Enter Solar Forecast Data
Keep the page in a single vertical flow. The form fields below expand into 3 columns on large screens, 2 on tablets, and 1 on mobile.
Formula Used
This calculator compares forecasted and actual solar output interval by interval, then summarizes the error pattern with engineering performance indicators.
- Error: Error = Forecast − Actual
- Absolute Error: |Forecast − Actual|
- MAE: Average of absolute errors
- RMSE: √(average of squared errors)
- MBE: Average signed error
- MAPE: Average of |Error| / Actual × 100, excluding low-output intervals
- sMAPE: Average of 2 × |Error| / (|Forecast| + |Actual|) × 100
- nMAE and nRMSE: MAE or RMSE divided by plant capacity × 100
- R²: 1 − SSE / SST, showing how closely forecasts track actual changes
- Forecast Skill: 1 − RMSEforecast / RMSEpersistence
- Within Tolerance: Percentage of intervals whose absolute error stays below the selected capacity-based limit
- Band Coverage: Percentage of actual values inside the forecast confidence band
The accuracy index is a weighted engineering score built from normalized RMSE, normalized MAE, bias, and MAPE. Higher values mean better forecast quality.
How to Use This Calculator
- Enter the solar site name and a short study label.
- Provide plant capacity in kW and the interval duration in minutes.
- Set a tolerance percentage for acceptable forecast deviation.
- Set the exclusion percentage to ignore very low actual output during MAPE calculation.
- Set the confidence band percentage to test coverage around the forecast.
- Paste the forecast output series and actual output series using matching interval counts.
- Submit the form to view result cards, a Plotly graph, and a detailed error table.
- Use the CSV and PDF buttons to export the calculated results.
Example Data Table
| Interval | Forecast (kW) | Actual (kW) | Error (kW) |
|---|---|---|---|
| Interval 1 | 120 | 110 | 10 |
| Interval 2 | 145 | 150 | -5 |
| Interval 3 | 180 | 175 | 5 |
| Interval 4 | 210 | 205 | 5 |
| Interval 5 | 250 | 245 | 5 |
| Interval 6 | 285 | 278 | 7 |
You can replace the sample values with day-ahead, intraday, or hour-ahead solar forecasts from your site historian, SCADA log, or forecasting platform.
FAQs
1. What does this calculator measure?
It measures how closely forecasted solar output matches actual observed output. It reports bias, absolute error, square error, normalized scores, tolerance hit rate, confidence band coverage, and forecast skill against a persistence baseline.
2. Why is RMSE important for solar forecasting?
RMSE emphasizes larger misses because it squares each error before averaging. That makes it useful when major forecast misses can disrupt reserve scheduling, battery dispatch, curtailment decisions, or market participation.
3. What is the difference between MAE and MBE?
MAE shows typical error size without direction. MBE keeps the error sign, so it shows systematic over-forecasting or under-forecasting. Using both helps separate random spread from directional bias.
4. Why does the calculator exclude low-output intervals from MAPE?
MAPE can explode when actual output is very small. Excluding low-output points reduces misleading percentages around sunrise, sunset, clipping recovery, or cloudy ramp periods with near-zero production.
5. What does forecast skill mean?
Forecast skill compares your model against a persistence baseline. A positive value means your forecast beats the naive approach. A negative value means the baseline performed better for the submitted intervals.
6. What is band coverage used for?
Band coverage checks whether actual output stays inside a selected percentage band around the forecast. It is useful when operators want quick confidence checks on uncertainty and reserve adequacy.
7. Can I use energy values instead of power values?
Yes, but the two series must use the same unit and interval basis. If you input interval energy values, interpret the results accordingly and keep plant capacity consistent for normalized metrics.
8. What indicates a strong forecast result?
A strong result usually combines low nRMSE, low MAE, small bias, positive forecast skill, high tolerance hit rate, and good band coverage. Context still matters because weather volatility differs by site and season.