摘要
电力负荷具有非线性和时序性的特点,为了深入研究各特征变量对于电力负荷预测的重要性,进而获得更高的电力负荷预测精度,提出了基于随机森林(random forest,RF)算法及长短期记忆网络(long short-term memory,LSTM)的混合负荷预测模型。首先根据时间日期因素及气候因素建立高维特征数据集作为随机森林模型的输入,通过随机森林算法筛选出重要特征量,并使其与历史负荷结合作为LSTM模型的输入,经过粒子群算法对LSTM模型进行参数寻优后得到RF-LSTM混合模型及负荷预测结果。使用该方法对河北电网某台区的电力负荷进行预测,结果表明该混合模型的预测精度比未经特征变量筛选的传统单一的随机森林算法、LSTM模型以及BP神经网络更为理想。
Power load has the characteristics of non-linearity and timing.In order to dig deeper into the importance of characteristic variables for power load forecasting and obtain higher accuracy of power load forecasting,in this paper,a hybrid load forecasting model based on random forest(RF)algorithm and long-term and short-term memory(LSTM)neural network model is applied to load forecasting.First,according to the time and date factors and climate factors,a high-dimensional characteristic data set is established as the input of the random forest model,and then the important features selected by the random forest algorithm and combined with historical load as the input of the LSTM network model.The particle swarm optimization algorithm is used to optimize the parameters of the LSTM network model to get the final RF-LSTM hybrid model and the load forecasting results.This method is applied to predict the power load of a certain station in Hebei.The results show that the hybrid prediction model proposed in this paper has better prediction accuracy than random forest model,LSTM network model and BP neural network without characteristic variable screening.
作者
董彦军
王晓甜
马红明
王立斌
李梦宇
岳凡丁
袁欢
DONG Yanjun;WANG Xiaotian;MA Hongming;WANG Libin;LI Mengyu;YUE Fanding;YUAN Huan(State Grid Hebei Electric Power Co.,Ltd.,Shijiazhuang 050081,Hebei Province,China;Marketing Service Center of State Grid Hebei Electric Power Co.,Ltd.,Shijiazhuang 050081,Hebei Province,China;School of Electrical Engineering,Xi’an Jiaotong University,Xi’an 710049,Shaanxi Province,China)
出处
《全球能源互联网》
CSCD
2022年第2期147-156,共10页
Journal of Global Energy Interconnection
基金
国网河北省电力有限公司科技项目(SGHEDK00DYJS1900303)
国家自然科学基金项目(51877170)。
