摘要
随着经济发展,对城市用电量进行精确预测变得越来越重要。以泰州市为实验区,以泰州市2011—2016年的用电量为实验数据,使用Tensorflow为人工神经网络框架,采用了分段和多参数2种方法提高预测电量的精度,分别是针对月份进行分段训练,将跨年和夏季用电高峰区分开来,同时,也加入经济和地理因素作为人工训练神经网络的参数,得到较好的精度结果:跨年月份为89.30%,夏季7、8月份为90.02%,其他月份为93.60%。实验证明,采用分段和多参数方法,能够切实提高用电量预测的精度,具有较好的实用性。
With the development of economy,accurate prediction of urban electricity consumption is becoming more and more important. In this paper,Taizhou city is taken as experimentation area and its electricity consumption from 2011 to 2016 as the experimental data,and two methods are adopted to improve the accuracy of prediction with tensorflow used as the artificial neural network frame. The data is trained for each month respectively with peak areas for the new year period and summer distinguished. In addition,the economic and geographical factors are added as parameters of artificial training neural network. As a result,better precision results are obtained: 89.30% in New Year period;90.02% in summer,and 93.60% in the other months. The experimental results show that the method of subsection and multi parameter can effectively improve the accuracy of power consumption prediction,and has better practicability.
作者
冯伟
蒋玮
杨乐
姚建光
郭亮
吴倩
汤海波
FENG Wei;JIANG Wei;YANG Le;YAO Jianguang;GUO Liang;WU Qian;TANG Haibo(Taizhou Power Supply Company,State Grid Jiangsu Electric Power Co.,Ltd.,Taizhou 225300,Jiangsu,China;Southeast University,Nanjing 210096,Jiangsu,China)
出处
《电网与清洁能源》
2018年第6期19-25,32,共8页
Power System and Clean Energy
基金
国家重点研发计划项目(2016YFB0901104)
国网江苏省电力有限公司科技项目(J2017112)~~
