On This Page

    Dynamic Strategy v2

    Lab feature — experimental strategy upgrade. Behavior may change between releases. Feedback welcome at GitHub Issues.

    02 strategy-dynamic-2.json is a redesigned Dynamic strategy that evaluates every price interval individually instead of working with fixed “cheapest N hours” and “expensive M hours” blocks. This gives the algorithm more flexibility to react to multiple peaks and troughs during a day.

    • Guarantees a minimum spread per cycle, so each marked cycle stays profitable.
    • Think of v2 as an interim upgrade — better than v1, while a more advanced algorithm is still in development.

    Quick start

    1. First-time setup (data source, Cheapest Energy Hours, dashboard wiring): follow the Dynamic Strategy Setup guide.
    2. Import 02 strategy-dynamic-2.json into Node-RED and deploy.
    3. In your Home Battery Control dashboard, switch to the Lab features tab.
    4. Pick a use-case configuration and set the controls accordingly.

    Defaults that work out of the box:

    Period Default action
    Low (cheap intervals) Charge
    Neutral (unmarked) Charge PV
    High (expensive intervals) Self-consumption

    Sensible starting values

    • min_delta = 6¢/kWh
    • cheapest_hrs = 0 (no cap), expensive_hrs = 0 (no cap).

    See Dashboard controls for what each does.

    Dashboard controls

    The v2 controls live in the Lab features tab of the Home Battery Control dashboard.

    Control Description
    Energy supplier / Data source Same data-source selector as the standard Dynamic strategy.
    Minimum price difference (min_delta) Spread threshold in cents/kWh — pairs with a spread below this are not marked. Default 6¢/kWh (covers round-trip losses for a ~€1250 battery at 6000 cycles, 80% RTE; full derivation in the setup guide). Lower towards 0 if you mostly charge from solar.
    Max cheap hours / day (cheapest_hrs) Cap on low marks per day, in hours. 0 = no limit (mark every viable pair). Useful when your battery can only absorb so much per day.
    Max expensive hours / day (expensive_hrs) Cap on high marks per day, in hours. 0 = no limit.
    Low period strategy Action during low-marked intervals. Choices: Charge, Charge PV, Self-consumption, Full stop.
    Regular period strategy Action during neutral (unmarked) intervals. Choices: Charge PV, Self-consumption, Full stop.
    High period strategy Action during high-marked intervals. Choices: Self-consumption, Sell, Charge PV, Full stop.

    Tip: the underlying Charge / Sell strategies can be configured with solar forecast, peak-shaving reserves, etc. Dynamic v2 respects those settings.

    The price-data table (today and tomorrow with marks) is visible in the same tab when debug/insights mode is enabled.

    Tweaker note — how hour caps map to intervals. With 15-min price data, 1 hour = 4 intervals; the cap is rounded to whole intervals. Asymmetric caps (e.g. cheapest_hrs = 2, expensive_hrs = 1) produce asymmetric marks — a “free sell” without a paired buy is allowed when the cap suppresses one side. Net spread only deducts round-trip loss for complete pairs.

    Use-case configurations

    What the marks mean across all three configurations:

    Mark Meaning
    low Cheap interval — economically attractive to charge
    high Expensive interval — economically attractive to discharge
    null (neutral) Not marked — neither cheap nor expensive enough relative to the other

    1. Self-Consumption Maximizer

    Goal: Reduce grid imports for household loads; protect against price peaks; minimal grid interaction.

    Mark Action
    low Charge from grid
    neutral Charge PV only
    high Self-consumption (discharge to home loads, no export)
    Suggested starting values:

    min_delta ≈ 5-6¢/kWh, cheapest_hrs ≈ 2–4, expensive_hrs ≈ 4–6 — tune to your battery’s daily cycle budget.

    Caveats:

    • On sunny summer days, the battery may already be full from PV before a low interval begins. The grid-charge slot is then wasted.
    • high only saves the retail price you would have paid for grid import.
      • If PV is already covering loads during a high hour, discharging stored energy on top of that provides no additional benefit.
    • If a long stretch of high intervals occurs (cold evening, no sun), the battery can deplete before prices drop.

    2. Arbitrage Trader

    Goal: Buy cheap, sell expensive, profit on the spread. Treat the battery as a wholesale-market participant.

    Mark Action
    low Charge from grid (charge until full)
    neutral Full stop / minimal Charge PV top-up
    high Sell (until empty)

    Suggested starting values: min_delta ≈ 8–10¢/kWh (must cover round-trip losses plus some profit margin), cheapest_hrs and expensive_hrs matched to your battery’s daily energy budget so you don’t over-cycle.

    Caveats:

    • Requires a dynamic contract
    • In the Netherlands, the salderingsregeling phase-out (from 2027) progressively reduces the feed-in value; this configuration has a limited shelf life under current law.
    • Daily cycle count increases — faster return on investment — battery degradation may also accelerate.
    • Grid-export caps and dynamic transport tariffs can claw back profit; the algorithm does not see them.
    • House loads during high hours may be drawn from the grid at high prices after the battery is discharged — this can feel counter-intuitive even when the overall math is positive.

    3. PV-First Hybrid

    Goal: Run on own solar as much as possible; only use the grid when the spread genuinely beats PV self-consumption.

    Mark Action
    low Charge from grid (solar forecast; PV on sunny days, grid otherwise)
    neutral Charge PV
    high Self-consumption (no import/export)

    Suggested starting values: min_delta ≈ 0–2¢/kWh (since PV charging is essentially free, almost any spread is profitable), cheapest_hrs ≈ 1 (avoid charging more than needed when PV is plentiful), expensive_hrs = 0 (no limit).

    Caveats:

    • Requires solar forecast to be configured. Easy, see.
    • This is the most conservative on battery wear of the three configurations — fewest grid-charge cycles — at the cost of missing some spread opportunities.

    What’s different from v1

    v1 uses fixed blocks: “find the cheapest N consecutive hours; find the most expensive M consecutive hours.” This works on a clean daily curve but misses opportunities on flat-price days or days with multiple peaks and troughs.

    Key difference: the fixed-block approach can only pick one expensive block — morning peak or evening peak, whichever scores higher. v2 identifies both peaks independently because it works on individual intervals without requiring them to be adjacent.

    On a flat-price day where no pair meets the threshold, v2 marks nothing — and that’s the correct behavior.

    How the algorithm works

    Under the hood — algorithm internals (for tweakers)

    The algorithm is called Extreme-Pair Matching and runs per local calendar day on the available price data. Per-day pairing means today’s marks don’t shift when tomorrow’s prices arrive.

    Step-by-step (per day):

    1. Take all intervals for the day and sort them by price.
    2. Point to the cheapest interval (low pointer) and the most expensive interval (high pointer).
    3. Calculate the spread: price_high − price_low.
    4. If the spread is ≥ min_delta:
      • Mark the cheap interval as low (unless this day’s cheapest_hrs cap is already reached).
      • Mark the expensive interval as high (unless this day’s expensive_hrs cap is already reached).
      • Move both pointers one step inward (next cheapest / next most expensive).
      • Repeat from step 3.
    5. If the spread falls below min_delta, or both per-day caps are reached: stop. No further intervals are marked for this day.

    The result is a set of low / high marks where each detected pair has a guaranteed minimum spread. With cheapest_hrs = 0 and expensive_hrs = 0, marks come in matched pairs — every high is backed by a low. With non-zero caps, one side of a detected pair may be suppressed, so the final mark counts can be asymmetric.

    The algorithm pairs by price only; it doesn’t check whether the cheap interval comes before the expensive one in time. See Current limitations for what that implies in practice.

    Why this algorithm? It marks the truly extreme hours (highest peaks as high, lowest troughs as low) — the intervals a user would intuitively expect. It is O(n log n), straightforward to implement, and easy to explain. It does not attempt a globally optimal pairing (which would require backtracking).

    Current limitations

    The algorithm identifies economically valid candidates; it does not enforce physical feasibility. Three things it cannot see:

    • Time direction — a high interval at 02:00 cannot be served by charging at a low interval at 11:00 of the same day; the battery would have needed to be charged the previous evening. Mitigation: use Solar Forecast charge or a Timed strategy to pre-charge for early-morning high intervals.
    • Battery capacity — if 8 intervals are marked low and 8 high but the battery only holds enough for 3 charge/discharge cycles, 5 of those marks are aspirational. Mitigation: the cheapest_hrs / expensive_hrs caps directly address this — set them to your battery’s daily cycle budget.
    • PV contribution — extra charge from solar during neutral intervals can make a planned low grid-charge unnecessary. Mitigation: set Low strategy to Charge PV instead of Charge on sunny days, or use a forecast-aware automation to switch dynamically.

    Feedback

    Lab features are experimental — your feedback shapes the next iteration. Report issues, ideas, or unexpected behavior at our Discord or the project’s GitHub Issues.

    ← Back to Home