Understanding digital twin data requirements before you start modelling
A municipal water utility may have years of operating records and still be unable to tell whether they can support a defensible rehabilitation model. Records collected for daily operations do not automatically describe the same conditions, locations or periods needed for modelling. The starting decision is which data can build the model, which measurements can test it, and which records or feeds can support the estimates the utility intends to make.
What a digital twin is — and what it isn’t
A simulation calculates how a system behaves under specified conditions. A digital twin adds a continuing connection to real-world data: the digital-twin definition describes a model that continually receives data from the physical asset or system. That connection, not artificial intelligence, is the defining distinction. Nor does the definition establish who coined the term.
A simulation remains useful before that connection exists. HeatNetSim, for example, is an open-source simulation tool for heating and cooling networks, while DHNx documentation includes simulation models. For a water utility assessing a proposed trunk line, the immediate need may be a model that can compare scenarios against known operating conditions. Calling a digital twin platform a substitute for that modelling and checking work would obscure the harder decision: what the available data can actually support.
Measurements for building and checking the model
Data used to construct a model and measurements used to check it have different jobs. A UK heat study examined heat consumption recorded by smart meters in different dwelling types. It also describes creating bespoke Modelica models and validating them through a digital twin of the experiment. Consumption records helped describe demand; comparison with the experiment gave the model a test against observed behaviour.
The same distinction matters when you assess pressure differences across a water distribution system. We would first identify what describes the system being modelled—such as its topology and operating boundaries—then identify observations from conditions the model must reproduce. Pressure or flow readings are more useful for validation when their location and time can be matched to those conditions. Otherwise, a mismatch may reflect the records being compared rather than the proposed pipe size or rehabilitation priority.
A calibrated heat-network study explicitly addresses validation using measured data. It illustrates the role of measurements without supplying a collection list for water distribution. The defensible list follows the utility’s question: a model for a capital decision needs observations that can challenge its predictions at the locations and operating conditions that matter to that decision.
From validated model to operational estimates
Once a model has been checked against measurements, its relationship to time determines the next data requirement. Historical replay uses recorded observations to examine past conditions. A current estimate needs an ongoing connection to operating data, with enough context to interpret when and where each reading was taken. A forecast instead depends on stated assumptions about future demand and operation. None of these calculations confirms what has happened in the field without a corresponding observation.
The choice of physics also follows the question. District-heating research addresses quasi-dynamic load flow, and thermal-transient modelling can account for heat stored in pipe walls. These are modelling considerations for a district-heating digital twin, not a ready-made water-distribution data specification or a promise of operational performance. For water rehabilitation planning, the useful test is whether the model’s governing equations, boundary conditions and measured checks are fit for the investment decision being made.
The intended decision determines the data a digital twin needs
For a municipal water utility, the settled starting point is therefore the decision and its evidence: data to construct the model, observations to validate it, and either historical records, current feeds or future assumptions to support the intended estimate. If you are weighing those inputs against a planned use case, request a Fluidit Heat demo to discuss what you have and what the proposed model would need.
