摘要
家庭用电是能源市场的一个重要组成部分,预测家庭用电需求能够实现智能供电,可以有效地提高供给率,但目前预测方法大多效果不佳。针对此,提出了一种基于LSTM的面向家庭智能用电预测算法,建立了端到端的智能家庭用电预测模型。其在Boruta特征筛选的基础上设计了特征选择方法,对多个特征进行重要性计算,选取其中重要性高的部分进行建模,然后利用LSTM网络与全连接层对时间序列数据进行训练,得到预测模型。实验结果表明,所提方法的预测效果明显优于其他三种模型,能与真实数据较好地拟合。
Household power consumption is an important part of the energy market.Predicting household power demand can effectively improve energy supply efficiency,but most of the current prediction methods are not effective.To address the problem,this paper proposes an LSTM(Long Short-term Memory)-based power consumption prediction algorithm for smart home,where an end-to-end smart home power prediction model is established based on LSTM.A feature selection method that utilizes Boruta feature screening is also designed.The importance of multiple features is calculated,the most important part of those features is selected for modeling,and then LSTM network and the full connection layer are used to train the time series data to obtain the prediction model.The experimental results show that the prediction effect of the proposed method is significantly better than the other three models,and it can fit the real data well.
作者
周游
徐丹
赵灿
谭宇渲
ZHOU You;XU Dan;ZHAO Can;TAN Yuxuan(Suzhou Power Supply Branch,State Grid Jiangsu Electric Power Limited Company,Suzhou 215004,China)
出处
《软件工程》
2022年第2期39-41,38,共4页
Software Engineering
关键词
智慧能源
用电预测
特征选择
时间序列
LSTM网络
smart energy
power consumption prediction
feature selection
time series
LSTM network