基于多粒度泄漏积分回声状态网络的日供水量预测  

Prediction of Daily Water Supply Based on Multi-granularity Leakage Integral Echo State Network

作　　者：白云[1] 严政杰 张晋[2] 李川[1] BAI Yun;YAN Zheng-jie;ZHANG Jin;LI Chuan(School of Management Science and Engineering,Chongqing Technology and Business University,Chongqing 400067,China;State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering,Yangtze Institute for Conservation and Development,Hohai University,Nanjing 210098,China)

机构地区：[1]重庆工商大学管理科学与工程学院,重庆400067 [2]河海大学长江保护与绿色发展研究院水文水资源与水利工程科学国家重点实验室,江苏南京210098

出　　处：《中国给水排水》2023年第9期50-56,共7页China Water & Wastewater

基　　金：国家自然科学基金资助项目(72271036、71801044、42077156、52121006);重庆市自然科学基金资助项目(CSTB2022NSCQ-MSX0510)。

摘　　要：针对多粒度因子耦合对城市日供水量产生的不确定性影响,提出一种基于多粒度挖掘与泄漏积分型回声状态网络(LiESN)的组合预测模型X11+LiESN,以提高城市日供水量预测精度。利用重庆市某水厂2018年1月1日—2020年12月31日的日供水量数据对该方法进行有效性验证。结果表明,所提出模型的平均绝对百分比误差(MAPE)为3.42%,决定系数(R^(2))为0.862。与单一的LiESN、极限学习机(ELM)和BP神经网络(BPNN)相比,该模型预测精确度高,能够更好地描述日供水量变化趋势,显示出了其有效性和应用潜力。Aiming at the uncertain impact of multi-granularity factor coupling on urban daily water supply,a combined prediction model X11+LiESN based on multi-granularity mining and leakage integral echo state network(LiESN)was proposed to improve the prediction accuracy of urban daily water supply.The effectiveness of the model was verified by using daily water supply data of a water treatment plant in Chongqing from January 1,2018 to December 31,2020.The mean absolute percentage error(MAPE)of the proposed model was 3.42%,and the coefficient of determination(R^(2))was 0.862.Compared with single models of LiESN,extreme learning machine(ELM)and BP neural network(BPNN),the model exhibited higher prediction accuracy and better description of daily water supply trend,and thus showed its effectiveness and application potential.

关 键 词：X11分解算法 粒度挖掘 回声状态网络 水量预测 

分 类 号：TU991[建筑科学—市政工程]