基于ARIMA和SVR的短期电力负荷预测  

Short-term Power Load Forecasting based on ARIMA and SVR

作　　者：余志成 孙皓月[1] 张碧宁 YU Zhicheng;SUN Haoyue;ZHANG Bining(Hebei University of Architecture,Zhangjiakou,Hebei 075000)

机构地区：[1]河北建筑工程学院,河北张家口075000

出　　处：《河北建筑工程学院学报》2023年第3期189-196,共8页Journal of Hebei Institute of Architecture and Civil Engineering

基　　金：河北省高校基本科研业务费研究项目,基于机器学习的办公楼用电设备非侵入式监测研究(2022CXTD09);河北建筑工程学院研究生创新基金项目,基于机器学习的办公楼用电设备非侵入式监测研究(XY202236)。

摘　　要：为提高用电量的预测精确度,将自回归差分移动平均(ARIMA)与支持向量机(SVR)模型相结合来进行预测。以用电数据为研究对象,借助ARIMA模型实现对电力数据的线性趋势预测,通过SVR回归模型对ARIMA模型的预测残差进行数据修正,得到ARIMA_SVR模型的预测值。实验结果表明,相较于传统ARIMA模型,ARIMA与SVR组合模型的性能和预测精度都得到大幅提升,具有一定的学术价值和应用意义。In order to improve the prediction accuracy of power consumption,autoregressive differential moving average(ARIMA)and support vector machine(SVR)models are combined in the prediction.Taking the power consumption data as the research object,the linear trend prediction of power data is realized by using ARIMA model,and the prediction residual error of ARIMA model is corrected by SVR regression model to obtain Arima_Predicted value of SVR model.The experimental results show that compared with the traditional ARIMA model,the performance and prediction accuracy of ARIMA and SVR combined model have been greatly improved,which has certain academic value and application significance.

关 键 词：负荷预测 自回归差分移动平均模型 支持向量机算法 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]