基于小波变换和LSTM模型的城市天然气负荷预测  

CITY NATURAL GAS LOAD FORECASTING BASED ON WAVELET TRANSFORM AND LSTM MODEL

作　　者：那幸仪 贾俊铖[1,2] 赵晓筠 张莉[1,2] 李凡长[1,2] Na Xingyi;Jia Juncheng;Zhao Xiaojun;Zhang Li;Li Fanzhang(School of Computer Science and Technology,Soochow University,Suzhou 215000,Jiangsu,China;Provincial Key Laboratory for Computer Information Processing Technology,Soochow University,Suzhou 215000,Jiangsu,China;Suzhou Gas Limited Liability Company,Suzhou 215000,Jiangsu,China)

机构地区：[1]苏州大学计算机科学与技术学院,江苏苏州215000 [2]苏州大学计算机信息处理技术省级重点实验室,江苏苏州215000 [3]苏州燃气集团有限责任公司,江苏苏州215000

出　　处：《计算机应用与软件》2021年第12期61-66,共6页Computer Applications and Software

基　　金：江苏高校优势学科建设工程项目(PAPD);中国博士后科学基金项目(2017M611905);江苏省高等学校自然科学研究面上资助经费项目(17KJB520034);苏州市科技项目(SS201701,SYSD20192152)。

摘　　要：城市燃气公司为防止出现储气量过剩或不足等问题,必须要对天然气负荷量进行预测。对此,提出一种基于小波变换和长短期记忆网络的深度学习模型(wavelet_LSTM),对未来一年的天然气日负荷量进行预测。在某市燃气公司真实数据集上的实验结果表明,该模型有较高的准确性,可以较好地满足城市燃气公司的业务需求。In order to prevent problems such as excess or insufficient gas reserves,city gas companies must forecast the natural gas load.This paper proposes a deep learning model based on wavelet transform and long short term memory network(wavelet_LSTM)to predict the daily natural gas load in the coming year.The experimental results on a real data set of a city gas company show that this model has higher accuracy and can better meet the business needs of city gas companies.

关 键 词：长短期记忆网络 小波变换 天然气 负荷预测 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]