GEODA: Digital twin for wastewater treatment plant control based on advanced hydrodynamic simulation

Development of a Hydrodynamic Digital Twin for WWTP control.

Or in other words… biological reactor control in a wastewater treatment plant using a digital twin based on an artificial neural network trained through computational fluid dynamics simulations for energy savings and improved wastewater purification performance.

The main objective of the project is the development of a digital twin in wastewater treatment plant (WWTP) bioreactors based on Computational Fluid Dynamics (CFD) that allows for precise and efficient real-time control. The DT will be a control tool that will use an Artificial Neural Network (ANN) trained by CFD simulation to determine the distribution of hydrodynamic parameters inside the reactors, thus solving the problem of computation time and computational resources of CFD simulations, but retaining the necessary distribution accuracy for the new control.



Specifically, the neural network will be trained to learn the hydraulic behaviour inside the biological reactor, allowing to solve the distribution of the hydrodynamic variables in a very short time from results generated by CFD simulations of the reactor for different operation scenarios of this treatment stage. The results of the CFD simulation will be transferred to the neural network, providing a high level of detail for learning and verification. The Hybrid Digital Twin will then calculate the biochemistry on the hydrodynamic solution, making this double distribution (hydrodynamic and biochemical variables) available throughout the reactor in real time. This huge amount of information calculated inside the digital twin of the WWTP reactor and validated by field probes, opens up new ways to carry out the control and operation of the reactors, in a more reliable and efficient way, and even in a predictive way. By means of this tool based on the prediction of the ANN by means of CFD simulation, a new paradigm in WWTP control is proposed.



Main results obtained: As main results, GEODA has allowed the development of a Digital Twin (DM) that contains an artificial neural network (ANN), trained with CFD simulation results, capable of solving the hydrodynamics of the biological reactor in less than a minute without renouncing the precision and 3D detail of the CFD. The DG is connected to the WWTP SCADA, so that the DG is updated with real data from its physical twin. As a result, the updated data constitute the input parameters for the updated real-time hydrodynamics calculation with the ANN. After the ANN hydrodynamics calculation, the DG calculates the reactor biochemistry on the hydrodynamics data, showing a degree of detail and accuracy in 3D not yet provided by other DGs. The DG allows to simulate times of more than 2 hours and whose results can be compared with real plant data so that the DG becomes an Early Warning System (EWS) and Decision Support System (DSS) for a more sustainable and responsible control of the WWTP.