Model-based opportunities for the optimization of the dosage of hydrogen peroxide in the photo Fenton process

Abstract

Advanced Oxidation Processes (AOP) have been proposed as alternative water treatments coping with recalcitrant organic pollutants [1]. AOPs are based on in-situ generation of highly oxidant hydroxyl radicals. Particularly, in the photo-Fenton process they are produced from ferrous salts (Fe) and hydrogen peroxide (HP). However, this process has been acknowledged to suffer inefficient reactions scavenging HP; which has motivated a large amount of research aimed to determine efficient ratios for the initial concentrations of reactants (Fe/HP). Dosage is also reported to reduce these side reactions and improve the performance of these processes. Certainly, since they are operated batchwise, the most efficient ratio Fe/HP should not be regarded as an initial value, but as a profile that may undergo optimization. Yet, such optimization problem has not been attempted. A large experimental effort has produced empirical models that cannot be scaled up and do not address the process dynamics, while some first-principle kinetic models that can be found on the literature [2] require a high computational cost for too simple reactions. Therefore, a first issue towards optimization is model selection. This work adopts the kinetic model by Cabrera Reina [3] based on aggregated components. This model focuses on practical observable variables such as dissolved oxygen and total organic carbon (TOC), and provides a simplified modelling of delayed response of TOC and scavenging reactions. Hence, this work expands it to semi-batch operation and addresses simulation and subsequent optimization of the dosage profile using the Python and Modelica open-source programing languages. Python is used as the core providing functions that Modelica lacks, while the model is implemented in Modelica to take advantage of its model-based language. The optimization of the HP dosage profile addressed two different scenarios and objective functions. Thus, Pareto frontiers were determined to analyse trade-offs and opportunities, and to aid decision making: i) The TOC reduction to be achieved under time and HP limitations, and ii) The total HP required to attain a given conversion (TOC) within a given time horizon. Continuous and piecewise optimization approaches were tested and discussed. Results validate the importance of determining efficient dosage profiles for the photo-Fenton process. Model based optimization allows exploring opportunities and trade-offs, and aids decision-making. Hence, this study fosters further work on model fitting for specific applications.