广东电力2025，Vol.38Issue(3)：8-17,10.DOI:10.3969/j.issn.1007-290X.2025.03.002
基于改进AP聚类和双重注意力机制的区域级新能源超短期出力预测方法
Regional-level New Energy Ultra-short-term Output Prediction Method for New Energy Sources Based on Improved AP Clustering and Dual-attention Mechanism
摘要
Abstract
In order to improve the accuracy of new energy ultra-short-term power prediction,a new energy ultra-short-term power prediction method at the regional level is proposed based on the improved affinity propagation(AP)clustering and the dual attention mechanism,with full consideration of the spatial and temporal complementary characteristics of the power sources and the key meteorological information.Firstly,the evaluation indexes of complementarity between power stations are established,and the complementarity matrix of regional power stations is calculated,and the spatial clustering of regional power stations is carried out by using the improved AP clustering algorithm.Then,the attention mechanism in both temporal and feature dimensions is introduced to capture the key meteorological features of the aggregation area.Finally,the new energy output ultra-short-term prediction model based on the bi-directional long and short-term memory(Bi-LSTM)is established on the basis of the proposed approach.Through actual data verification,the prediction method proposed has higher accuracy than the overall regional prediction and the traditional AP clustering prediction.Meanwhile,compared with the traditional correlation coefficient method,the prediction model incorporating the attention mechanism in this paper is more effective in capturing the meteorological characteristics of the convergence area.关键词
新能源出力/超短期预测/近邻传播聚类/双向长短期记忆/注意力机制Key words
new energy output/ultra-short-term forecasting/affinity propagation(AP)clustering algorithm/bi-directional long and short-term memory(Bi-LSTM)/attention mechanism分类
动力与电气工程引用本文复制引用
苏华英,林晨,张俨,王融融,程春田,张俊涛..基于改进AP聚类和双重注意力机制的区域级新能源超短期出力预测方法[J].广东电力,2025,38(3):8-17,10.基金项目
贵州电网有限责任公司电力调度控制中心高比例新能源电网水电灵活性量化动态评估及智慧调控技术研究项目(GZKJXM20220086) （GZKJXM20220086）
国家自然科学基金重点项目(52239001) （52239001）
国家自然科学基金项目(52409010) （52409010）
