SOLAR PREDICTION STATS
MAE
—
kWh mean absolute error
CALIBRATION
—
actual / theoretical
PREDICTED vs ACTUALiHow well the solar forecast matches actual production. Bars close together = good. Consistent over-prediction (blue above orange) may mean your calibration factor needs lowering. Under-prediction means conservative forecasts — reserves may be set higher than needed.
Click a day to see hourly breakdown
ERROR TRENDiDaily prediction error as a percentage. The orange moving average should trend downward as the calibration factor adapts to your system. Spikes are normal on partly cloudy days. Persistent high error (>30%) suggests the system profile needs attention.
Percentage error with 7-day moving average
CONSUMPTION: PREDICTED vs ACTUALiCompares predicted home energy use against actual. Accurate consumption estimates improve reserve calculations. If actual consistently exceeds predicted, consider raising your default consumption setting. Large spikes may indicate unusual loads (EV charging, HVAC, etc).
Daily home consumption — per-day predicted vs actual
CONSUMPTION ERROR TRENDiTracks how far off consumption predictions are over time. A declining trend means the system is learning your usage patterns well. High variability is expected if your household load changes a lot day to day (e.g. work-from-home vs away days).
Percentage error with 7-day moving average
EXTENDED STATS
CONSUMPTION MAE
—
kWh mean absolute error
SURPLUS MAE
—
kWh net balance error
PROJECTION MAE
—
projected vs actual peak %
FULL PREDICTION
—
will-reach-full accuracy
FULL TIME ERROR
—
avg prediction error
RESERVE MARGIN
—
avg min% above reserve
SURPLUS/DEFICIT: PREDICTED vs ACTUALiNet energy balance: solar production minus consumption. Positive = surplus (battery fills, car can charge). Negative = deficit (grid needed). This directly drives reserve decisions — if the predicted deficit is consistently wrong, reserves will be over- or under-set.
Net energy balance (solar - consumption) — drives reserve decisions
PROJECTED vs ACTUAL PEAK %iCompares the projected battery peak (how full the battery was expected to get) against the actual peak reached. If projected consistently exceeds actual, solar is being over-estimated. If actual exceeds projected, you may have headroom to charge the car earlier.
Morning battery projection accuracy over time
RESERVE vs MIN BATTERYiShows the reserve level set each evening vs the battery's actual daily minimum. Ideally the minimum stays above the reserve line. Red markers mean the battery dipped below reserve — if this happens often, increase the safety margin. A large gap means reserves are overly conservative.
Applied reserve % with actual daily minimum — red marks days below reserve
PREDICTED vs ACTUALiScatter plot showing correlation between predicted and actual solar kWh. Points along the diagonal line = perfect predictions. R² close to 1.0 means strong correlation. Points below the line mean over-prediction; above means under-prediction. Clusters away from the line reveal systematic bias.
HOURLY BIASiShows which hours of the day tend to be over- or under-predicted. Solar-colored bars = over-prediction, green = under-prediction. Consistent bias at specific hours (e.g. morning shade, afternoon clouds) helps explain where the forecast struggles most. The hourly profile adapts over time to correct these.
Average prediction error by hour of day
OVER-PREDICTED
UNDER-PREDICTED
CALIBRATION FACTOR TRENDiThe calibration factor scales raw weather forecasts to match your system's real output. It accounts for panel efficiency, shading, orientation, and inverter losses. A stable value around 0.7-0.85 is typical. Large swings suggest variable shading or panel issues. The factor auto-adjusts based on recent actuals.
System efficiency modifier (actual ÷ theoretical) — how raw GHI projection is scaled to match your system's real output