SolarEdge Prometheus Exporter

Having just installed a SolarEdge inverter and not happy with the 15 minute delay and low resolution of the monitoring data provided by the monitoring service/api, I created this exporter to connects directly to SolarEdge inverter over ModBus TCP to export (near) real time data to Prometheus.

Status

The code could use some clean up but I have had it running for a weeks scraping data from the inverter every 5 seconds without any issues.

Requirements

• SolarEdge Inverter that supports SunSpec protocol (Tested with SE5000 w. CPU version 3.2221.0)
• ModBus TCP Enabled on the inverter
• Local network connection to the inverter (No ZigBee/GSM support)

Modbus TCP is a local network connection only and does not interfere or replace your connection to the SolarEdge monitoring service. As per the SolarEdge documentation, the two monitoring methods can be used in parallel without impacting each other.

More information on how to enable ModBus TCP can be found in the SolarEdge Documentation here

Quick Start

1. Download the binary from the Releases section for your platform

2. Configure the exporter using one of the two methods available.

   Replace the IP address in these samples with the address of your inverter

   • Environment Variables:

   	INVERTER_ADDRESS=192.168.1.189
   	EXPORTER_INTERVAL=5
   	INVERTER_PORT=502
   	NUMBER_METERS=1
   	LOG_PATH=SolarEdge-Exporter.log
   

   • config.yaml: Create a config file named config.yaml in the same location that you downloaded the executable with the following contents:

   SolarEdge:
     InverterAddress: "192.168.1.189"
     InverterPort: 502
     # Number of meters to read (0-3). Meters that are not detected are skipped.
     NumMeters: 1
   Exporter:
     # Update Interval in seconds
     Interval: 5
   Log:
     # Path of the log file
     Path: "SolarEdge-Exporter.log"

3. Add the target to your prometheus server with port 2112

Multiple inverters

InverterAddress (or INVERTER_ADDRESS) accepts a comma separated list of addresses, each optionally with its own port (host:port, defaulting to InverterPort):

SolarEdge:
  InverterAddress: "192.168.1.189,192.168.1.190:1502"

All metrics carry an inverter label with the host:port of the inverter they were read from, so the data of each inverter can be queried separately, e.g. AC_Power{inverter="192.168.1.189:502"}.

Meters

Up to three SunSpec meters are supported via NumMeters / NUMBER_METERS (default 1). The first meter is exported with the historical M_* metric prefix, the second and third as M2_* and M3_*. Meters that are not installed/enabled on the inverter are detected automatically and skipped, so no zero-valued meter metrics are exported on systems without a meter.

Grafana

A sample Grafana dashboard is available in grafana/SolarEdgeExporter.json.

Session recovery

On connection, protocol, parsing, or detected-meter read failures, the exporter closes and discards the entire Modbus session. Each retry uses a fresh handler, re-reads inverter metadata, and rediscovers enabled meters before polling. Transaction-ID validation remains strict: only a reply matching the current request can become a measurement. The transport discards up to three complete unmatched replies per request, logging each discard, then waits for the matching reply within the original 10-second deadline. Requests are not resent and stale replies do not extend the deadline. Exceeding the limit, timing out, or receiving an invalid frame triggers session reinitialization. Protocol and unit IDs, frame lengths, and matching-response function codes are still validated.

Consecutive failed attempts wait 2, 4, 8, 16, then 30 seconds, capped at 30 seconds. Backoff resets only after a complete successful collection cycle, not merely a successful TCP connection. Polling and retry waits are interruptible on shutdown; cancellation also closes active sockets and interrupts connection attempts. Network operations retain their 10-second timeout. This improves recovery but cannot guarantee recovery from an inverter or network fault. Use the health metrics below to detect ongoing outages.

Collection health and stale data

Prometheus up only indicates that /metrics is reachable, not that inverter polling works. Two additional metrics carry the same inverter="host:port" label as the measurements:

• solaredge_collection_success: 1 when the latest cycle read the inverter, completed meter detection, and read all detected meters; 0 on failure or before the first completed cycle. An absent/disabled meter is not a failure.
• solaredge_collection_last_success_timestamp_seconds: time of the last successful complete cycle, or 0 if none has succeeded since startup. Failed cycles leave this timestamp unchanged.

Existing measurements remain cached on failures for compatibility. Check these health metrics before trusting them. They describe complete-cycle health, so a meter failure can mark collection unhealthy even if inverter values updated. Check both success and age: during a blocked read, success can still describe an earlier cycle. Choose an age threshold above your polling interval and expected read duration (particularly with multiple meters).

Example Prometheus alert rules (adjust thresholds for your configuration):

groups:
  - name: solaredge
    rules:
      - alert: SolarEdgeCollectionFailed
        expr: solaredge_collection_success == 0
        for: 30s
        labels:
          severity: warning
        annotations:
          summary: "SolarEdge collection failing for {{ $labels.inverter }}"
      - alert: SolarEdgeDataStale
        expr: time() - solaredge_collection_last_success_timestamp_seconds > 60
        for: 30s
        labels:
          severity: warning
        annotations:
          summary: "SolarEdge data stale for {{ $labels.inverter }}"
      - alert: SolarEdgeExporterDown
        expr: up{job="solaredge-exporter"} == 0
        for: 30s
        labels:
          severity: warning
        annotations:
          summary: "SolarEdge exporter scrape failing"

Metrics

All metrics are labeled with the inverter they were read from. In addition to the inverter metrics below, each detected meter exports the SunSpec meter values (M_AC_Current, M_AC_Power, M_Exported, M_Imported, ... — prefixed M_, M2_ or M3_ depending on the meter number).

Metric	Type	Description/Help
SunSpec_DID	Guage	101 = single phase 102 = split phase1 103 = three phase
SunSpec_Length	Guage	Registers 50 = Length of model block
AC_Current	Guage	Amps AC Total Current value
AC_CurrentA	Guage	Amps AC Phase A Current value
AC_CurrentB	Guage	Amps AC Phase B Current value
AC_CurrentC	Guage	Amps AC Phase C Current value
AC_Current_SF	Guage	AC Current scale factor
AC_VoltageAB	Guage	Volts AC Voltage Phase AB value
AC_VoltageBC	Guage	Volts AC Voltage Phase BC value
AC_VoltageCA	Guage	Volts AC Voltage Phase CA value
AC_VoltageAN	Guage	Volts AC Voltage Phase A to N value
AC_VoltageBN	Guage	Volts AC Voltage Phase B to N value
AC_VoltageCN	Guage	Volts AC Voltage Phase C to N value
AC_Voltage_SF	Guage	AC Voltage scale factor
AC_Power	Guage	Watts AC Power value
AC_Power_SF	Guage	AC Power scale factor
AC_Frequency	Guage	Hertz AC Frequency value
AC_Frequency_SF	Guage	Scale factor
AC_VA	Guage	VA Apparent Power
AC_VA_SF	Guage	Scale factor
AC_VAR	Guage	VAR Reactive Power
AC_VAR_SF	Guage	Scale factor
AC_PF	Guage	% Power Factor
AC_PF_SF	Guage	Scale factor
AC_Energy_WH	Guage	WattHours AC Lifetime Energy production
AC_Energy_WH_SF	Guage	Scale factor
DC_Current	Guage	Amps DC Current value
DC_Current_SF	Guage	Scale factor
DC_Voltage	Guage	Volts DC Voltage value
DC_Voltage_SF	Guage	Scale factor
DC_Power	Guage	Watts DC Power value
DC_Power_SF	Guage	Scale factor
Temp_Sink	Guage	Degrees C Heat Sink Temperature
Temp_SF	Guage	Scale factor
Status	Guage	Operating State
Status_Vendor	Guage	Vendor-defined operating state and error codes. For error description, meaning and troubleshooting, refer to the SolarEdge Installation Guide.