Accidents almost happen every day
Follow one ESS plant through 24 hours and see, in six near-misses, how the common way and ENERBOT part ways.
A reconstruction for explanation. The times are made up, but every scene actually happened in the field.
11:20 When a cloud passes
A passing cloud does not pull power from the grid
Solar output shifts more than ten thousand times a day. Only when commands go out on change, not on a timer, does the battery stop pulling the shortfall from the grid.
If it still pulls, a red row appears; after 30 seconds in a row (or 60 seconds within 5 minutes) charging stops.
- 10-second timer
- Up to 13 s late
- 456 kW command held
- 216 kW pulled from the grid
- Read every 0.2 s
- Sent on change
- 228 kW within 0.4 s
- Zero grid pull
Illustrative example. While the orange line is above the yellow one, the battery charges from the grid.
14:05 When one of three PCS units fails
One failed PCS does not stop the plant
Only the failed unit is taken out; the healthy ones share the command and keep running. Automatic resets go out at the set count and interval.
Exception · an emergency stop or fire stops the whole plant, and no automatic reset is sent.
- PCS #2 fails
- Whole plant stops
- No charge or discharge
- Someone drives out to reset
- PCS #2 fails
- Only that unit is excluded
- Two units share the load
- Automatic reset
Illustrative example. For the common way, “after reset” means after someone visited the site.
19:40 When battery-room humidity rises
Battery-room limits differ per plant, and crossing them stops operation first
Condensation and insulation set different ranges; the low-temperature and high-humidity limits move together, and in-cell batteries have their own. The battery room decides first whether charging is allowed.
It stops when a limit is crossed, or when the limits cannot be read. Only plants without their own limits use the defaults of 18–28 ℃ and 80 %.
- Humidity rises
- Nothing until someone looks
- Risk of condensation
- Per-plant limits
- Stop charging when crossed
- Adjust the air conditioner remotely
Illustrative example
22:10 When a device stops answering
What we do not know is never shown as normal
We ask in this order: not installed · switched off by us · asked but no answer · asked but could not tell. Turning “could not ask” into “nothing wrong” opens the safety gate.
Decide on the safe side; show it honestly.
- Sensor stops answering
- Last value 25 ℃ stays on screen
- Screen says normal
- Sensor stops answering
- “Disconnected” in red
- Stop and alert
| 1Not installed | Left out of decisions | None |
| 2Switched off by us | Nothing to do now | Not collected |
| 3Asked, no answer | Someone goes on site | Disconnected |
| 4Asked, could not tell | Look for the cause | - |
Reverse the order and a plant without the device is stopped as “disconnected”, or a device we switched off is read as “failed”.
02:30 When the line to the server drops
When the line drops, the plant stops safely and the record survives
If the server or internet drops, the ESS stops charging and discharging; it does not keep running on the last setting. Instead it alerts within a minute and loses no records.
While cut off, the field RTU keeps readings (ESS every minute, solar every 10 minutes) for up to 7 days and fills them into history at their own times once reconnected.
- Line drops
- Records lost meanwhile
- Line drops
- Charging stops
- Phone alert within 1 min
- Kept 7 days, then filled in
Illustrative example. Once reconnected, readings go into history only, at their own times (screens and alerts untouched).
06:00 When many plants go quiet at once
When many plants go quiet together, we look at our side, not the plants
If 14 of 40 plants go quiet at the same moment, fourteen plants did not break at once. At 3 or more plants and 30 % or more, we treat it as a server, VPN or line issue and send one alert to the administrator.
Picking the plants that need hands now, out of hundreds, is the job of monitoring.
- 14 plants quiet
- 14 alarms
- Blamed on equipment
- 3 plants · 30 %+
- Judged as our side
- One alert to the admin
Illustrative example
The answer to each of the six scenes came from 198 sites
| Scene | Common way | ENERBOT |
|---|---|---|
| 11:20 Cloud | Timer commands · up to 13 s late | Sent on change · 0.4 s |
| 14:05 PCS failure | Whole plant stops | Runs without the failed PCS · auto reset |
| 19:40 Humidity | Nothing until someone looks | Stops on per-plant limits · remote A/C |
| 22:10 No answer | Last value shown as normal | Not installed / not collected / disconnected / unknown · stop |
| 02:30 Line down | Records lost meanwhile | Stop · 1-minute alert · 7 days kept, then filled in |
| 06:00 Many quiet | A flood of alarms, one per plant | Read as our side · one alert |