Predict Remaining Time

Predict remaining time answers the question: how long until a telemetry value reaches a critical threshold? Instead of reacting after the event occurs, you get a live countdown that you can display, trend over time, and attach alarm rules to.

  Telemetry history
  ──────────────────────────────────────────────────
  ●──●──●──●──●──●──●  ← historical values
                        ↓ trained prediction model
                       ●──●──●──●──●──●  ← forecast
                                    ↑
                              threshold crossing
                                    │
                       ← remaining ─┘
                          time

The approach combines two Trendz features:

A Prediction Model that extends your telemetry into the future.
A Batch Calculated Field that scans the forecast, finds the first crossing of your threshold, and returns the time distance from now to that crossing point.

The resulting countdown value is written back to ThingsBoard as a regular telemetry key, so it works with alarms, dashboards, and widgets — no special integration required. If the threshold is not crossed within the prediction range, nothing is written and any previously stored countdown value is left unchanged.

Multiple thresholds on the same entity (e.g. low reserve and critical reserve) are supported by creating one countdown calculation field per threshold.

Use cases

Scenario	Telemetry	Threshold	Direction
Fuel depletion	Tank level (%)	Reserve level, e.g. 10%	Drops below (≤)
Battery discharge	Battery charge (%)	Minimum, e.g. 5%	Drops below (≤)
Energy limit	Cumulative consumption (kWh)	Monthly quota	Exceeds (≥)
Pressure overload	Boiler pressure (bar)	Safety limit	Exceeds (≥)
Time-to-Maintenance	Operating hours or vibration level	Maintenance threshold	Exceeds (≥)
Storage capacity	Used storage (GB)	Capacity limit	Exceeds (≥)
Temperature runaway	Sensor temperature (°C)	Critical temperature	Exceeds (≥)

How it works

  Step 1 — Prediction model
  ─────────────────────────────────────────────────
  historical telemetry
        │
        ▼  train
  ML model (ARIMA / Holt-Winters / custom Python)
        │
        ▼  generate
  future data points  →  stored in ThingsBoard
                          under _EPD_… key

  Step 2 — Countdown field (Batch CF)
  ─────────────────────────────────────────────────
  future data points
        │
        ▼  scan forward in time
  first point where value crosses threshold
        │
        ▼  compute
  distance from now → that timestamp
        │
        ▼  write
  countdown value  →  stored in ThingsBoard
                       under _ECD_… key

  Step 3 — Alarm rule (ThingsBoard)
  ─────────────────────────────────────────────────
  countdown ≤ alert horizon  →  alarm created
  countdown  > alert horizon  →  alarm cleared

The countdown value decreases over time. When it drops below your chosen alert horizon (e.g. 3 days), ThingsBoard fires an alarm so you can take action before the event occurs.

Implementation guide

For full step-by-step instructions — building the prediction model, writing the batch script, enabling the field, and configuring alarm rules — see:

Calculate Time to Event