Real-time event detection: A model-based approach

Security issues have become increasingly important within distribution systems, which has led to the development of event detection algorithms (EDAs) to provide timely detection of intrusion events. The current study develops a model-based EDA utilizing non-specific water quality sensors to iden- tify water quality anomalies, which incorporates both the localized water quality information and operational changes. The proposed EDA focuses on estimating the likelihood of an observed error signal time series using a moving time-window of error statistics. The likelihood of the error signals are estimated based on two formulations of the underlying probability density function (pdf): 1) a Normal pdf estimation, which assumes the errors follow a normal distribution, and 2) a kernel density estimation (KDE), which is type of non-parametric representation of the error distribution. A prelim- inary analysis was performed using chlorine as the water quality parameter. Results suggest that the proposed EDA, using KDE to estimate the error pdf, performed reasonably well in differentiating a true water quality anomaly from the modeling error time series.