A Comprehensive Review of Artificial Intelligence Techniques for Urban Water Demand Forecasting: From Classical Time-Series Models to Transformer-Based Approaches (۲۰۱۰–۲۰۲۵)

The escalating demands on urban water systems, driven by population expansion, industrial growth, and climate fluctuations, necessitate advanced forecasting techniques to ensure sustainable resource allocation. Traditional statistical models have long been employed for water demand prediction, but their limitations in handling nonlinear patterns and vast datasets have paved the way for artificial intelligence (AI) innovations. Over the past ۱۵ years, AI has revolutionized this domain by enabling precise, adaptive predictions that integrate diverse variables like weather, demographics, and historical usage. Despite these advancements, no systematic and exhaustive review exists to map the progression from classical time-series methods to cutting-edge Transformer-based models. This paper conducts a rigorous systematic literature review of AI-driven urban water demand forecasting techniques from ۲۰۱۰ to ۲۰۲۵, categorizing them into evolutionary phases and evaluating their performance across key metrics. A detailed comparative analysis highlights Transformer architectures as a transformative leap, offering superior handling of long-range dependencies and real-time adaptability. Recommendations for future research emphasize hybrid integrations and regional adaptations to enhance resilience in water-stressed urban environments.

کلیدواژه ها:

Urban Water Demand Forecasting ، Artificial Intelligence ، Time-Series Modeling ، Deep Learning ، Transformer Architectures