Abstract:
Abstract To monitor water quality, utilities typically employ periodic manual sampling. However, when a contamination event occurs, it may require days before it is detected. To enhance monitoring, utilities employ sensors which monitor various water quality parameters. A common approach is the use of chlorine sensors for monitoring chlorine residuals at different locations in the network, in order to determine whether a contamination event has occurred. Unfortunately, due to significant variability in water demands, as well as the effect of hydraulic and quality control actions, the disinfectant residual at the sensor location may fluctuate significantly in time, and therefore, model-free event detection algorithms may not be able to detect certain contamination events, or they may cause false alarms. This work extends the work in by proposing a model-based method for contamination event detection using real-time concentration lower-bound estimations as well as multi-level thresholds, for enhancing detection and reducing detection delay while minimizing false positive alarms.
