复杂庞大的供水管网系统拥有众多监测点，在人工判断的情况下，各个监测点采集的海量数据无法被及时有效地处理，数据准确性无从保障，这对供水管网异常情况的判断造成极大阻碍。针对此情况，将北京市某供水管网监测站56个月的在线监测数据进行时段上和季节上的切分，构建自回归滑动平均（autoregressive moving average，ARMA）模型，并通过该模型建立的置信区间识别人工模拟序列中的异常值，从而实现独立节点自身数据的自识别。结果表明：经过数据反馈矫正，该自识别过程能够准确提取人工模拟监测数据中的异常值。ARMA模型的建立极大限度压缩了需人工处理的数据量，以便在异常数据中人工甄选无效数据，实现数据质量控制。
Humans cannot process the vast amounts of monitoring data produced every minute, so data faults may not been seen and the data can become unreliable. Systems are needed to recognize abnormalities and invalid data. This paper describes a data self-recognition approach based on independent monitoring sites. 56 months of monitoring data was collected from one water distribution network in Beijing. The data was cut into different time-slice series for an autoregressive moving average model (ARMA). This provided a prediction confidence interval to identify the outliers in the test data series. The results gave good outlier identification with data feedback correction. Thus, the system provides data quality control with self-recognition that easily and efficiently deals with invalid data.
[1] Chang N, Pongsanone N P, Ernest A. A rule-based decision support system for sensor deployment in small drinking water networks[J]. Journal of Cleaner Production, 2013, 60:152-162. [2] Perelman L, Ostfeld A. Operation of remote mobile sensors for security of drinking water distribution systems[J]. Water Research, 2013, 47(13):4217-4226. [3] Behboudian S, Tabesh M, Falahnezhad M, et al. A long-term prediction of domestic water demand using preprocessing in artificial neural network[J]. Journal of Water Supply Research and Technology-AQUA, 2014, 63(1):31-42. [4] Nazif S, Karamouz M, Tabesh M, et al. Pressure management model for urban water distribution networks[J]. Water Resources Management, 2010, 24(3):437-458. [5] 王丽娟, 张宏伟. 差分自回归移动平均模型预测管网漏损的研究[J]. 中国给水排水, 2010, 26(11):127-129.WANG Lijuan, ZHANG Hongwei. Leakage prediction of water distribution network by ARIMA model[J]. China Water & Wastewater, 2010, 26(11):127-129. (in Chinese) [6] 孙平, 王丽萍, 陈凯, 等. 基于时间序列模型ARMA的水厂逐日需水量过程预测方法[J]. 中国农村水利水电, 2013, 11:139-142.SUN Ping, WANG Liping, Chen Kai, et al. Water plant daily water demand forecasting model based on time series model ARMA[J]. China Rural Water and Hydropower, 2013, 11:139-142. (in Chinese) [7] Huang L, Zhang C, Peng Y, et al. Application of a combination model based on wavelet transform and KPLS-ARMA for urban annul water demand forecasting[J]. Journal of Water Resources Planning and Management, 2014, 140(8), 04014013. [8] Mounce S R, Mounce R B, Jackson T, et al. Pattern matching and associative artificial neural networks for water distribution system time series data analysis[J]. Journal of Hydroinformatics, 2014, 16(3):617-632. [9] Box G E P, Jenkins G M, Reinsel G C. Time Series Analysis, Forecasting and Control[M]. 4th Edition. Englewood Cliffs:Prentice Hall, 2008. [10] Bozdogan H. Model selection and Akaike's information criterion (AIC):The general theory and its analytical extensions[J]. Psychometrika, 1987, 52:345-370. [11] Montgomery D C, Peck E A, Vining G G. Introduction to Linear Regression Analysis[M]. 4th Edition. New York:John Wiley & Sons, 2006. [12] 田增尧, 张明理, 赵瑞. 短期电力负荷预报中异常负荷数据的识别和修正[J]. 吉林电力, 2004, 175:21-23.TIAN Zengyao, ZHANG Mingli, ZHAO Rui. Identification and manipulation of anomalous data of load sequence in short-term load forecasting[J]. Jilin Electric Power, 2004, 175:21-23. (in Chinese)
