基于多视角特征提取与多任务学习的光伏功率多步预测  

Multi-step Prediction of Photovoltaic Power Based on Multi-view Features Extraction and Multi-task Learning

作　　者：陈殿昊 臧海祥[1] 刘璟璇 卫志农[1] 孙国强[1] 李鑫鑫 CHEN Dianhao;ZANG Haixiang;LIU Jingxuan;WEI Zhinong;SUN Guoqiang;LI Xinxin(School of Electrical and Power Engineering,Hohai University,Nanjing 211100,China;Jiangsu Clean Energy Branch of Huaneng International Co.,Ltd.,Nanjing 210015,China)

机构地区：[1]河海大学电气与动力工程学院,南京211100 [2]华能国际电力江苏能源开发有限公司清洁能源分公司,南京210015

出　　处：《高电压技术》2024年第9期3924-3933,I0009-I0012,共14页High Voltage Engineering

基　　金：国家自然科学基金(52077062)。

摘　　要：准确的光伏功率多步预测结果对于电网的调度优化具有重要指导意义,针对现有光伏功率多步预测方法对历史数据特征提取不充分、忽略多步预测值之间的关联性而导致的预测精度不足等问题,提出了一种基于多视角特征提取与多任务学习的光伏功率多步预测方法。首先,为获得丰富且全面的特征信息,从时序、局部、全局3个不同的视角对输入数据进行特征提取;其次,将多步光伏功率预测任务转化为多个单步光伏功率预测子任务,使用基于注意力机制与专家网络的多任务学习模型进行多步预测,实现对多步预测值关联性的充分利用;最后,提出了一种改进的动态权重平均法对损失权重进行自适应优化调整,进一步提升模型性能。算例测试结果表明,该方法能够有效提高光伏功率多步预测的准确性。Accurate multi-step prediction results of photovoltaic(PV)power have important guiding significance for the scheduling optimization of power grid.To solve the problems of insufficient prediction accuracy caused by insufficient feature extraction of historical data and ignoring the correlation between multi-step prediction values,a multi-step predic-tion method of photovoltaic power based on multi-view feature extraction and multi-task learning was proposed.Firstly,in order to obtain rich and comprehensive feature information,feature extraction of input data is carried out from time se-ries,local,and global viewpoints.Subsequently,the multi-step PV power prediction task is transformed into multiple single-step PV power prediction sub-tasks,and multi-step PV power prediction is carried out by using the multi-task learning model based on the attention mechanism and expert network to realize the full utilization of the correlation be-tween multi-step prediction values.Lastly,an improved dynamic weight average method is proposed to adaptively optimize the loss weight to further improve the performance of the model.Experimental results show that the proposed method can be adopted to effectively improve the accuracy of photovoltaic power multi-step prediction.

关 键 词：光伏功率多步预测 多任务学习 特征提取 注意力机制 损失权重优化 深度学习 

分 类 号：TM615[电气工程—电力系统及自动化]