The term digital twin is already a trending topic for the water industry. However, there is no globally accepted definition of what exactly is a digital twin for a water distribution system. This blog post is another five-cent contribution to contextualize the idea.

A digital twin is a virtualized representation of a behavior of a physical system or process in a digital environment. It reproduces the behavior of the system or process under different working conditions and level of abstractions. When the physical system communicates its working conditions to the digital twin through sensor measurements, the digital twin is expected to reproduce exactly the same behavior of the real system at some abstraction level.

Simulations of water distribution system has been done for a long time already. I liked a post written by Keshvinder Singh where he explains very well the use of water network models in real time and their connection to data. However, can we consider as a digital twin, a model running simulations in EPANET or any other existing simulator? Is the connection of the running simulations with sensor data what makes it a digital twin? Did we failed on creating models and now we are pretending to solve the problem by creating a digital twin? Is the digital twin an upgrade of the model we had? What does exactly make a digital twin a digital twin?

In my opinion a digital twin leaves the border of simulators to enter what in other industries are known as emulators [a simulator mimics the basic behavior of a system; an emulator duplicates the thing “exactly” as it exists in real life]. This idea should be contextualized particularly for the case of water distribution systems. Simulation models have always used a simplified version of the real network [several client demands concentrated at single nodes, rough estimations of demands and consumption patterns, elasticity at pies and water compressibility not considered, etc.] and it is especially difficult to create something closer to an emulator. Emulating a water distribution system would imply to emulate the diversity of all the components taking part in it [pump, motors, valves, pipes, etc], and it would open the question about considering the stochastic components of the water consumption itself as part of the system or as a working condition. Even if it gets to be technically possible, it would be economically unfeasible for water utilities today. We defend the idea of a continuous improvement of network models connected to data to make them progressively closer to the reality. Both companies and researchers should be also aware of the existing economic limit for that, and this limit will depend on the utility conditions and the price of water. After some point it makes no sense in a single water network technically speaking to create a digital twin at a higher cost than the benefits you get from it. These are the reasons why I prefer to understand digital twins for Water Distribution Systems as a progressive improvements of network models connected to data scoped to the condition of the companies and water price. It is a journey, not a destiny.

After meeting Agnethe Peterson in a zoom session about digital twin organized by the SWAN-Forum she sends me her paper “Living and Prototyping Digital Twins for Urban Water Systems: Towards Multi-Purpose Value Creation Using Models and Sensors”. The paper was fantastic and very ambitious. The authors were dividing digital twins for WDS in two types: living and prototyping. Living was referring to a DT coupled with field observations and the prototyping version was referring to a scenario of the system without direct coupling with real time observations. In the approach we have been developing, we are making no distinctions of the digital twins. Once a Digital Twin is in use, coupling it to real data or not is more related to the application than to the digital twin itself. However, working on a digital twin for water distribution system will necessarily imply at some point, a coupling with real data in order to calibrate the model, no matter if at the application level it will be later used coupled with real data or to analyze decoupled what-if scenarios for planning or design. At the application level we defend the idea of running something similar to the concept of virtualization, where virtualized instances of the digital twin can assume different specific applications referring to one single physical system. Applications may go not only in the direction of real time representations or prototyping but also in the direction of look-ahead scenarios and what-if scenarios coupled with online data to evaluate the consequences of operation alternatives in real time.

The industry is living a great time of digitalization and improvement of a way water distribution system are being operated and managed today. I´m open to collaborate and discuss with other specialists and utilities about ideas and project developments. As a company, we are providing WATERing as a solution for water network analysis able to get connected to online data and to be extended with plugins for meeting any specific utility requirement. We offer development services, project consultancy and online courses subscriptions.  If you are interested in these areas just get in touch with us.