MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202621095540 A) filed by Mohanty, Ashrumochan; Mohanty, Roshan Suryakant; and Dora, Pratyush Kumar on August 06, 2026, for A Deep Learning-Based Transformer System For Short-Term Rainfall Forecasting.
Inventors include Mohanty, Ashrumochan; Mohanty, Roshan Suryakant; and Dora, Pratyush Kumar.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: The present invention relates to a deep learning-based transformer system for short-term rainfall forecasting. The system comprises a data acquisition unit (101) for multi-modal meteorological data, a preprocessing module (102), a spatiotemporal tokenization and embedding module (103), a hierarchical Transformer encoder (104) for multi-scale feature extraction, a cross-modal fusion module (105), a decoder module (106) generating rainfall intensity maps at multiple lead times within 0–6 hours, and a post-processing module (107). The architecture fuses radar, satellite, surface observations and numerical weather prediction guidance through hierarchical selfattention and cross-attention mechanisms adapted to the sparsity of precipitation fields. The invention provides improved forecast skill, efficient parallel computation, and robustness for operational nowcasting applications. (Figure 1)
Disclaimer: Curated by HT Syndication.
