118 - Adaptive Chemical Mixing Resolution to Accelerate Water Quality Simulations without Accuracy Loss

The release of EPANET 2.2 in 2020 included an important improvement to its water quality module, in which the transport of chemical species is simulated sequentially from pipe to pipe following flow directions. While this no longer allows for the mixing computations at junctions to be performed in parallel, it explicitly enables species to traverse multiple pipes in a single time step, leading to more accurate results even with coarse temporal resolutions. However, as we demonstrate here, the lengthening of time steps with this scheme still leads to large errors when the plumes in the upstream pipes present heterogeneities (e.g., with pulsing chemical injections or sudden biofilm detachments). Instead of averaging the concentrations of all incoming water parcels together, and thus flattening the response downstream, here we propose to adaptively mix and release incoming parcels in partial groups to produce a non-homogeneous response from junctions. In this way, the strategy virtually increases the temporal and spatial resolution of the model, on demand, whenever significant heterogeneities are detected. This proposal is implemented in InfoWater Pro, which allows us to demonstrate in two test networks the much slower deterioration of accuracy as we increase the length of the computational steps. While the new mixing scheme requires additional memory and chemical reaction calculations due to the additional parcels produced, our results show that these overheads are overcompensated by the gains in reducing the number of computation steps without loss of accuracy.