Inverter monitoring: what stays at home

A local design has a short path: inverter, ESP32, home network, browser display. ESP32 can serve HTTP requests using the ESP-IDF HTTP Server. This is a platform capability, not evidence of compatibility with a particular inverter.

Before connecting anything, check the model, firmware version, interface, communication settings and register map against the manufacturer’s documentation. The Modbus specification describes the protocol but does not by itself define your inverter’s register meanings. This article is not an electrical installation guide.

A connection does not prove freshness

Store the value, unit, time of the last valid response and connection status separately for each source. An HTTP response from the hub does not prove it has just read the inverter: it may return a cached value.

Update the reading timestamp only after validating the device response. A timeout, protocol error or decoding failure must not make an old value new. An unchanged value does not prove failure either: the load may simply be unchanged.

Use monotonic elapsed time to measure age within one boot. esp_timer_get_time() returns microseconds since initialization and starts again from zero after deep sleep. Calendar time can be corrected through SNTP, so keep it separate from freshness checks.

Example: when 800 W becomes stale

Assumptions: polling every 1 second, a 5-second threshold, and a last valid power reading of 800 W at hub uptime 100 seconds.

At uptime 103 seconds, age is 103 − 100 = 3 seconds: fresh under this rule. At uptime 106 seconds, age is 106 − 100 = 6 seconds; 6 > 5, so the reading is stale.

Keep 800 W visible as the last known value, with its age and a text warning. Do not replace it with zero, which would imply a different physical measurement. The 5-second threshold is an example, not an IVORX specification or a universal standard.

If the browser stops receiving responses, it should continue increasing the displayed age using its own monotonic timer. Do not directly compare the browser and ESP32 clocks. After the hub restarts, reset the state to “no data” until a new valid reading arrives.

How to test failures without the cloud

  • Disconnect the router’s internet access while keeping local Wi-Fi running. Check local-address access, data updates and loading of required resources.
  • In a test environment, simulate missing inverter responses. Check that age increases and cached values are not labelled as new measurements.
  • Interrupt the browser’s connection to the hub. The display should show a lost connection rather than readings that remain indefinitely “live”.
  • Restart the hub and restore communication. Fresh status should return only after a valid read, separately for the inverter and BMS.

Record actual intervals and delays before choosing a threshold. The router and hub also need power: a local architecture alone does not guarantee operation during an outage.

The limits of a local architecture

Start with read-only monitoring and do not expose hub ports to the internet. A local network still needs access control; having no cloud does not automatically make it secure. Remote access needs a separately designed secure channel.

Stale telemetry is insufficient for a confident control decision. Safe automation behavior must be defined and tested for the equipment; this design does not replace inverter or BMS protection. This is a recommended approach, not a tested IVORX feature list. Check compatibility on the product page and ask questions through the contact form.

Frequently asked questions

Can monitoring work without internet access?

A local design can work without internet if devices have power, the network is available and required resources do not depend on external services. Verify this on the actual installation.

Is a Wi-Fi indicator enough?

No. Check the age of the inverter’s last valid response and, separately, whether the browser can reach the hub.

Does a 5-second threshold suit every inverter?

No. It is an assumption for an example with polling once per second. Choose the threshold based on the protocol, measured delays and installation requirements.

Primary sources

Explore IVORXAll articles