• CSCD核心库收录期刊
  • 中文核心期刊
  • 中国科技核心期刊

电力建设 ›› 2022, Vol. 43 ›› Issue (9): 117-124.doi: 10.12204/j.issn.1000-7229.2022.09.012

• 新型电力系统下配电网规划与运行优化关键技术研究及应用·栏目主持 王守相教授、赵倩宇博士· • 上一篇    下一篇

基于考虑气温影响的门限自回归移动平均模型居民日用电负荷预测

孙玉芹, 王亚文, 朱威(), 李彦   

  1. 上海电力大学数理学院,上海市 200090
  • 收稿日期:2021-11-25 出版日期:2022-09-01 发布日期:2022-08-31
  • 通讯作者: 朱威 E-mail:zhuwei_ok@126.com
  • 作者简介:孙玉芹(1971),女,博士,教授,主要研究方向为组合数学与图论、智能电网优化;
    王亚文(1994),男,硕士研究生,主要研究方向为电力大数据分析;
    李彦(1979),男,博士,讲师,主要研究方向为概率论与随机过程、电力大数据分析。
  • 基金资助:
    国家自然科学基金资助项目(11871377);国家自然科学基金资助项目(12071274)

Residential Daily Power Load Forecasting Based on Threshold ARMA Model Considering the Influence of Temperature

SUN Yuqin, WANG Yawen, ZHU Wei(), LI Yan   

  1. College of Mathematics and Physics, Shanghai University of Electric Power, Shanghai 200090, China
  • Received:2021-11-25 Online:2022-09-01 Published:2022-08-31
  • Contact: ZHU Wei E-mail:zhuwei_ok@126.com
  • Supported by:
    National Natural Science Foundation of China(11871377);National Natural Science Foundation of China(12071274)

摘要:

由于气温突变点的影响,负荷序列存在门限效应,导致传统线性时间序列模型的负荷预测效果较差。将气温突变点作为门限,建立了以气温为协变量的门限自回归移动平均(threshold autoregressive moving average with exogenous variable,TARMAX)模型,提高了预测精度。首先,应用马尔科夫链蒙特卡洛(Markov chain Monte Carlo,MCMC)方法对气温突变点进行搜寻得到模型参数。然后,采用随机搜索变量的方法快速选择出最优模型,有效降低选择时间序列模型的计算量。最后,对不同季节下的居民日用电负荷进行预测。实例表明,与线性时间序列模型、长短期记忆网络(long short-term memory network,LSTM)和多层感知机(multilayer perceptron, MLP)相比,TARMAX模型提高了电力负荷的预测精度。

关键词: 居民日用电负荷预测, 门限自回归移动平均(TARMA)模型, 气温突变点, 门限, 协变量

Abstract:

Due to the influence of the abrupt change-point of temperature, the load sequence has a threshold effect, which leads to poor load forecasting effects of traditional linear time series models. This paper uses the abrupt change-point of the temperature as the threshold and establishes a threshold autoregressive moving average model with temperature as the exogenous variable (TARMAX). The forecasting accuracy is improved. In this paper, the Markov Chain Monte Carlo (MCMC) method is firstly applied to search for the abrupt change-point of the temperature, and the model parameters are obtained. Then, the method of random search variables is used to quickly select the optimal model, which effectively reduces the amount of calculation for selecting the time series model. Finally, the residential daily power load in different seasons is forecasted. The example shows that, compared with the linear time series models, the long short-term memory network (LSTM), and the multi-layer perceptron (MLP), the TARMAX model improves the forecasting accuracy of the power load.

Key words: residential daily power load forecasting, threshold autoregressive moving average (TARMA) model, abrupt change-point of temperature, threshold, exogenous variable

中图分类号: