大数据管理概念技术与挑战 (4).pdfVIP

44 卷 9 期 电 网 技 术 Vol. 44 No. 9 2020 年9 月 Power System Technology Sep. 2020 文章编号：1000-3673 （2020 ）09-3409-07 中图分类号：TM 715 文献标志码：A 学科代码：470 40 基于CNN-QRLightGBM 的短期 负荷概率密度预测 Short-term Load Density Prediction Based on CNN-QRLightGBM 1 关键词：Copula 函数；卷积神经网络；分位数回归； ABSTRACT: In order to cope with the active and unstable demand of electric power in the era of smart grid ，a new LightGBM ；核密度估计：TPE method of short-term public building load probability density DOI 10.13335/j.1000-3673.pst.2019.2243 prediction is proposed based on convolutional neural network 0 引言 and combined with quantile regression Light Gradient Boosting Machine (QRLightGBM). Firstly, the Copula model is used to 负荷预测能为电力系统的决策提供重要依据， analyze the correlation between variables and the feature 是智能电网中的一个关键环节。传统的点负荷预测 extraction is performed by the convolutional neural network. 方法只能提供未来某一时刻的单个值，而概率密度 Then the parameters of the regression prediction model are 预测通过构建概率密度函数提供对未来需求的完 tuned by the Tree-structured Parzen Estimator (TPE). Finally, [1] 整概率分布描述 ，这对智能电网的智能化部署提 the tuned model is used to predict the different quantile points, 供有利的数据支撑。 and the prediction results are estimated by kernel density to 基于机器学习的电力负荷预测算法可提供良 obtain the probability density function. The simulation results on the Maine dataset in New England, USA show that the 好的预测精度，如长短期记忆神经网络(long short- proposed method can better quantify the load probability term memory ，LSTM) 、集成学习算法随机森林 density prediction