Model-Based Analysis of Increased Loads on the Performance of Activated Sludge and Waste Stabilization Ponds

Abstract

In a way to counter criticism on low cost-effective conventional activated sludge (AS) technology, waste stabilization ponds (WSPs) offer a valid alternative for wastewater treatment due to their simple and inexpensive operation. To evaluate this alternative with respect to its robustness and resilience capacity, we perform in silico experiments of different peak-load scenarios in two mathematical models representing the two systems. A systematic process of quality assurance for these virtual experiments is implemented, including sensitivity and identifiability analysis, with non-linear error propagation. Moreover, model calibration of a 210-day real experiment with 31 days of increased load was added to the evaluation. Generally speaking, increased-load scenarios run in silico showed that WSP systems are more resilient towards intermediate disturbances, hence, are suitable to treat not only municipal wastewater, but also industrial wastewater, such as poultry wastewater, and paperboard wastewater. However, when disturbances are extreme (over 7000 mg COD·L−1), the common design of the natural system fails to perform better than AS. Besides, the application of sensitivity analysis reveals the most influential parameters on the performance of the two systems. In the AS system, parameters related to autotrophic bacteria have the highest influence on the dynamics of particulate organic matter, while nitrogen removal is largely driven by nitrification and denitrification. Conversely, with an insignificant contribution of heterotrophs, the nutrient removal in the pond system is mostly done by algal assimilation. Furthermore, this systematic model-based analysis proved to be a suitable means for investigating the maximum load of wastewater treatment systems, and from that avoiding environmental problems and high economic costs for cleaning surface waters after severe overload events.

A Practical Protocol for the Experimental Design of Comparative Studies on Water Treatment

“…To be valid in experimental design, it is important to note that preliminary models need to be optimized and evaluated first. Model optimization and evaluation are the link between virtual experiments and real experiments as they require real data to calibrate and validate the simulating models [35,36]. This data requirement is one of the major challenges of virtual experiments.…”

Machine learning applications in river research: Trends, opportunities and challenges

“…Having no output features, unsupervised algorithms are often applied to explore and describe data; hence, it is difficult to assess their performance. Regarding supervised ML, linear models and ensemble methods are two classes of powerful predictive models that are widely applied in practice (Géron, 2017; Ho et al, 2018). Simply put, linear models use a linear function to predict output feature, while ensemble methods make use of the ‘wisdom of the crowd’ by aggregating the predictions of a group of predictors, that is, regressors or classifiers, to predict new incoming instances (Ho et al, 2018; Krawczyk et al, 2017).…”

“…Regarding supervised ML, linear models and ensemble methods are two classes of powerful predictive models that are widely applied in practice (Géron, 2017; Ho et al, 2018). Simply put, linear models use a linear function to predict output feature, while ensemble methods make use of the ‘wisdom of the crowd’ by aggregating the predictions of a group of predictors, that is, regressors or classifiers, to predict new incoming instances (Ho et al, 2018; Krawczyk et al, 2017). As new methods in river ML modelling, reinforcement learning, Naïve Bayes, associate rule and multiclass and multilabel algorithms have ranked the lowest in the spectrum of ML applications in river research.…”

Sustainability of Lakes and Reservoirs: Multiple Perspectives Based on Ecosystem Services

“…The most known and broadly used WSP layout is composed of a sequence of anaerobic (AP), facultative (FP), and (a series of) maturation ponds (MP). Anaerobic ponds are normally located at the primary treatment stage to remove organic matter, due to their robustness against a high loading rate [16,17]. Subsequently, taking advantage of photosynthetic oxygenation, FPs are applied for further organic matter and nutrient removal with minimal operational costs [18].…”