基于集合经验模式分解的ARIMA行业售电量预测模型

ARIMA model based on ensemble empirical mode decomposition for industry electricitysales prediction

售电量的准确预测是电力市场课题研究的重要内容之一,目前已有许多模型用于售电量预测。在此背景下,考虑售电量时间序列的非线性、波动性和周期性,提出基于集合经验模式分解和自回归积分滑动算法的预测模型。该模型首先对售电量时间序列进行集合经验模态分解,通过添加白噪声得到不同时间尺度分布的售电量时间序列,分解后得到一系列相对平稳的本征模态函数和趋势项,然后利用自回归积分滑动算法对各平稳化本征模态函数和趋势项分别进行预测,得到各分量的预测结果,最后将分量预测结果叠加得到最终的售电量预测值。基于历史统计售电量数据的预测结果分析表明,基于集合经验模式分解的 ARIMA模型具有良好的预测精度。

The accurate prediction of electricity sales is an essential part of electricity market resrearch.This paper presents an ARIMA model based on the ensemble empirical mode decomposition in consideration of the nonlinearity,volatility and periodicity of electricity sales.Firstly,the ensemble empirical mode decomposition is utilized to decompose the time series of electricity sales with different time scales by adding white noise.After decomposition,a series of relatively stable intrinsic mode functions and trend items are obtained.Then,the ARIMA model is utilized to predict the steady-state intrinsic mode functions and trend items respectively in order to get the prediction results of each component.Finally,the result of component forecasting is added together to get the final forecast of electricity sales.The analysis of prediction results based on the historical sales data shows that ARIMA model based on set empirical mode decomposition has good prediction accuracy.