摘要
为提高地区负荷预测的运算效率和预测精度,提出了一种基于卷积神经网络支持向量回归机的地区负荷聚类集成预测方法。首先,通过聚类模型对地区内大量用户的真实负荷数据进行分组并分析了不同聚类模型的效果。其次,使用得到的聚类分组标签将用户数据分组集成并构建训练数据。然后,基于改进的卷积神经网络构建了卷积神经网络支持向量回归机模型。最后,分组进行负荷预测并将预测结果求和得到地区最终预测月负荷,并与卷积神经网络模型、长短期记忆神经网络模型、决策树模型、支持向量回归机模型进行对比。文中使用扬中市高新区的负荷数据作为算例进行分析,结果表明文中所提方法相较于现有算法具有更高的负荷预测精度和运算效率。
In order to improve computation efficiency and prediction accuracy of regional load forecasting,this paper proposes a regional load clustering integrated forecasting method based on convolutional neural network support vector regression machine(CNN-SVR).Firstly,clustering model is applied to group the real load data of the users within the region and analyze the effects of different clustering models.Secondly,the cluster classification labels are obtained,and the user data groups are integrated to construct training data.Then,a convolutional neural network support vector regression model is constructed based on improved convolutional neural network.Finally,the prediction results are saved and summed to obtain the final predicted load of the region.Packet load forecasting is used and the prediction results are summed to obtain the final monthly load of the region,and compared with the convolutional neural network model,long short-term memory(LSTM)model,decision tree model and support vector regression model.Simulation using the data of Yangzhong High-tech Zone confirms that this proposed method has higher efficiency and accuracy of load prediction than currently used algorithms.
作者
沈兆轩
袁三男
SHEN Zhaoxuan;YUAN Sannan(School of Electronics and Information Engineering,Shanghai University of Electric Power,Yangpu District,Shanghai 200090,China)
出处
《电网技术》
EI
CSCD
北大核心
2020年第6期2237-2244,共8页
Power System Technology
关键词
负荷预测
卷积神经网络
支持向量回归机
聚类
load forecasting
convolutional neural network
support vector regression machine
clustering
