Introduction

To address the growing challenges in the drinking water sector, such as the global aging of supply systems, risk management, climate emergencies, and the need for resilient management to adapt to climate change and variability, the USEPA Agency has launched the Water Network Tool for Resilience, known by its English acronym WNTR (Klise et al., 2017). This tool is focused on the management of Drinking Water Systems, WSS in its English acronym, and uses EPANET (Rossman et al., 2021) as its main calculation engine.

On the other hand, the computational development company Novastudio-DCKStudios has developed the Boorie initiative, which consists of a set of tools, or Toolbox, to assist in the management of Drinking Water and Sanitation Systems (APyS) through the use of Artificial Intelligence (AI). The goal is to increase efficiency in service delivery, emphasizing efficiency in water usage, aligning with the Sustainable Development Goal (SDG 6), and in energy efficiency, thereby contributing to the reduction of the carbon footprint in line with SDG 13.

In this context, the purpose of this article is to present the main advantages of the freely accessible Boorie-WNTR product, seeking to disseminate knowledge among various users and decision-makers to promote effective ownership of management by all stakeholders involved in the provision of Drinking Water services.

Given that WNTR is developed in Python 3, its use is limited to those users who are familiar with this programming language, restricting its reach to a small group of experts. In light of this situation, Novastudio proposes the development of Boorie-WNTR, with the intention of making the tool more accessible and more powerful for socializing and appropriating knowledge among the different water stakeholders.

Methodology

1. First, the advantages and limitations of the WNTR development are evaluated with the aim of establishing initial proposals for its dissemination with Boorie-WNTR. 1.1 Advantages of WNTR: It has been identified that WNTR is capable of: a) building a water network model from scratch and executing a supply network, b) adding or removing elements in the network, c) graphically representing the original network, d) displaying graphical results, e) iterating over elements to simulate different scenarios, f) implementing controls and advanced rules to simulate real operating conditions, g) manipulating Network-type graphs or directed multigraphs, h) presenting simulation results for different elements, such as pressures in nodes, demands, flow in pipes, levels in storage tanks, and water quality variables, i) simulating disaster scenarios, highlighting: i.1 contamination in networks, i.2 earthquakes (seismic resilience and fragility curves), the last aspect having been specifically developed for the City of Mérida, Venezuela (Astorga and Mora, 2014), i.3 breaks or leaks in pipes (service interruption), among others, j) skeletonization of networks using algorithms proposed by Walski or others (2003), generating an Essential Twin from a Digital Twin, k) rescaling and coordinate changes for integration with Geographic Information Systems (GIS). 1.2 Limitations: Among the limitations of WNTR are its inability to a) simulate variable-speed pumps, b) work with general-purpose valves, c) the pump curves only accept three points and are limited to the use of the Hazen-Williams (HW) formula for head loss, and d) do not allow the use of volume curves for tanks.

2. After identifying the advantages and limitations, the main features are selected to share with the general public: 2.1) General simulation of networks, adjustments and presentation of results, 2.2) Scenarios of critical pipes and affected nodes, 2.3) Break and service suspension scenarios, including percentages of affected nodes and the corresponding population, 2.4) Skeletonization of the network and time savings in computation depending on the platform used for simulation.

3. Selection of the platform for the execution of WNTR by non-specialized users. For this purpose, STREAMLIT® is chosen, which provides an interactive Web environment.

4. The user can input the network of their choice in [.inp] format

5. In the final phase of dissemination, a Generative Artificial Intelligence based on a large language model (LLM), such as ChatGPT 3.5 turbo and 4, is trained so that BoorieWNTR acts as a Hydraulic Agent capable of answering specific questions about the analyzed cases. LANGCHAIN is selected as the working platform.

Results and Discussion.

The following results are presented extracted from STREAMLIT® Presentations. In Figure 1, the nodes with affected service capacity (Demands lower than 90% of the target value) due to a pipe break or service denial are shown.