Benchmark Simulation Model No. 1 Research Articles

Modelling anaerobic co-digestion in Benchmark Simulation Model No. 2: Parameter estimation, substrate characterisation and plant-wide integration

Anaerobic co-digestion is an emerging practice at wastewater treatment plants (WWTPs) to improve the energy balance and integrate waste management. Modelling of co-digestion in a plant-wide WWTP model is a powerful tool to assess the impact of co-substrate selection and dose strategy on digester performance and plant-wide effects. A feasible procedure to characterise and fractionate co-substrates COD for the Benchmark Simulation Model No. 2 (BSM2) was developed. This procedure is also applicable for the Anaerobic Digestion Model No. 1 (ADM1). Long chain fatty acid inhibition was included in the ADM1 model to allow for realistic modelling of lipid rich co-substrates. Sensitivity analysis revealed that, apart from the biodegradable fraction of COD, protein and lipid fractions are the most important fractions for methane production and digester stability, with at least two major failure modes identified through principal component analysis (PCA). The model and procedure were tested on bio-methane potential (BMP) tests on three substrates, each rich on carbohydrates, proteins or lipids with good predictive capability in all three cases. This model was then applied to a plant-wide simulation study which confirmed the positive effects of co-digestion on methane production and total operational cost. Simulations also revealed the importance of limiting the protein load to the anaerobic digester to avoid ammonia inhibition in the digester and overloading of the nitrogen removal processes in the water train. In contrast, the digester can treat relatively high loads of lipid rich substrates without prolonged disturbances.

Control structure design for resource recovery using the enhanced biological phosphorus removal and recovery (EBP2R) activated sludge process

Nowadays, wastewater is considered as a set of resources to be recovered rather than a mixture of pollutants that should be removed. Many resource recovery schemes have been proposed, involving the use of novel technologies whose controllability is poorly studied. In this paper we present a control structure for the novel enhanced biological phosphorus removal and recovery (EBP2R) process, which is currently under development. The aim of the EBP2R is to maximize phosphorus recovery through optimal green micro-algal cultivation, which is achieved by controlling the nitrogen to phosphorus ratio (N-to-P ratio) fed to the algae. Process control structures are developed for a sequencing batch reactor (SBR) and a continuous flow reactor system (CFS). Results, obtained using the Benchmark Simulation Model No. 1 (BSM1) dynamic input disturbance time series, suggest that the SBR can maintain a stable N-to-P ratio in the effluent (16.9±0.07) and can recover about 72% of the influent phosphorus. The phosphorus recovered by the CFS is limited by the influent nitrogen (65% of the influent phosphorus load). Using the CFS configuration the effluent N-to-P ratio cannot be effectively controlled (16.45±2.48). Therefore, it is concluded that the SBR is the most effective reactor configuration for the EBP2R process. Importantly, the designed control structures rely on control loops that do not require chemical dosing for nutrient management, thereby reducing the environmental footprint of the EBP2R process. The proposed control strategies can be applied to other phosphorus recovery schemes where short sludge age EBPR systems play an important role.

Modelling phosphorus (P), sulfur (S) and iron (Fe) interactions for dynamic simulations of anaerobic digestion processes

This paper proposes a series of extensions to functionally upgrade the IWA Anaerobic Digestion Model No. 1 (ADM1) to allow for plant-wide phosphorus (P) simulation. The close interplay between the P, sulfur (S) and iron (Fe) cycles requires a substantial (and unavoidable) increase in model complexity due to the involved three-phase physico-chemical and biological transformations. The ADM1 version, implemented in the plant-wide context provided by the Benchmark Simulation Model No. 2 (BSM2), is used as the basic platform (A0). Three different model extensions (A1, A2, A3) are implemented, simulated and evaluated. The first extension (A1) considers P transformations by accounting for the kinetic decay of polyphosphates (XPP) and potential uptake of volatile fatty acids (VFA) to produce polyhydroxyalkanoates (XPHA) by phosphorus accumulating organisms (XPAO). Two variant extensions (A2,1/A2,2) describe biological production of sulfides (SIS) by means of sulfate reducing bacteria (XSRB) utilising hydrogen only (autolithotrophically) or hydrogen plus organic acids (heterorganotrophically) as electron sources, respectively. These two approaches also consider a potential hydrogen sulfide (ZH2S) inhibition effect and stripping to the gas phase (GH2S). The third extension (A3) accounts for chemical iron (III) (SFe3+) reduction to iron (II) (SFe2+) using hydrogen (SH2) and sulfides (SIS) as electron donors. A set of pre/post interfaces between the Activated Sludge Model No. 2d (ASM2d) and ADM1 are furthermore proposed in order to allow for plant-wide (model-based) analysis and study of the interactions between the water and sludge lines. Simulation (A1 – A3) results show that the ratio between soluble/particulate P compounds strongly depends on the pH and cationic load, which determines the capacity to form (or not) precipitation products. Implementations A1 and A2,1/A2,2 lead to a reduction in the predicted methane/biogas production (and potential energy recovery) compared to reference ADM1 predictions (A0). This reduction is attributed to two factors: (1) loss of electron equivalents due to sulfate (SSO4) reduction by XSRB and storage of XPHA by XPAO; and, (2) decrease of acetoclastic and hydrogenotrophic methanogenesis due to ZH2S inhibition. Model A3 shows the potential for iron to remove free SIS (and consequently inhibition) and instead promote iron sulfide (XFeS) precipitation. It also reduces the quantities of struvite (XMgNH4PO4) and calcium phosphate (XCa3(PO4)2) that are formed due to its higher affinity for phosphate anions. This study provides a detailed analysis of the different model assumptions, the effect that operational/design conditions have on the model predictions and the practical implications of the proposed model extensions in view of plant-wide modelling/development of resource recovery strategies.

Study on Nonlinear Model Predictive Control of Activated Sludge Process

The model predictive control (MPC) with a nonlinear back propagation (BP) was used to design activated sludge process control system, and then a new improved dynamic matrix controller (IDMC) design method was proposed. The method was tested with the digital simulation analysis in a small-scale wastewater treatment plant based on benchmark simulation model No.1 (BSM1). The results showed that IDMC designed by the method has good dynamic characteristics and strong adaptability to stock load, and thus improves stability of the control system.

Fault detection and isolation of sensors in aeration control systems.

In this paper, we consider the problem of fault detection (FD) and isolation in the aeration system of an activated sludge process. For this study, the dissolved oxygen in each aerated zone is assumed to be controlled automatically. As the basis for an FD method we use the ratio of air flow rates into different zones. The method is evaluated in two scenarios: using the Benchmark Simulation Model no. 1 (BSM1) by Monte Carlo simulations and using data from a wastewater treatment plant. The FD method shows good results for a correct and early FD and isolation.

A plant-wide aqueous phase chemistry module describing pH variations and ion speciation/pairing in wastewater treatment process models

There is a growing interest within the Wastewater Treatment Plant (WWTP) modelling community to correctly describe physico–chemical processes after many years of mainly focusing on biokinetics. Indeed, future modelling needs, such as a plant-wide phosphorus (P) description, require a major, but unavoidable, additional degree of complexity when representing cationic/anionic behaviour in Activated Sludge (AS)/Anaerobic Digestion (AD) systems. In this paper, a plant-wide aqueous phase chemistry module describing pH variations plus ion speciation/pairing is presented and interfaced with industry standard models. The module accounts for extensive consideration of non-ideality, including ion activities instead of molar concentrations and complex ion pairing. The general equilibria are formulated as a set of Differential Algebraic Equations (DAEs) instead of Ordinary Differential Equations (ODEs) in order to reduce the overall stiffness of the system, thereby enhancing simulation speed. Additionally, a multi-dimensional version of the Newton–Raphson algorithm is applied to handle the existing multiple algebraic inter-dependencies. The latter is reinforced with the Simulated Annealing method to increase the robustness of the solver making the system not so dependant of the initial conditions. Simulation results show pH predictions when describing Biological Nutrient Removal (BNR) by the activated sludge models (ASM) 1, 2d and 3 comparing the performance of a nitrogen removal (WWTP1) and a combined nitrogen and phosphorus removal (WWTP2) treatment plant configuration under different anaerobic/anoxic/aerobic conditions. The same framework is implemented in the Benchmark Simulation Model No. 2 (BSM2) version of the Anaerobic Digestion Model No. 1 (ADM1) (WWTP3) as well, predicting pH values at different cationic/anionic loads. In this way, the general applicability/flexibility of the proposed approach is demonstrated, by implementing the aqueous phase chemistry module in some of the most frequently used WWTP process simulation models. Finally, it is shown how traditional wastewater modelling studies can be complemented with a rigorous description of aqueous phase and ion chemistry (pH, speciation, complexation).

Removing violations of the effluent pollution in a wastewater treatment process

This paper presents different control strategies for biological wastewater treatment plants, with the goal of avoiding violations of effluent pollution limits while, at the same time, improving effluent quality and decreasing operational costs. The control strategies are based on Model Predictive Control (MPC) and functions that relate the input and manipulated variables. The Benchmark Simulation Model No.1 (BSM1) is used for evaluation. A hierarchical structure regulates the dissolved oxygen (DO) of the three aerated tanks based on the ammonium and ammonia nitrogen concentration (NH) in the fifth tank (NH5). An MPC with feedforward compensation is proposed for the lower level and an affine function is selected for the higher level. A tuning region is determined modifying the tuning parameters of the higher level, in which the effluent quality and operational costs are simultaneously improved in comparison with the default control strategy of BSM1. To eliminate violations of total nitrogen in the effluent (Ntot,e), an affine function, implemented with a sliding window, adds external carbon flow rate in the first tank based on nitrate nitrogen in the fifth tank (NO5) plus NH5. To avoid violations of NH in the effluent (NHe), a combination of a linear function and an exponential function that manipulates the internal recirculation flow rate based on NH5 and NH in the influent is proposed. As a result, Ntot,e violations and NHe violations are avoided for dry, rain and storm weather conditions. In addition, an improvement of effluent quality and a reduction of operational costs are achieved at the same time, except in the cases of rain and storm weathers for NHe violations removal, in which the costs increase.

Fuzzy Control and Model Predictive Control Configurations for Effluent Violations Removal in Wastewater Treatment Plants

In this paper the following new control objectives for biological wastewater treatment plants (WWTPs) have been established: to eliminate violations of total nitrogen in the effluent (Ntot,e) or ammonium and ammonia nitrogen concentration (NH) in the effluent (NHe) and at the same time handle the customary requirements of improving effluent quality and reducing operational costs. The Benchmark Simulation Model No. 1 (BSM1) is used for evaluation, and the control is based on Model Predicitive Control (MPC) and fuzzy logic. To improve effluent quality and to reduce operational costs, a hierarchical control structure is implemented to regulate the dissolved oxygen (DO) on the three aerated tanks. The high level of this hierarchical structure is developed with a fuzzy controller that adapts the DO set points of the low level based on the NH concentration in the fifth tank (NH5). The low level is composed of three MPC controllers with feedforward control (MPC + FF). For avoiding violations of Ntot,e, a second fuzzy controller is used to manipulate the external carbon flow rate in the first tank (qEC1) based on nitrate nitrogen in the fifth tank (NO5) plus NH5. For avoiding violations of NHe, a third fuzzy controller is applied to manipulate the internal recirculation flow rate (Qrin) based on NH5 and NH in the influent. Simulation results show the benefit of the proposed approach.

Data-Driven Approach of KPI Monitoring and Prediction with Application to Wastewater Treatment Process

In this paper, a data-driven scheme of key performance indicator (KPI) monitoring, prediction and KPI related fault detection is applied to the wastewater treatment process (WWTP). By means of a data-driven realization of the so-called left coprime factorization (LCF) of the process, the efficient monitoring and prediction of chemical oxygen demand (COD) concentration in the effluent flow are realized both for the situation that COD is measurable and unmeasurable. The well established Benchmark Simulation Model no. 1 (BSM1) is utilized for the demonstration of the effectiveness of this approach.

Challenges encountered when expanding activated sludge models: a case study based on N2O production.

It is common practice in wastewater engineering to extend standard activated sludge models (ASMs) with extra process equations derived from batch experiments. However, such experiments have often been performed under conditions different from the ones normally found in wastewater treatment plants (WWTPs). As a consequence, these experiments might not be representative for full-scale performance, and unexpected behaviour may be observed when simulating WWTP models using the derived process equations. In this paper we want to highlight problems encountered using a simplified case study: a modified version of the Activated Sludge Model No. 1 (ASM1) is upgraded with nitrous oxide (N2O) formation by ammonia-oxidizing bacteria. Four different model structures have been implemented in the Benchmark Simulation Model No. 1 (BSM1). The results of the investigations revealed two typical difficulties: problems related to the overall mathematical model structure and problems related to the published set of parameter values. The paper describes the model implementation incompatibilities, the variability in parameter values and the difficulties of reaching similar conditions when simulating a full-scale activated sludge plant. Finally, the simulation results show large differences in oxygen uptake rates, nitritation rates and consequently the quantity of N2O emission (GN2O) using the different models.

Multi-objective controller for enhancing nutrient removal and biogas production in wastewater treatment plants

A multi-objective genetic algorithm (MOGA) aimed at solving a contradictory problem that occurs while combining several single-loop controllers into a multi-loop multi-objective controller (ML-MOC) was proposed to enhance the nutrient removal and biogas production in wastewater treatment plants at a low operational cost. First, an ML-MOC that consists of three proportional–integral (PI) controllers for improving nutrient removal and biogas production was developed. Then, a multi-objective optimization (MOO) was performed using the MOGA in order to determine the optimal set-points of the ML-MOC. The conflicting objective functions are (1) to minimize the effluent loads, (2) to minimize the operational cost and (3) to maximize the biogas production. The proposed ML-MOC was applied to benchmark simulation model no. 2 (BSM2) which is a benchmark system for wastewater treatment plants for evaluating the performances of the control strategies. The results of this study demonstrates that the ML-MOC showed a better nutrient removal performance than the reference controller in BSM2 maintaining economic operational costs, where both nutrient treatment removal rate and biogas production were increased by 3.9% and 3.6%, respectively.

Modification of the activated sludge model for chemical dosage

Full-scale experiments have been carried out to adapt the activated sludge model ASM2d to include the influence of metal dosage (Fe3+ and Al3+) for phosphorus removal. Phosphorus removal rates, nitrification rates, as well as pH and sludge settling performance, were evaluated as functions of the metal dosages. Furthermore, models relating certain parameters to the dosage of chemicals have been derived. Corresponding parameters in the ASM2d and the secondary settler models, included in the Benchmark Simulation Model No 1 (BSM1), have been modified to take the metal influence into consideration. Based on the effluent limits and penalty policy of China, an equivalent evaluation method was derived for the total cost assessment. A large number of 300-day steady-state and 14-day open-loop dynamic simulations were performed to demonstrate the difference in behavior between the original and the modified BSM1. The results show that 1) both in low and high mole concentrations, Fe3+ addition results in a higher phosphorus removal rate than Al3+; 2) the sludge settling velocity will increase due to the metal addition; 3) the respiration rate of the activated sludge is decreased more by the dosage of Al3+ than Fe3+; 4) the inhibition of Al3+ on the nitrification rate is stronger than that of Fe3+; 5) the total operating cost will reach the minimum point for smaller dosages of Fe3+, but always increase with Al3+ addition.

Calibration and validation of a phenomenological influent pollutant disturbance scenario generator using full-scale data

The objective of this paper is to demonstrate the full-scale feasibility of the phenomenological dynamic influent pollutant disturbance scenario generator (DIPDSG) that was originally used to create the influent data of the International Water Association (IWA) Benchmark Simulation Model No. 2 (BSM2). In this study, the influent characteristics of two large Scandinavian treatment facilities are studied for a period of two years. A step-wise procedure based on adjusting the most sensitive parameters at different time scales is followed to calibrate/validate the DIPDSG model blocks for: 1) flow rate; 2) pollutants (carbon, nitrogen); 3) temperature; and, 4) transport. Simulation results show that the model successfully describes daily/weekly and seasonal variations and the effect of rainfall and snow melting on the influent flow rate, pollutant concentrations and temperature profiles. Furthermore, additional phenomena such as size and accumulation/flush of particulates of/in the upstream catchment and sewer system are incorporated in the simulated time series. Finally, this study is complemented with: 1) the generation of additional future scenarios showing the effects of different rainfall patterns (climate change) or influent biodegradability (process uncertainty) on the generated time series; 2) a demonstration of how to reduce the cost/workload of measuring campaigns by filling the gaps due to missing data in the influent profiles; and, 3) a critical discussion of the presented results balancing model structure/calibration procedure complexity and prediction capabilities.

Sensitivity Analysis of Optimal Operation of an Activated Sludge Process Model for Economic Controlled Variable Selection

This paper describes a systematic sensitivity analysis of optimal operation conducted on an activated sludge process model based on the test-bed benchmark simulation model no. 1 (BSM1) and the activated sludge model no. 1 (ASM1). The objective is to search for a control structure that leads to optimal economic operation, while promptly rejecting disturbances at lower layers in the control hierarchy avoiding thus violation of the more important regulation constraints on effluent discharge. We start by optimizing a steady-state nonlinear model of the process. Here, a new steady-state secondary settler mathematical model is developed based on the theory of partial differential equations applied to the conservation law with discontinuous fluxes. The resulting active constraints must be chosen as economic controlled variables. These are the effluent ammonia from the bioreaction section and the final effluent total suspended solids at their respective upper limits, in addition to the internal recycle flow rate ...

Calibration and validation of an activated sludge model for greenhouse gases no. 1 (ASMG1): prediction of temperature-dependent N2O emission dynamics

An activated sludge model for greenhouse gases no. 1 was calibrated with data from a wastewater treatment plant (WWTP) without control systems and validated with data from three similar plants equipped with control systems. Special about the calibration/validation approach adopted in this paper is that the data are obtained from simulations with a mathematical model that is widely accepted to describe effluent quality and operating costs of actual WWTPs, the Benchmark Simulation Model No. 2 (BSM2). The calibration also aimed at fitting the model to typical observed nitrous oxide (N₂O) emission data, i.e., a yearly average of 0.5% of the influent total nitrogen load emitted as N₂O-N. Model validation was performed by challenging the model in configurations with different control strategies. The kinetic term describing the dissolved oxygen effect on the denitrification by ammonia-oxidizing bacteria (AOB) was modified into a Haldane term. Both original and Haldane-modified models passed calibration and validation. Even though their yearly averaged values were similar, the two models presented different dynamic N₂O emissions under cold temperature conditions and control. Therefore, data collected in such situations can potentially permit model discrimination. Observed seasonal trends in N₂O emissions are simulated well with both original and Haldane-modified models. A mechanistic explanation based on the temperature-dependent interaction between heterotrophic and autotrophic N₂O pathways was provided. Finally, while adding the AOB denitrification pathway to a model with only heterotrophic N₂O production showed little impact on effluent quality and operating cost criteria, it clearly affected N2O emission productions.

Adaptive optimal control for a wastewater treatment plant based on a data-driven method

In order to optimize the operating points of the dissolved oxygen concentration and the nitrate level in a wastewater treatment plant (WWTP) benchmark, a data-driven adaptive optimal controller (DDAOC) based on adaptive dynamical programming is proposed. This DDAOC consists of an evaluation module and an optimization module. When a certain group of operating points is given, first the evaluation module estimates the energy consumption and the effluent quality in the future under this policy, and then the optimization module adjusts the operating points according to the evaluation result generated by the evaluation module. The optimal operating points will be found gradually as this process continues repeatedly. During the optimization, only the input-output data measured from the plant are needed, while a mechanistic model is unnecessary. The DDAOC is tested and evaluated on BSM1 (Benchmark Simulation Model No.1), and its performance is compared to the performance of a proportional-integral-derivative (PID) controller with fixed operating points under the full range of operating conditions. The results show that DDAOC can reduce the energy consumption significantly.

Energy consumption model for wastewater treatment process control

Wastewater treatment must satisfy discharge requirements under specified constraints and have minimal operating costs (OC). The operating results of wastewater treatment processes (WWTPs) have significantly focused on both the energy consumption (EC) and effluent quality (EQ). To reflect the relationship between the EC and EQ of WWTPs directly, an extended Elman neural network-based energy consumption model (EENN-ECM) was studied for WWTP control in this paper. The proposed EENN-ECM was capable of predicting EC values in the treatment process. Moreover, the self-adaptive characteristic of the EENN ensured the modeling accuracy. A performance demonstration was carried out through a comparison of the EC between the benchmark simulation model No.1 (BSM1) and the EENN-ECM. The experimental results demonstrate that this EENN-ECM is more effective to model the EC of WWTPs.

Application of multivariate virtual reference feedback tuning for wastewater treatment plant control

The objective of this paper is to apply the Virtual Reference Feedback Tuning (VRFT) to Multiple-Input Multiple-Output (MIMO) control strategies in wastewater treatment plants (WWTPs). Using the Benchmark Simulation Model No. 1 (BSM1) as a case study, the proposed data-driven approach provides reduced-order controllers only using a batch of input–output data points (i.e. manipulated variable—controlled variable), obtained from an open-loop experiment. The methodology also includes a pre-processing step that subtracts the impact of influent variability and sensor noise from the output signals. The possible interactions amongst different control loops are handled using a decoupling approach where each control signal is computed depending on the error signal of all the loops at the same time. In order to test the methodology, several control strategies are evaluated via simulation. The results show that substantial improvements in the plant performance can be obtained when controllers are implemented.

Uncertainty analysis of WWTP control strategies made feasible

The control of wastewater treatment plants can help to achieve good effluent quality, in a complex, highly non-linear environment. A key but time-demanding component of such modelling studies is uncertainty analysis (UA). The general aims of this paper are (a) to evaluate methods for reduction of the time necessary to conduct an UA, and (b) to evaluate the sensitivity of parameters and model subsystems. Two UA studies on the Benchmark Simulation Model no. 2 (BSM2) are used to illustrate how the above mentioned aims can be achieved: (1) robustness of performance evaluations against changing operation and design conditions; and (2) uncertainty of performance evaluations for a given plant layout and operation. The main conclusions are: (1) solver settings have a large impact on simulation speed and require proper attention; (2) to reach convergence in Monte Carlo simulations with Latin Hypercube Sampling, the number of simulations should be at least 50 times the number of sampled parameters, which is more than what is reported in similar studies; and (3) the number of uncertain parameters that needs to be considered to make a proper uncertainty assessment of a model can be reduced significantly by omitting parameters that have little influence.

EVALUATION OF DIFFERENT CONTROL STRATEGIES OF THE WASTE WATER TREATMENT PLANT BASED ON A MODIFIED ACTIVATED SLUDGE MODEL NO. 3

The first part of this work presents the development and implementation in the Benchmark Simulation Model No 1 (BSM1) of a modified Activated Sludge Model No 3 (ASM3). The enhanced ASM3 presented in this study has three modifications compared to the original ASM3. The first modification is the representation of the simultaneous heterotrophic biomass growth on the primary substrate and on the internal storage products. Second, nitrification is modeled as a two-step process with nitrite as an intermediate product, bringing an increased degree of complexity to the mathematical model. Third, the denitrification process is modeled as a three step process with nitrite and nitric oxide as intermediates. The nitric oxide is introduced in the model to account for the inhibition of some enzymes that are responsible for the growth of the heterotrophic bacteria under aerobic conditions. For a better representation of the real plant behavior, the secondary settler is modeled to be reactive. The built reactive settler model is the combination of the settler model described by Takacs in 1991 and the enhanced ASM3. The second part of this research consists of the investigation of five control strategies applied to the waste water treatment plant (WWTP). The control architectures studied in this research are multi-input-multi-output (MIMO) Model Predictive Controllers (MPC). The assessment of these strategies is made from three points of view: control performance, cost evaluation and quality of the effluent. The simulation results show that operational costs can be reduced using automatic control.