Irradiance Ramp Rate

A foundational technique for intra-hour ramp prediction that derives wind velocity fields by tracking cloud feature displacement in consecutive sky images or satellite frames. CMV algorithms advect cloud fields forward in time to anticipate when a cumulus cloud will occlude the solar disk, directly forecasting the onset and magnitude of a ramp event before it impacts irradiance sensors.

A normalization metric defined as the ratio of actual GHI to theoretical irradiance under cloudless conditions. By dividing measured irradiance by the clear sky model, the index isolates the attenuation caused solely by clouds. A rapid drop in the Clear Sky Index from near 1.0 to below 0.3 over a few minutes is the direct signature of a high-magnitude ramp event, making it essential for ramp detection algorithms.

The naive baseline model that assumes current irradiance remains constant over the forecast horizon. For ramp rate analysis, persistence serves as the minimum skill threshold—any intelligent forecasting system must outperform the assumption that the sky condition will not change. During clear sky periods, persistence is highly accurate; during broken cloud conditions, it fails catastrophically at precisely the moments ramp prediction is most critical.

A prediction that outputs a full probability distribution or quantile range rather than a single deterministic value. For ramp events, probabilistic forecasts communicate the likelihood of exceeding critical ramp thresholds, enabling grid operators to hold dynamic operating reserves proportional to the quantified risk. Key outputs include:

• Prediction intervals (e.g., 10th–90th percentile)
• Exceedance probabilities for specific ramp magnitudes
• Temporal uncertainty around ramp onset timing

A strictly proper scoring rule that evaluates the full predictive distribution against the eventual observation. CRPS measures the integrated squared difference between the cumulative distribution function of the forecast and the empirical observation, penalizing both miscalibration (systematic bias) and lack of sharpness (excessive spread). It is the preferred metric for verifying probabilistic ramp forecasts because it rewards forecasts that are simultaneously confident and accurate.

A real-time, hourly-updating numerical weather prediction model operated by NOAA that provides 3-km resolution atmospheric forecasts over the continental United States. HRRR resolves mesoscale cloud features that coarser global models miss, making it a critical input for intraday ramp prediction. Its frequent update cycle captures evolving cloud fields, though its 15-minute output interval may still miss sub-minute ramp events that sky-imager-based methods can detect.