Interpretable Event Diagnosis in Water Distribution Networks

TL;DR

This work addresses the need for interpretable event diagnosis in water distribution networks by introducing counterfactual explanations to connect algorithm outputs with operator expertise. It develops counterfactual event fingerprints (CDF) for detecting events and counterfactual event isolation fingerprints (CIF) for classifying the event type, leveraging a residual-based data-driven detector trained on normal operation and a calibrated digital model. The approach is evaluated on two benchmarks, Hanoi and L-Town, demonstrating strong detection and interpretable isolation performance, with clear distinctions between leakages and sensor faults reflected in the fingerprints. The framework enhances trust, enables operator collaboration, and suggests future extensions such as handling unknown events, physics-informed learning, and interval-based explanations for practical deployment in critical water infrastructure.

Abstract

The increasing penetration of information and communication technologies in the design, monitoring, and control of water systems enables the use of algorithms for detecting and identifying unanticipated events (such as leakages or water contamination) using sensor measurements. However, data-driven methodologies do not always give accurate results and are often not trusted by operators, who may prefer to use their engineering judgment and experience to deal with such events. In this work, we propose a framework for interpretable event diagnosis -- an approach that assists the operators in associating the results of algorithmic event diagnosis methodologies with their own intuition and experience. This is achieved by providing contrasting (i.e., counterfactual) explanations of the results provided by fault diagnosis algorithms; their aim is to improve the understanding of the algorithm's inner workings by the operators, thus enabling them to take a more informed decision by combining the results with their personal experiences. Specifically, we propose counterfactual event fingerprints, a representation of the difference between the current event diagnosis and the closest alternative explanation, which can be presented in a graphical way. The proposed methodology is applied and evaluated on a realistic use case using the L-Town benchmark.

Figures (10)

• Figure 1: Illustration of the proposed Interpretable Event Diagnosis methodology: The blue block indicates the sensor component for collecting data, the purple blocks indicate the modeling component for the WDN, the green blocks indicate the event detection components, and the orange blocks indicate the event isolation components.
• Figure 2: L-Town-Network ($\Delta t = 5min$): Pressure forecasts in normal vs. leaky times -- a leakage is present from time $170$ onwards (starting point indicated by the vertical dashed red line).
• Figure 3: The Hanoi network including four ($4$) pressure sensors and one ($1$) flow sensor at the inlet.
• Figure 5: The L-Town network (Area A) (Vrachimis et al., 2022) and the locations of $29$ pressure sensors.
• Figure : (a) Sensor fault at sensor 13.

Theorems & Definitions (5)

• Definition 1: (Closest) Counterfactual Explanation CounterfactualWachter
• Definition 2
• Definition 3: Closest Counterfactual Event Detection Fingerprint
• Definition 4: Counterfactual Event Isolation Fingerprint
• Definition 5: Closest Counterfactual Event Isolation Fingerprint