Experiments and simulations of direct and reverse electrodialysis: power generation, fouling prediction and optimization routes

This PhD thesis focused on the investigation and advancement of two ion-exchange membrane-based technologies, namely Reverse Electrodialysis (RED) and Electrodialysis (ED). Within the framework of circular economy approaches, these technologies offer valuable alternatives for energy generation through reverse electrodialysis (RED), as well as for desalination and resources recovery applications (ED).In the first research area, RED technology was extensively studied at laboratory and pilot scale. Within this framework, an upscaled RED unit (10 × 80 cm²) was experimentally investigated to assess the feasibility of salinity gradient energy recovery at industrial scale and to analyse the influence of electrode segmentation. Several electrode configurations and operating conditions were explored, including the use of hypersaline solutions (up to 5.0 mol/L NaCl) to reproduce real brine and industrial effluents concentration. This study contributed to filling a gap in the RED literature, where industrial scale channel lengths and electrode segmentation have so far received limited attention. A key outcome was the successful implementation of electrode segmentation, which enabled up to a 21% increase in power output compared to undivided electrodes. The RED unit achieved energy yield values of 0.5 kWh/m³, exceeding values typically reported in the literature, and its adoption demonstrated the advantages of upscaled units. Long-term continuous operation (7 days) and the use of real process feedwaters confirmed stable performance and positive net power output. RED experimental activities also demonstrated the potential benefits of the integration of RED technology into circular processes for sustainable energy recovery and critical raw material valorisation.The second research area addressed fouling phenomena in ED systems, particularly focusing on colloidal fouling mechanisms. A hybrid modelling simulation platform, combining Computational Fluid Dynamics (CFD) and a fouling growth model, was developed to simulate flow and concentration fields within spacer-filled channels of ED units subject to fouling. This work represented the first CFD model for fouling prediction in IEM-based processes. Data-driven models based on Artificial Intelligence (AI) algorithms, including Recurrent Neural Networks (RNN), Ensemble Decision Tree methods, Temporal Convolutional Network (TCN), and Temporal Fusion Transformer (TFT), were also used to predict fouling occurring in more complex conditions, closer to real applications. Within these activities, two wide experimental investigations were conducted, examining also dynamic operating conditions in ED, to evaluate fouling magnitude under a vast range of operating variables.The two datasets coming from the experimental activities were lately employed to train and validate the developed machine learning/deep learning models. Both phenomenon-driven and data-driven approaches accurately predicted fouling behaviour and offered complementary insights into its temporal evolution and effects, thereby addressing the existing gap in modelling tools for fouling prediction in electrodialysis.Overall, the results of this PhD thesis demonstrated the scalability of RED for salinity gradient energy harvesting and enhanced the modelling tools availability for predicting and understanding fouling in ED processes.