Abstract
As AI technology gains momentum, utilities and professionals in the water sector are excited and curious about its potential applications. The presentation addresses questions like the relevance of AI to the water sector, current and future AI applications in water management, and how utilities can adopt AI in a measured way. The presentation will demystify AI and guide water-sector professionals, regardless of their expertise or organization size. The importance of adopting AI models and tools to avoid being left behind can't be overstated. OpenAI's GPT-4 has revolutionized generative AI with its multimodal capabilities, handling text and image inputs. The advanced abilities of these models make them adaptable and flexible, capable of learning new behaviors through last-mile data training and fine-tuning. ChatGPT generates human-like text from natural language and image prompts, while DALL-E-2 generates images from textual prompts. As these models evolve, the water sector must utilize large, diverse, and representative data that reflect industry-specific vocabulary, jargon, and concepts to create industry-specific AI solutions. AI tools offer the promise of democratization to utilities of all sizes through access to powerful tools without the need for specialized knowledge. However, industry experts must test these models for accuracy and supervise them to ensure they align with the water sector's unique needs. Collaboration on last-mile data can help utilities leverage AI for more effective water management strategies and improved customer and environmental outcomes. To reach this goal, water industry professionals must understand how AI works, its different types, and its potential applications. Most current AIs are artificial narrow intelligence (ANI), which focuses on solving a specific task or problem and gives an example of using computer vision to analyze images and video in the water sector. The difference between machine learning and deep learning is that the latter doesn't require human intervention-it can improve by learning from its own mistakes. However, deep learning requires vast amounts of training data, processing power, and time to learn to produce accurate results. This presentation will educate the attendees about different AI terminology and how it may change their jobs in the future. Opportunities and barriers to AI adoption in the water industry include regulatory pressures and risk aversion, but AI can work collaboratively with humans through digital twins to make informed decisions. Digital twins, which are digital replicas of utility assets and performance, can help engineers and operators test the outcome of decisions before making them in the real world. Using AI with digital twins will allow utility operators to focus on other aspects of their workload that require human intelligence. Computer vision can analyze video and photographs to detect anomalies, leaks, and faults faster and more accurately than traditionally human-manned analysis. Image-based deep learning models using closed-circuit television (CCTV) footage can even analyze the density of raindrops to recognize the intensity of rainfall, which has important implications for quick flood management responses. One of the most relevant applications of AI in a utility's daily processes is fault detection, which can use machine learning to detect faults in a water system before they cause problems for customers. This AI-driven analysis and alert system, combined with a generative natural language model like ChatGPT, could benefit utilities with limited resources, especially smaller utilities without an in-house data analytics department. One area where AI could significantly impact the water industry is customer service. An AI-assisted chatbot, like ChatGPT, could answer customer questions about issues like low water pressure or discolored water and even alert crews to issues in the customer's area. Another area where AI could improve water management is an evolutionary computation for discovering optimal designs for water distribution systems. This subfield of AI can solve complex problems with too many variables for traditional algorithms, making it useful for planning initiatives and optimizing future expansion, maintenance, and repair costs. Tools like these are currently used in the water sector to provide alerts for anomalies like abnormal pressure, tank levels, flows, and other important daily operations. Looking to the future, it's easy to envision autonomous robots crawling through a pipe network and continuously collecting and sending data to be analyzed with AI analytics. Although devices have been developed for leak detection in pipes, the water sector hasn't fully leveraged robotics. Pipebots could revolutionize buried pipe infrastructure management by developing microrobots designed to work in underground pipe networks. Utilities need to identify their specific problems before implementing AI solutions. They shouldn't adopt AI just because it seems like a trendy technology; instead, they must carefully evaluate the problems that must be solved and select appropriate tools for the job. Utilities can also collaborate by providing certain similar datasets that can be used to help train LLMs on specific problems that most utilities are challenged with. These large data sets can provide the required training to develop a few generic tools that many water utilities could have access to. Utility data are valuable assets, but data quality varies and may be distributed between business units, making it difficult to use effectively. Utilities should improve data hygiene and preparation to leverage AI tools better. We can also share how utilities can use LLMs with plugins such as Code Interpreter to help clean and analyze existing and historical data. Change management is another critical consideration when implementing AI. Change can be challenging, and a lack of clear communication and identification of why new technologies are needed can exacerbate resistance to change, especially around AI's perceived threat of job elimination. However, AI can also attract new talent to the water sector. Although larger utilities may have more data and resources, smaller utilities could be more agile in adopting new technology due to fewer people aligning toward technology changes. Utilities should approach AI with a focus on problem-solving and carefully consider their unique goals and priorities when evaluating AI solutions. Water professionals should be optimistic about the exciting potential of future AI solutions in the water sector. This presentation is based on the Journal of AWWA (Volume 115, Issue 8, October 2023, pp. 48-49.) article of the same name by Gigi Karmous-Edwards and Sasa Tomic, and 'AI & Water Management: What Utilities Need to Know Now,' article posted on the Qatium website (www.qatium.com) and written by Gigi Karmous-Edwards, Sasa Tomic, Dragan Savic, and Paul Fleming.
This paper was presented at the WEF/AWWA Utility Management Conference, February 13-16, 2024.
Author(s)S. Tomic1, G. Karmous-Edwards2, P. Fleming3
Author affiliation(s)Burns & McDonnell 1; Karmous-Edwards Consulting LLC 2; WaterValue 3;
SourceProceedings of the Water Environment Federation
Document typeConference Paper
Print publication date Feb 2024
DOI10.2175/193864718825159238
Volume / Issue
Content sourceUtility Management Conference
Word count8