer-plant credentials. Historical + live.
- Sungrow iSolarCloudFull device-tree ingestion.
- Fronius Solar.webLive + archive via OAuth.
- SolarEdgeMonitoring API with per-inverter resolution.
- Enphase EnlightenMicroinverter-level data.
- OxelUtility-scale portal integration.
- Other brandsOn request — typical integration 2–3 weeks.
Time-Series Databases
Read directly from your existing telemetry store.
- InfluxDB 1.x / 2.xFlux and InfluxQL supported.
- TimescaleDBNative hypertable queries.
- PrometheusRemote-read API for metrics shared with infra.
- PI HistorianOSIsoft connector (enterprise).
Files & Manual Upload
For bootstrapping or archive backfills.
- CSVSchema auto-detection with column mapping wizard.
- ParquetBulk imports for historical analysis.
- SFTP dropScheduled pickups from your archive server.
- S3 / GCS / Azure BlobCloud bucket subscribers.
From kick-off to live data in 2–3 weeks
Scoping
Share an inverter model list, SCADA diagram (if any), existing credentials, and network topology. 30-minute call.
Connect
We stand up the connector, open a secure tunnel if needed, and auto-discover your devices. First registers start flowing.
Validate
Side-by-side comparison against your existing monitoring for a sanity check. Fix register scaling, unit mismatches, timezone drift.
Go live
Alerts armed, dashboards populated, reports scheduled. Historical backfill continues in the background.
Security by default
Outbound-only tunnels. Credentials encrypted at rest. No inbound ports on your plant network. Deploy in our managed cloud or fully on-premise behind your firewall — the platform runs the same either way. SOC 2 Type II controls available for enterprise deployments.
Don't see your stack? Tell us what you run.
If your inverter brand, SCADA, or data warehouse isn't above, drop the details below. Most new connectors are 2–3 weeks to go-live, and the scoping call is free.