文章摘要
阎洁,李宁,刘永前,等.短期风电功率动态云模型不确定性预测方法[J].电力系统自动化,2019,43(3):17-23. DOI: 10.7500/AEPS20180213009.
YAN Jie,LI Ning,LIU Yongqian, et al.Short-term Uncertainty Forecasting Method for Wind Power Based on Real-time Switching Cloud Model[J].Automation of Electric Power Systems,2019,43(3):17-23. DOI: 10.7500/AEPS20180213009.
短期风电功率动态云模型不确定性预测方法
Short-term Uncertainty Forecasting Method for Wind Power Based on Real-time Switching Cloud Model
DOI:10.7500/AEPS20180213009
关键词: 风电功率预测  不确定性  概率预测  动态建模  云模型
KeyWords: wind power forecasting  uncertainty  probabilistic forecasting  real-time switching modeling  cloud model
上网日期:2018-12-26
基金项目:国家自然科学基金青年科学基金资助项目(51707063);国家重点研发计划资助项目(2016YFB0900100);中央高校基本科研业务费专项资金资助项目(2017MS024)
作者单位E-mail
阎洁 新能源电力系统国家重点实验室(华北电力大学), 北京市 102206
华北电力大学可再生能源学院, 北京市 102206 
yanjie_freda@163.com 
李宁 新能源电力系统国家重点实验室(华北电力大学), 北京市 102206
华北电力大学可再生能源学院, 北京市 102206 
 
刘永前 新能源电力系统国家重点实验室(华北电力大学), 北京市 102206
华北电力大学可再生能源学院, 北京市 102206 
 
李莉 新能源电力系统国家重点实验室(华北电力大学), 北京市 102206
华北电力大学可再生能源学院, 北京市 102206 
 
孔德明 国家电投集团东方新能源股份有限公司, 河北省石家庄市 050031  
龙泉 中国大唐集团新能源科学技术研究院有限公司, 北京市 100040  
摘要:
      高比例风电并网场景下,电力系统优化运行势必对风电功率预测精度及其不确定性分析结果的可靠性提出更高要求。现有的不确定性预测研究中大多为整体性的误差分析与建模,难以满足模型在各个时刻和各类天气下的适应性。因此,提出了动态云模型的短期风电功率不确定性预测方法。首先,建立各个预测功率区间段内的单点预测误差云模型,利用云模型数字特征(期望、熵、超熵)生成云滴分布图,以此量化预测不确定性态势。然后,计算给定置信水平下的云滴分位点,以及与之相对应的预测功率可能发生波动的置信范围,即风电功率预测不确定性分析结果。根据实时条件更新云模型,可以提高各个运行时刻点不确定性预测结果的可靠性。以中国北方某风电场运行数据为例进行验证,结果表明与传统的分位数回归方法相比,所提方法可靠性有所提升,能够为电力系统调度决策、备用安排等提供更为可靠的指导信息。
Abstract:
      Power system with high penetration of wind power requires much better performance for the accuracy of wind power forecasting (WPF) and reliability of its uncertainty analysis. Most of studies on the WPF uncertainty analysis mainly focus on the static error analysis and modelling for the overall span, and this will limit the model performance and adaptabilities at various time slots and weather conditions. Therefore, a time-switching cloud model is presented for the short-term WPF uncertainty analysis. First, the cloud models of the deterministic forecasting deviation for each selected predictive power range are established in a real-time updating manner. Then, the distribution of cloud drops can be generated according to three mathematical characteristics of the cloud, i.e. expectation, entropy and ultra-entropy. In this way, the uncertainty conditions of given predictive power ranges can be quantified. Moreover, by calculating the quantile of generated cloud drops, the uncertainty forecasting results can be achieved and expressed as the possible power range at given confidence level. The model performance at each time slot could be improved by updating the cloud model according to the current conditions. To take a Chinese wind farm as an example, the results show that the proposed method achieves more reliable uncertain intervals compared with the traditional quantile regression model and it can provide more reliable information for the dispatch and reserve of the power system.
查看全文(Free!)   查看/发表评论  下载PDF阅读器