基于图结构学习和负荷周期模式的短期电力负荷预测方法
By employing multi-scale temporal fine-grained input, similar graph learning, and dynamic non-stationary graph learning, combined with graph attention convolution and LSTM networks, spatial and temporal features are decoupled. This addresses the issues of poor graph structure quality and neglect of periodic patterns in short-term power load forecasting, achieving higher forecast accuracy and stability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG UNIV OF FINANCE & ECONOMICS
- Filing Date
- 2025-06-11
- Publication Date
- 2026-07-17
AI Technical Summary
Existing short-term power load forecasting methods face challenges such as poor graph structure quality, neglect of periodic patterns in load data, and coupling of spatial and temporal characteristics, resulting in insufficient forecast accuracy and reliability.
We employ a multi-scale temporal fine-grained input strategy, combining similar graph learning and dynamic non-stationary graph learning methods. Spatial features are extracted through graph attention convolution, temporal features are extracted through stacked LSTM networks, and spatial and temporal features are separated using a decoupling mechanism. Finally, prediction is performed through a multilayer perceptron.
It improves the accuracy and robustness of short-term power load forecasting, especially demonstrating strong robustness and real-time performance under non-stationary conditions.
Smart Images
Figure CN120706619B_ABST