摘要
目前我国的分布式光伏短期功率预测多缺乏功率与气象要素的历史实测和预报数据,难以直接复制集中式光伏的成熟技术路线,现有无辐照预测路线误差较大。本系列论文分为上下两篇阐述一种基于数据增强的分布式光伏电站群预测技术。此文为上篇,研究数据增强技术,包括基于时空相关性的缺失功率数据重构方法和基于三维神经网络的辐照预测加密模型,实现了功率数据、关键气象变量的高分辨率网格覆盖,为下篇网格化精细预测奠定了基础。仿真结果表明,相较于现有方法精度有显著提高,并适用于小样本训练。
The short-term forecast of distributed solar generation in China is usually faced with the lack of historical and meteorological data on the power curve.As a result,it is almost impossible to copy the mature technology that has been widely used in the centralized PV stations.The pre-existing methods are designed without radical prediction,and not good enough in preciseness.This series of papers include two parts to propose the distributed PV cluster forecast technology based on data augmentation.Part Ⅰ here focuses on the data augmentation method,including the power curve reconfiguration based on the spatial-temporal relationship and the radical prediction density improvements using the 3 D convolution neutral network.In this way,the power curves and the key meteorological factors are covered matched on the to-be-predicted area grid,which provides a possibility for the elaborate prediction in Part II.Simulation has shown there are advances in accuracy comparing with the pre-existing methods,and the proposed method is suitable for small sample set training.
作者
乔颖
孙荣富
丁然
黎上强
鲁宗相
QIAO Ying;SUN Rongfu;DING Ran;LI Shangqiang;LU Zongxiang(State Key Lab of Control and Simulation of Power Systems and Generation Equipments(Dept.of Electrical Engineering,Tsinghua University),Haidian District,Beijing 100084,China;State Grid Jibei Electric Power Co.,Ltd.,Xicheng District,Beijing 100054,China)
出处
《电网技术》
EI
CSCD
北大核心
2021年第5期1799-1808,共10页
Power System Technology
基金
国网冀北电力有限公司科技项目“冀北电网分布式电源监测、预测及运行分析技术研究”。
关键词
分布式光伏电站群
短期功率预测
数据增强
曲线重构
辐照加密
distributed PV stations
short-term power forecasting
data augmentation
curve reconfiguration
improving radical prediction density
