Water Demand Forecasting of District Metered Areas through Learned Consumer Representations

2025-09-09

AI Summary

To address the low accuracy and strong interference from multi-source uncertainties in short-term regional water demand forecasting under climate change, this paper proposes a DMA-level water demand forecasting framework integrating contrastive learning and cross-attention mechanisms. Methodologically, it first applies unsupervised contrastive learning to hourly smart meter data to discover and cluster user consumption behavior patterns; then employs a wavelet-transform-based convolutional network to extract temporal features and leverages cross-attention to jointly fuse heterogeneous multi-source information—including meteorological variables, historical water consumption, and socioeconomic attributes. Experiments on six months of real-world DMA data demonstrate that the proposed model reduces MAPE by up to 4.9% over benchmark methods. Moreover, it identifies user groups significantly influenced by socioeconomic factors, thereby enabling interpretable, differentiated water demand management.

Technology Category

• Data Mining & Knowledge Management: Knowledge Acquisition from the Web
• Machine Learning: Time-Series/Data Streams
• Search and Optimization: Metareasoning and Metaheuristics

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

Advancements in smart metering technologies have significantly improved the ability to monitor and manage water utilities. In the context of increasing uncertainty due to climate change, securing water resources and supply has emerged as an urgent global issue with extensive socioeconomic ramifications. Hourly consumption data from end-users have yielded substantial insights for projecting demand across regions characterized by diverse consumption patterns. Nevertheless, the prediction of water demand remains challenging due to influencing non-deterministic factors, such as meteorological conditions. This work introduces a novel method for short-term water demand forecasting for District Metered Areas (DMAs) which encompass commercial, agricultural, and residential consumers. Unsupervised contrastive learning is applied to categorize end-users according to distinct consumption behaviors present within a DMA. Subsequently, the distinct consumption behaviors are utilized as features in the ensuing demand forecasting task using wavelet-transformed convolutional networks that incorporate a cross-attention mechanism combining both historical data and the derived representations. The proposed approach is evaluated on real-world DMAs over a six-month period, demonstrating improved forecasting performance in terms of MAPE across different DMAs, with a maximum improvement of 4.9%. Additionally, it identifies consumers whose behavior is shaped by socioeconomic factors, enhancing prior knowledge about the deterministic patterns that influence demand.

Problem

• Forecasting short-term water demand in District Metered Areas
• Addressing non-deterministic factors like meteorological conditions
• Categorizing consumers by distinct consumption behaviors for prediction

Innovation

• Unsupervised contrastive learning categorizes consumption behaviors
• Wavelet-transformed convolutional networks with cross-attention mechanism
• Combines historical data and derived representations for forecasting