Forecast confidence

The plain-English version

Weather forecasts are never perfect. Tomorrow’s predicted solar is a best guess; the real day might come in higher or lower depending on cloud cover. Forecast confidence captures how seriously you want the simulator to take that prediction. What “more” or “less” cautious means depends on whether you have a grid connection or not — the same number drives two different conversations.

• 0.95 — very optimistic. The simulator plans as if tomorrow’s forecast is almost exactly right. Grid-tied: battery and EV strategies discharge more aggressively because they trust they’ll get a good refill tomorrow. Off-grid: shortfall-risk numbers come in low, headroom comes in high — only flagged-risky days show up as a problem.
• 0.85 — default. A sensible middle ground for typical conditions. The simulator hedges a little by assuming tomorrow only delivers ~85 % of its forecasted irradiance.
• 0.70 — conservative. The simulator plans like only 70 % of tomorrow’s forecast shows up. Grid-tied: battery is preserved more aggressively; export-revenue strategies hold back more reserve. Off-grid: more days are flagged as potentially short, so the Action plan recommends planning a generator run or a powered site.
• 0.50 or lower — very cautious / paranoid. Simulator behaves as if tomorrow might barely deliver anything. Grid-tied: almost no battery export; all surplus stays banked. Off-grid: most days flagged as risky — useful when you’re heading somewhere with no fallback (remote site, far-from-town off-grid home) and want the most pessimistic reading you can get.

On the chart, the same number controls the width of the shaded confidence band around the predicted solar curve. Lower confidence = wider band. The band gets wider further into the future too, because the further out a forecast looks, the less reliable it gets — with decreasing confidence at later dates.

If you’re off-grid

For caravan, RV, off-grid home, and other off-grid users the forecast-confidence number translates directly into shortfall risk — the chance of a day where the battery doesn’t have enough to cover load. The Week-ahead card’s three columns (Mean, Low, High) bracket how the day count of surplus vs shortfall days changes if the weather lands at the “cloudier than expected” or “clearer than expected” end of the band.

Practical reading rule: if the Low (cloudier) column shows shortfall days that Mean doesn’t, those are the marginal days — days where the forecast says you’ll be fine but a cloudier-than-expected weather pattern would tip you into shortfall. For caravan / RV trips, plan a powered site or a generator run for those days. For an off-grid home, plan to fire the backup generator long enough to cover the gap, or shift heavy loads to a known-surplus day.

A lower forecast-confidence value pushes Low further from Mean (wider band, more days flagged as marginal). A higher value tightens the band and trusts the nominal forecast. Set it based on how much fallback you have available — trips to remote country with no generator and no powered sites available warrant a lower setting; weekend caravan-park hops with a shore-power option warrant a higher one.

When to look at it

Most users can leave it at the default and never touch it. But if you find yourself in any of these situations, it’s worth tweaking:

• You’re running a Charge EV strategy and want the simulator to be more or less aggressive about draining the battery into the EV. Lower confidence = battery preserved; higher = more EV charge delivered.
• You’re looking at a forecast for a week with unsettled weather and the simulator’s default optimism feels wrong. Drop to 0.70 to see the conservative scenario.
• You’re planning an off-grid trip and want to know which days are at risk of shortfall under cloudier-than- expected weather. Run once at 0.85 to see the most-likely picture, then drop to 0.65 and watch which days flip from green to amber in the Action plan — those are the marginal days you should plan a generator or powered site for.
• You want to compare best-case vs worst-case daily outcomes for trip planning — run once at 0.95 and once at 0.65, and you bracket the realistic range.
• Your real production keeps coming in below predictions. Lowering forecast confidence pulls the simulator’s assumptions back toward what your roof actually delivers.

How the maths works (technical)

If you want the detail, here’s what the number does:

On day 1, the likely range on the predicted solar is ±(1 − c) of the predicted value, where c is the confidence. At c = 0.85 that’s a ±15 % band; at c = 0.70 it’s ±30 %.

For multi-day forecasts (Date range, Week ahead, Calendar year), the per-day uncertainties combine via the standard root-sum-square rule for independent measurements — band widens as √N across N days. In plain English: a 4-day-out forecast has roughly twice the uncertainty of day 1; a 9-day-out forecast has three times.

The band is also clamped to physical limits. No matter how low the confidence, the lower bound never goes below the maximally-cloudy floor (~10 % of clear-sky output, the diffuse-only contribution that even a fully overcast day delivers). The upper bound never goes above the clear-sky ceiling (the theoretical maximum at the same sun angle and panel orientation, with no atmosphere losses). These clamps prevent the band from claiming numbers that physics doesn’t allow.

Grid-tied: for look-ahead-driven strategies (Default with peak-export bonus, Charge EV with bonus battery discharge, Maximise Charge), the confidence factor also discounts tomorrow’s predicted solar when we work out the reservation, the charge the battery must keep at the end of the day to avoid buying power tomorrow. Lower confidence = higher reservation = less aggressive discharge today.

Off-grid: there’s no import-vs-export trade-off to reason about, so the confidence factor just sets the band width on solar irradiance. The simulator chains battery state day by day from the (perturbed) solar values; the Week-ahead card’s Low / Mean / High columns are re-simulations of that chain at the band’s lower / nominal / upper edges. A wider band — lower confidence — therefore produces more days where the battery’s minimum reserve gets touched in at least one of the three runs.