Skip to content

Feature request: forecast-aware battery reserve in real-time control (Balanced) #73

Open

Description

@MonsterD1

Feature request: forecast-aware battery reserve in real-time control (Balanced)

Current behavior (0.4.0rc1)

  • Target SoC and Target Time are only used by the planner (planner.py, _balanced_strategy splits slot excess 50/50). With Plan Influence = None this has no effect on real-time decisions.
  • In real-time control, _power_for_optimizer() reserves battery charging power only for Battery First (battery_first_budget()). That reservation is a hard SoC threshold: excess is held back until SoC ≥ target, regardless of the solar forecast or the target time.

So in real time there are two options: Balanced/Appliance First, which never protects the battery target, or Battery First, which always protects it, even on a sunny morning when the forecast is far above what the battery still needs.

Request

A real-time mode that reserves battery charging only when the remaining forecast is not enough to reach the target SoC by the target time. Roughly:

needed_kwh   = (target_soc - soc) / 100 * battery_capacity
usage_kwh    = hours_until_target_time * avg_unmanaged_load_kw
reserve      = forecast.remaining_today_kwh < needed_kwh + usage_kwh + buffer_kwh
  • reserve == False: appliances get the full excess (as Appliance First today).
  • reserve == True: reserve charging power as battery_first_budget() does today.
  • Missing forecast, SoC or capacity: fall back to the current Battery First behavior. Never assume a sunny forecast.

This could be the real-time behavior of Balanced, or a separate strategy, so existing setups don't change.

Why

This is how the original pyscript blueprint (nicorusti/ha-advanced-blueprints, _force_charge_battery()) behaved: min_home_battery_level acted as an end-of-day target with a forecast check and a fixed 2 kWh buffer, not as a start threshold. Users migrating from the blueprint lose this. Today they must choose between appliances never waiting for the battery and appliances always waiting until the target is reached.

Related: #44 (battery not prioritized), #58 (SoC ceiling for appliances).

Points to consider

  • dmacwalter's fork tried a time-to-target feasibility gate (0.3.11–0.3.14) and removed it in 0.3.15. It was focused on cheap-tariff grid supplementation, and its simulation found regressions when the gate blocked appliances during cheap grid windows or while the battery was already being grid-charged. A solar-only reserve without tariff interaction should be simpler, but the same cases (grid charging active, minimum runtime pending) should be exempt.
  • Hysteresis is needed so the reserve does not toggle every cycle when the forecast is close to the threshold (e.g. stay engaged until the margin is clearly positive again).
  • The generic forecast provider gives only remaining_today_kwh. That is enough for this check; hourly data is not required.

I can prepare a PR if the approach is acceptable.

Activity

  1. MonsterD1 commented on Nov 13, 2026

    @MonsterD1
    Author

    I opened #76 implementing this for Balanced: real-time reservation via the existing battery_first_budget() only when remaining_today_kwh - needed_kwh - usage_kwh falls below a 0.5 kWh buffer (1 kWh to release). Without an available forecast Balanced behaves as before. Base load (0.5 kW) and buffer are constants for now; happy to make them options or the behavior opt-in if you prefer.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions