UtilityTwin

An Intelligent Digital Twin for Utility Networks with Heterogenous Data & AI algorithms

This project was funded by the Bavarian State Ministry of Digital Affairs as part of the “IuK Bayern” funding program – Grant Number DIK0325/01.

On this page:

 

Abstract

The UtilityTwin project investigates the creation of an intelligent digital twin to bridge the gap between fragmented utility data silos and real-time network behavior. Traditional asset management systems often lack the integration of heterogeneous data sources, resulting in isolated solutions. UtilityTwin proposes an integrated Knowledge Graph approach (Neo4j) to synchronize Metering (AMI), GIS, SCADA, and meteorological data into a unified, spatially intuitive representation of water and heat supply networks.

The project focuses on three main pillars: (1) High-resolution communication technology, including the development of solar-powered gateways for comprehensive network coverage; (2) AI-based network analysis, leveraging Convolutional Neural Networks (CNN) and Graph Neural Networks (GNN) for demand forecasting and leak localization; and (3) Automated digitization of historical network documents to capture the historical evolution of the infrastructure. The developed framework enables utilities to move from reactive maintenance to proactive, data-driven management.

Key Research Areas

  • Graph-Based Digital Twin: Migration from relational databases to Knowledge Graphs to enable complex recursive queries, shortest path analysis, and structural integrity modeling of utility networks.
  • Predictive Demand Forecasting: Implementation of a novel framework using Continuous Wavelet Transform (CWT) and Cross-Attention mechanisms to forecast heat and water demand at the District Metered Area (DMA) level.
  • Hybrid Leak Localization: A synergized approach combining real-world AMR data with hydraulic simulations (EPANET/WNTR) and Deep Learning to identify and localize leaks within subzones with high accuracy.
  • AI Document Digitization: Utilizing Image-to-Image translation (CycleGAN/CUT) and Vision Transformers (ViT) to geo-reference and digitize historical paper-based maps into modern GIS systems.

Data and Code Availability

Following the principles of Open Science, the UtilityTwin project provides public access to the curated datasets used for training and benchmarking our demand forecasting and anomaly detection models.

Public Dataset

The labeled district heating dataset, encompassing diverse consumption patterns and anomaly scenarios, is available on Zenodo:

  • UtilityTwin: District Heating Demand & Anomaly Dataset
    Zenodo Record: 17398331

Publications

2026

Journal Articles

2025

Conference Contributions

2024

Journal Articles

Conference Contributions

2022

Conference Contributions

Awards & Recognition

National Student Prize for Social Innovations (StiPS)


— Recognition for innovative research contributions for climate protection and sustainability by Bundesministerium für Forschung, Technologie und Raumfahrt.

Project Details

2021

  • UtilityTwin


    (Third Party Funds Group – Overall project)
    Project leader: , ,
    Term: September 1, 2021 - August 31, 2024
    Acronym: UtilityTwin
    Funding source: Bayerisches Staatsministerium für Wirtschaft, Landesentwicklung und Energie (StMWi) (seit 2018)
    Abstract

    In the UtilityTwin research project, an intelligent digital twin for any energy or water supply network is to be researched and developed on the basis of adaptive high-resolution sensor data (down to the sub-second range) and machine learning techniques. Overall, the technology concepts BigData and AI are to be combined in an innovative way in this research project in order to make positive contributions to the implementation of the energy transition and to counteract climate change.