摘要
本文通过对时间序列的研究分析,提出基于ARMA模型和基于Holt-Winters指数平滑模型进行企业用电量预测的方法。对采集的中山市某企业用电量数据样本分别用上述两种模型建模和后续30天的用电量数据预测,分析样本数量对预测准确度的影响、用累计短期高频数据预测中长期数据的效果,比较两种模型的预测准确度。结果显示:基于ARMA模型的预测方法对于企业未来30天逐日用电量预测能够达到平均95%以上的拟合度,基于Holt-Winters指数平滑模型累计逐日预测数据预测下1个月用电量数据具有高达99%以上的拟合度,二者组合应用具有较高的可行性和推广价值。
With the research and analysis of time series,two electricity-consumption forecasting methods for enterprises based on ARMA model and Holt-Winters index smoothing model were put forward.Firstly,the two models above were respectively carried out on the collected electricity-consumption data samples in order to forecast the next 30 daily electricity-consumption of an enterprise from Zhongshan City.Secondly,the impact of the sample size on the prediction accuracy,the effect of accumulating short-term and high-frequency data on the forecasting of medium and long-term data were analyzed.Finally,the prediction accuracy of the two models were compared.The results demonstrated that the proposed method based on ARMA model can achieve a high-level fitting degree,more than 95%in average,for the enterprise′s daily electricity-consumption forecasting in the next 30 days.To forecast the next month′s electricity-consumption data,that accumulating daily forecasting data based on Holt-Winters index smoothing model hit a much higher fitting degree,up to 99%.The combination of the two methods had high feasibility and a great value of application potentials.
作者
陈云浩
周冬
CHEN Yunhao;ZHOU Dong(Hunan Tianyu Energy Technology Co.,Ltd.,Changsha 410205,Hunan,China;Sichuan Academy of Social Sciences,Chengdu 610072,Sichuan,China)
出处
《资源信息与工程》
2021年第4期131-136,共6页
Resource Information and Engineering
