Add HRDS dehumidifier implementation plan

[mikopp]

Summary

This commit adds a comprehensive implementation plan for integrating the HRDS+ dehumidifier with the existing airflow cooling system. The plan documents the design, coordination strategy, and migration path for adding humidity control via the dehumidifier unit while maintaining compatibility with the current Comfo ventilation controller.

Key Changes

• New plan document (plans/hrds-dehumidifier.md): Complete specification covering:
  • System context and how the HRDS coordinates with existing airflow logic
  • Analysis of existing airflow states and HRDS interaction points
  • Design decisions (all pre-resolved with the user)
  • Detailed package architecture for three new/refactored files:
    • packages/climate_common.yaml — shared climate parameters and primitives
    • packages/energy.yaml — PV surplus detection and free-energy signals
    • packages/hrds_dehumidifier.yaml — the HRDS control package
  • Refactoring of packages/airflow_cooling.yaml to use shared climate entities
  • Helper entities, decision sensors, and automation logic
  • Pre-implementation hardware verification checklist
  • History-preserving migration runbook using ha_mcp
  • Comprehensive test strategy and verification steps

Notable Implementation Details

• Coordination model: The HRDS dehumidifies on free solar PV surplus (above target humidity) or unconditionally (above max humidity). It uses recirculation-fan top-up to maintain minimum combined airflow (150 m³/h baseline, 250 m³/h during boost window).
• Shared thresholds: Both airflow and HRDS packages use dew-point-based decision logic with asymmetric Schmitt hysteresis, eliminating command contention.
• Anti-flip-flop protection: Minimum runtime timer and hysteresis bands prevent oscillation.
• Three airflow targets: 150 m³/h (baseline), ~200 m³/h (opportunistic PV), 250 m³/h (serious moisture + boost schedule).
• No Comfo edits required: Existing flush/boost logic already gates on outdoor dew point; HRDS recirculation is independent.
• Entity rename strategy: Template sensors preserve history via unique_id reuse; YAML helpers use snapshot-and-restore.

This plan is not yet implemented — it documents the design and serves as the specification for the actual implementation phase.

https://claude.ai/code/session_01A89KSdA4V9j6JzdfdTEyvK