水资源监测异常数据模态分解-支持向量机重构方法

发布时间：2018-07-03 14:12

本文选题：水资源监测 + 异常数据　；参考：《农业机械学报》2017年11期

【摘要】：完备真实的水资源监测数据是支撑数据分析与决策的基本前提。在梳理现阶段水资源监测异常数据的基础上,提出运用移动平均拟合初筛来直观辨识异常监测数据,进而选取集合模态分解对非可直观辨识异常监测数据进行挖掘的方法。将剔除异常监测值后的时序数据作为基于粒子群优化最小二乘支持向量机模型的模拟样本,并利用其恢复所剔除的异常监测数据。对水务公司日取水量监测数据的实证分析结果表明,通过移动平均拟合与模态分解可较大限度地保留含有异常数据的特征向量并实现数据的有效重构,相比传统的统计方法其具有更好的适用性;运用粒子群优化的最小二乘支持向量机可进一步提高对剔除异常值数据的拟合效果,且符合水资源监测数据的季节波动规律特征及对实际取用水状态的客观反映,据此可相对合理地达到恢复所剔除异常监测数据的目的。
[Abstract]:Complete and real monitoring data of water resources is the basic premise to support data analysis and decision-making. On the basis of combing the abnormal data of water resources monitoring at the present stage, this paper puts forward a method of using moving average fitting initial sieve to identify the abnormal monitoring data intuitively, and then select the set mode decomposition to mine the non-intuitively identifiable anomaly monitoring data. The time series data after eliminating the abnormal monitoring values are used as the simulation samples based on the particle swarm optimization least squares support vector machine model, and the outlier monitoring data are recovered by using them. The results of empirical analysis on the daily water intake monitoring data of water supply companies show that the feature vectors containing abnormal data can be retained to a large extent by moving average fitting and modal decomposition, and the effective reconstruction of the data can be realized. Compared with the traditional statistical method, it has better applicability, and the least squares support vector machine (LS-SVM) based on particle swarm optimization can further improve the fitting effect of eliminating outliers. In accordance with the seasonal fluctuation characteristics of water resources monitoring data and the objective reflection of the actual water use state, the purpose of recovering the abnormal monitoring data eliminated can be achieved relatively reasonably.
【作者单位】：中国航天系统科学与工程研究院研究生部;纽约州立宾汉姆顿大学汤玛士·华生工程与应用科学学院;泰山学院商学院;水利部水资源管理中心;
【基金】：国家自然科学基金委员会-广东联合基金项目(U1501253) 广东省省级科技计划项目(2016B010127005)
【分类号】：TP18;TV213.4

【相似文献】