Cutting-edge innovations in drinking water management

The 19th Computing and Control for the Water Industry Conference (CCWI23) was held in Leicester (UK) from 4th to 7th September 2023. During this event, several diversified innovations in modelling and analysis of urban water systems were shared by water researchers and practitioners from 27 countries.

This Special Issue includes eight peer-reviewed papers reflecting the diversity of cutting-edge innovations in drinking water management in water distribution systems (WDSs). Specifically, the contributions collected in this Special Issue are organized into the following two main categories: (1) system status evaluation and (2) operational control and management with novel methods. Four papers are included in each category.

The first group of papers focuses on the evaluation of the WDS status with new perspectives or new techniques. Satish et al. (2024a) provided a new perspective that uncoordinated drinking water hoarding (simultaneous withdrawal of water by a large number of customers) can lead to a short-term performance decrease in water supply from the WDS. Based on analyses using socio-technical indicators, critical thresholds that govern the WDS's ability to handle simultaneous water hoarding were identified.

As for the application of new techniques, Satish et al. (2024b) developed a graph-based approach to effectively identify critical combinations of pipes whose failure could significantly impact the functionality of WDSs. The approach utilizes structural and topological characteristics of WDSs as well as spatial demand distribution to replicate hydraulic behaviour. Compared with the state-of-the-art hydraulic-based method, the graph-based method exhibited a significant computational gain factor of greater than 30 without significant loss of accuracy in results.

In addition to the hydraulic status, water quality is also critical; yet, there is comparatively less robust water quality data in drinking water services, particularly in many low-income settings. In this regard, Jantarakasem et al. (2024) developed an entire smartphone camera-based application to measure turbidity in drinking water, removing both the need for external equipment and skilled labour. By leveraging convolutional neural network technologies, the turbidity classification accuracy reached 98.7 and 90.9% for formazine samples and kaolin clay samples, respectively, in the laboratory.

Apart from the evaluation of the current status of WDSs, Oberascher et al. (2024) developed a model to create future projections of water demand, resource availability and drinking water quality for the city of Klagenfurt (Austria). The model uses open-access data and easy formulas to allow good transferability to other WDSs. This study found that additional water sources will be needed in 2050 or 2080 to continue to maintain the high quality of water supply in the city, which is important information for the WDS operator to plan 20–30 years in advance due to the time needed to put a new source into operation. Another meaningful finding from this study is that linear and polynomial regression are suitable approaches to model and estimate water temperature and system input, respectively, whereas the groundwater level requires additional non-linear terms to consider the non-linear decrease over the depth.

The second group of papers focuses on the improvement of the operational control and management of WDSs with novel methods. For water leak management, Park et al. (2024) and Doss et al. (2024) applied machine learning approaches for leak detection in WDSs. Park et al. (2024) applied the eXtreme Gradient Boosting (XGBoost) model to analyse 40,000 sets of leakage vibration data of water supply pipes collected through leak detection sensors. The developed model showed high accuracy with the ability to classify whether a leak occurs indoors or outdoors. Doss et al. (2024) compared three different encoder–decoder neural network models and tested the models using the benchmark WDS L-Town and a real-world WDS in Norway. For the benchmark WDS, the tested models performed on par with the existing methods reported in the literature. For the real-world WDS, the performance was significantly lower due to the quantity and quality of data available as well as the positioning of pressure sensors in the WDS. In addition, this study found that the long short-term memory autoencoder showed better detection performance.

For intervention planning, Minaei et al. (2024) formulated the dynamic phased asset rehabilitation planning into an optimization problem to minimize the net present value of rehabilitation costs by optimizing the number of phases and type of intervention activities. The developed approach is verified via a real-world case study in Montreal, Canada, which not only includes WDSs but also spatially interdependent urban infrastructure systems including sewer systems and road networks. The case study shows that the timing of the rehabilitation can play a major role in finding the minimum investment cost, e.g. 25% cost savings by using a four-phase plan with each phase lasting for about 6 years.

Regarding operational control, Janus et al. (2024) extended the ability of mixed integer linear programming to include variable speed pumps for pump scheduling optimization. The method was demonstrated on a two-variable speed pump single-tank network to facilitate readers to easily understand the principle of the methodology. The average calculation time is approximately 1 s, indicating that the developed approach can be adapted to solving more complex WDSs. The method was also proven to be robust and able to reach global optimal solutions in a similar time for a range of operating points.