#modbus #telemetry #pcap #inverter #serial #wifi #sunsynk

bin+lib sunsniff

Intercept and store telemetry from a Sunsynk inverter

8 releases

new 0.4.0 Dec 7, 2024
0.3.2 Dec 25, 2023
0.3.1 Oct 22, 2023
0.3.0 May 13, 2023
0.1.1 Nov 13, 2022

#292 in Network programming

Download history 14/week @ 2024-09-21 12/week @ 2024-09-28

498 downloads per month

GPL-3.0-or-later

59KB
1K SLoC

Inverter telemetry capture

This program collects data from a Sunsynk/Deye router and makes it available for use. It can collect the data in two ways (referred to as "frontends"):

  1. By running on a router sitting between an inverter with an Inteless WiFi dongle and the remote server. In this mode it is a completely passive observer, so it cannot interface with the inverter's operation. This is called the pcap frontend.

  2. By connecting a serial cable to the inverter, it is possible to query it interactively. This requires additional hardware, but allows the query interval be set (and made much faster than the 5 minute interval the dongle uses), and the dongle can be removed for better privacy and security. In this mode commands are sent to your inverter, but they only read (not write) the registers, so it is still pretty safe. This is the modbus frontend. See this guide for information on how to wire the RS485 cable. There are reports that the RS232 connection works too.

There are also currently two "backends", which determine what to do with the data.

  1. Store the values in an Influxdb database (requires Influxdb2).
  2. Broadcast the values over MQTT.

This is alpha software (although I am using it every day). All the schemas may change. The data you collect might vanish, or leak onto the internet (but it's already being sent unencrypted, which is why this project works in the first place). The config file format may change. It may hang your router.

Compilation

  1. Install Rust e.g. using these instructions.
  2. Ensure that you have a C compiler and linker, and libpcap installed.
  3. Run cargo install sunsniff to install the binary. Alternatively, check out the repository and run cargo build --release. This will compile the binary to target/release/sunsniff.

If you want to cross-compile:

  1. Install and set up cross e.g. using these instructions.
  2. Run cross build --release --target=armv7-unknown-linux-gnueabihf (replace with your target architecture).
  3. Find the binary in target/<arch>/release/target.

I had problems because the resulting binary needed a newer glibc than the host I was targeting. To build a static binary, set the environment variable RUSTFLAGS to -C target-feature=+crt-static -lpcap. I also found that DNS wasn't working with glibc, so I ended up using a target of armv7-unknown-linux-musleabihf instead.

Configuration

Configuration is stored in a TOML file, which is passed on the command line.

Configure one of the possible frontends (do not try to configure more than one), and least one backend. It's possible to have more than one instance of the same backend (the doubled square brackets are the TOML syntax that allows for this).

Pcap frontend

Create a [pcap] section. It has the following fields:

  • device (required): the Ethernet device to capture. Note that the any device is not currently supported.
  • filter (optional but recommended): A pcap filter to select the traffic to inspect. If the device handles data for any other devices on the network then setting filter is necessary to prevent other data from being accidentally interpreted as sensor readings.
  • file (optional): if set to true, then device is interpreted as a pcap file rather than a device. Note that the pcap file is fully loaded into memory, so it should not be used with very large files.
  • timezone (required): The timezone name used by the inverter. This is used to convert the timestamps to UTC.

I have the following setup:

[pcap]
device = "br0"
filter = "src host 192.168.0.21"
timezone = "Africa/Johannesburg"

Modbus frontend

Create a [modbus] section. It has the following fields:

  • device (required): the serial device, or the address for Modbus over TCP in the format host:port (the port is required even when using the Modbus default).
  • interval (required): time (in seconds) between samples
  • baud (optional): baud rate for the serial port. Defaults to 9600.
  • modbus_id (optional): Modbus slave number of the inverter. Check your inverter settings. Defaults to 1.

I have the following configuration:

[modbus]
device = "/dev/ttyUSB0"
baud = 9600
interval = 20

Influxdb2 backend

The readings are inserted into an Influxdb 2.x bucket. Note that the schema is not final.

The configuration section looks like this:

[[influxdb2]]
host = "http://192.168.0.123:8086/"
org = "my_org"
bucket = "my_bucket"
token = "..."

The implementation tries very hard to deal with intermittent connections to Influxdb, buffering messages until it is able to deliver them (but only in memory; if the service is stopped, any pending messages are lost). Since the updates are only sent every 5 minutes is can be quite practical to buffer messages for hours or days, and I'm currently running the Influxdb server on my home PC which is switched off at night.

The downside of this robustness is that if you get the configuration wrong, the server won't stop with an error. It will just keep trying to deliver, and use more and more memory to buffer the incoming messages.

MQTT backend (Home Assistant)

This backend publishes sensor values to an MQTT broker. The topics are specifically designed for use with Home Assistant and provide the appropriate discovery information, but this doesn't prevent other use cases. You will need to install an MQTT broker (Home Assistant supports Mosquitto as an add-on) and configure Home Assistant to use it. A typical configuration then looks like this:

[[mqtt]]
url = "mqtt://192.168.0.123:1883"
username = "my_username"
password = "my_password"

The username and password can be omitted if the broker doesn't require authentication.

Unfortunately the MQTT library I'm using doesn't support MQTT last will messages, so there is no availability information to indicate that the service is running.

Supported hardware

So far I've only tested this with my personal setup. I'm hoping other devices will work too. If it works for you, please let me know. Note that it's unlikely to work with the 3-phase inverters, as they use different registers.

  • Inverter: Sunsynk 5 kW (Sunsynk-5K-SG01LP1)
  • Dongle: unbranded Inteless dongle (it has red and green lights). Apparently the Sunsynk-branded dongle is the same thing.

Troubleshooting

Logging is done with env_logger, so you can enable debugging by setting the environment variable RUST_LOG=debug. There isn't very much logging yet though.

TODO:

  • Explain what to look for in a packet capture
  • Explain that missing pcap filter can cause bogus data

Changelog

0.4.0

  • Fix handling of systems with two PV strings. Previously the PV Power metric reported only the power for the first string. There are now PV Power 1 and PV Power 2 metrics, and PV Power reports their sum.
  • Support newer dongle firmware in the pcap backend, which uses a different packet size and layout.
  • Update dependencies

0.3.2

  • Update dependencies
  • Use modbus-robust so that restarting mbusd will be handled robustly

0.3.1

  • Update dependencies
  • Fix cross builds for aarch64-linux-gnu

0.3

  • Add grid_connected sensor
  • Add sensors for programmed time blocks
  • Make pcap frontend optional
  • Fix cross compiling of pcap for some architectures

0.2

  • Add modbus support.
  • Rename pv_voltage 1 to pv_voltage_1 (the space was a typo).
  • Change the offsets used for grid_power, inverter_power, load_power. This makes no difference on my inverter, but may give you more correct values if you have additional connections.
  • Hopefully fix the kWh sensors to include the upper 32 bits, so that they can support values above 32767 kWh (untested).

0.1.2

  • Attempt to connect to InfluxDB on startup and show a warning if it's not reachable.
  • Bump versions of dependencies (this seems to fix InfluxDB with TLS).
  • Improve documentation.

0.1.1

Add more fields.

0.1.0

First release.

Dependencies

~18–35MB
~547K SLoC