Using well-pump power as an irrigation flow indicator

Method

1. Read the pump drive's power, pressure and operating state into Home Assistant every 15 seconds, through Home Assistant's Modbus support until October 20, 2026 and through my Pentek IntelliDrive integration after that.
2. Calibrated each irrigation zone's typical pump power during known runs on September 26, 2026. Flow is estimated from the power above the pump's idle floor, scaled by that calibration. Pump power is a flow indicator calibrated against known runs, not a flow measurement.
3. Built Home Assistant automations that compare what the pump did with what the irrigation controller reported: a valve-open check, a flow-estimate check, an overnight count of pump starts, an hourly short-cycling count, an overpressure alert and a Fault Recovery alert.
4. Checked the alerts against a manual run of every zone and against incidents found in Home Assistant's history.

Setup

Pump drive

Pentek IntelliDrive PID50 variable-frequency drive on the well pump that supplies both the house and the irrigation. Its own over-pressure trip is 80.1 psi.

Irrigation

A Rachio controller, with some zones driven through a Zigbee relay that mirrors Rachio. My Rachio guide covers that arrangement.

Integration

Pentek IntelliDrive, read-only, through a Modbus TCP gateway, polling every 15 seconds.

Evidence

My notes from September 22 and 25, 2026 (overnight starts), September 26 (calibration), October 6, 14 and 17 (the three fault loops and the alert tests), the alert settings and last-triggered times in my November 7, 2026 audit, and my own confirmation of the November 7 alert. None was measured again for this note.

Results

Thresholds in use on November 7, 2026

Valve-open check: pump above 650 W within about 90 seconds of a zone starting. Overnight leak alert: 10 or more pump starts in an overnight window with no zone running. Short-cycling alert: 10 or more starts in an hour with no zone running. Overpressure alert: above 80 psi for 10 seconds. These are values observed on this one installation, not recommendations. The 80 psi alert closely follows this drive's own 80.1 psi trip and should not be copied to another system.

Valve-open check, October 17, 2026

During a manual run of every zone, one zone's controller and relay both reported on, but the pump stayed asleep and pressure decayed normally. The check alerted correctly. The fault was somewhere between the relay and the valve: wiring, the solenoid or the relay channel.

Flow estimate, October 17, 2026

Every other zone's estimate was within 15 percent of its September 26 calibration, so no broken or blocked lines showed up. A first version of the check, with a narrower tolerance and one averaging window, false-alarmed on a small zone, where the drive hunts more. Widening the tolerance and averaging twice stopped that.

Fault loops, October 5 and 11, 2026

Scheduled runs of one lawn zone looped through the drive's Fault Recovery state: 21 entries in 26 minutes on October 5 and 19 on October 11. Pressure collapsed toward zero, restarts spiked power to 5.5 and 5.7 kW and pressure to 81 and 87 psi, and the zone got little water while the controller counted a full run. Neither was caught by an alert. Both were found in reviews of Home Assistant's history: the October 5 loop about an hour after the run, and the October 11 loop on October 14.

Fault loop, October 17, 2026

A third loop on the same zone happened on a manual start, so it was not tied to the schedule. Restart attempts lasted 5 to 9 seconds, and one restart into the closed system read 93.3 psi. The Fault Recovery alert, built earlier that day, and the overpressure alert were both triggered during this loop, according to their last-triggered times in my November 7 audit.

Working hypothesis for the loops

With both bladder tanks failed, a zone that opens while the pump sleeps near the bottom of its pressure band drains the pipes before the drive ramps up, and each restart then spikes past the drive's 80.1 psi trip. This fits the recorded behavior. It is not a confirmed cause, and the fault type itself is not readable from Home Assistant.

Overnight pump starts, September 2026

Starts between midnight and 5 AM with nothing watering, measured on September 22, 2026 from Home Assistant's history: 34 on September 16, 58 on September 18, 57 on September 20 and 66 on September 22. A separate entry on September 25 recorded 73 starts in the same window. The measurements were taken only from nights when no irrigation was running.

Short-cycling alert, November 7, 2026

The hourly short-cycling alert fired. When I checked after the alert, I found water escaping from the plumbing, and after the source was fixed the pump's starts with no demand returned to normal.

Failures

• The Fault Recovery alert did not detect the October 5 and 11 loops live. It was created on October 17, after both had been found in reviews of the history. It has not been shown to prevent damage.
• On October 17, my notes record that the overpressure alert still sent its phone notification to a phone registration that no longer existed, so that alert may not have reached me.
• Fifteen-second polling missed the 5 to 9 second restart attempts; only the longer states and the readings between them were recorded.
• Pump power cannot show which outlet the water reached, and household water use at the same time changes the reading.
• An automation description contains two overnight counts that could not be reconciled with the dated measurements, so they are omitted.
• This note does not establish a cause for the September overnight starts.
• All of this comes from one installation. Other pumps, wells, drives, irrigation systems and plumbing need their own calibration and safe limits, and the same behavior may not occur on them.

Removed for privacy

• Zone names and the areas each zone waters
• Irrigation schedules and the times of runs and alerts
• Network addresses and Home Assistant entity names
• The pump and well details beyond the drive model
• Household water use and where the November 7 leak was
• The flow formula and per-zone calibration values

Without zone names, times or the calibration values, a reader cannot repeat the flow estimate or tie an incident to a particular area. The thresholds are given only as values from this installation.