Инвертор Deye с украинской спецификой

Summary

Sergey Petrenko explains how he monitors and automates a Deye hybrid solar system locally, rather than relying on the manufacturer’s cloud or the inverter’s built-in settings alone.

Solar installation

• The system has roughly 7 kW of solar panels and a 12 kW Deye inverter.
• Two Anern batteries from different models are connected electrically but are not integrated as a single coordinated battery system. Their combined capacity is about 30 kWh.
• The house also has a 6.5 kW three-phase generator for backup.
• The inverter is configured for zero export to the grid.

Why he avoids relying on Deye Cloud

The inverter needs a separate logger to send data. Petrenko says his newer logger sends data only to Deye’s cloud, whereas older Solarman loggers could expose data locally. He prefers local control because:

• Cloud data can be delayed, sometimes by around 20 minutes.
• API request limits may restrict access or potentially lead to paid service.
• Essential home power management should not depend on internet access or a remote service, including possible firmware changes.

Local monitoring and control

An Elwin EW41 RS-485-to-Wi-Fi adapter connects to the inverter’s local communications interface, providing access to inverter data without relying on the cloud.

A server-side setup polls inverter registers about every 30 seconds, publishes readings over MQTT, and stores selected data in a database. Scripts use the local data to:

• Change inverter settings.
• Send Telegram notifications about grid outages and charging modes.
• Populate monitoring dashboards.

Most inverter data is available this way, though some battery BMS information is not exposed through the inverter’s registers.

Custom battery-management logic

Petrenko’s main concern with the inverter’s built-in settings is that its battery thresholds do not match his intended use. A configured threshold can act both as the point at which grid charging is permitted and as the level to which the inverter charges. He instead wants the battery to be replenished when needed and then return to battery operation.

His scripts temporarily change the inverter’s threshold: when battery charge falls to about 50%, the script changes the setting to charge from the grid up to 100%. Once charging is complete, it restores the 50% threshold.

The automation also accounts for electricity tariffs and expected solar production:

• At 23:00, when the cheaper night tariff begins, it checks the battery level. If charge is below about 80%, it schedules charging to 100%.
• Around 05:00, it checks a generation forecast from Open-Meteo, based on the solar installation’s location and panel configuration. If the forecast suggests a shortfall, it can charge from the grid before the night tariff ends at about 07:00.
• The goal is to store enough energy to get through the night while avoiding unnecessary charging at the daytime rate.

Practical setup

Petrenko shows the inverter communications hardware and describes deploying the monitoring and automation software on a Raspberry Pi. He recommends a Pi 4 with wired Ethernet and roughly 3–4 GB of RAM. A Pi Zero may work where Wi-Fi is reliable.

He also notes that open-source inverter-MQTT software and register information can help, but register mappings may vary between inverter models and firmware versions.

Guide and tutorial content

The video is an overview and practical demonstration of:

1. The solar-and-battery installation and its backup options.
2. Connecting a local RS-485/Wi-Fi adapter to avoid cloud-only monitoring.
3. Using MQTT, scripts, a database, and notifications for custom inverter automation.
4. Applying tariff-aware battery charging and solar-forecast-based decisions.

It is not a complete, step-by-step register configuration guide. Petrenko emphasizes that the setup required testing and adjustment.

Main speaker and sources

• Main speaker: Sergey Petrenko.
• Other sources and tools mentioned: Open-source Deye inverter/MQTT software, Open-Meteo forecasts, inverter register information, and an AI coding assistant referred to in the subtitles as “Cloud Code” (likely Claude Code).