Outliers Identification Of Water Intake Monitoring Data In Industrial And Domestic Service Industry Based On LOF And CEEMD

It is an important support to strengthen the construction of water resources monitoring capacity in Hebei Province that the outliers identification of on-line measurement and monitoring data of water intake in industrial and domestic service industry.With the rapid growth of monitoring coverage in Hebei Province,the amount of water intake monitoring data is increasing.Under the current massive data with strong randomness and noise interference,the traditional methods are faced with problems such as low manual identification efficiency,low identification accuracy and insufficient applicability.Therefore,the research and development of scientific methods that can effectively identify outlier in monitoring data is very necessary for the follow-up work of total water consumption control and water resources tax collection.According to the characteristics of large volume and many types of water intake monitoring data,based on the theoretical research of artificial intelligence and modal decomposition method,combined with the characteristics of outliers of monitoring data,this dissertation makes an in-depth analysis and research on the abnormal value identification of monitoring data.The main research contents are as follows:Firstly,based on a large number of water intake monitoring data,the characteristics of outliers of water intake monitoring data are summarized and analyzed,and they are divided into"intuitive identification"and"non intuitive identification"according to the difficulty of identification;Secondly,according to the characteristics of short-term fluctuation of monitoring data and uneven spatial distribution of normal and abnormal samples,a method based on local outlier factor(LOF)is proposed to identify intuitive outliers;Thirdly,based on the comparative analysis of various modal decomposition methods,according to the characteristics of noise interference,nonlinearity and nonstationary of the monitoring data,the complementary ensemble empirical mode decomposition(CEEMD)method is proposed to accurately identify the initially recognized data;Finally,the proposed method combining LOF and CEEMD is applied to the analysis of daily water intake monitoring data of a waterworks in Hebei Province,and its recognition effect is compared with the abnormal value recognition effect of traditional methods.The research results show that the annual water intake data is revised from 512700m~3 monitored to 411900 m~3 after initially identification,and finally revised to 411400 m~3after accurately identification.The error between the annual water intake and consumption data and the approved annual water intake and consumption data is small,with a total reduction of 60%.It shows that this method has high effectiveness in the abnormal value identification of water intake monitoring data,and the corrected results are reasonable and accurate,which can better improve the reliability of monitoring data.Compared with the results of traditional methods,the research results show that the recognition accuracy and detection rate of the automatic recognition method combining LOF and CEEMD are better than the traditional methods.