Burst detection in district metering areas using a data driven clustering algorithm.

Area of Science:

• Hydraulic Engineering
• Water Resource Management
• Data Science

Background:

• Water distribution systems are susceptible to bursts, leading to significant water loss and infrastructure damage.
• Accurate and timely burst detection is crucial for efficient water management and minimizing service disruptions.

Purpose of the Study:

• To develop and validate a novel, two-stage methodology for detecting bursts in water distribution systems.
• To address limitations of existing methods, particularly in handling non-stationary data and multiple data sources.

Main Methods:

• A two-stage approach combining clustering for outlier detection and subsequent burst identification.
• Utilizes simultaneous measurements from multiple flow meters within a district metering area (DMA).
• The clustering algorithm automatically adapts to non-stationary historical data without prior selection.

Main Results:

• Successfully detected simulated bursts (e.g., flushing) with short durations.
• Achieved a significantly low false positive rate compared to previous studies.
• Detected abnormal water usage patterns correlated with weather changes, indicating robustness.

Conclusions:

• The proposed methodology offers an effective and robust solution for burst detection in water distribution networks.
• Demonstrates high potential for real-world application in complex multi-inlet and multi-outlet DMAs.
• The method's ability to handle non-stationary data and reduce false positives marks a significant advancement.