Timely deployment of contaminant warning systems
requires on-line sensors and advancement of data
analysis and decision support systems to accurately
detect water quality changes. Many utilities are beginning
to install on-line water quality sensors or are
expanding their current on-line monitoring capabilities.
As more and more data from these systems become
available, various algorithms to extract useful information
from them are being developed.
These new algorithms must be quantitatively evaluated
in order to understand their capabilities and limitations
as well as to determine whether they provide
added value. This study was prompted by the need to
evaluate the performance of new algorithms for the
detection of changes in water quality. In addition to
being assessed on how well they identified water quality
changes, algorithms were also assessed on how few
false alarms they generated.
The results of this study are likely to affect
approaches to distribution system security with respect
to a possible accidental or deliberate contamination
event. The steps necessary for evaluating detection
tools are detailed using both previously collected data
and simulated events. Includes 28 references, tables, figures.