An improved method of power generation forecasting

By using multidimensional heterogeneous data processing and deep learning networks, the problems of insufficient feature extraction and weak spatiotemporal correlation in the prediction of new energy power generation have been solved. High-precision prediction under extreme weather conditions has been achieved, which is highly adaptable and suitable for power grid dispatch.

CN122244706APending Publication Date: 2026-06-19ZHEJIANG UNIV OF TECH
0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
ZHEJIANG UNIV OF TECH
Filing Date
2026-03-17
Publication Date
2026-06-19

AI Technical Summary

⚠Technical Problem

Existing power generation forecasting technologies suffer from insufficient feature extraction, weak spatiotemporal correlation capture, and lack of online correction mechanisms when faced with complex and ever-changing real-world conditions, leading to decreased forecast accuracy and poor adaptability, especially under extreme weather conditions.

⚗Method used

By acquiring multidimensional heterogeneous data, performing multi-scale signal decomposition and spatiotemporal fusion, extracting features using deep learning networks, capturing spatiotemporal dynamic correlations by combining graph convolution and temporal attention modules, and introducing extreme weather recognition and incremental learning mechanisms for online correction.

🎯Benefits of technology

It significantly improves the stability and accuracy of new energy power generation forecasting, can maintain efficient forecasting under extreme weather conditions, adapts to data distribution drift, and provides reliable support for grid dispatching.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

This invention provides an improved power generation prediction method, relating to the fields of deep learning and new energy data processing technology. Its key feature is the acquisition of multidimensional heterogeneous data of the target new energy power station and its surrounding areas, including historical power generation data, numerical weather prediction data, and satellite cloud imagery data. The advantages of this invention are: through multimodal data fusion and multi-scale decomposition, it deeply mines the implicit features of meteorological conditions, images, and equipment status; utilizing dynamic graph convolution and temporal attention mechanisms, it accurately simulates the spatial propagation and long- and short-term temporal dependencies of meteorological systems; it introduces an extreme weather identification and incremental learning online correction mechanism, significantly improving prediction stability under severe conditions such as typhoons; and it possesses adaptive update capabilities, effectively addressing seasonal drift in data distribution, providing reliable technical support for the safe and stable operation of the power grid.
Need to check novelty before this filing date? Find Prior Art