EV Charger Fault and Diagnostic Guide
Organised by symptom, because that is what you have when something goes wrong. Each entry gives what the symptom usually means and the first thing worth checking, which is frequently not the thing people check first.
The recurring theme is that charging faults cross layers. Hardware, firmware, protocol and backend can each demonstrate they are behaving correctly while the fault persists in the interaction between them.
The charger will not come online
The unit has power but the platform does not see it, or sees it and will not use it.
Charger connects but never becomes usable
Usually means. Registration is being refused or left pending, frequently because the identity presented does not exactly match what the platform expects.
Check first. Read the identity from the charger's own configuration rather than the label, and read the registration response status rather than assuming an error.
BootNotification and heartbeat debuggingCharger reconnects in a loop every few seconds
Usually means. It is being rejected below the application layer: a failed connection upgrade, a protocol version mismatch, or a certificate problem.
Check first. Look at the transport layer before the protocol. A unit that never sends its first message leaves no protocol evidence, which is itself the signal.
BootNotification and heartbeat debuggingPlatform says offline, unit looks normal on site
Usually means. No network path, a path that does not reach the platform, or the unit is pointed at a different backend than the one being checked.
Check first. Confirm the configured backend address first. Indicators report power and readiness, not platform connectivity.
Offline in the CMS, normal on siteSessions will not start
The charger is online and a driver cannot begin charging.
Remote start is accepted and nothing happens
Usually means. Acceptance means the request was understood, not that a session began. A condition at the charger is unmet, or the connector referenced is not the one in use.
Check first. Look for a StartTransaction message. If it is absent, the failure is at the charger and the connector number is worth checking specifically.
RemoteStartTransaction failuresConnector appears occupied but no vehicle is connected
Usually means. A previous transaction never closed properly, leaving the connector held from the charger's point of view.
Check first. Check for an open transaction on that connector and whether interrupted sessions are being reconciled at all.
Offline transaction recoveryCard or app is rejected at one charger only
Usually means. The charger is authorising locally against a cached list that is out of date, rather than asking the platform.
Check first. Establish where the authorisation decision is being made, and how often the cached list refreshes.
Offline authorisation and store-and-forwardSessions start and then stop
Energy flows briefly or intermittently and the session ends without a driver action.
Session ends shortly after starting, repeatedly
Usually means. Supply conditions outside the acceptable window, thermal protection, or a contactor that no longer closes reliably.
Check first. Check whether other bays on the same site stopped at the same moment. That separates a site supply event from a unit fault immediately.
Stops mid-session: voltage, thermal, contactorCharger reports available, vehicle is not drawing
Usually means. A vehicle-side stop: charge limit reached, a departure timer, or a battery management decision. Not a charger fault.
Check first. Check which suspended state was reported. A charger offering energy that the vehicle declines is behaving correctly.
StatusNotification mappingProtection device trips during charging
Usually means. A genuine insulation fault, water ingress, or cumulative leakage across several vehicles on a shared device.
Check first. Do not reclose without establishing the condition has cleared. Record every trip, because a rising rate at one bay is a strong predictive signal.
Why RCDs trip and when self-recovery is appropriateEnergy and billing do not add up
Sessions complete and the numbers afterwards are wrong or missing.
Session totals do not match the meter
Usually means. Unit mismatch, cumulative readings treated as per-interval, or readings included that should have been filtered by context.
Check first. Check the unit the charger reports against the unit the platform assumes, then whether readings are cumulative or interval.
MeterValues errors that break billingSessions delivered energy and produced no payment
Usually means. Offline sessions whose records were rejected as too old, or authorisations that were never captured.
Check first. Reconcile charger records against platform records against settlement. Comparing only two of the three misses this case.
Payment failures and revenue leakageA driver was billed twice
Usually means. The charger resent a transaction after an uncertain acknowledgement and the platform did not detect the duplicate.
Check first. Confirm duplicate detection exists. Resending is normal and expected behaviour, so the platform must handle it.
Offline transaction recoveryFleet-level symptoms
Problems that appear across many units rather than at one.
The same fault keeps recurring after being fixed
Usually means. Tickets are being closed without a recorded root cause, so the underlying condition is never addressed.
Check first. Require a cause at closure from a defined list. Without it a network learns nothing from its own faults.
Automating fault tickets from telemetryFault patterns make no sense across the fleet
Usually means. Firmware drift. The same symptom has different causes on different builds, so grouping by symptom produces a mixture.
Check first. Establish the actual firmware version distribution rather than the intended one. It is usually surprising.
Firmware version driftUptime looks good and drivers complain
Usually means. Uptime measures the charger's self-assessment. Failed session attempts leave no ticket because the driver leaves rather than reports.
Check first. Start counting authorisation attempts that produced no transaction, and report session success alongside uptime.
Online is not the same as workingBefore you disturb the unit
The most common reason an investigation cannot conclude is that the evidence was destroyed before anyone looked. Power cycling clears volatile state, disassembly destroys thermal and mechanical evidence, and replacing the suspect part removes the only sample.
- Capture logs and backend records before touching the unit.
- Photograph the installation as found, including cable routing and enclosure condition.
- Record firmware version, configuration and recent update history.
- Retain the failed part rather than scrapping it.
- Note environmental and supply conditions at the time of failure.