Extending intraday solar forecast horizons with deep generative models

TL;DR

This paper introduces SHADECast, a deep generative diffusion model for probabilistic, high-resolution, near real-time solar irradiance nowcasting, significantly improving forecast accuracy and reliability for energy grid management.

Contribution

The paper presents SHADECast, a novel deep generative diffusion model that enhances spatiotemporal solar irradiance forecasting by conditioning on deterministic cloud evolution, outperforming existing methods.

Findings

SHADECast improves forecast CRPS by 15% over state-of-the-art.

Conditioning on deterministic forecasts boosts reliability by over 7%.

Model provides realistic, consistent predictions up to 120 minutes lead time.

Abstract

Surface solar irradiance (SSI) plays a crucial role in tackling climate change - as an abundant, non-fossil energy source, exploited primarily via photovoltaic (PV) energy production. With the growing contribution of SSI to total energy production, the stability of the latter is challenged by the intermittent character of the former, arising primarily from cloud effects. Mitigating this stability challenge requires accurate, uncertainty-aware, near real-time, regional-scale SSI forecasts with lead times of minutes to a few hours, enabling robust real-time energy grid management. State-of-the-art nowcasting methods typically meet only some of these requirements. Here we present SHADECast, a deep generative diffusion model for the probabilistic spatiotemporal nowcasting of SSI, conditioned on deterministic aspects of cloud evolution to guide the probabilistic ensemble forecast, and based on near real-time satellite data. We demonstrate that SHADECast provides improved forecast quality, reliability, and accuracy in different weather scenarios. Our model produces realistic and spatiotemporally consistent predictions outperforming the state of the art by 15% in the continuous ranked probability score (CRPS) over different regions up to 512 km x 512 km with lead times of 15-120 min. Conditioning the ensemble generation on deterministic forecasts improves reliability and performance by more than 7% on CRPS. Our approach empowers grid operators and energy traders to make informed decisions, ensuring stability and facilitating the seamless integration of PV energy across multiple locations simultaneously.