Conformal Prediction-Driven Adaptive Sampling for Digital Twins of Water Distribution Networks

2025-11-06

AI Summary

This work addresses the state estimation problem in digital twins of water distribution networks, where uniform sampling proves resource-inefficient under heterogeneous node-level uncertainties. To overcome this limitation, we propose a lightweight adaptive sampling framework that integrates LSTM-based time-series forecasting with marginal Conformal Prediction (CP). The method dynamically allocates sensors at the node level according to real-time uncertainty estimates, enabling efficient uncertainty quantification and low-overhead decision-making while maintaining high coverage. Experiments on benchmark networks—Hanoi, Net3, and CTOWN—demonstrate that, at 40% sensor coverage, our approach reduces demand prediction error by 33–34% compared to uniform sampling, achieves empirical coverage rates of 89.4–90.2%, and incurs only 5–10% additional computational overhead. The key contribution lies in the first integration of marginal CP into an adaptive sampling mechanism, thereby reconciling statistical reliability with engineering real-time constraints.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Intelligent Robots: State Estimation
• Reasoning under Uncertainty: Stochastic Optimization

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Security and Privacy: Data transparency and provenance

Abstract

Digital Twins (DTs) for Water Distribution Networks (WDNs) require accurate state estimation with limited sensors. Uniform sampling often wastes resources across nodes with different uncertainty. We propose an adaptive framework combining LSTM forecasting and Conformal Prediction (CP) to estimate node-wise uncertainty and focus sensing on the most uncertain points. Marginal CP is used for its low computational cost, suitable for real-time DTs. Experiments on Hanoi, Net3, and CTOWN show 33-34% lower demand error than uniform sampling at 40% coverage and maintain 89.4-90.2% empirical coverage with only 5-10% extra computation.

Problem

• Estimating water network states with limited sensors efficiently
• Reducing resource waste from uniform sampling uncertainty
• Developing real-time digital twins with low computational cost

Innovation

• Adaptive sampling uses LSTM and Conformal Prediction
• Marginal CP enables low-cost real-time uncertainty estimation
• Focuses sensing on nodes with highest uncertainty levels