Integrating Process‐Based Reactive Transport Modeling and Machine Learning for Electrokinetic Remediation of Contaminated Groundwater

Advanced reactive transport models of fluid flow and solute in subsurface porous media are instrumental for the assessment contaminant environmental fate design situ remediation interventions. However, increasing complexity process-based simulators often leads to long runtimes, which poses severe restrictions tasks that require numerous model evaluations. To overcome this limitation, we demonstrate how machine learning surrogate models, trained on outputs a limited number simulations, can predict evolution complex systems. We focus electrokinetic enhanced bioremediation chlorinated solvents low-permeability media, is an technology entailing suite coupled physical, chemical, biological processes. A process-based, multicomponent model, capable describing key mechanisms transport, setup two-dimensional domain. The accounts electromigration electroosmosis, electrostatic interactions between charged species, chemistry pore water solution, microbially mediated degradation organic compounds, dynamics different degraders. develop response surface framework using artificial neural network as approximation function show has capability flexibility capture allows computationally efficient exploration, sensitivity analysis, uncertainty quantification.