Research On DMA Leakage Analysis Based On Big Data

Water loss in water distribution systems has always been a hot issue of common interest.As an effective measure to control water loss,district metered area(DMA)has been widely used in recent years.With the establishment and improvement of DMA,more and more DMA data can be obtained,but the actual data is rarely used.Therefore,it is necessary to perform data mining on the data in the DMA.In order to make reasonable use of the large amount of data obtained from DMA and find the law of water use to make reasonable decisions,this study uses data-driven approaches to analyze the water loss in DMA.This study focuses on cluster analysis for water loss detection based on the law of water use.Prior to cluster analysis,the data is visualized to visually understand the overall flow data and determine the steps of data preprocessing.Firstly,the data is dimension-reduced,the abnormal data is judged and deleted.Besides,the normal data is cleaned and normalized.The missing data is interpolated,and the instantaneous flow data at 1 minute is converted into the average flow data of 15 minutes.The average flow data of 15 minutes was then used to construct the data-driven algorithm model.In the model construction,the k-means algorithm is used to select the window size.The X-means algorithm and OPTICS algorithm in the clustering algorithm are compared.Finally,the OPTICS algorithm is selected for the detection of water loss.In order to reduce false alarms,after detecting the water loss,the method is performed by tuning the epsilon,judging the special date,and analyzing the window date interval in the process of detecting water loss.Moreover,to evaluate the performance of OPTCIS algorithm for water loss detection,correct rate,true positive rate and false positive rate are used in this paper.New water loss analysis is performed on six different areas of CM,BS,BYT,BDS,YT and YLY by using the established data-driven algorithm model.Since the algorithm does not easily detect the inventory water loss,the analysis of the inventory water loss is performed later on.The results show that the appropriate window size different between workdays and holidays.In addition,the OPTICS algorithm is superior to the X-means algorithm in detecting leakage,so the OPTICS algorithm is chosen.Besides,the minimum water loss that can be detected is independent of the regional scale and is related to the standard deviation of water use at various times in the region.We also found the minimum night flow in the YT region is obviously high,and the region has inventory water loss.At the same time,in the process of data pre-processing,it was found that the normal water use data of the Spring Festival was easily classified as abnormal data.Therefore,the particularity of the water used in the New Year was found.