基于EMD-LSTM的用电负荷预测

Electricity Load Forecast Based on EMD-LSTM

摘要: 基于电力数据分析处理的系统决策正在不断地推进电力系统优化，我国对于构建有效的电力大数据模型有着强烈的现实诉求。本文运用经验模态分解及小波变换来降低该序列中的环境噪声，并根据各个分解信号的样本熵值进行序列重组，以降低重复建模的工作量，进而提高模型预测精度。实证结果表明，经过降噪处理和数据重组后，长短期记忆网络模型的拟合效果较好，所建模型具有稳健性。

Abstract: System decision-making based on power data analysis and processing is continuously advancing power system optimization, which appeals a strong practical application for building an effective power big data model in our country. In this paper, empirical mode decomposition and wavelet transform are used to reduce the environmental noise in the sequence, and the series recombina-tion is performed according to the sample entropy of each decomposition signal to reduce the workload of repeated modeling and improve the prediction accuracy of the model. The empirical results show that the fitting effect of the long and short-term memory network model is better after reducing the noise and reorganization data, which shows our theoretical model effective and robust.