基于随机森林与卷积神经网络的电力负荷预测研究

肖玉东?覮

基于随机森林与卷积神经网络的电力负荷预测研究

引用本文：肖玉东?覮.基于随机森林与卷积神经网络的电力负荷预测研究[J].计算技术与自动化,2020,(3):13-16

摘要点击次数: 597

全文下载次数: 0

作者	单位
肖玉东?覮	（吐鲁番职业技术学院，新疆吐鲁番 838000）

中文摘要:针对电力负荷预测模型中变量冗余与拟合性能不佳问题，提出了应用随机森林（RF）筛选最优输入变量并结合卷积神经网络（CNN）的电力负荷预测模型。实证显示，经RF变量优选后模型平均MAE减少2.49%，EMSE减少3.40%；基于CNN神预测模型的平均MAE与RMSE分别降低了1.33%、2.46%。采用RF与CNN集成的方法具有最高的预测精度，其MAE为3.46%，RMSE为4.08%，该模型性能优于其他组合方案，是电荷预测精准建模的一种可靠方案。

中文关键词:随机森林卷积神经网络变量优选建模预测电力负荷

Research on Power Load Forecasting Based on Random Forest and Convolution Neural Network

Abstract:In order to solve the problem of variable redundancy and poor fitting performance in power load forecasting model，a power load forecasting model is proposed，which uses random forest（RF） to select the optimal input variables and convolution neural network （CNN）. The empirical results show that the average Mae and EMSE decrease by 2.49% and 3.40% respectively after RF variable optimization，and the average Mae and RMSE decrease by 1.33% and 2.46% respectively based on CNN prediction model. The integration method of RF and CNN has the highest prediction accuracy，with MAE of 3.46% and RMSE of 4.08%. The performance of the model is better than other combination schemes，and it is a reliable scheme for accurate charge prediction modeling.

keywords:random forest convolution neural network variable optimization modeling and forecasting power load

查看全文 查看/发表评论 下载pdf阅读器