水电站机电技术2024，Vol.47Issue(7)：40-46,7.DOI:10.13599/j.cnki.11-5130.2024.07.013
基于图注意力网络的短期负荷预测模型
Short-term load forecasting model based on graph attention network
摘要
Abstract
Existing methods have limitations in dealing with the nonlinearity,dynamic nature,and spatial and temporal dependence of load data,resulting in low forecast accuracy.In order to solve the above problems,this paper proposes a hybrid model based on the sequential variational mode decomposition(SVMD),graph attention network(GAT)and long short-term memory network(LSTM),and optimizes the parameters through beluga whale optimization(BWO).The model aims to improve the accuracy and robustness of short-term load forecasting by combining the advantages of all components.For the model,SVMD was first applied to effectively decompose load data and extract key modal components;then,GAT captured the correlation between the load fluctuation and weighted space to enhance the perception of the model on spatial characteristics of load data;LSTM extracted and forecasted time series characteristics;finally,BWO was used to optimize model parameters and improve the convergence speed and forecast performance of the model.The test results show that the SVMD-GAT-BWO-LSTM model performs best in terms of the mean absolute percentage error(EMAPE),root mean square error(ERMSE)and degree of fitting(R2)and shows high stability and robustness.The ablation tests further verified the importance of all components in the model and their contributions to improving forecast performance.关键词
负荷预测/图注意力/SVMD/白鲸优化/LSTMKey words
load forecasting/graph attention/SVMD/beluga whale optimization/LSTM分类
信息技术与安全科学引用本文复制引用
何群英,谷卫..基于图注意力网络的短期负荷预测模型[J].水电站机电技术,2024,47(7):40-46,7.基金项目
浙江同济科技职业学院科研项目(FRF23YB009) （FRF23YB009）
国家自然科学基金资助项目(62172363). （62172363）
