【摘要】：針對月度風電曲線預測存在的預測變量多且關系復雜、可利用天氣信息少以及不確定性強等問題,提出了一種基于因子分析和神經網絡分位數(shù)回歸的月度風電曲線概率預測方法。采用因子分析對日內小時級風電功率序列向量降維,提取出相互獨立的風電公共因子作為預測變量,分別建立以日天氣特征為輸入的神經網絡分位數(shù)條件概率模型;利用中期天氣預報信息,預測未來30日各公共因子的概率分布;最后通過模擬服從預測分布的風電公共因子和各時刻特殊因子,并代入因子模型逐日還原風電預測曲線,生成未來月風電曲線的隨機場景。兩個實際風電場的預測結果驗證了所提風電曲線概率預測方法的準確性、適應性和高效性,為中長期風電功率概率預測提供了一種可行的解決思路。
[Abstract]:There are many forecasting variables and complex relationships in monthly wind power curve prediction, such as low availability of weather information and strong uncertainty, and so on. A probabilistic forecasting method of monthly wind power curve based on factor analysis and neural network quantile regression is proposed. Factor analysis is used to reduce the dimension of wind power series vector, and independent common wind power factors are extracted as prediction variables. The conditional probability models of neural network quantiles are established with daily weather characteristics as input. Using the medium-term weather forecast information, the probability distribution of the common factors in the next 30 days is predicted. Finally, the wind power forecasting curve is reduced day by simulating the common and special wind power factors which obey the forecast distribution, and the random scene of the future monthly wind power curve is generated by using the factor model to restore the wind power forecast curve day by day. The prediction results of two practical wind farms verify the accuracy, adaptability and efficiency of the proposed wind power curve probability prediction method, which provides a feasible solution for the medium-and long-term wind power probability prediction.
【作者單位】： 重慶大學電氣工程學院;南方電網科學研究院;
【基金】：國家自然科學基金項目(51177178,51677012) 重慶市科委基礎與前沿研究計劃項目(cstc2013jcyj A90001)~~
【分類號】：TM614

【相似文獻】

相關期刊論文 前2條

1 蘇方林;宋幫英;;中國電力消費影響因素權重的分位數(shù)回歸研究[J];河北科技大學學報;2010年04期

2 閻潔;劉永前;韓爽;王勃;;分位數(shù)回歸在風電功率預測不確定性分析中的應用[J];太陽能學報;2013年12期

相關碩士學位論文 前3條

1 羅小青;基于分位數(shù)回歸的中國GDP與電力消費量關系研究[D];華南理工大學;2015年

2 梁沛;風電功率短期預測方法研究[D];重慶大學;2016年

3 劉瑞;基于支持向量分位數(shù)回歸與智能電網的短期電力負荷概率密度預測方法[D];合肥工業(yè)大學;2017年

，

本文編號：2436193