基于宽度回声状态网络的PCCP断丝智能预测  

Intelligent Prediction for PCCP Wire Breakage Based on Broad Echo State Network

作　　者：张海鹏 王建慧 邵青 张立 ZHANG Hai-peng;WANG Jian-hui;SHAO Qing;ZHANG Li(Beijing South-to-North Water Transfer Trunk Project Management Office,Beijing 100097,China;Beijing Water Science&Technology Institute,Beijing 100048,China)

机构地区：[1]北京市南水北调干线管理处,北京100097 [2]北京市水科学技术研究院,北京100048

出　　处：《计算机仿真》2024年第11期529-533,538,共6页Computer Simulation

基　　金：北京市科技计划课题(Z201100008220004)。

摘　　要：预应力钢筒混凝土管PCCP在长距离供水、市政管道等工程中应用广泛,但逐渐累积的断丝会威胁管道安全。为提前防范PCCP断丝事故,针对PCCP断丝的提前预测问题提出基于智能技术的预测方法。首先,对PCCP断丝数据进行统计分析,总结断丝数据规律;其次,提出基于宽度回声状态网络的PCCP断丝预测方法;最后,对断丝数据进行预测实验,结果表明该网络结构相对经典深度学习网络可以提高预测精度,实现对复杂时间序列数据的可靠预测。所提方法有助于提前感知断丝趋势,进而为PCCP管道预防式检修及风险预警提供重要参考信息。Pre stressed steel cylinder concrete pipe(PCCP)is widely used in long-distance water supply,municipal pipelines,and other engineering projects,but the accumulation of broken wires can threaten pipeline safety.To prevent the wire breakage accident of PCCP in advance,a prediction method based on intelligent technology was proposed for PCCP wire breakage.Firstly,the PCCP wire breaking data was statistically analyzed and the law of wire breaking data was summarized.Secondly,a prediction method based on broad echo state network was proposed.Finally,the prediction experiment was carried out,and the results show that the proposed network can improve the prediction accuracy compared with the classical deep learning network,and realize the reliable prediction of complex time series data.The proposed method can help perceive the tendency of wire breakage in advance and provide important reference information for preventive maintenance and risk warning of PCCP.

关 键 词：预应力钢筒混凝土管断丝 时序预测 宽度学习 回声状态网络 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]