电力学报2025，Vol.40Issue(1)：50-58,9.DOI:10.13357/j.dlxb.2025.006
基于气候分类和并行神经网络的短期负荷预测
Short-term Load Forecasting based on Climate Classification and Parallel Neural Networks
摘要
Abstract
Accurate load forecasting is an effective measure to reduce the waste caused by difficult storage of elec-tricity.Considering the obvious seasonal characteristics of regional loads,a short-term load forecasting method based on climate classification and parallel neural network is proposed in this paper.In the climate classifica-tion,K-means is used to cluster the strongly correlated features,and then the segmentation strategy is devel-oped to determine the boundaries,which ensures the continuity of the classification in time.In parallel net-work,the feature information of input data is extracted by CNN path and the long-term dependence relationship between data is learned by LSTM path.Then,CBAM assigns different weights to the input features to im-prove the feature extraction capability of the network.Finally,taking the load forecasting of a region in Shanxi as an example,the results show that the proposed method has higher prediction accuracy under various climate classifications than CNN-LSTM,LSTM,CNN and other forecasting models.关键词
负荷预测/气候分类/并行神经网络/空间聚类/时间分段/注意力机制Key words
load forecasting/climate classification/parallel neural network/spatial clustering/time slicing/atten-tion mechanism分类
信息技术与安全科学引用本文复制引用
牛天聪,武晓冬,王麟斌,席鹏辉,王正..基于气候分类和并行神经网络的短期负荷预测[J].电力学报,2025,40(1):50-58,9.基金项目
2023年山西省研究生教育创新计划支持项目(2023AL05). （2023AL05）
