Work

Multimodal Fusion for Short-Term Solar Irradiance Forecast With Satellite and Extraterrestrial Irradiance Data

Year: 2026
Type: article
Abstract: The rapid growth of photovoltaic systems has created an urgent need for high-precision short-term Global Horizontal Irradiance (GHI) forecast to ensure power system stability and optimize dispatch operations. While existing models have shown promising results, many still fail to fully exploit the complementary nature of different data modalities or incorporate meaningful physical constraints. To address these limitations, this article proposes a novel short-term GHI forecast model that implements a sophisticated multimodal data fusion strategy. We integrate three critical data sources: satellite image for capturing spatiotemporal cloud dynamics, historical GHI values for identifying temporal patterns, and an innovative auxiliary extraterrestrial horizontal irradiance layer that encodes temporal, geographic, and extraterrestrial irradiance information. The model employs the parallel dual-branch architecture to effectively extract and process features from image modality data and time-series modality data, respectively. These heterogeneous features are then fused by a channel-attention-like feature fusion block and ultimately output. Comparative experiments demonstrate that the proposed model surpasses existing approaches in both accuracy and robustness, validating its effectiveness and practicality.
Language: English
Field: Energy
Open Access status: hybrid
Estimated APC paid: $2,645
OpenAlex ID: W7204188581
FWCI:
Cites: 32
Cited by:
Related to: 10