Making Regional Probabilistic Solar Forecasts from Deterministic Forecasts: Performance with Improved Deterministic Forecasts

This work updates previous work on making probabilistic solar forecasts for a fleet of solar plants, demonstrating performance improvements from an improved deterministic forecasts from a commercial vendor. Using the same open-source methods, reliability and sharpness of the probabilistic forecasts improved notably — the current forecast is about 40% sharper (narrower) than the previous forecast for the realized 95% and 99% prediction intervals relative to benchmark persistence ensembles. Grid operators with existing deterministic forecasts could replicate these methods to produce usable probabilistic forecasts.