基于Elman神经网络算法的用户短期用电量预测  

Short-Term Electricity Consumption Forecast of Users Based on Elman Neural Network Algorithm

作　　者：余红平 YU Hongping(State Grid Hubei Electric Power Co.,Ltd.,Laohekou Power Supply Company,Laohekou 441800,China)

机构地区：[1]国网湖北省电力有限公司老河口市供电公司,湖北老河口441800

出　　处：《通信电源技术》2023年第5期38-41,共4页Telecom Power Technology

摘　　要：针对用户短期用电量预测能力低下的问题,提出了神经网络算法模型,实现用电预测系统的设计。用电预测评定的功能设计,通过完全集合经验模态分解(Complete Ensemble Empirical Mode Decomposition with Adaptive Noise,CEEMDAN)算法和经验模态分解(Empirical Mode Decomposition,EMD)信号对用电预测数据进行评估和计算,进而实现对用电预测终端、电网负荷的评定。采用双层极限梯度提升(Extreme Gradient Boosting,XGBoost)算法构建弱学习器,提取用电预测数据的特征变量,调用权重和增益完成特征选择,建立好预测模型后进行负荷预测。实验表明,在进行用电预测的精确度测试时,用电预测的准确度可达97%。For the problem of short-term power consumption prediction ability is low,put forward the neural network algorithm model,in order to realize the electricity prediction system design,the design of electricity prediction evaluation function,also through Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN)algorithm using Empirical Mode Decomposition(EMD)signal evaluation and calculation of electricity forecast data,and realize the prediction of electricity terminal,power grid load evaluation.The two-layer algorithm is used to construct the weak learners,extract the feature variables of the power consumption prediction data,call the weight and gain to complete the feature selection,and make the load prediction after establishing the prediction model.The experiments show that the accuracy of electricity prediction achieves the accuracy of 97%.

关 键 词：用电预测 神经网络 负荷预测 极限梯度提升(XGBoost)算法 弱学习器 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]