基于Evolutionary Bagging集成学习的超短期风功率预测  

Ultra-short Term Wind Power Prediction Based on Evolutionary Bagging Ensemble Learning

作　　者：康英哲 田宇航 梁世昌 唐振浩[3] KANG Yingzhe;TIAN Yuhang;LIANG Shichang;TANG Zhenhao(Jilin Power Share Co.,Ltd,Changchun Jilin 132001;Siping first thermal power company,Jilin Electric Power Co.,LTD.Siping Jilin 136000;School of Automation Engineering,Northeast Electric Power University,Jilin Jilin 132012)

机构地区：[1]吉林电力股份有限公司,吉林长春132001 [2]吉林电力股份有限公司四平第一热电公司,吉林四平136000 [3]东北电力大学自动化工程学院,吉林吉林132012

出　　处：《东北电力大学学报》2024年第5期24-32,72,共10页Journal of Northeast Electric Power University

基　　金：吉林省科技发展计划项目(20220203102SF)

摘　　要：为实现风电功率的高精度超短期预测,文中基于Wasserstein距离和随机森林(Random Forest,RF)进行跨域特征选择,并将其与进化Bagging集成学习(Evolutionary Bagging,EvoBagging)相结合,提出了一种超短期风电功率预测的新方法。首先,将局部离群因子(Local Outlier Factor,LOF)算法用于异常值检测,并使用最近邻插值法(K-Nearest Neighbors Interpolation,K-NNI)替换原始数据中的异常值点;其次,对异常值处理后的数据使用经验模态分解(Empirical Mode Decomposition,EMD)算法分解并进行统计学计算以构建重构数据,使用Wasserstein距离和RF跨域特征选择对重构数据进行特征降维;最后,结合各个模型的优势提高模型的预测精度,构建以深度置信网路(Deep Belief Network,DBN)、深度神经网络(Deep Neural Networks,DNN)、轻量级梯度提升机(Light Gradient Boosting Machine,LGBM)、极限梯度提升算法(eXtreme Gradient Boosting,XGBoost)为基学习器的EvoBagging集成学习超短期风功率预测模型。经验证该模型对比单一模型预测误差平均减少5%,能够对超短期风功率实现高精度预测。In order to achieve high-precision ultra-short-term prediction of wind power,this study conducted cross-domain feature selection based on Wasserstein distance and Random Forest(RF),and combined it with Evolutionary Bagging(EvoBagging).A new method for ultra-short term wind power prediction is proposed.Firstly,the Local Outlier Factor(LOF)algorithm is used for outlier detection,and K-Nearest Neighbors InterpolationK-NNI is used to replace outlier points in the original data.Secondly,the data after outliers were decomposed by Empirical Mode Decomposition(EMD)algorithm and statistically calculated to build the reconstructed data,and Wasserstein distance and RF cross-domain feature selection were used to reduce the feature dimension of the reconstructed data.Finally,in order to combine the advantages of each model to improve the prediction accuracy of the model,It is constructed with Deep Belief Network(DBN),Deep Neural Networks(DNN),Light Gradient Boosting Machine(LGBM)and eXtreme Gradient EvoBagging ensemble learning ultra-short term wind power prediction model based on Boosting(XGBoost)learner.It is proved that the prediction error of this model is reduced by 5%on average compared with that of a single model,and it can achieve high precision prediction of ultra-short term wind power.

关 键 词：超短期风功率 异常值处理 数据重构 数据降维 集成学习 

分 类 号：TM614[电气工程—电力系统及自动化]