摘要
针对水利算据在归集过程中存在随机误差和噪声数据,导致水利模型的稳定性、可靠性和预测能力受到很大影响甚至失效的问题,提出一种基于SG平滑处理的水利算据有效性改进方法。抽取钱塘江流域开化县水文测站的洪水数据作为数据样本,通过多种不同平滑算法对比试验发现SG算法平滑效果最优,故用动态时间规整(DTW)算法计算洪水相似度时,采用SG算法平滑后的算据进行计算,根据计算结果寻找历史相似洪水,再根据相似洪水的洪峰演进时长的平均误差判定算据的有效性。开化县水文测站的60场洪水试验表明:利用SG算法平滑处理后的算据求得的相似洪水具有较高的匹配度,可保证DTW结果的稳定性、可靠性和预测能力,说明基于SG平滑处理的水利算据可以改进模型的有效性。
There are random errors and noise data in the collection process of water conservancy dataset,which greatly affect the stability,reliability,and predictive ability of water conservancy models,and even lead to failure.This paper proposes a method for improving the effectiveness of water conservancy dataset based on SG smoothing processing.Extracting the flood data from the hydrological station in Kaihua County of the Qiantang River Basin as data sample,comparative experiments with various smoothing algorithms reveal that the SG algorithm results in the best smoothing effect.Therefore,SG smoothed is used for data pre-processing when the Dynamic Time Warping(DTW)algorithm is applied to calculate flood similarity.According to calculation results,historical similar floods can be found,and the validity of the calculation data is determined based on the average error of the flood peak evolution duration of similar floods.Through 60 floods tests conducted at hydrological station in Kaihua County,the findings suggest that similar floods obtained from SG algorithm smoothed processing data have a higher degree of matching,which can ensure the stability,reliability,and predictive ability of DTW results.It indicates that water conservancy dataset based on SG smoothing processing can improve the effectiveness of the model.
作者
李军
傅媛媛
潘跃建
陈日剑
LI Jun;FU Yuanyuan;PAN Yuejian;CHEN Rijian(Zhejiang Design Institute of Water Conservancy and Hydroelectric Power Co.,Ltd.,Hangzhou 310002,China)
出处
《水利信息化》
2023年第6期51-55,61,共6页
Water Resources Informatization
关键词
水利算据
有效性
SG算法
平滑处理
DTW计算
water conservancy dataset
effectiveness
SG algorithm
smoothing processing
DTW calculation
作者简介
李军(1979—),男,上海人,高级工程师,主要从事水利信息化和数字孪生工程领域的研究和实践工作。E-mail:lj7906@mail.zdwp.com。
