Anomaly detection and discrimination using real-time monitoring data in a water distribution system |
---|
학술지명 IWA Aspire 2017
저자 김경필,최두용,유도근
발표일 2017-09-11
|
The number of installation rate of real-time monitoring sensors such as flow rate, pressure, and water quality sensors is increasing in water distribution system due to drop in prices of digital infrastructure like a wireless communication, computing and storage devices. The real-time monitoring system provides the operational status information to operator and it enable decision maker to efficient real-time operation management based on time series data. In real-time data, various abnormal points or intervals can occur due to abnormal operating conditions, measurement error of sensors, and data loss on the time series flow, so it is important to refine and utilize the data in a better manner. This study suggests a method to detect abnormal situation and to identify its type through statistical analysis of monitoring data. For the detection and judgement, the cumulative sum control chart (CUSUM) of each sensor, the cross-correlation between two sensors that are hydraulically connected, and the CUSUM for the correlation time series were used. The proposed method can solve the problem in case of determining abnormal situations by using only single sensor data. Through the pilot application study, it shows effective detection results even if where frequent changes of operating conditions can occur such as water supply boundary area. |