Solar Panel Generation Forecast

This dashboard estimates hourly solar panel generation from weather-model shortwave radiation and air temperature. It starts with a normal residential-style default so nobody has to fill in a full PV design form: Denver, a 6 kW DC array, a fixed-tilt array-geometry proxy, 14 percent total PV losses, 96 percent inverter efficiency, and a 1.2 DC/AC ratio.

The forecast uses GFS, HRRR, or NBM data through the standard GribStream /api/v2/[dataset]/timeseries endpoint. The core weather input is DSWRF at the surface, NOAA's downward shortwave radiation flux field, combined with 2 m air temperature for a simple panel-temperature derate. The dashboard keeps the lead window to 24 or 36 hours so the returned rows stay hourly across the available model choices.

What the controls do

• Site label, latitude, and longitude: define the forecast point. The default is Denver, Colorado.
• Weather model: switches between GFS, HRRR, and NBM when the selected time window is compatible with that model.
• DC size: sets the PV array nameplate capacity in kW DC.
• Array geometry: applies a transparent scenario multiplier to horizontal shortwave radiation. It is useful for comparing horizontal, fixed-tilt, east-west, and tracker-style assumptions without pretending to be a full transposition model.
• Total losses: changes the non-weather PV derate applied after irradiance and temperature. The default follows the common PVWatts-style 14 percent assumption.
• DC/AC ratio: changes inverter sizing and clipping behavior by setting DC nameplate divided by AC nameplate.
• Max lead time: keeps the request inside a forecast horizon where returned rows are hourly. Actual coverage can be shorter than the dashboard time picker when the latest model cycle does not reach the full window.

How to read the panels

• Hourly AC output: estimated inverter-side power in kW. Because the dashboard constrains the query to hourly rows, the same value is also used as a one-hour energy estimate in kWh.
• Energy forecast: total estimated kWh across the returned hourly forecast rows in the selected time range.
• Peak and thermal stats: peak AC output, inverter limit, and the lowest temperature factor in the selected period.
• GHI and array-plane proxy: shows the model's horizontal shortwave radiation beside the selected array-geometry scenario.
• Forecast audit table: exposes the raw and derived fields so the assumptions are inspectable.

The math is intentionally lightweight. It is useful for comparing weather-driven production patterns, screening a proposed install, and explaining how GribStream expressions can turn forecast fields into application metrics. It is not a bankable PV design model and it does not replace PVWatts, SAM, site shade modeling, measured soiling, or an engineering-grade plane-of-array irradiance model.

The array-geometry control is deliberately conservative. A real tilt and azimuth calculation needs solar position, direct normal irradiance, diffuse horizontal irradiance, albedo, module orientation, row geometry, and shading. This dashboard only requests DSWRF and 2 m temperature from GribStream, so it exposes the geometry assumption as a named multiplier instead of hiding a fake precision model behind a tilt field.

Math and constants

The dashboard computes every derived field in GribStream expressions. With the default controls, the equations are:

1. temp_c = temp_k - 273.15
2. poa_wm2 = ghi_wm2 * 1.0800
3. cell_temp_c = temp_c + (poa_wm2 / 800.0) * 25.0000
4. temperature_factor = max(0, 1 + (-0.3500 / 100) * (cell_temp_c - 25.0))
5. loss_factor = max(0, 1 - 14.0000 / 100)
6. dc_power_kw = 6.0000 * poa_wm2 / 1000.0 * temperature_factor * loss_factor
7. ac_limit_kw = 6.0000 / 1.2000
8. ac_power_kw = min(dc_power_kw * 96.0000 / 100, ac_limit_kw)
9. hourly_energy_kwh = ac_power_kw, because the dashboard limits the query to hourly rows.

The most important constants are:

• 6.0000 is the default DC array size in kW. It is user-editable.
• 1.0800 is the default array-geometry multiplier, a fixed-tilt proxy applied to horizontal DSWRF. Other exposed scenarios use 1.0000, 0.9300, and 1.1800.
• 800.0 W/m2 and 25.0000 C form the simple cell-temperature rise estimate. They are hidden defaults, not a site-calibrated module thermal model.
• -0.3500 is the hidden power temperature coefficient in percent per C, applied relative to the 25.0 C reference temperature.
• 14.0000 is the default total non-weather loss percentage. It is user-selectable.
• 1.2000 is the default DC/AC ratio, so a 6 kW DC array has a 5 kW AC limit. It is user-selectable.
• 96.0000 is the hidden inverter efficiency percentage.
• 1000.0 W/m2 normalizes irradiance to a PV nameplate-style reference level.

These constants make the forecast explainable and easy to vary, but they are not a replacement for a module datasheet, measured albedo, measured soiling, shade geometry, or a full PVWatts/SAM model run.

Representative GribStream request

This example uses the default 6 kW fixed-tilt proxy at Denver with GFS. The dashboard fills the numeric defaults from Grafana variables and computes the PV estimates in one standard GribStream /timeseries call.

curl -X POST 'https://gribstream.com/api/v2/gfs/timeseries' \
  -H 'Content-Type: application/json' \
  -H 'Accept: text/csv' \
  -H 'Authorization: Bearer [API_TOKEN]' \
  -d '{
    "fromTime": "2026-05-03T12:00:00Z",
    "untilTime": "2026-05-05T00:00:00Z",
    "minLeadTime": "0h",
    "maxLeadTime": "36h",
    "coordinates": [
      { "lat": 39.7392, "lon": -104.9903, "name": "Denver, CO" }
    ],
    "variables": [
      { "name": "DSWRF", "level": "surface", "alias": "ghi_wm2" },
      { "name": "TMP", "level": "2 m above ground", "alias": "temp_k" }
    ],
    "expressions": [
      { "expression": "temp_k - 273.15", "alias": "temp_c" },
      { "expression": "ghi_wm2 * 1.0800", "alias": "poa_wm2" },
      { "expression": "temp_c + (poa_wm2 / 800.0) * 25.0000", "alias": "cell_temp_c" },
      { "expression": "func.Max(0.0, 1 + (-0.3500 / 100.0) * (cell_temp_c - 25.0))", "alias": "temperature_factor" },
      { "expression": "func.Max(0.0, 1 - 14.0000 / 100.0)", "alias": "loss_factor" },
      { "expression": "6.0000 * poa_wm2 / 1000.0 * temperature_factor * loss_factor", "alias": "dc_power_kw" },
      { "expression": "6.0000 / 1.2000", "alias": "ac_limit_kw" },
      { "expression": "func.Min(dc_power_kw * 96.0000 / 100.0, ac_limit_kw)", "alias": "ac_power_kw" },
      { "expression": "ac_power_kw", "alias": "hourly_energy_kwh" },
      { "expression": "func.Max(0.0, (6.0000 - ac_power_kw) / 6.0000 * 100.0)", "alias": "unused_capacity_pct" }
    ]
  }'

Why the defaults are reasonable

• Total losses: the default uses a PVWatts-style 14 percent loss assumption and treats it as a single transparent derate.
• Temperature derate: the preset uses a typical crystalline-silicon style negative power coefficient, applied against estimated cell temperature above 25 C.
• DC/AC ratio: the default residential value allows mild inverter clipping while avoiding an overcomplicated inverter model.
• Array geometry: v1 folds tilt, orientation, or tracking assumptions into one named multiplier because DSWRF is a horizontal radiation field.

Background and references: GribStream expressions · GFS model inventory · HRRR model inventory · NBM model inventory · NOAA GRIB2 shortwave radiation table · PVWatts V8 API · SAM irradiance definitions · Grafana variables.