Live Ops (real-time intelligence)
Live Ops turns a live telemetry feed into decisions an operator can act on. It runs the feed against your latest model run - the digital twin - so you can see where reality is drifting from the model, and get a control action that has been simulated before it is advised, not guessed.
Open it from Operations & reports ▸ Live Ops (real-time).
The workflow
Section titled “The workflow” TELEMETRY FEED MODEL RUN ( = the "twin" ) SCADA replay ─┐ │ or synthetic ─┘ │ │ │ └───────────────┬───────────────────────┘ ▼ ┌──────────────┐ observed vs simulated · live NSE │ LIVE OPS │ residual anomalies (drift/spike/stuck) └──────┬───────┘ │ "Recommend action" ▼ simulate candidate control actions (lower pump start levels) on the twin │ pick the biggest flooding reduction ▼ ┌───────────────────────┐ │ DECISION (auditable) │ trigger · before→after evidence │ rationale · setpoints│ · provenance · status └───────────┬───────────┘ ┌─────────────────┼──────────────────┐ ▼ ▼ ▼ Apply to model Send to SCADA Audit ledger (re-run to verify)(advisory webhook) → export PDF (persisted)
24/7: CSO monitoring ▸ create monitor ▸ Decisions: on → hourly loop logs decisions → "Sync server decisions"Step by step
Section titled “Step by step”-
Run the twin. Open your model and Simulate. That run is the baseline the live feed is compared against, so Live Ops needs it first.
-
Watch the feed. Pick a sensor and a feed - Imported SCADA (your telemetry, see Monitoring data) or Synthetic (perturbed from the twin, with an inject-fault toggle to demonstrate detection). Press play: the chart shows observed vs simulated, a live NSE fit score, and flags residual anomalies (a sensor drifting, sticking or spiking away from the model).
-
Recommend an action. Click Recommend action. The engine tries candidate control actions (lowering the pump start levels to open wet-well headroom), simulates each one on the SWMM engine, and recommends the one that buys down the most flooding. It shows the flooding before → after, the exact setpoint changes, and a plain-language rationale.
-
Act on it (advisory). Apply to model writes the setpoints into the model so you can re-run and verify the improvement; or, if you have configured a webhook, Send to SCADA POSTs the recommended setpoints to your control system. Every decision is Accept / Reject / Execute and is kept in an audit ledger you can Export as a regulator-ready PDF. The ledger is saved with the project.
-
Run it 24/7. Open CSO monitoring & alerts, create a monitor for the (cloud-saved) project, and turn Decisions: on. An hourly server loop then runs the engine and logs recommendations automatically - pull them into the ledger with Sync server decisions.
What drives the live loop
Section titled “What drives the live loop”The server loop drives your model with the live signal that actually moves it, chosen automatically from the model’s shape:
- Rainfall-runoff models (with subcatchments) are driven by the live weather forecast. More predicted rain means more predicted spill, so the hourly recommendation tracks the weather. This is the CSO compliance case (Buzet, Oldham, Lebanon, Watertown).
- Inflow-driven models (a sanitary or combined system driven by
[INFLOWS]hydrographs, with no subcatchments - like the Warkworth live model) are driven by the measured SCADA inflow. The weather cannot move such a model, so the loop reads the live inflow at the works and bias-corrects the model’s inflow to match it (peak-scaling) before recommending.
For an inflow-driven model, configure a SCADA connector on its monitor (CSO monitoring
& alerts ▸ SCADA): map the inflow tag (e.g. LSPS_Inflow) to its model node (e.g.
LSPS_WW), and choose a live REST historian (PI Web API / InfluxDB / an MQTT bridge -
a URL template plus an auth header) or a recorded replay for a pilot. The
Warkworth pilot preset fills this in against the bundled Warkworth SCADA. See
SCADA connector for the full setup, and
Worked example: Warkworth live operations to run it end to
end on a real model.
The panel makes the choice visible: the driver badge shows Weather-forecast driven or Inflow-driven (needs SCADA), and if a run logs nothing an Explain button hands the outcome to the Copilot so a quiet loop never reads as a failure.
When a run cannot finish
Section titled “When a run cannot finish”Each monitor in CSO monitoring & alerts shows the loop’s last run: when it ran and what it found (a new recommendation, the same one again, or nothing to act on). A model with no pump start/stop rules has nothing for the loop to tune, and the monitor says so.
Every scheduled check gets up to 150 seconds on the server. A model that needs longer stops at that limit and records how far it got and roughly how long a full run would take, so you can bring it inside: a shorter forecast window is the quickest lever. Nothing waits behind a slow model; it moves to the back of the queue. The morning CSO check works the same way, and Run now uses the same limit, so it shows you what the scheduled check will do.
Ask the Copilot for a control fix
Section titled “Ask the Copilot for a control fix”Beyond the one-click Recommend action, you can work the problem conversationally. After a run, ask the Copilot “how can I reduce this flooding using the control setpoints?”. It reads your rules as structured data, finds the relief setpoint most relevant to the flooding, proposes a specific one-click change to a control threshold, and you re-run to confirm the reduction - the same simulate-before-recommend discipline, driven by conversation. Change gradually: too large a drop can make pumps chatter and worsen flooding, so only a simulated result is trustworthy.
Why “simulate before recommend”
Section titled “Why “simulate before recommend””This is the part a dashboard cannot do. On the Warkworth model, lowering the pump start levels by ~20% cuts predicted flooding by ~93% - but lowering them ~35-50% makes it worse (the pumps chatter near the off setpoint). No rule of thumb catches that; you have to run the model. Live Ops simulates each option and recommends the one that actually works, with the evidence attached.
Advisory, designed for actuation
Section titled “Advisory, designed for actuation”Live Ops is advisory: it recommends, a person approves, and the setpoints are applied in the model or sent to your control system. The decision-as-auditable-object (rationale + simulated evidence + provenance + who did what) is the compliance record a regulator-facing utility needs. A direct, one-click SCADA write is a deliberate, gated step - the interface is built for it, but nothing is written to a control system without an explicit, configured endpoint.