摘要
根据一个大城市供水管网连续八年的漏损事故数据资料的调查,对镀锌钢管出现的漏损事故特点进行详细分析;选择遗传程序设计和进化多项式回归两种方法,分别建立了两类镀锌钢管的漏损预测模型,利用实际数据对预测模型的预测精度进行了验证;根据结果对两种方法所建立的模型进行多角度对比评价,提出了最终推荐的模型公式。研究表明:两类预测模型均能较好地反映镀锌钢管漏损水平和不同管径漏损事故数量的变化趋势;其中,进化多项式回归建模效率更高,且公式更为简洁,适合不同管径组合建模,而遗传程序设计构建的模型应用选择范围更广,适合单独管径建模。对于不同管径的管道,建模数据采用综合或单一分组取决于漏损事件所表现出的规律是否具有一致性。
According to investigation of a large city water supply network for eight consecutive years, leakage accident data of galvanized steel pipe leakage accident characteristics were analyzed in detail. Genetic programnfing （GP） and evolution polynomial regression （EPR） were selected to build galvanized steel pipe failure prediction models respectively. Actual data of the same network was used to verify resulting prediction models in order to improve model accuracy. Based on the results, multiple perspectives comparison was made to propose the model of final recommended formula. It was determined that both two prediction models could reflect actual situation and variation trend of leakage at different diameters. Among them, the model based on EPR method had higher efficiency in model building and more concise in formula, which was suitable for different diameter combination modeling. Another model established by GP was broader in range selection, applicable to separate diameter modeling. For pipes with different diameters, the use of synthetic or single diameter data for modeling depended on whether the regular pattern presented by the leakage event was consistent.
作者
李岚
吴珊
侯本伟
李云峰
Li Lan;Wu Shan;Hou Benwei;Li Yunfeng(College of Civil Engineering,Beijing University of Technology,Beijing 100124,China;Beijing Waterworks Group Co.,Ltd.,Beijing 100031,China)
出处
《净水技术》
CAS
2018年第9期23-31,共9页
Water Purification Technology
基金
国家水体污染控制与治理科技重大专项资助(2017ZX07108)
关键词
漏损
镀锌钢管
遗传程序设计
进化多项式回归
预测模型
water leakage
galvanized steel pipe
genetic programming（GP）
evolutionary polynomial regression（EPR）
prediction medel
