摘要
为实现“双碳”发展目标和满足新型电力系统应用需求,亟需对用电进行精准预测。为了应对周期长、变化幅度大的数据,将KTR模型应用于电能负荷预测的实际场景中。该模型在时变系数回归的方法上进行改进,能够应对较长的时间序列,避免出现过拟合的情况;以及根据不同数据变化情况自适应地使用不同的核函数,保证模型学习与数据特征匹配。实验结果表明,使用通过最佳参数构建的KTR模型进行预测,其总体的电能负荷数据预测值和原始值的SMAPE为8.46%。此外,将文中方法与Prophet和SARIMA模型预测结果进行了对比,结果表明,文中方法的预测精度比另外两种模型分别高2.57%和9.23%,验证了该方法电能预测的准确性。
In order to realize the"double carbon"development goal and meet new power system applications require,accurate forecasting of power consumption is need to conduct.In order to cope with the long period and large variation of data,the KTR(kernel-based time-varying regression model)model is applied to the actual scenario of electric energy load forecasting.The model is improved on the method of time-varying coefficient regression,which can cope with longer time series and avoid overfitting.The different kernel functions are used adaptively according to different data changes to ensure model learning and data feature matching.The experimental results show that the overall SMAPE of the predicted and original electric energy load data is 8.46% by means of the KTR model constructed with the best parameters for prediction.The proposed method was compared with the prediction results of Prophet and SARIMA models,and the results show that the prediction accuracy of this method is 2.57% and 9.23% higher than that of other two models,respectively.It verifies the accuracy of the proposed method in electricity prediction.
作者
田野
王大鹏
刘荣权
钟佳晨
TIAN Ye;WANG Dapeng;LIU Rongquan;ZHONG Jiachen(NARI Technology Nanjing Control Systems Co.,Ltd.,Nanjing 211106,China;State Grid Inner Mongolia East Electric Power Co.,Ltd.,Power Supply Service Supervision and Support Center,Tongliao 028000,China;College of Artificial Intelligence,Nanjing Agricultural University,Nanjing 210095,China)
出处
《现代电子技术》
2023年第24期109-114,共6页
Modern Electronics Technique
基金
上海市大数据管理系统工程研究中心开放基金项目(HYSY21022)。
作者简介
田野(1988-),男,山西人,助理工程师,研究方向为电能计量管理;王大鹏(1988-),男,山东人,中级工程师,研究方向为营销计量、反窃降损;刘荣权(2001-),男,四川人,研究方向为智能数据分析;钟佳晨(1999-),男,江苏人,硕士研究生,研究方向为分布式优化、智能系统设计。
