风电功率概率预测模型的构建方法和风电功率概率的预测方法
By constructing a wind power probability prediction model, wind farm data is decoupled into three types of heterogeneous feature streams and processed using feature mapping and asymmetric interaction modules. This solves the timeliness and accuracy problems in wind power prediction and achieves efficient and reliable wind power prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUAZHONG UNIV OF SCI & TECH
- Filing Date
- 2026-04-14
- Publication Date
- 2026-07-17
AI Technical Summary
Existing wind power forecasting technologies suffer from poor timeliness and forecasting performance, especially when dealing with multi-source heterogeneous data, where accuracy is significantly lacking. Furthermore, long-sequence computational costs are high, and they cannot effectively capture the dynamic characteristics of meteorological data and the non-stationary features under abrupt changes.
A wind power probability prediction model is constructed, which decouples the historical operation data of wind farms into three types of heterogeneous feature streams. The model is trained through a feature alignment module, a feature mapping module, an asymmetric interaction module, and a dimension recovery module. Feature processing is performed using a gated residual extension network and a conditional variable attention network to generate a monotonically increasing cumulative distribution function model.
It effectively reduces the computational cost of multi-source external covariates, improves the accuracy and reliability of prediction results, ensures the consistency of distributed prediction results in physical and mathematical logic, and provides highly reliable data support for risk assessment and dispatch decisions of power grids under complex operating conditions.
Smart Images
Figure CN122412908A_ABST