Benchmark Simulation Model No. 1 Research Articles

Development of semi-supervised multiple-output soft-sensors with Co-training and tri-training MPLS and MRVM

Soft sensors are the most commonly used tools to estimate the hard-to-measure variables in the chemical processes and other industries, mainly due to unknown mechanism, significant measurement delay and highly unacceptable costs. However, a small number of labeled data and a large number of unlabeled data are not fully investigated and coordination of them to train the models and to improve the prediction accuracy is even rare. Multiple tasks learning or multiple outputs learning adds more complexity to this problem. In this light, this paper proposed semi-supervised multiple-output learning soft sensor models with co-training MPLS (Multiple-output partial least squares), co-training MRVM (Multiple-output relevance vector machines), tri-training MPLS and tri-training MRVM. Co-training MPLS model is developed by extending the traditional co-training PLS model. Co-training MRVM is promoted by replacing MPLS with MRVM. Tri-training MPLS and tri-training MRVM are built by combining tri-training algorithm with MPLS and MRVM. The proposed four models can make full use of appropriate unlabeled data to optimize the regression model, and then to directly strengthen multiple-output variables prediction. These models were firstly demonstrated by a numerical example, then accessed by a wastewater plant (WWTP) simulated with well-established WWTP validation platform, Benchmark Simulation Model No. 1 (BSM1). The results proved that the proposed models were able to significantly improve the prediction performance and efficiency.

Robust Data-Driven Soft Sensors for Online Monitoring of Volatile Fatty Acids in Anaerobic Digestion Processes

The concentration of volatile fatty acids (VFAs) is one of the most important measurements for evaluating the performance of anaerobic digestion (AD) processes. In real-time applications, VFAs can be measured by dedicated sensors, which are still currently expensive and very sensitive to harsh environmental conditions. Moreover, sensors usually have a delay that is undesirable for real-time monitoring. Due to these problems, data-driven soft sensors are very attractive alternatives. This study proposes different data-driven methods for estimating reliable VFA values. We evaluated random forest (RF), artificial neural network (ANN), extreme learning machine (ELM), support vector machine (SVM) and genetic programming (GP) based on synthetic data obtained from the international water association (IWA) Benchmark Simulation Model No. 2 (BSM2). The organic load to the AD in BSM2 was modified to simulate the behavior of an anaerobic co-digestion process. The prediction and generalization performances of the different models were also compared. This comparison showed that the GP soft sensor is more precise than the other soft sensors. In addition, the model robustness was assessed to determine the performance of each model under different process states. It is also shown that, in addition to their robustness, GP soft sensors are easy to implement and provide useful insights into the process by providing explicit equations.

Model Based Control of a Full-Scale Biological Wastewater Treatment Plant

In the present contribution, different combinations of control strategies based on PI, MPC and Fuzzy controllers are implemented with a prime concern of improving the quality of the discharge and reducing the operational expenditure in a full scale wastewater treatment plant. The Benchmark Simulation Model No.2 (BSM2) is used as the working platform. The control objective is to maintain the dissolved oxygen concentration in the fourth tank by manipulating the oxygen transfer coefficients of third, fourth and fifth tanks in the biological wastewater treatment. PI and MPC are used at the lower level whereas MPC and Fuzzy controllers are used at the higher level. The effluent violations are reduced for all the proposed control schemes. It is also observed that PI-MPC results in best Effluent Quality Index (EQI) and PI-Fuzzy results in best Operational Cost Index (OCI).

Effect of Temperature on Effluent Quality in a Biological Wastewater Treatment Process

Abstract In the present study, the temperature effects on the organic carbon and nitrogen removal in an activated sludge process are evaluated. Benchmark Simulation Model No.1 (BSM1) based on activated sludge process is used for all the simulation purposes. A steady state simulation is performed to analyze the effluent concentrations with varying kinetic parameters obtained from different temperature coefficients over a wide range of temperatures from 15 °C to 35 °C. The temperature coefficient ‘a’ is assumed to have different set of values specific to the kinetic parameters, namely, Maximum heterotrophic growth rate μ m H \left( {{\mu _{mH}}} \right) , Maximum autotrophic growth rate μ m A \left( {{\mu _{mA}}} \right) , Heterotrophic decay rate b H \left( {{b_H}} \right) , Autotrophic decay rate b A \left( {{b_A}} \right) . The effluent concentration defined in terms of Chemical Oxygen Demand (COD), Biochemical Oxygen Demand (BOD), Total Suspended Solids (TSS), Total Nitrogen (TN) and Ammonia are observed to be significantly changing with a change in the kinetic parameters which are in turn a strong function of temperature coefficient. Emphasis is laid on the temperature range of 25–30 °C as it is commonly the most operated temperature range in a WWTP in India. It is also noticed that at temperatures <20 °C and >30 °C, the effluent limitations are violated from the standard values.

Cooperative Optimal Controller and Its Application to Activated Sludge Process.

With the increasing complexity and scale of activated sludge process (ASP), it is quite challenging to coordinate the performance indices with different time scales. To address this problem, a cooperative optimal controller (COC) is proposed to improve the operation performance in this paper. First, a cooperative optimal scheme is developed for designing the control system, where the different time-scale performance indices are formulated by two levels. Second, a data-driven surrogate-assisted optimization (DDSAO) algorithm is provided to optimize the cooperative objectives, where a surrogate model is established for evaluating the feasibility of optimal solutions based on the minimum squared error. Third, an adaptive predictive control strategy is investigated to derive the control laws for improving the tracking control performance. Finally, the proposed COC is tested on benchmark simulation model No. 1 (BSM1). The results demonstrate that the proposed COC is able to coordinate the multiple time-scale performance indices and achieve the competitive optimal control performance.

Dynamic MOPSO-Based Optimal Control for Wastewater Treatment Process.

To achieve excellent treatment performance of complex and time-varying characteristics, the operation of wastewater treatment process (WWTP) has been considered as a dynamic multiobjective control problem. In this paper, an optimal controller, based on a dynamic multiobjective particle swarm optimization (DMOPSO) algorithm, is developed to deal with the dynamic multiple conflicting criteria [i.e., effluent quality (EQ), operation cost, and operation stability]. The novelties and advantages of this proposed DMOPSO-based optimal controller (DMOPSO-OC) include the following two aspects. First, an integrated optimization framework, where the multiple objectives not only conflict with each other but also change over time, is able to catch more characteristics of WWTP than the existing works. Second, a DMOPSO algorithm, with an adaptive global best selection mechanism, is designed to solve the multiobjective optimization problem (MOP) for the proposed optimal controller, thus leading to a significant improvement of optimal synthesis for performance. Finally, the proposed DMOPSO-OC is tested in the benchmark simulation model No. 1 (BSM1) and implemented in a real WWTP to evaluate its effectiveness. The experimental results demonstrate that this proposed DMOPSO-OC can achieve a significant improvement in optimal control performance and obey the requirement of multiple conflicting criteria.

Meta-modeling based efficient global sensitivity analysis for wastewater treatment plants – An application to the BSM2 model

Global sensitivity analysis (GSA) is a powerful tool for quantifying the effects of model parameters on the performance outputs of engineering systems, such as wastewater treatment plants (WWTP). Due to the ever-growing sophistication of such systems and their models, significantly longer processing times are required to perform a system-wide simulation, which makes the use of traditional Monte Carlo (MC) based approaches for calculation of GSA measures, such as Sobol indices, impractical. In this work, we present a systematic framework to construct and validate highly accurate meta-models to perform an efficient GSA of complex WWTP models such as the Benchmark Simulation Model No. 2 (BSM2). The robustness and the efficacy of three meta-modeling approaches, namely polynomial chaos expansion (PCE), Gaussian process regression (GPR), and artificial neural networks (ANN), are tested on four engineering scenarios. The results reveal significant computational gains of the proposed framework over the MC-based approach without compromising accuracy.

Plant-wide model-based analysis of iron dosage strategies for chemical phosphorus removal in wastewater treatment systems

Stringent phosphorus discharge standards (i.e. 0.15–0.3 g P.m−3) in the Baltic area will compel wastewater treatment practice to augment enhanced biological phosphorus removal (EBPR) with chemical precipitation using metal salts. This study examines control of iron chemical dosing for phosphorus removal under dynamic loading conditions to optimize operational aspects of a membrane biological reactor (MBR) pilot plant. An upgraded version of the Benchmark Simulation Model No. 2 (BSM2) with an improved physico-chemical framework (PCF) is used to develop a plant-wide model for the pilot plant. The PCF consists of an equilibrium approach describing ion speciation and pairing, kinetic minerals precipitation (such as hydrous ferric oxides (HFO) and FePO4) as well as adsorption and co-precipitation. Model performance is assessed against data sets from the pilot plant, evaluating the capability to describe water and sludge lines across the treatment process under steady-state operation. Simulated phosphorus differed as little as 5–10% (relative) from measured phosphorus, indicating that the model was representative of reality. The study also shows that environmental factors such as pH, as well operating conditions such as Fe/P molar ratios (1, 1.5 and 2), influence the concentration of dissolved phosphate in the effluent. The time constant of simultaneous precipitation in the calibrated model, due to a step change decrease/increase in FeSO4 dosage, was found to be roughly 5 days, indicating a slow dynamic response due to a multi-step process involving dissolution, oxidation, precipitation, aging, adsorption and co-precipitation. The persistence effect of accumulated iron-precipitates (HFO particulates) in the activated sludge seemed important for phosphorus removal, and therefore solids retention time plays a crucial role according to the model. The aerobic tank was deemed to be the most suitable dosing location for FeSO4 addition, due to high dissolved oxygen levels and good mixing conditions. Finally, dynamic model-based analyses show the benefits of using automatic control when dosing chemicals.

WWTP MODEL CALIBRATION BASED ON DIFFERENT OPTIMIZATION APPROACHES

There is a significant interest in modelling the wastewater treatment plants (WWTP) because with a calibrated model a large variety of operational strategies can be evaluated, the control system design may be improved and enhanced constructive alternatives may be proposed. The Activated Sludge Model No. 1 (ASM1) has the potential of describing the biochemical processes taking place in WWTPs, in association with the Benchmark Simulation Model No. 1 (BSM1) which adds the description of the involved physical processes. An important challenge for reliable simulators development is the dynamically changing composition of the influent wastewater, also varying from one WWTP to another. Due to this reason, it is necessary to fit the model for each WWTP. This paper proposes a new methodology for the dynamic simulator calibration of a municipal WWTP. Different optimization methods are used for the simulator calibration and results of the different approaches are comparatively presented. Based on WWTP measurements a set of influent variables, bioreactor parameters and settler parameters are calibrated. Optimization was carried out based on classical, genetic, hybrid and Pareto multiobjective algorithms and their performance is discussed, revealing their strengths and limitations.

Multiobjective optimal control for wastewater treatment process using adaptive MOEA/D

Through the analysis of the biological wastewater treatment process (WWTP), a multiobjective optimal control strategy is developed with the usage of energy consumption (EC) and effluent quality (EQ) as objectives to be optimized. To effectively handle the multiobjective optimization problem (MOP) with complex Pareto-optimal front (POF), an adaptive multiobjective evolutionary algorithm based on decomposition (AMOEA/D) is proposed in this paper. Since the efficiency of the multiple reference points and two-phase optimization strategies in solving MOPs with complex POFs has been proved. In the proposed AMOEA/D, an auto-switching strategy based on the aggregation function enhancement is designed to automatically make the algorithm switch from the first phase to the second phase. Besides, an adaptive differential evolution strategy is introduced into AMOEA/D to balance exploration and exploitation during the evolutionary process. Finally, the dynamic optimization, intelligent decision and bottom tracking control of the set-points of the dissolved oxygen and nitrate nitrogen in the WWTP are achieved via the combination of AMOEA/D with the self-organizing fuzzy neural network approximator and the self-organizing fuzzy neural network controller. The international benchmark simulation model No. 1 (BSM1) is utilized for experimental verification. Simulation results demonstrate that the proposed AMOEA/D can effectively reduce the EC of the WWTP under the premise of ensuring effluent parameters to meet the effluent discharge standards.

Multi‐Variable Direct Self‐Organizing Fuzzy Neural Network Control for Wastewater Treatment Process

AbstractA multi‐variable direct self‐organizing fuzzy neural network control (M‐DSNNC) method is proposed for the multi‐variable control of the wastewater treatment process (WWTP). In this paper, the proposed control system is an essential multi‐variable control method for the WWTP. No exact plant model is required, which avoids the difficulty of establishing the mathematics model of WWTP. The M‐DSNNC system is comprised of a fuzzy neural network controller and a compensation controller. The fuzzy neural network is used for approximating the ideal control law under a general nonlinear system. Moreover, the neural network is designed in a self‐organizing mode to adapt the uncertainty environment. Simulation results, based on the international benchmark simulation model No.1 (BSM1), demonstrate that the control accuracy is improved under the proposed M‐DSNNC method, and the controller has a much stronger decoupling ability.

Multiobjective design of fuzzy neural network controller for wastewater treatment process

In this paper, an improved multiobjective optimal control (MOOC) strategy is developed to improve the operational efficiency, satisfy the effluent quality (EQ) and reduce the energy consumption (EC) in wastewater treatment process (WWTP). First, the adaptive kernel function models of the process, which can describe the complex dynamics of EQ and EC, are developed for the proposed MOOC strategy. Meanwhile, a multiobjective optimization problem is constituted to account for WWTP. Second, an improved multiobjective particle swarm optimization (MOPSO) algorithm, using a self-adaptive flight parameters mechanism and a multiobjective gradient (MOG) method, is designed to minimize the established objectives. And then the optimal set-points of dissolved oxygen (SO) and nitrate (SNO) are obtained in the treatment process. Third, an adaptive fuzzy neural network controller (FNNC) is applied for realizing the tracking control of the obtained set-points in the proposed MOOC strategy. Finally, Benchmark Simulation Model No.1 (BSM1) is introduced to evaluate the effectiveness of the proposed MOOC strategy. Experimental results show the efficacy of the proposed method.

State estimation of wastewater treatment plants based on reduced-order model

In this paper, we consider state estimation of wastewater treatment plants based on model approximation. A wastewater treatment plant described by the Benchmark Simulation Model No. 1 (BSM1) is used. We use the proper orthogonal decomposition approach with re-identification of output equations to obtain a reduced-order model for the original system and then use the reduced-order model in state estimation. An approach on how to determine an appropriate minimum measurement set is also proposed. A continuous-discrete extended Kalman filtering algorithm is used to design an estimator based on the reduced-order model with re-identified output equations. The estimator gives good state estimates for the actual process.

Modelling an industrial anaerobic granular reactor using a multi-scale approach

The objective of this paper is to show the results of an industrial project dealing with modelling of anaerobic digesters. A multi-scale mathematical approach is developed to describe reactor hydrodynamics, granule growth/distribution and microbial competition/inhibition for substrate/space within the biofilm. The main biochemical and physico-chemical processes in the model are based on the Anaerobic Digestion Model No 1 (ADM1) extended with the fate of phosphorus (P), sulfur (S) and ethanol (Et−OH). Wastewater dynamic conditions are reproduced and data frequency increased using the Benchmark Simulation Model No 2 (BSM2) influent generator. All models are tested using two plant data sets corresponding to different operational periods (#D1, #D2). Simulation results reveal that the proposed approach can satisfactorily describe the transformation of organics, nutrients and minerals, the production of methane, carbon dioxide and sulfide and the potential formation of precipitates within the bulk (average deviation between computer simulations and measurements for both #D1, #D2 is around 10%). Model predictions suggest a stratified structure within the granule which is the result of: 1) applied loading rates, 2) mass transfer limitations and 3) specific (bacterial) affinity for substrate. Hence, inerts (XI) and methanogens (Xac) are situated in the inner zone, and this fraction lowers as the radius increases favouring the presence of acidogens (Xsu,Xaa, Xfa) and acetogens (Xc4,Xpro). Additional simulations show the effects on the overall process performance when operational (pH) and loading (S:COD) conditions are modified. Lastly, the effect of intra-granular precipitation on the overall organic/inorganic distribution is assessed at: 1) different times; and, 2) reactor heights. Finally, the possibilities and opportunities offered by the proposed approach for conducting engineering optimization projects are discussed.

Online adaptive dynamic programming based on echo state networks for dissolved oxygen control

To solve the control problem of dissolved oxygen concentration in wastewater treatment process (WWTP), an online adaptive dynamic programming (ADP) scheme based on echo state networks (ESN) is proposed. The ESN-based ADP controller contains three modules composed of echo state networks, and the three modules approximate respectively the critic index, the system model and the optimal control policy. Further, an online learning algorithm on the basis of recursive least squares (RLS) is designed to train the ESN-based ADP controller, and also the convergence of the online learning algorithm is investigated. The design and operation of the ESN-based ADP controller depend mainly on the online data observed from controlled system, and minor prior knowledge is required. The proposed ESN-based ADP controller’s performance is tested and evaluated on Benchmark Simulation Model no. 1 (BSM1)—a general WWTP benchmark. Experimental results demonstrate that the proposed controller can achieve high performance on dissolved oxygen level control in WWTP.

Plant-wide modelling of phosphorus transformations in wastewater treatment systems: Impacts of control and operational strategies

The objective of this paper is to report the effects that control/operational strategies may have on plant-wide phosphorus (P) transformations in wastewater treatment plants (WWTP). The development of a new set of biological (activated sludge, anaerobic digestion), physico-chemical (aqueous phase, precipitation, mass transfer) process models and model interfaces (between water and sludge line) were required to describe the required tri-phasic (gas, liquid, solid) compound transformations and the close interlinks between the P and the sulfur (S) and iron (Fe) cycles. A modified version of the Benchmark Simulation Model No. 2 (BSM2) (open loop) is used as test platform upon which three different operational alternatives (A1, A2, A3) are evaluated. Rigorous sensor and actuator models are also included in order to reproduce realistic control actions. Model-based analysis shows that the combination of an ammonium (SNHX) and total suspended solids (XTSS) control strategy (A1) better adapts the system to influent dynamics, improves phosphate (SPO4) accumulation by phosphorus accumulating organisms (XPAO) (41%), increases nitrification/denitrification efficiency (18%) and reduces aeration energy (Eaeration) (21%). The addition of iron (XFeCl3) for chemical P removal (A2) promotes the formation of ferric oxides (XHFO−H, XHFO−L), phosphate adsorption (XHFO−H,P, XHFO−L,P), co-precipitation (XHFO−H,P,old, XHFO−L,P,old) and consequently reduces the P levels in the effluent (from 2.8 to 0.9 g P.m−3). This also has an impact on the sludge line, with hydrogen sulfide production (GH2S) reduced (36%) due to iron sulfide (XFeS) precipitation. As a consequence, there is also a slightly higher energy production (Eproduction) from biogas. Lastly, the inclusion of a stripping and crystallization unit (A3) for P recovery reduces the quantity of P in the anaerobic digester supernatant returning to the water line and allows potential struvite (XMgNH4PO4) recovery ranging from 69 to 227 kg.day−1 depending on: (1) airflow (Qstripping); and, (2) magnesium (QMg(OH)2) addition. All the proposed alternatives are evaluated from an environmental and economical point of view using appropriate performance indices. Finally, some deficiencies and opportunities of the proposed approach when performing (plant-wide) wastewater treatment modelling/engineering projects are discussed.

Generation of synthetic influent data to perform (micro)pollutant wastewater treatment modelling studies

The use of process models to simulate the fate of micropollutants in wastewater treatment plants is constantly growing. However, due to the high workload and cost of measuring campaigns, many simulation studies lack sufficiently long time series representing realistic wastewater influent dynamics. In this paper, the feasibility of the Benchmark Simulation Model No. 2 (BSM2) influent generator is tested to create realistic dynamic influent (micro)pollutant disturbance scenarios. The presented set of models is adjusted to describe the occurrence of three pharmaceutical compounds and one of each of its metabolites with samples taken every 2–4h: the anti-inflammatory drug ibuprofen (IBU), the antibiotic sulfamethoxazole (SMX) and the psychoactive carbamazepine (CMZ). Information about type of excretion and total consumption rates forms the basis for creating the data-defined profiles used to generate the dynamic time series. In addition, the traditional influent characteristics such as flow rate, ammonium, particulate chemical oxygen demand and temperature are also modelled using the same framework with high frequency data. The calibration is performed semi-automatically with two different methods depending on data availability. The ‘traditional’ variables are calibrated with the Bootstrap method while the pharmaceutical loads are estimated with a least squares approach. The simulation results demonstrate that the BSM2 influent generator can describe the dynamics of both traditional variables and pharmaceuticals. Lastly, the study is complemented with: 1) the generation of longer time series for IBU following the same catchment principles; 2) the study of the impact of in-sewer SMX biotransformation when estimating the average daily load; and, 3) a critical discussion of the results, and the future opportunities of the presented approach balancing model structure/calibration procedure complexity versus predictive capabilities.

Evaluation of Control Strategies in Activated Sludge Process for Biological Wastewater Treatment

Abstract In this paper, benchmark simulation model no.1 (BSM1) of an activated sludge process is used to evaluate various control strategies. Control configurations such as feedback control and feed-forward plus feedback (FF-FB) control are applied and compared with respect to effluent discharge requirements under specified constraints and operating costs. Feed-forward control is incorporated in the PI control configuration for preventing the influent loading disturbance affecting the process. No case studies of BSM1 model have been reported in the literature for the Indian wastewater. In this work, the dynamic simulation of an activated sludge process is performed using the data collected from the sewage treatment plant, located in India. The influent load data are collected during the dry weather period. The influent fractionation is carried out using the activated sludge model no.1 (ASM1). The results of the dynamic simulation indicate that FF-FB control of the activated sludge process is more effective than feedback control in meeting the constraints, especially effluent ammonia concentration which is considered as very important. From the comparison of performance evaluation criteria, it is observed that FF-FB control has achieved almost the same operating costs as with feedback control.

Self-organizing fuzzy control for dissolved oxygen concentration using fuzzy neural network1

In order to improve the accuracy and adaptive ability of dissolved oxygen concentration control in the wastewater treatment process (WWTP), a self-organizing fuzzy control (SOFC) method is developed in this paper. The main feature of this control system is that the fuzzy controller can extract fuzzy rules automatically using a self-organizing fuzzy neural network (FNN), which can adjust the network structure during the process based on the growing-pruning-combining algorithm. Furthermore, to ensure the convergence of the system, a compensation controller is designed to dispel the FNN approximation error, and the parameter compensation is also considered while adjusting the network structure. Finally, simulation results, based on the international benchmark simulation model No.1 (BSM1), demonstrate that the proposed method can achieve better control performance and superior adaptive ability compared with PID, model predictive control and conventional fuzzy logic controller.

Joint Environmental and Economical Analysis of Wastewater Treatment Plants Control Strategies: A Benchmark Scenario Analysis

In this paper, a joint environmental and economic analysis of different Wastewater Treatment Plant (WWTP) control strategies is carried out. The assessment is based on the application of the Life Cycle Assessment (LCA) as a method to evaluate the environmental impact and the Benchmark Simulation Model No. 1 (BSM1). The BSM1 is taken as the benchmark scenario used to implement the control strategies. The Effluent Quality Index (EQI) and the Overall Cost Index (OCI) are two indicators provided by BSM1 and used to evaluate the plant’s performance from the effluent quality and the economic points of view, respectively. This work conducts a combined analysis and assessment of ten different control strategies defined to operate a wastewater treatment plant. This analysis includes the usual economic and performance indexes provided by BSM1 joined with the LCA analysis that determines the environmental impact linked to each one of the considered control strategies. It is shown how to get an overall evaluation of the environmental effects by using a normalized graphical representation that can be easily used to compare control strategies from the environmental impact point of view. The use of only the BSM1 indexes provides an assessment that leads to a clustering of control strategies according to the cost/quality tradeoff they show. Therefore, regarding the cost/quality tradeoff, all strategies in the same group are almost equal and do not provide an indication on how to proceed in order to select the appropriate one. It is therefore shown how the fact of adding a new, complementary, evaluation (LCA based) allows either to reinforce a decision that could be taken solely on the basis of the EQI/OCI tradeoff or to select one control strategy among the others.