摘要
针对历史气象数据较少、天气波动较大时光伏出力预测精确度较低的问题,提出一种适用于小样本和多种天气下的分时段光伏出力综合预测法。该方法结合了分时段神经网络模型和相似时段筛选法,将分时段神经网络模型作为相似时段筛选法在相似度不够时的补充:分时段神经网络模型利用光伏出力历史数据对预测模型进行训练,采用近相似时段神经网络进行预测,摆脱了历史气象数据的制约。多种气象条件下对光伏出力的训练与预测验证了所提方法的有效性。
Aiming at the problem of low forecasting accuracy of photovoltaic output with inadequate historical meteorological data and severe weather fluctuations, a comprehensive segmented forecasting method suitable for small sample and various weather conditions is proposed, which combines the segmented neural network model and the similar period screening method. The segmented neural network model is used as a supplement to the similar period screening method when the similarity is not enough, which uses the historical photovohaic output data to train the forecasting model. The near-similar period neural network is adopted for forecasting, getting rid of the constraints of historical meteorological data. The effectiveness of the proposed method is verified by the training and prediction of photovoltaic output under various weather conditions.
作者
李建文
焦衡
刘凤梧
王雪莹
LI Jianwen;JIAO Heng;LIU Fengwu;WANG Xueying(State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources,North China Electric Power University,Baoding 071003,China)
出处
《电力自动化设备》
EI
CSCD
北大核心
2018年第8期183-188,共6页
Electric Power Automation Equipment
基金
河北省自然科学基金资助项目(E2017502053)
中央高校基本科研业务费专项资金资助项目(2017MS104)~~
关键词
光伏出力预测
分时段预测
相似时段
神经网络
photovoltaic output forecasting
segmented forecasting
similar period
neural network
