基于Kmeans++-Bi-LSTM的太阳辐照度超短期预测  

ULTRA-SHORT-TERM FORECAST OF SOLAR IRRADIANCE BASED ON KMEANS++-BI-LSTM

作　　者：官松泽 唐钰本 蔡争 吴凌涛 郑含博 覃团发[3,4] Guan Songze;Tang Yuben;Cai Zheng;Wu Lingtao;Zheng Hanbo;Qin Tuanfa(College of Optical Science and Engineering,University of Electronic Science and Technology of China,Chengdu 611731,China;College of Electrical Engineering,Guangxi University,Nanning 530004,China;College of Computing and Electronic Information,Guangxi University,Nanning 530004,China;Guangxi Key Laboratory of Multimedia Communication and Network Technology,Guangxi University,Nanning 530004,China)

机构地区：[1]电子科技大学光电科学与工程学院,成都611731 [2]广西大学电气工程学院,南宁530004 [3]广西大学计算机与电子信息学院,南宁530004 [4]广西多媒体通讯与网络技术重点实验室(广西大学),南宁530004

出　　处：《太阳能学报》2023年第12期170-174,共5页Acta Energiae Solaris Sinica

基　　金：2020年度南宁市创新创业领军人才(团队)“邕江计划”项目(2020006);广西重点研发计划(AB23026037)。

摘　　要：针对地表太阳辐射的不确定性和随机波动性,进而对大型光伏发电并网对电力系统的稳定性造成冲击,提出一种新的太阳辐照度超短期预测方案。该方案通过使用皮尔逊相关性分析和无监督学习中的Kmeans++算法,对多种气象数据进行筛选,找出关键气象数据并进行划分以及添加标签,接着将带有标签的关键气象数据输入双向长短期记忆网络预测模型中,以达到10 min时间间隔的太阳辐照度超短期预测。结果表明所提预测模型相较于目前常用的模型提高了预测精度。A new ultra-short-term prediction scheme for solar irradiance is proposed to address the uncertainty and stochastic fluctuations of surface solar radiation,which in turn has an impact on the stability of large-scale photovoltaic power grid connection to the power system.The scheme uses Pearson correlation analysis and the Kmeans++algorithm in unsupervised learning to filter multiple meteorological data,identify and classify key meteorological data and add labels to them,and then feed the labelled key meteorological data into a bi-directional long-short term memory network prediction model to achieve a 10-minute ultra-short-term forecast of solar irradiance.The results show that the proposed prediction model has lower root mean square error and lower mean absolute error than the currently used models.

关 键 词：太阳辐射 预测 聚类分析 超短期 双向长短期记忆网络 

分 类 号：TM615[电气工程—电力系统及自动化]