Ground it in geography
Load public GIS into PostGIS. Derive pipe connectivity and place synthetic meters on the real network geometry.
Configure a region ↗THE UTILITY NETWORK FIELD GUIDE
See the network.
Understand
the behaviour.
A framework for exploring utility digital twins. Connect public GIS, explicitly synthetic telemetry and hydraulic modelling in one traceable workflow.
IMPLEMENTED FOCUSWater networks · EPANET
NETWORK EXPLORERCONCEPT / NOT LIVE
Pipe topologySynthetic meterSupply
Illustrative topology. No actual GIS or telemetry values.
Real geography.
Visible assumptions.
REAL
Public network geometry
and source observations
SYNTHETIC
Derived zones, meter locations
and generated readings
MODELLED
Hydraulic calculations
under stated assumptions
01 / THE METHOD
A twin you can
trace back to its source.
Follow each layer from published geometry to a residual worth investigating. No hidden switch from generated data to measured truth.
Connect the observations
Stream adapter and Datagen feeds through Kafka and Flink, preserving the distinction between real and synthetic telemetry.
Follow the pipeline ↗Make the model explainable
Run EPANET with current meter-derived demands. Assess pressure residuals against each DMA’s own history.
Understand the method ↗02 / SYSTEM MAP
One workflow.
Clear boundaries.
FOUNDATION
→GIS → PostGIS
Geometry, connectivity
and device registries
OBSERVATIONS
→Kafka → Flink
Real and synthetic feeds
into TimescaleDB
CALCULATION
→EPANET
Hydraulic solves
and pressure residuals
INSPECTION
Console & GIS
Map, freshness, flags
and ELK read copy
Postgres is the system of record.
Storage observations use a direct batch path. Elasticsearch holds a derived copy for inspection.
03 / FIND YOUR PATH
From curious
to configured.
A practical guide for understanding the framework, running a local stack and adapting it to a region.
SCOPE / READ BEFORE YOU BUILD
Useful exploration.
Honest limits.
The current hydraulic implementation is for water networks using EPANET. Electricity and transport asset schemas provide an extension foundation; their simulation models require further work.
Published topology may have gaps. Model defaults need review. Residual flags are investigation cues, and synthetic observations do not validate field detection accuracy.
Read assumptions and limits ↗OPEN THE FIELD GUIDE