基于波动信息优选及切换输入机制的短期延长期风电集群功率预测  

POWER FORECAST OF WIND POWER CLUSTER IN SHORT-TERM EXTENSION BASED ON FLUCTUATION INFORMATION OPTIMIZATION AND SWITCHING INPUT MECHANISM

作　　者：杨茂[1] 鞠超毅 张薇[1] 苏欣 Yang Mao;Ju Chaoyi;Zhang Wei;Su Xin(Key Laboratory of Modern Power System Simulation Control and New Green Energy Technology of Ministry of Education(Northeast Electric Power University),Jilin 132012,China)

机构地区：[1]现代电力系统仿真控制与绿色电能新技术教育部重点实验室(东北电力大学),吉林132012

出　　处：《太阳能学报》2025年第3期546-558,共13页Acta Energiae Solaris Sinica

基　　金：国家重点研发计划项目(2022YFB2403000)。

摘　　要：在风电功率预测领域,现有短期时间尺度研究和应用的预见期最长为7d,缺乏对8~15d短期延长期时间尺度下的预测研究。针对上述问题,提出基于天气过程挖掘和切换机制的8~15d短期延长期预测框架,着重对未来出力水平进行预测,将历史选择分为波动性优先历史选择和稳定性优先历史选择,在波动性优先历史选择效果较差时,利用稳定性优先历史选择进行误差平衡。所提框架在甘肃省某风电集群进行验证,结果表明,所提框架均方根误差在8~15d所有时间尺度下平均降低0.84%~1.45%,在未来数值天气预报(NWP)可用性匮乏的情况下实现了8~15d预测,有效提高短期延长期预测的可靠性。In the field of wind power forecast,the longest forecast period for existing short-term time scale research and applications is 7 days,and there is a lack of forecast research on the 8-15 day short-term extension period time scale.To solve the above problems,this study proposed an 8-15 day short-term extension forecast framework based on the weather process mining and switching mechanism,focusing on the forecast of future output levels,and divided historical selection into volatility first historical selection and stability first historical selection,and used stability first historical selection to balance errors when the performance of volatility first historical selection was poor.The proposed framework was verified in a wind power cluster in Gansu Province.The results show that the root-meansquare error of the proposed framework decreases by 0.84 to 1.45 percentage points on average on all timescales of 8-15 days,and the 8-15 days forecast is realized under the condition of lack of availability of NWP,and the reliability of short-term extension forecast is improved.

关 键 词：风电功率 预测 切换机制 优选 短期 短期延长期 

分 类 号：TM614[电气工程—电力系统及自动化]