Computer Engineering and Applications ›› 2022, Vol. 58 ›› Issue (22): 291-296.DOI: 10.3778/j.issn.1002-8331.2105-0111

• Engineering and Applications • Previous Articles     Next Articles

Gas Load Forecasting with Three-Way Residual Correction

WANG Bing, WU Siqi, FANG Yu   

  1. School of Computer Science, Southwest Petroleum University, Chengdu 610500, China
  • Online:2022-11-15 Published:2022-11-15

三支残差修正的燃气负荷预测

王兵,吴思琪,方宇   

  1. 西南石油大学 计算机科学学院,成都 610500

Abstract: Accurate prediction of gas load is very important to dispatch and plan gas usage. A single forecasting model can not achieve good forecasting effects in gas load forecasting. Therefore, this paper designs a combined forecasting model with three-way residual correction according to the characteristics of gas load data. The model firstly decomposes load sequence with robust locally weighted regression, then the combination of ARIMA(autoregressive integrated moving average) and LightGBM(light gradient boosting machine) is designed for the decomposed trend item, periodic item and remaining item, and three-way residual correction is designed to correct the prediction of LightGBM based on three-way decision. The experiment shows that the proposed combined model performs well and the prediction is better than common single models.

Key words: gas load forecasting, three-way decision, time series, ARIMA, LightGBM, robust locally weighted regression

摘要: 燃气负荷的准确预测对于燃气调度、规划燃气使用有着重要的意义。单一的预测模型在燃气负荷预测中不能取得很好的预测效果,故基于燃气负荷数据的特点设计了一种三支残差修正的燃气负荷组合预测模型。首先基于燃气负荷数据特点,采用鲁棒局部加权回归对负荷序列进行了分解,针对分解后的趋势项、周期项、余项设计了ARIMA(autoregressive integrated moving average)和LightGBM(light gradient boosting machine)的组合预测模型,然后结合三支决策理论设计了三支残差修正法对LightGBM的预测结果进行修正。实验结果表明该组合模型的表现良好,预测效果优于常见单一模型。

关键词: 燃气负荷预测, 三支决策, 时序预测, ARIMA, LightGBM, 鲁棒局部加权回归