| 基于CNN-GRU分位数回归的短期母线负荷概率密度预测 |
| |
| 引用本文： | 臧海祥,刘冲冲,滕俊,孔伯骏,孙国强,卫志农.基于CNN-GRU分位数回归的短期母线负荷概率密度预测[J].陕西电力,2020,0(8):24-30,69. |
| |
| 作者姓名： | 臧海祥 刘冲冲 滕俊 孔伯骏 孙国强 卫志农 |
| |
| 作者单位： | （1. 河海大学 能源与电气学院， 江苏 南京 211100； 2. 国网扬州供电公司， 江苏 扬州 225009） |
| |
| 摘 要： | 随着分布式电源大规模并网，母线负荷的波动性和不确定性日益增加，给母线负荷预测带来新的挑战。传统的点预测方法难以对母线负荷的不确定性进行描述，为此提出一种基于卷积神经网络和门控循环神经网络分位数回归的概率密度预测方法。该方法通过卷积神经网络提取反映母线负荷动态变化的高阶特征，门控循环神经网络基于提取的高阶特征、天气、日类型等因素进行分位数回归建模，预测未来任意时刻不同分位数条件下的母线负荷值，最后利用核密度估计得到母线负荷概率密度曲线。以江苏省某市220 kV母线负荷数据进行测试，结果表明本文所提方法能够有效刻画未来母线负荷的概率分布，为配电网安全运行提供更多的决策信息。
|
| 关 键 词： | 母线负荷预测 概率密度 卷积神经网络 门控循环神经网络 分位数回归 |
Short-term Bus Load Probability Density Forecasting Based on
CNN-GRU Quantile Regression |
| |
| ZANG Haixiang,LIU Chongchong,TENG Jun,KONG Bojun,SUN Guoqiang,WEI Zhinong.Short-term Bus Load Probability Density Forecasting Based on
CNN-GRU Quantile Regression[J].Shanxi Electric Power,2020,0(8):24-30,69. |
| |
| Authors: | ZANG Haixiang LIU Chongchong TENG Jun KONG Bojun SUN Guoqiang WEI Zhinong |
| |
| Affiliation: | （1. College of Energy and Electrical Engineering， Hohai University， Nanjing 211100， China；2. State Grid Jiangsu Yangzhou Power Supply Company，Yangzhou 225009， China） |
| |
| Abstract: | With the large-scale grid connection of distributed power，the fluctuation and uncertainty of bus load are increasing，which brings new challenges to bus load forecasting. Traditional point forecasting methods are difficult to describe the uncertainty of bus load， therefore the probability density forecasting method based on CNN （convolution neural network）-GRU（gated recurrent unit） quantile regression is proposed in this paper. The CNN is used to extract the high-order features reflecting the dynamic changes of bus load. Based on the extracted high-order features，weather，day type and other factors，the GRU performs quantile regression modeling to predict the bus load value under different quantiles at any time in the future. Finally， the probability density curve of bus load is obtained by kernel density estimation. 220 kV bus load data of certain city in Jiangsu province is used for testing，the results show that the proposed method can effectively describe the probability distribution of future bus load and provide more decision-making information for the safe operation of distribution network. |
| |
| Keywords: | bus load forecasting probability density convolution neural network gated recurrent neural network quantile regression |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《陕西电力》浏览原始摘要信息 |
|
点击此处可从《陕西电力》下载全文 |
