Journal of Water Resources Planning and Management

Large earthquakes often damage water pipelines and disrupt the functionality of water distribution systems (WDS). This paper investigates the hydraulic performance of WDS following an earthquake, taking into account the postearthquake firefighting capacity of the system and the corrosion deterioration of pipelines. The seismic failure probability for corroded water pipes is calculated using a modified American Lifeline Alliance fragility function, and hydraulic simulations are performed using the water network tool for resilience. Water serviceability and the generalized resilience index are used as measures to determine the hydraulic performance of the WDS, which is evaluated under various conditions of network ages and seismic intensities. The framework developed in this study is exemplified for the Zhi Jiang (ZJ) WDS. The results indicate that, although firefighting does not significantly reduce performance measures, it is necessary to consider it in recovery plans since more customers are likely to lose a water supply due to postearthquake fires. The hydraulic behavior is worst in older WDS or those subjected to higher seismic magnitudes. Due to the scenario earthquake event, water serviceability reduces as the WDS ages, and the older networks require longer recovery times. Therefore, corrosion deterioration has a significant influence on the system-level seismic hydraulic performance.

Abstract

Deep reinforcement learning (DRL) has undergone a revolution in recent years, enabling researchers to tackle a variety of previously inaccessible sequential decision problems. However, its application to the control of water distribution systems (WDS) remains limited. This research demonstrates the successful application of DRL for pressure control in WDS by simulating an environment using EPANET version 2.2, a popular open-source hydraulic simulator. We highlight the ability of DRL-EPANET to handle large action spaces, with more than 1 million possible actions in each time step, and its capacity to deal with uncertainties such as random pipe breaks. We employ the Branching Dueling Q-Network (BDQ) algorithm, which can learn in this context, and enhance it with an algorithmic modification called BDQ with fixed actions (BDQF) that achieves better rewards, especially when manipulated actions are sparse. The proposed methodology was validated using the hydraulic models of 10 real WDS, one of which integrated transmission and distribution systems operated by Hidralia, and the rest of which were operated by Aigües de Barcelona.

Practical Applications

This research presents the DRL-EPANET framework, which combines deep reinforcement learning and EPANET to optimize water distribution systems. Although the focus of this paper is on pressure control, the approach is highly versatile and can be applied to various sequential decision-making problems within WDS, such as pump optimization, energy management, and water quality control. DRL-EPANET was tested and proven effective on 10 real-world WDS, resulting in as much as 26% improvement in mean pressure compared with the reference solutions. The framework offers real-time control solutions, enabling water utility operators to react quickly to changes in the network. Additionally, it is capable of handling stochastic scenarios, such as random pipe bursts, demand uncertainty, contamination, and component failures, making it a valuable tool for managing complex and unpredictable situations. This method can be developed more with the use of model-based deep reinforcement learning for enhanced sample efficiency, graph neural networks for better representation, and the quantification of agent action uncertainty for improved decision-making in uncharted situations. Overall, DRL-EPANET has the potential to revolutionize the management and operation of water distribution systems, leading to more-efficient use of resources and improved service for consumers.

Tracing disinfectant (e.g., chlorine) and contaminants evolution in water networks requires the solution of one-dimensional (1D) advection-reaction (AR) partial differential equations (PDEs). With the absence of analytical solutions in many scenarios, numerical solutions require high-resolution time and space discretizations, resulting in large model dimensions. This adds complexity to the water quality control problem. In addition, considering multispecies water quality dynamics rather than the single-species dynamics produces a more accurate description of the reaction dynamics under abnormal hazardous conditions (e.g., contamination events). Yet, these dynamics introduce a nonlinear reaction formulation to the model. To that end, solving nonlinear 1D AR PDEs in real time is critical to achieving monitoring and control goals for various scaled networks with a high computational burden. In this work, we propose a novel comprehensive framework to overcome the large-dimensionality issue by introducing different approaches for applying model order reduction (MOR) algorithms to the nonlinear system followed by applying a real-time water quality regulation algorithm that is based on an advanced model to maintain desirable disinfectant levels in water networks under multispecies dynamics. The performance of this framework is validated using rigorous numerical case studies under a wide range of scenarios demonstrating the challenges associated with regulating water quality under such conditions.

Abstract

In recent years, machine learning (ML) approaches have been used widely for water pipe condition assessment and failure prediction. These methods require a considerable amount of data from water distribution networks (WDNs). Imbalanced and missing data, whether asset or failure data, compromise a model’s prediction performance. In this research, using only 2 years of failure data in a real WDN, three ML methods—XGBoost, random forest and logistic regression—were used to prioritize asset rehabilitation. To address the issue of imbalanced data, a novel method of semisupervised clustering is proposed to leverage the domain knowledge in combination with unsupervised learning to divide the data set into homogeneous categories and enhance the classification accuracy. The introduced approach performed better than well-known data science class imbalance treatment techniques. Furthermore, analysis of the results indicated that classification evaluation metrics struggled to assess practically the effectiveness of various methods. To address this, an economic indicator is proposed to rank the pipes for rehabilitation based on their cost and likelihood of failure (LoF). Preventive maintenance using the results of an economic indicator reduces the number of failures with a small fraction of the total replacement cost. Moreover, another indicator was developed to consider the consequence of the failures and LoF simultaneously. This indicator mitigates in a cost-effective manner the flow capacity reductions in WDNs caused by failures. The results of this study provide asset managers with a powerful tool to prioritize assets for rehabilitation.

Practical Applications

In recent years, machine learning algorithms have gained popularity for assessing water pipe conditions and predicting failures. However, their effectiveness relies on substantial data from water distribution networks (WDNs). Challenges arise with limited (imbalanced) data, affecting prediction accuracy. This study focused on a specific WDN with only 2 years of failure data, aiming to identify priority assets for rehabilitation. Three ML methods (XGBoost, random forest, and logistic regression) and a novel semisupervised clustering approach were employed. This method combines expert knowledge with traditional techniques, significantly improving predictive accuracy. By applying ML algorithms within these homogenous clusters, predictive accuracy was enhanced notably. Two novel metrics were introduced for prioritizing pipe rehabilitation: one combining failure likelihood and replacement costs, and the other evaluating pipes based on their significance within the WDN and associated rehabilitation expenses. These models empower asset managers to optimize pipe replacement budget allocation and enhance the network performance.

The Rio Grande–Rio Bravo’s flow regime has been highly altered for more than 130 years, yet the river ecosystem still supports important biodiversity including numerous endangered species. More than 80% of water consumed in the basin goes to irrigating farms, but in recent decades, farmers have repeatedly experienced severe water shortages. Given this water-scarce condition, any plans for enhancing environmental flows must be carefully designed to minimize impacts or provide benefits to agriculture. This study describes the development of the Rio Grande–Rio Bravo’s first whole-basin hydrologic model—representing both the United States and Mexico portions of the basin—to enable exploration of environmental flow restoration needs and options for meeting these needs. We then demonstrate an analytical process in which environmental flow needs are compared to existing flow conditions to quantify gaps, and then evaluate how those gaps can be filled by reducing farm irrigation needs by shifting to less water-intensive crops and fallowing a portion of existing farmland while maintaining or improving net revenues. In our pilot assessment we find that an improvement of ${2.2\mathrm{m}}^3/\mathrm{s}$ would fill the environmental flow gap for late-summer low-flow conditions at Albuquerque, New Mexico. This flow enhancement is attainable by fallowing 18%–26% of cropland and shifting to more profitable and less water-intensive crops to sustain overall farm revenues.

Designing urban drainage systems under deep uncertainty and for extreme climate conditions is a challenging task because various interrelated socioeconomic policies, urban characteristics, and hydraulics must be considered. To address such problems, scenario planning is the most well-known alternative solution for managing a wide range of future scenarios for any planning process under potential uncertainty. In the relevant literature, few efforts have been devoted to providing a long-term urban drainage system (UDS) planning solution to prevent several different levels of scenarios simultaneously. This study proposes a multiperiod optimization framework based on scenario planning. The proposed framework determines various system components and minimizes the regret costs (RCs) and total investment while satisfying a series of constraints: (1) scenario-optimal solutions are obtained from the multiperiod single scenario model; (2) common elements are identified across all scenarios; and (3) a compromise solution is determined to minimize the total regret cost considered in the multiperiod multiscenario model. The proposed framework outperformed a traditional design methodology in realizing long-term planning depending on the climate conditions considered in four different levels of scenarios. The RCs can provide flexible responses to future strategies under deep uncertainty and reduce the total investment in the UDS planning projection. In addition, aspects of the common elements process, identified in the initial period (i.e., Period 0) will be valid for a wide range of plausible future scenarios; therefore, identifying common elements is shown to be necessary for system planning under future scenarios. This study can be extended to different civic infrastructures with different standards and goals, thereby widening its applicability and providing additional insights for achieving a sustainable infrastructure service. This study has the potential to support the public and private sectors under uncertainty.

Iraq has been plagued with extreme droughts in the last two decades, resulting in a significant decrease in rainfall. Water harvesting has become essential for addressing these droughts. Building dams for rainwater harvesting can help mitigate the effects of drought and provide water resources. Rainwater conservation is a commonly employed method in drought-prone regions. The objective of this research is to identify suitable locations for rainwater harvesting and dam construction in the Sulaimanyah Governorate. The study relied on several criteria, including rainfall data, runoff, slope, land cover/use, stream order, soil quality, and hydrological soil group, to assess the area’s suitability for water harvesting through dam utilization. To obtain the land use and land cover map, a Sentinel-1 image from 2020 was used in conjunction with unsupervised classification using ERDAS Imagine software version 9.1. Additionally, a digital elevation model was utilized to extract stream order, contour lines, and slope maps. The rainfall data were spatially interpolated using inverse distance weighting (IDW). The findings indicated that suitable water storage sites are primarily located in the north, northeast, and some parts of the center of the study area The study categorized 27.5% of the regions as highly suitable and 7.6% as medium-high suitable. Additionally, 0.7% and 10.1% of the study area were classified as moderately suitable and low suitable, respectively, while 54.1% was deemed unsuitable.