Impact of influent data frequency and model structure on the quality of WWTP model calibration and uncertainty

Application of activated sludge models (ASMs) to full-scale wastewater treatment plants (WWTPs) is still hampered by the problem of model calibration of these over-parameterised models. This either requires expert knowledge or global methods that explore a large parameter space. However, a better balance in structure between the submodels (ASM, hydraulic, aeration, etc.) and improved quality of influent data result in much smaller calibration efforts. In this contribution, a methodology is proposed that links data frequency and model structure to calibration quality and output uncertainty. It is composed of defining the model structure, the input data, an automated calibration, confidence interval computation and uncertainty propagation to the model output. Apart from the last step, the methodology is applied to an existing WWTP using three models differing only in the aeration submodel. A sensitivity analysis was performed on all models, allowing the ranking of the most important parameters to select in the subsequent calibration step. The aeration submodel proved very important to get good NH(4) predictions. Finally, the impact of data frequency was explored. Lowering the frequency resulted in larger deviations of parameter estimates from their default values and larger confidence intervals. Autocorrelation due to high frequency calibration data has an opposite effect on the confidence intervals. The proposed methodology opens doors to facilitate and improve calibration efforts and to design measurement campaigns.     

Parameter sensitivity analysis for activated sludge models No. 1 and 3 combined with one-dimensional settling model.

The aim of this study was to suggest a sensitivity analysis technique that can reliably predict effluent quality and minimize calibration efforts without being seriously affected by influent composition and parameter uncertainty in the activated sludge models No. 1 (ASM1) and No. 3 (ASM3) with a settling model. The parameter sensitivities for ASM1 and ASM3 were analyzed by three techniques such as SVM-Slope, RVM-SlopeMA, and RVM-AreaCRF. The settling model parameters were also considered. The selected highly sensitive parameters were estimated with a genetic algorithm, and the simulation results were compared as deltaEQ. For ASM1, the SVM-Slope technique proved to be an acceptable approach because it identified consistent sensitive parameter sets and presented smaller deltaEQ under every tested condition. For ASM3, no technique identified consistently sensitive parameters under different conditions. This phenomenon was regarded as the reflection of the high sensitivity of the ASM3 parameters. But it should be noted that the SVM-Slope technique presented reliable deltaEQ under every influent condition. Moreover, it was the simplest and easiest methodology for coding and quantification among those tested. Therefore, it was concluded that the SVM-Slope technique could be a reasonable approach for both ASM1 and ASM3.     

Experimental and model-based evaluation of the role of denitrifying polyphosphate accumulating organisms at two large scale WWTPs in northern Poland.

The capabilities of denitrifying Polyphosphate Accumulating Organisms (DPAOs) in two large-scale plants in northern Poland performing enhanced biological phosphorus removal (EBPR) were evaluated in this study. A series of batch tests with the process biomass aimed at the measurements of phosphate release (with artificial substrate and real wastewater) and subsequent phosphate uptake under anoxic/aerobic conditions. The process kinetics were predicted using ASM2d implemented in the GPS-X ver. 4.0.2 simulation package. The results from one experimental series (summer) were used for the model calibration, whereas the results from another series (spring) were used for the model validation. The model parameters were also accurately confirmed by predictions of the accompanying field measurements in the full-scale bioreactors. The experimental and simulation results revealed that a relatively small fraction of PAO could denitrify (eta(NO3,PAO) = 0.32). The denitrification rates associated with the anoxic storage of PP and the anoxic growth of PAO only constituted 16.0-21.0% of the denitrification rates associated with the anoxic activity of "ordinary" heterotrophs.     

Calibrating a side-stream membrane bioreactor using Activated Sludge Model No. 1.

Membrane bioreactors (MBRs) are attracting global interest but the mathematical modeling of the biological performance of MBRs remains very limited. This study focuses on the modeling of a side-stream MBR system using the Activated Sludge Model No. 1 (ASM1), and compares the results with the modeling of traditional activated sludge processes. ASM1 parameters relevant for the long-term biological behaviour in MBR systems were calibrated (i.e. Y(H) = 0.72 gCOD/gCOD, Y(A) = 0.25 gCOD/gN, b(H) = 0.25 d(-1), b(A) = 0.080 d(-1) and f(p) = 0.06), and generally agreed with the parameters in traditional activated sludge processes, with the exception that a higher autotrophic biomass decay rate was observed in the MBR. Influent wastewater characterization was proven to be a critical step in model calibration, and special care should be taken in characterizing the inert particulate COD (X(I)) concentration in the MBR influent. It appeared that the chemical-biological method was superior to the physical-chemical method. A sensitivity analysis for steady-state operation and DO dynamics suggested that the biological performance of the MBR system (the sludge concentration, effluent quality and the DO dynamics) are very sensitive to the parameters (i.e. Y(H), Y(A), b(H), b(A) micro(maxH) and micro(maxA), and influent wastewater components (X(I), S(s), X(s) and S(NH)).     

Specific SBR population behaviour as revealed by comparative dynamic simulation analysis of three full-scale municipal SBR wastewater treatment plants.

Three full-scale municipal sequential batch reactor (SBR) wastewater treatment plants (WWTPs) were investigated by dynamic simulation studies using ASM1. All three WWTPs showed similar kinetic and stoichiometric conditions in the SBR population behaviour after calibration of the models. The simulation results detected only a discrepancy to the ammonia online data during and shortly after shock loading under anoxic and anaerobic conditions that so far could not be adjusted by the ASM1 model. However, these differences did not severely affect the quality of the simulations nor the effluent flows. Additionally, in all cases a dynamic alpha factor curve occurred during the aeration phases that was verified by further oxygen transfer measurements. This might reveal new aspects for the process control, design and simulation of SBR WWTPs. A short lag phase was detected in many cases at the beginning of the first aeration phase.     

Biological nutrient removal model No.1 (BNRM1).

This paper presents the results of the work carried out by the CALAGUA Group on Mathematical Modelling of Biological Treatment Processes: the Biological Nutrient Removal Model No.1. This model is based on a new concept for dynamic simulation of wastewater treatment plants: a unique model can be used to design, simulate and optimize the whole plant, as it includes most of the biological and physico-chemical processes taking place in all treatment operations. The physical processes included are: settling and clarification processes (flocculated settling, hindered settling and thickening), volatile fatty acids elutriation and gas-liquid transfer. The chemical interactions included comprise acid-base processes, where equilibrium conditions are assumed. The biological processes included are: organic matter, nitrogen and phosphorus removal; acidogenesis, acetogenesis and methanogenesis. Environmental conditions in each operation unit (aerobic, anoxic or anaerobic) will determine which bacterial groups can grow. Thus, only the model parameters related to bacterial groups able to grow in any of the operation units of a specific WWTP will require calibration. One of the most important advantages of this model is that no additional analysis with respect to ASM2d is required for wastewater characterization. Some applications of this model have also been briefly explained in this paper.     

Impact of reactive settler models on simulated WWTP performance.

Including a reactive settler model in a wastewater treatment plant model allows representation of the biological reactions taking place in the sludge blanket in the settler, something that is neglected in many simulation studies. The idea of including a reactive settler model is investigated for an ASM1 case study. Simulations with a whole plant model including the non-reactive Takács settler model are used as a reference, and are compared to simulation results considering two reactive settler models. The first is a return sludge model block removing oxygen and a user-defined fraction of nitrate, combined with a non-reactive Takács settler. The second is a fully reactive ASM1 Takács settler model. Simulations with the ASM1 reactive settler model predicted a 15.3% and 7.4% improvement of the simulated N removal performance, for constant (steady-state) and dynamic influent conditions respectively. The oxygen/nitrate return sludge model block predicts a 10% improvement of N removal performance under dynamic conditions, and might be the better modelling option for ASM1 plants: it is computationally more efficient and it will not overrate the importance of decay processes in the settler.     

Modification of ASM3 for the determination of biomass adsorption/storage capacity in bulking sludge control.

The selector activated sludge (SAS) systems are known to prevent excessive growth of filamentous microorganisms responsible for bulking sludge, but these systems were hardly ever modelled. This study aimed to develop a model capable of predicting rapid substrate removal in the SAS systems. For this purpose, the Activated Sludge Model No. 3 (ASM3) was extended with three processes (adsorption, direct growth on the adsorbed substrate under aerobic or anoxic conditions). The modified ASM3 was tested against the results of batch experiments with the biomass originating from two full-scale SAS systems in Germany. The endogenous biomass was mixed with various readily biodegradable substrates (acetate, peptone, glucose and wastewater) and the utilisation of substrate (expresses as COD) and oxygen uptake rates (OURs) were measured during the experiments. In general, model predictions fitted to the experimental data, but a considerable number of kinetic (5) and stoichiometric (2) parameters needed to be adjusted during model calibration. The simulation results revealed that storage was generally a dominating process compared to direct growth in terms of the adsorbed substrate utilisation. The contribution of storage ranged from 65-71% (Plant A) and 69-92% (Plant B).     

Model-based evaluation of nitrogen removal in a tannery wastewater treatment plant.

Computer modelling has been used in the last 15 years as a powerful tool for understanding the behaviour of activated sludge wastewater treatment systems. However, computer models are mainly applied for domestic wastewater treatment plants (WWTPs). Application of these types of models to industrial wastewater treatment plants requires a different model structure and an accurate estimation of the kinetics and stoichiometry of the model parameters, which may be different from the ones used for domestic wastewater. Most of these parameters are strongly dependent on the wastewater composition. In this study a modified version of the activated sludge model No. 1 (ASM 1) was used to describe a tannery WWTP. Several biological tests and complementary physical-chemical analyses were performed to characterise the wastewater and sludge composition in the context of activated sludge modelling. The proposed model was calibrated under steady-state conditions and validated under dynamic flow conditions. The model was successfully used to obtain insight into the existing plant performance, possible extension and options for process optimisation. The model illustrated the potential capacity of the plant to achieve full denitrification and to handle a higher hydraulic load. Moreover, the use of a mathematical model as an effective tool in decision making was demonstrated.     

New mathematical procedure for the automatic estimation of influent characteristics in WWTPs.

This paper proposes a new methodology for the automatic characterization of the influent wastewater in WWTP. With this methodology, model components are automatically estimated by means of optimization algorithms combining a-priori knowledge of the expected wastewater composition with experimental information from the available measurement data. The characterization is carried out based on an extended model components list in which components are described by means their elemental mass fractions. This allows an easy establishment of relationships between model components with experimental data and also, to obtain a general methodology applicable to any model used for wastewater biological treatments. The characterization of the wastewater influent of Galindo-Bilbao according this methodology has demonstrated its validity and the easy application to the ASM1 model influent characterization.     

Application of Activated Sludge Model No. 1 to biological treatment of pure winery effluents: case studies.

The practical applicability of computer simulation of aerobic biological treatment systems for winery effluents was investigated to enhance traditional on-site evaluation of new processes. As there is no existing modelling tool for pure winery effluent, a model widely used for municipal activated sludge (ASM1) was used. The calibration and validation steps were performed on extended on-site data. The global soluble COD, DO and OUR were properly reproduced. Possible causes for the remaining discrepancies between measured and simulated data were identified and suggestions for improvement directions were made to adapt ASM1 to winery effluents. The calibrated model was then used to simulate scenarios to evaluate the plant behaviour for different operation or design. In combination with on-site observations, it allowed us to establish useful and justified improvement suggestions for aeration tank and aeration device design as well as feed, draw and aeration operation.     

Calibration and verification of models of organic carbon removal kinetics in Aerated Submerged Fixed-Bed Biofilm Reactors (ASFBBR): a case study of wastewater from an oil-refinery

The article presents a case-study on the calibration and verification of mathematical models of organic carbon removal kinetics in biofilm. The chosen Harrem?es and Wanner & Reichert models were calibrated with a set of model parameters obtained both during dedicated studies conducted at pilot- and lab-scales for petrochemical wastewater conditions and from the literature. Next, the models were successfully verified through studies carried out utilizing a pilot ASFBBR type bioreactor installed in an oil-refinery wastewater treatment plant. During verification the pilot biofilm reactor worked under varying surface organic loading rates (SOL), dissolved oxygen concentrations and temperatures. The verification proved that the models can be applied in practice to petrochemical wastewater treatment engineering for e.g. biofilm bioreactor dimensioning.     

Combined anaerobic digestion and biological nitrogen removal for piggery wastewater treatment: a modelling approach

In order to deal with the environmental problems associated with animal production industrialization and at the same time considering energy costs increasing, a piggery wastewater treatment process consisting of combined anaerobic digestion and biological nitrogen removal by activated sludge was developed. This contribution presents a modelling framework in order to optimize this process. Modified versions of the well established ASM1 and ADM1 models have been used. The ADM1 was extended with biological denitrification. pH calculation and liquid gas-transfer were modified to take into account the effect of associated components. Finally, two interfaces (ADMtoASM and ASMtoADM) were built in order to combine both models. These interfaces set up the COD, nitrogen, alkalinity and charge fractionation between both models. However, for the mass balances between both models, some hypotheses were considered and might be evaluated.     

Partial Least Squares local calibration of a UV-visible spectrometer used for in situ measurements of COD and TSS concentrations in urban drainage systems.

Recent UV-visible spectrometers deliver on line and in situ absorbance spectra in wastewater or stormwater transported in urban drainage systems. After calibration with local data sets, spectra can be used to estimate pollutant concentrations. Calibration methods are usually based on PLS (Partial Least Squares) regression. Their most important difficulty lies in the identification of the number of both i) the latent vectors and ii) the independent variables. A method is proposed to identify these variables, based on an exhaustive tests procedure (Jackknife cross validation and matrix of prediction indicator). It was applied to estimate TSS (total suspended solids) or COD (chemical oxygen demand) concentrations at the inlet of a storage-settling tank in a stormwater separate sewer system, and compared to three other calibration methods used either for turbidity meters or UV-visible spectrometers. With the available calibration data set: i) the spectrometer gives results with better prediction quality than the turbidity meter, ii) for the spectrometer, local calibration gives better results than global calibration, iii) the proposed PLS method gives results with a similar order of magnitude in uncertainties as the manufacturer local calibration method, but is more open and transparent for the user. Similar results were obtained for a second data set.     

A Parsimonious Hydrological Model for a Data Scarce Dryland Region

Inapplicability of state of the art hydrological models due to scarce data motivates the need for a modeling approach that can be well constrained to available data and still model the dominant processes. Such an approach requires embedded model relationships to be simple and parsimonious in parameters for robust model selection. Simplicity in functional relationship is also important from water management point of view if these models are to be coupled with economic system models for meaningful policy assessment. We propose a similar approach, but rather than selecting (through calibration) processes from a set of processes predefined in terms of functionalities or modules, we model already known dominant processes in dryland areas (evaporation, Hortonian overland flows, transmission loses and subsurface flows) in a simple manner by explicitly programming them as constraints and obtain parameters by minimizing a performance based objective function. Such use of mathematical programming allows flexible model calibration and simulation in terms of available data and constraints. The model results of the approach are however not perfect given its infancy. Nonetheless its imperfections can guide us to further improvements, in particular with regards to model structure improvement.     

Calibration and simulation of two large wastewater treatment plants operated for nutrient removal.

Control and optimisation of plant processes has become a priority for WWTP managers. The calibration and verification of a mathematical model provides an important tool for the investigation of advanced control strategies that may assist in the design or optimization of WWTPs. This paper describes the calibration of the ASM2d model for two full scale biological nitrogen and phosphorus removal plants in order to characterize the biological process and to upgrade the plants' performance. Results from simulation showed a good correspondence with experimental data demonstrating that the model and the calibrated parameters were able to predict the behaviour of both WWTPs. Once the calibration and simulation process was finished, a study for each WWTP was done with the aim of improving its performance. Modifications focused on reactor configuration and operation strategies were proposed.     

Modelling carbon oxidation in pulp mill activated sludge systems: calibration of Activated Sludge Model No 3.

Activated Sludge Model No 3 (ASM3) was chosen to model an activated sludge system treating effluents from a mechanical pulp and paper mill. The high COD concentration and the high content of readily biodegradable substrates of the wastewater make this model appropriate for this system. ASM3 was calibrated based on batch respirometric tests using fresh wastewater and sludge from the treatment plant, and on analytical measurements of COD, TSS and VSS. The model, developed for municipal wastewater, was found suitable for fitting a variety of respirometric batch tests, performed at different temperatures and food to microorganism ratios (F/M). Therefore, a set of calibrated parameters, as well as the wastewater COD fractions, was estimated for this industrial wastewater. The majority of the calibrated parameters were in the range of those found in the literature.