The primary objective of this research is to model and optimize wastewater treatment process in a wastewater treatment plant (WWTP). As the treatment process is complex, its operations pose challenges. Traditional physics-based and mathematical- models have limitations in predicting the behavior of the wastewater process and optimization of its operations. Automated control and information technology enables continuous collection of data. The collected data contains process information allowing to predict and optimize the process. Although the data offered by the WWTP is plentiful, it has not been fully used to extract meaningful information to improve performance of the plant. A data-driven approach is promising in identifying useful patterns and models using algorithms versed in statistics and computational intelligence. Successful data-mining applications have been reported in business, manufacturing, science, and engineering. The focus of this research is to model and optimize the wastewater treatment process and ultimately improve efficiency of WWTPs. To maintain the effluent quality, the influent flow rate, the influent pollutants including the total suspended solids (TSS) and CBOD, are predicted in short-term and long-term to provide information to efficiently operate the treatment process. To reduce energy consumption and improve energy efficiency, the process of biogas production, activated sludge process and pumping station are modeled and optimized with evolutionary computation algorithms. Modeling and optimization of wastewater treatment processes faces three major challenges. The first one is related to the data. As wastewater treatment includes physical, chemical, and biological processes, and instruments collecting large volumes of data. Many variables in the dataset are strongly coupled. The data is noisy, uncertain, and incomplete. Therefore, several preprocessing algorithms should be used to preprocess the data, reduce its dimensionality, and determine import variables. The second challenge is in the temporal nature of the process. Different data-mining algorithms are used to obtain accurate models. The last challenge is the optimization of the process models. As the models are usually highly nonlinear and dynamic, novel evolutionary computational algorithms are used. This research addresses these three challenges. The major contribution of this research is in modeling and optimizing the wastewater treatment process with a data-driven approach. The process model built is then optimized with evolutionary computational algorithms to find the optimal solutions for improving process efficiency and reducing energy consumption.

Aerated treatment wetlands are an increasingly recognized nature–based technology for thetreatment of domestic and industrial wastewater. As biodegradation is the most importanttreatment mechanism in aerated wetlands, these systems heavily rely on mechanical aerationmediated oxygen transfer to supply the dissolved oxygen demand of the associated microbialcommunity. In the last decade, research on aerated wetlands has evolved, however, majorquestions on aeration, the associated oxygen transfer and the quantitative link to treatmentperformance still remain unknown. Answering these questions can further improve aeratedwetland design to optimize treatment efficacy and economical efficiency. This dissertation investigated the link of oxygen transfer to the air flow rate of aerationand elucidated the associated impact on treatment performance for organic carbon and nitrogenin horizontal flow aerated wetlands. Therefore, a numerical process model includingone dimensional reactive transport was developed. This model describes the main processesinvolved in horizontal flow aerated wetlands: water flow, heat transport, transport of solubleand particulate wastewater pollutants, biodegradation by a network of bacterial communitiesand oxygen transfer through mechanical aeration. For model calibration and validation, pilot–scale experiments in horizontal flow aerated wetlands treating real wastewater were conducted.These included conservative tracer experiments as well as monitoring steady–state operationat variable air flow rates and aeration interruption. In general, the model was able to simulate conservative tracer transport as well as treatmentperformance for organic carbon and nitrogen at steady–state operation and aeration interruptionwith sufficient accuracy. A local sensitivity analysis of the calibrated parameters revealedporosity, hydraulic permeability and dispersion length as well as the oxygen transfer coefficientkLa as most important. When operating the wetland systems at steady–state, aeration provideda mostly aerobe environment, except at the influent zone. However, when aeration wasinterrupted, anaerobe process started to take over and treatment performance declined within3–4 days. The modeling elucidated that methanogenic and sulphate reducing bacteria can playa significant role for organic carbon removal during aeration interruption. Moreover, the modelrevealed a non–linear declining relationship of the air flow rate with oxygen transfer coefficientkLa and of kLa with treatment performance. The alteration of oxygen transfer by wastewaterpollutant concentration was then investigated in a laboratory–scale column experiment. Basedon this experiment, an empirical equation describing the inhibitory effect of soluble chemicaloxygen demand (CODs) on the oxygen transfer coefficient kLa was derived and incorporatedinto the process model. With the extended model several simulation scenarios were analyzedto quantify the impact of the inhibited oxygen transfer on treatment performance. It turnedout that the reduction of oxygen transfer by CODs will, most likely, be relevant only at highinfluent wastewater strength (CODs 300 mg L-1), low aeration (air flow rate 50 L m-2h-1) or when the aerated wetland design includes zoned aeration. With respect to secondarytreatment of domestic effluents at similar strength using a spatially uniform aeration, an airflow rate of approximately 150–200 L m-2 h-1 can be recommended as a reasonable compromisebetween treatment efficiency and robustness. If zoned aeration is intended (e.g. to create a redox zonation), however, the air flow rate should be increased to approximately 400 L m-2 h-1 to supress the inhibition of oxygen transfer by CODs concentration. Furthermore, the air flow rate at steady–state operation (50–500 L m-2 h-1) did not substantially affect the response in effluent concentrations for organic carbon and nitrogen. This means that at steady–state air flow rates of 50–500 L m-2 h-1 operation, treatment efficacy during aeration interruption will deteriorate and recover in a similar time. In conclusion, this dissertation provides quantitative insights into the mechanisms of aeration and treatment performance for organic carbon and nitrogen in horizontal flow aerated treatment wetlands. The findings obtained can support aerated treatment wetland design for research experiments and engineering applications. Therefore, this dissertation represents a significant advancement in the field of aerated treatment wetland research.

L’objectif de ce travail est d’une part (i) étudier le comportement de la biomasse notamment la production d’EPS en présence des composés pharmaceutiques (un agent anticancéreux et cinq antibiotiques); et d’autre part, (ii) étudier les EPS dans le contexte de décantation des boues en présence d’agents fongiques et en situations réelles dans des stations d’épuration. L’étude en présence de l’agent anticancéreux a été réalisée dans des bioréacteurs à membranes. La présence de l’agent anticancéreux a induit l’augmentation de la production d’EPS agissant comme un mécanisme de protection microbienne qui était à l’origine du colmatage des membranes. L’effet de cinq antibiotiques a été évalué en réacteur batch. La famille des macrolides a montré un effet plus important sur l’activité microbienne avec une augmentation significative de la production d’EPS associée à un mécanisme de protection. La décantation des boues en présence des cultures fongiques a été conduite en réacteur pilote. Une amélioration spectaculaire de la décantation a été liée à une meilleure cohésion au sein des flocs imputable en grand partie à l’augmentation de la production d’EPS. Enfin, le diagnostic du procédé de traitement des eaux a été établi sur trois stations d’épuration des papeteries grâce à une double approche d’une part l’analyse physico-chimique des boues et d’autre part, l’exploitation statistique d’analyse en composantes principales (ACP) des paramètres technologiques enregistrées dans chaque station. Nous avons tenté d’exprimer sous forme de régressions linéaires ou polynomiales de deuxième degré, la décantation en fonction d’une quantité réduite des paramètres mesurés / The objective of this work is firstly, i) to study the microbial behaviour of the biomass especially the production of the EPS in the presence of pharmaceutical compounds (an anticancer product and five antibiotics); and secondly, ii) to study EPS in the context of the sludge settling in wastewater treatment plants. The study in the presence of the anticancer product was done in membrane bioreactors. The presence of the anticancer product provoked the production of the EPS as the protection mechanism which is at the origin of the membrane fouling.The effect of five antibiotics was evaluated in batch reactors. The family of macrolides showed the most important effect on the microbial activity with a significant increase of the EPS production which was associated with a protection mechanism.Sludge settling in the presence of fungi was carried out in a pilot reactor. The spectacularly improvement of the sludge settleability was related with a better cohesion inside of the flocs attributed to an increase of the EPS production.Finally, the diagnosis of different wastewater treatment processes was established in three paper mill wastewater treatment plants thanks to the double approach used here, the physico-chemical analysis of the sludge and the statistical analysis by principal components analysis of the different parameters recorded in each plant. We tried to describe the parameter related to the settling behaviour by linear or polynomial regressions of second degree in function of a reduced number of the measured parameters

Les systèmes réels sont généralement de nature non-linéaire, et leurs modélisations etsurveillance restent une tâche difficile à accomplir. Néanmoins, avec les progrès technologiqueson dispose maintenant d'un atout de taille sur ces systèmes qui est les données.Ce travail présente une technique de diagnostic de défaut et de modélisation basée en grandepartie sur la méthode d'apprentissage automatique « Les Machines à Vecteurs de Support,SVM » qui est basée sur les données. La méthodologie proposée est appliquée à différentessystèmes multivariables et non linéaires, à savoir : un procédé de traitement des eaux usées, unsystème éolien et un réacteur chimique parfaitement agité.L'objectif de cette thèse de doctorat est d'examiner la possibilité d'extraire le maximumd'information à partir de données afin de surveiller efficacement le comportement de systèmesréels et de détecter rapidement tout défaut qui peut compromettre leur bon fonctionnement. Lamême méthode est utilisée pour la modélisation des différents systèmes. Plusieurs défis ont étérelevés tels que la complexité du comportement des systèmes, le grand nombre de mesuresvariant à différentes échelles de temps, la présence de bruit et les perturbations. Une méthodegénérique de diagnostic de défauts est proposée par la génération des caractéristiques de chaquedéfaut suivie d’une étape d'évaluation de ces caractéristiques avec une amélioration du transfertde connaissances en modélisation.Dans cette thèse ont a démontré l'utilité de l'outil Machines à Vecteurs de Support, enclassification par la construction de modèles de décision SVM dédiés à l'évaluation descaractéristiques de défaut, et aussi en tant qu'estimateur non linéaire/ou pour la modélisation parl'utilisation des machines à vecteurs de support dédiés pour la régression (SVR).La combinaison de SVM et d’une méthode basée sur le modèle "observateur" a été aussi étudiéeet a été nécessaire dans certains cas pour garantir un bon diagnostic de défauts. / Real systems are usually nonlinear and their modeling and monitoring remains adifficult task. However, with advances in technology and the availability of big amounts of data,we have a facility to operate these systems.This work presents a methodology for fault diagnosis and modeling which is in large part basedon the method of Support Vector Machines (SVM) which data-based. The proposedmethodology is applied to various nonlinear multivariable systems including: wastewatertreatment processes, wind turbines and stirred tank reactors.The objective of this PhD is to examine the possibility of extracting the maximum of informationfrom data to effectively monitor the behavior of real systems and rapidly detect any faults whichmay impair their proper functioning. The same method is used for modeling the differentsystems. Several challenges were identified and surmounted such as the complexity of thesystem behavior, large amount of data varying at different time scales, the presence of noise anddisturbances. A generic method of fault diagnosis is proposed for the generation of the faultcharacteristics followed by an evaluation of these characteristics as well as an improved transferof knowledge in modeling.In this thesis the usefulness of the tool Support Vector Machines in Classification has beendemonstrated by the construction of decision models dedicated to evaluating the characteristicsof faults, and also its usefulness for modeling/ or as estimator for the nonlinear systems usingsupport vector machines dedicated for regression (SVR).The combination of SVM and a method based on models “observer” was also considered andwas found to be interesting in some cases to ensure proper fault diagnosis.