Quantum-inspired software for water modelling
- Chris Cade
- Nov 26
- 3 min read
Our Aquaduct project was recently selected by the Dutch ministries of Defence and Infrastructure & Water Management for continued development as a part of the Purple NECtar Quantum Challenge. The project brought together experts in water modelling (Deltares), quantum-inspired and quantum algorithms (Fermioniq), and quantum hardware (Quix Quantum) in order to boost water modeling capabilities with quantum technology.
Project highlights
We established the use of quantum-inspired techniques for water simulations through rigorous scientific validation in collaboration with Deltares – marking the first application of its kind.
We achieved up to 1000x speedups on the project’s test models using GPU-accelerated quantum-inspired algorithms – demonstrating that quantum-inspired software can already begin to deliver added value.
Alongside the project, we began work on a platform that will get our technology into the hands of users – paving the way for real impact from quantum technology in a short time-frame.
Validation
We validated our quantum-inspired software on several test models to ensure that the simulations were behaving as expected and producing correct results. Such validation is an extremely important process which is necessary if the software is to be deployed in real-world scenarios.
We used three test models in the project:
Flow along a 1D sloped channel (Belanger channel).
Flow over a floor (bathymetry) with ‘wavy’ topology.
A model of a 2D water droplet in a bath.
For our real-world test case, we used a model of the river Waal in the Netherlands, pictured below.
Speedups
Quantum-inspired algorithms achieve speedups by introducing a compression layer underneath the simulation algorithm. This compression can significantly reduce the number of floating-point operations (FLOPs) required by the computation, resulting in memory reductions and runtime speedups. When applying these algorithms to modelling problems, a crucial step in development is a ‘compression study’ – a systematic analysis of the amount of compression that can be introduced while maintaining accuracy of the simulation. From these studies, we can infer what level of compression is appropriate for each model, and from those arrive at concrete runtime and speedup predictions.
As the models become very large, runtime improvements become increasingly significant. In essence, the quantum-inspired algorithms allow for a decoupling between model size and runtime – instead the runtime scales with the compressibility (or lack thereof) of the model.
From our compressibility studies, we can predict memory reductions and runtime improvements for all four models at different model sizes (numbers of grid cells). These speedups reflect the current runtime of the quantum-inspired software, which can be expected to improve substantially with additional engineering effort.
These speedups may seem extreme, but they highlight the entirely new paradigm that quantum-inspired algorithms unlock for physics simulations. In particular, the runtime scaling of simulations can be decoupled from the grid size, and are tied instead to the compressibility (in the tensor-network sense) of the simulation. Simply put: conventional methods follow the rule "large model/grid→ high runtime", compared to quantum-inspired methods which follow the rule "high compressibility → low runtime".
Impact
To understand the impact this technology can have on real-world applications, we need to consider two questions:
How much compression can be achieved for a real-world simulation (e.g. of a river or sea model)?
What size grid is necessary and useful for that simulation?
When we encounter applications that require large grids and are amenable to compression, we can expect quantum-inspired software to be the fastest and most efficient solution. Identifying these use-cases, and then building software tailored to tackling them, is precisely what we will do in the next phase of the Purple NECtar project.
Beyond water modelling
The algorithms (quantum-inspired and quantum) that we developed and will develop in the Purple NECtar challenge can also be applied to problems outside of water modelling. In particular, the algorithms that we use are tailored towards solving systems of partial differential equations (PDEs). These appear throughout physics and engineering, and simulations of many physical phenomena typically boil down to solving a particular set of PDEs. For instance: Maxwell's equations can be used to simulate electrodynamics, which is relevant to chip and radar design; and the Navier-Stokes equations can be used to simulate fluid dynamics, relevant to car, plane, and ship design. At Fermioniq, we are currently developing quantum-inspired software for computational fluid dynamics (CFD), in collaboration with the Dutch aerospace centre (NLR).
