Monitoring data identification for a water distribution system based on data self-recognition approach

Detecting the occurrence of hydraulic accidents or contamination events in the shortest time has always been a significant but difficult task. The simple and efficient way is to identify the sudden changes or outliers hidden in the vast amounts of monitoring data produced minute by minute, which is unpractical for human. A new method, which employs a data self-recognition approach to achieve that automatically, has been proposed in this paper. The autoregressive moving average (ARMA) model was employed in this research to construct the self-recognition model. 56 months monitoring data from Changping water distribution network in Beijing, which was firstly cut into different time-slice series, was used to establish the ARMA model. This provided a prediction confidence interval in order to identify the outliers in the test data series. The results showed a good performance in outlier identification and the accuracy ranges from 90% to 95%.Thus, the ARMA model showed great potential in dealing with monitoring data and achieving the expected performance of data self-recognition technology.