Open Source Projects

  • wlcrs/huawei_solar — Home Assistant integration for Huawei Solar inverters via Modbus RTU/TCP, with 882 stars. Gives you direct access to inverter telemetry (production, consumption, grid export/import) from a widely deployed residential inverter brand — useful if you want to prototype energy monitoring pipelines or build on top of real solar data.

  • Giorgio866/lumina-energy-card — A 3D dashboard card for Home Assistant that visualizes house consumption, battery state of charge, EV charging status, and multiple solar arrays in a single view. A good reference for frontend engineers exploring how to present multi-source energy data in a digestible, real-time format.

  • mtrossbach/noah-mqtt — Bridges Growatt Noah 2000 home batteries to Home Assistant over MQTT, exposing charge/discharge control and state-of-charge monitoring. Demonstrates a clean IoT integration pattern (MQTT broker → HA entity) that’s directly transferable to commercial battery management or microgrid monitoring work.

  • CharlesGillanders/homeassistant-alphaESS — Integrates Alpha ESS battery systems with Home Assistant using the vendor’s official API, covering generation, storage, and consumption data. Shows how to work within a manufacturer’s documented API contract — a pattern that scales to commercial and industrial energy storage integrations.

  • Matthew1471/Enphase-API — Unofficial API wrapper and documentation for Enphase microinverter systems, including local LAN Gateway access to per-panel generation data. Valuable if you’re interested in ML applications at the edge — per-panel telemetry enables anomaly detection, degradation modeling, and shade analysis that are core to improving PV system performance.

Today’s Synthesis

If you’re looking for a concrete weekend project that ties together real data ingestion with ML, consider building a home energy self-consumption optimizer. Start by pulling inverter telemetry from wlcrs/huawei_suawei_solar or per-panel generation data from Matthew1471/Enphase-API , then layer in battery charge/discharge state from mtrossbach/noah-mqtt via MQTT. With production, consumption, and storage timeseries flowing into a single data store, you can train a lightweight forecasting model — predicting solar output and household load over the next few hours and issuing charge/discharge decisions to minimize grid import. The MQTT-to-timeseries pipeline you build mirrors the exact architecture used in commercial microgrid monitoring, so this scales from a single home to a multi-site C&I prototype. The Enphase per-panel dataset is especially valuable if you want to add a second phase: anomaly detection or degradation modeling on individual strings, which is a core ML problem in PV performance analytics.