Modeling wastewater bioreactors is a challenging problem in environmental engineering because the processes are constantly changing and the microorganisms do not behave in a linear or predictable way. This makes it difficult to predict and control the system behavior. This research investigates three artificial intelligence approaches for modeling wastewater bioreactors: the Mamdani Fuzzy Inference System (FIS), the Adaptive Neuro-Fuzzy Inference System (ANFIS), and clustering-based fuzzy models. These models help us predict what is happening in the bioreactors. For instance they help us predict how the amount of substances in the water is changing over time like dS0/dt dSs/dt and dXH/dt. We used a lot of data, 15,000 points to train and test these models. The results of the models that employed neuro-clustering methods were excellent, showing high prediction accuracy. Moreover, data-driven self-training proved more efficient than manually defined rule-based training. This is especially important because the obtained models enable control of bioreactor performance by monitoring its behavior over time, thus saving energy and ensuring stable treatment process operation. Furthermore, we checked how efficient the developed models are in the case of changes in the conditions; specifically, when the properties of the influent changed. The models again showed satisfactory results, which allows for developing in the future their cooperation with control systems. This can help make the process more efficient and resilient. Since bioreactors are a core part of wastewater treatment, accurate modeling is crucial. This research contributes to improving wastewater treatment facility design and operation.
Kamala Najaf Mammadzada (Thu,) studied this question.