Improved real-time data anomaly detection using context classification

• DOI: 10.2166/hydro.2011.042
• Publication Date: 2011-05-06T16:15:10Z
• Authors: Nemanja Branisavl...; Zoran Kapelan; Dušan Prodanović
• Keywords: SCADA; Anomaly (physics)

The number of automated measuring and reporting systems used in water distribution sewer is dramatically increasing and, as a consequence, so the volume data acquired. Since real-time likely to contain certain amount anomalous values acquisition equipment not perfect, it essential equip SCADA (Supervisory Control Data Acquisition) system with automatic procedures that can detect related problems assist user monitoring managing incoming data. A different anomaly detection techniques methods exist be varying success. To improve performance, these must fine tuned according crucial aspects process monitored contexts which are classified. aim this paper explore if context classification pre-processing methods, especially fully systems. methodology developed tested on sets real-life data, using standard experimental including statistical, model-based data-mining approaches. results obtained clearly demonstrate effectiveness suggested methodology.