Journal of Cleaner Production 367 (2022) 132873 Contents lists available at ScienceDirect Journal of Cleaner Production journal homepage: www.elsevier.com/locate/jclepro Smart-Plant Decision Support System (SP-DSS): Defining a multi-criteria decision-making framework for the selection of WWTP configurations with resource recovery Živko Južnič-Zonta , Albert Guisasola , Juan Antonio Baeza * GENOCOV. Departament d’Enginyeria Química, Biològica i Ambiental, Escola d’Enginyeria. Universitat Autònoma de Barcelona, 08193, Cerdanyola del Vallès, Barcelona, Spain A R T I C L E I N F O A B S T R A C T Handling Editor: Mingzhou Jin The development of the Smart-Plant Decision Support System (SP-DSS) is presented, which helps to find the optimal water resource recovery facility (WRRF) configuration for a specific case. A general plant superstructure Keywords: is defined to allow the evaluation of different combinations among the available process unit options, considering DSS technical, economic, and environmental impact criteria. The complete evaluation is based on a five-step EBPR framework: (i) design problem set-up, (ii) wastewater inflow generation, (iii) superstructure generation and Resource recovery plant-wide model simulation under dynamic conditions, (iv) estimation of multi-criteria objective values Superstructure Water resource recovery facility (WRRF) (greenhouse gas emissions, accumulation of effluent violations, effluent quality index, net present value, system readiness level and plant land area), and (v) design configuration multi-criteria sorting with the technique for order of preference by similarity to ideal solution (TOPSIS). SP-DSS application is shown comparing the main­ stream shortcut enhanced phosphorus and polyhydroxyalkanoate (PHA) recovery (SCEPPHAR) technology with a conventional anaerobic/anoxic/aerobic (A2O) configuration for a medium size WWTP (Manresa, Spain). The comparison shows both configurations allow meeting discharge limits, and among other criteria, the SCEPPHAR novel configuration, with respect to A2O, shows higher pollutant contents in the effluent (3.30 vs 2.34 gPoll/m3), higher capital expenditures (CAPEX 0.191 vs 0.130 €/m3), lower operational expenditures (OPEX 0.397 vs 0.515 €/m3), lower required tariff (0.654 vs 0.674 €/m3) and higher greenhouse gas emissions (GHG 0.19 vs 0.14 kgCO2/m3). Overall, TOPSIS sorting indicates A2O is still the best configuration for this specific case. short chain fatty acids (Crutchik et al., 2018) and other valuable com­ pounds (Da Ros et al., 2020; Xue et al., 2019). In addition, there are 1. Introduction political initiatives to promote the circular economy with the develop­ ment of resource recovery technologies for the wastewater treatment The wastewater treatment sector has recently undergone an sector, while still meeting legal discharge limits and reducing other extraordinary increase in the variety of available technologies. The side-effects such as greenhouse gas (GHG) emissions (Vasilaki et al., choice of technologies and their benchmarking for the design of new 2020). However, one of the reasons hindering the promotion and plants or for the retrofitting of existing plants to meet new legal re­ implementation of new technologies for resource recovery is the quirements has become increasingly challenging due to the number of complexity of evaluating the different effects of implementing any new different designs possible for a given scenario. Moreover, the complexity process addition to a WWTP. A benchmarking analysis would provide a increases if one includes the new environmental paradigm of conven­ clear comparison on a case-by-case basis in many different criteria such tional wastewater treatment plants (WWTP) becoming water resource as economical costs of construction and operation, contaminants recovery facilities (WRRF). The resources to be recovered are energy removal performance, level of discharge limits accomplishment, and (Zarei, 2020), water (Jiménez-Benítez et al., 2020), nutrients (Guida emission of greenhouse gases. However, this type of multivariable et al., 2020; Larriba et al., 2020), bioplastics (Conca et al., 2020; Lorini analysis is usually complex and time demanding when the number of et al., 2021), biomass for methane production (Chan et al., 2020; alternatives to evaluate is high. This is a critical hurdle to the use of new Jimenez et al., 2015), cellulose (Palmieri et al., 2019; Zhou et al., 2019), * Corresponding author. E-mail addresses: jzivko@gmail.com (Ž. Južnič-Zonta), Albert.Guisasola@uab.cat (A. Guisasola), JuanAntonio.Baeza@uab.cat (J.A. Baeza). https://doi.org/10.1016/j.jclepro.2022.132873 Received 12 December 2021; Received in revised form 8 May 2022; Accepted 22 June 2022 Available online 28 June 2022 0959-6526/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Ž. Južnič-Zonta et al. Journal of Cleaner Production 367 (2022) 132873 Abbreviations OMEdit OpenModelica Connection Editor PAO Polyphosphate-Accumulating Organisms A2O Anaerobic, Anoxic, Oxic EBPR process PHA Poly-Hydroxy-Alkanoate AD Anaerobic Digestion PI Proportional Integral controller ADF Anaerobic Digestion with Fermentation OPEX Operational Expenditures ADM Anaerobic Digestion Model PLC Programmable Logic Controller AEV Accumulation of Effluent Violations PS Primary Sludge clarifier AOB Ammonia Oxidizing Bacteria PWM Plant-Wide Model ASM Activated Sludge Model RL Reporting Level BSM Benchmark Simulation Model SBR Sequential Batch Reactor CAPEX Capital Expenditures SCENA Short-Cut Enhanced Nutrient Abatement COD Chemical Oxygen Demand SCEPPHAR Short-Cut Enhanced Phosphorus and PHA Recovery DAE Differential-Algebraic Equations SOR Surface Overflow Rate DF Drum filter unit to remove all the effluent TSS SP-DSS SMART-Plant Decision Support System DSS Decision Support System SRL System Readiness Levels EBPR Enhanced Biological Phosphorus Removal SRT Sludge/Solids Retention Time EFN2O Emission Factor for N2O ST Smart Technologies of the SMART-Plant project EQI Effluent Quality Index ST1 Smartech 1. Dynamic fine-screen and post-processing of FOB Free On Board equipment cost cellulosic sludge GHG GreenHouse Gas ST2a Smartech 2a. Polyurethane-based anaerobic digestion bio- HET-SBR SBR for heterotrophic processes in SCEPPHAR filter configuration ST2b Smartech 2b. SCEPPHAR configuration HRAS High Rate Activated Sludge ST3 Smartech 3. Tertiary hybrid ion exchange for N and P HRT Hydraulic Retention Time recovery IQI Influent Quality Index ST4a Smartech 4a. SCENA and ordinary digestion INT Interchange vessel in SCEPPHAR configuration ST4b Smartech 4b. SCENA and CAMBI-enhanced digestion KPI Key Performance Indicator ST5 Smartech 5. SCEPPHAR side-stream process LCA Life Cycle Assesment TOPSIS Technique Of Preference ordering by Similarity to the Ideal LCCA Life Cycle Costing Analysis Solution MC Monte-Carlo method TRL Technology Readiness Levels MINLP Mixed Integer (Non)linear Programming TSS Total Suspended Solids MLSS Mixed Liquor Suspended Solids VER Volume Exchange Ratio NPV Net Present Value WISE Water Information System for Europe NRM Nutrient Recovery Model WRRF Water Resource Recovery Facility OMC OpenModelica Compiler WWTP Wastewater Treatment Plant resource recovery technologies, coupled with the understandable risk dynamic models (i.e. ADM1, ASM2d, chemical precipitation with pH aversion of WWTP managers to switch to new technologies. variation, etc.) that often require the use of control strategies. A static Benchmarking analysis of multiple alternative configurations can be model may be difficult to find for this case, or its validity would be poor facilitated by creating a decision support system (DSS) where the com­ for a wide range of scenarios. Influent wastewater has different dy­ parison of the different criteria is made straightforward. DSSs have been namics resulting in changes of flowrate and contaminants composition, proposed to assist a decision-maker in improving an existing plant such as diurnal and weekly variations, periodical or punctual industrial operation (de Faria et al., 2016; Torregrossa et al., 2017) or help in discharges, variable weather with rain periods of different intensity, and choosing an optimal WWTP design. For example, the knowledge-based seasonal variation of populations. All these variations result in variable DSS methodologies described by Comas et al. (2004) and Garrido-Ba­ effluent quality. However, WWTP designs are usually based on steady- serba et al. (2012) aim to reduce the number of potential WWTP con­ state operation with an additional security that often results in over­ figurations to a manageable subset. Once the reduced size subset has sizing, and the possibility to improve plant adaptability using a control been defined, the detailed design and selection of the WWTP configu­ system is not considered (Mussati et al., 2002). Moreover, if severe re­ ration, which is computationally very expensive, can be carried out. strictions on legal effluent limits are imposed not based on average Many of these DSS proposals are based on the development of a plant discharge limits, the scenario becomes more complex. superstructure defined as a general structure composed of several The shortcomings of steady-state WWTP designs for resource re­ sub-units where different technologies compatible with each sub-unit covery problems demand a more detailed analysis based on dynamic can be used. Bozkurt et al. (2016, 2015) proposed a WWTP super­ plant-wide models (PWM). PWMs including both water and sludge lines structure composed of static process unit models, which find the optimal (e.g. Grau et al., 2007; Nopens et al., 2010; Seco et al., 2020) have been plant design by using mixed integer (non)linear programming (MINLP). used for specific comparisons such as the operation of a membrane This methodology has the benefit of being based on robust MINLP bioreactor vs. an activated sludge (Mannina et al., 2020), assessment solvers and having a computing complexity that is relatively low. An and optimization of WWTPs (Dragan et al., 2017; Hvala et al., 2018), integration and demonstration of the knowledge-based DSS and the evaluation of energy balance and GHG emissions (Zaborowska et al., static-model based design MINLP-optimization is presented by Castillo 2021) and multi-objective performance assessment of WWTP including et al. (2017, 2016a, 2016b). However, one of the main limitations of life cycle assessment (LCA) (Arnell et al., 2017). However, optimization these methodologies is that only static process models for steady-state studies of superstructure design with ASM/ADM-type dynamic models designs can be used. are scarce. Guerrero et al. (2013) reported the benchmarking of five Resource recovery processes are generally described by complex different configurations for COD/N/P removal based on the ASM2d 2 Ž. Južnič-Zonta et al. Journal of Cleaner Production 367 (2022) 132873 model and typical design parameters. Rigopoulos and Linke (2002) built Dymola®, Wolfram SystemModeler® and MapleSim®. In this work, the an ASM1-based superstructure to find an optimal WWTP design for a open-source alternative OpenModelica (OSMC, 2022) is used, since it certain constant input condition. Fernández-Arévalo et al. (2017) eval­ provides a very useful tool for schematically assembling component uated three plant layouts for a theoretical scenario with a plant-wide models called OpenModelica Connection Editor (OMEdit). The indi­ modelling methodology and Machado et al. (2020) systematically vidual process units are modelled with Modelica and OMEdit is used to compared four EBPR configurations under dynamic conditions to assemble them into more complex WWTP process sub-systems called retrofit a specific WRRF. However, most of these works only focused on stages. The combination of process stages builds up a full WWTP model the water line, without considering the effect of return streams that can superstructure (see Fig. 1) that is compiled and simulated within impact on the overall plant performance. Regarding the important effect OpenModelica. The process of building plant configurations is auto­ of these side-streams, the analysis developed with a benchmark simu­ matically repeated by changing the specific models of each stage in the lation model (BSM) by Solís et al. (2022) showed that reject water pre-defined general WWTP superstructure. Fig. S1 in supplementary generated during dewatering of anaerobic sludge in an anaerobic/a­ information (SI) shows an image of the OMEdit editor with the imple­ noxic/oxic (A2O) WWTP configuration can contain an additional load mented superstructure and some additional details. The Modelica with respect to the influent around 20% for N and 110% for P. Then, it is StateGraph standard library is used to model discrete events and reac­ clear that models able to consider these side-stream effects, such as the tive systems based on the Grafchart method (Johnsson and Årzén, PWM BSM2-PSFe (Solon et al., 2017), the nutrient recovery model 1999), as needed for some phase-based technologies requiring control (NRM) library (Vaneeckhaute et al., 2018) and the BSM2-PSFe-GHG based on programmable logic controllers (PLC). Finally, results are (Solís et al., 2022) are desirable to build superstructures. post-processed in Python to produce an environmental and economic One problem with the above approaches is that they rely on a few assessment of the plant design. predefined initial configurations. To the authors’ knowledge, no attempt has been made to use these PWM in a DSS designed to build super­ 2.2. SP-DSS framework structures automatically. Note that for most project feasibility studies, the most expensive tasks in terms of time are relative to scenario defi­ The following steps describe the SP-DSS framework (Fig. 1) devel­ nition, model assembling and results reporting, while computation time oped for a plant superstructure with resource recovery units. is marginal and could be easily improved by high performance computing scaling up. Step1: Design problem set-up In this work, the Smart-Plant Decision Support System (SP-DSS) modelling framework is presented for the systematic building of dy­ Scope (new design or plant retrofit) and general parameters are namic PWM configurations with an integrated benchmark analysis, to considered (wastewater characterization, legal discharge limits, plant help water utilities understand the effect of integrating new treatment location …). process units. In this way, informed decisions can be made on a case-by- case basis, reducing decision-making risk. The numerous implications of Step2: Wastewater inflow generation introducing resource recovery units in the overall plant operation are studied using different criteria. The developed SP-DSS addresses a This step can be skipped if specific time-series for the plant are problem that can be formulated as follows: given a set of resource re­ available or if an average constant influent is considered. Otherwise, a covery and wastewater treatment process units, quickly determine the dynamic wastewater inflow time-series with a 1-h time step is created. best plant configuration to treat a given wastewater that considers the Python scraps information relative to the weather history from web- following criteria in a balanced manner: effluent quality index (EQI), databases of the nearest weather stations. Based on the data from the accumulation of effluent quality violations (AEV), GHG emissions, net user and the web, the wet/dry weather and sewer models are executed to present value (NPV), plant footprint and system readiness level (SRL). generate the time-series. The weather model is implemented in Python, To obtain a reliable response, the most important processes and dy­ while the sewer is described as a simple tank-in-series model in namics must be considered, but if a quick response is also required, the Modelica. automation of the methodology becomes necessary. The article is structured as follows: Section 2 presents the modeling framework and a Step3: Superstructure generation and simulation case study of WWTP design for resource recovery, Section 3 presents the results and the discussion of the case study. Finally concluding remarks The WWTP superstructure consists of six interconnected stages that with future development opportunities are presented in Section 4. form an overall plant structure into which the available processes and resource recovery units can be incorporated. Each stage is a section of 2. Methods the WWTP with a common operational objective (pretreatment, acti­ vated sludge, effluent polishing, anaerobic sludge digestion and reject This section shows the different elements that constitute the SP-DSS, water treatment). A complete list of technologies available for each the organization of the designed superstructure, the calculation of the stage is provided in Table 1. Any stage can be left empty, resulting in no different criteria, the dynamic models used and finally the details of the change in that stream. They comprise classical technologies and novel reported case study. resource recovery technologies developed inside the SMART-Plant project (Smart Techs, ST). 2.1. SP-DSS architecture Stages can be programmatically substituted using Modelica redeclare function to automatically generate all possible plant configurations. The The software architecture of SP-DSS is based on the programming Modelica code is translated into C code using the OpenModelica languages Modelica and Python. Modelica is an object-oriented and compiler (OMC), which compiles each specific configuration into a equation-based language supporting the simulation of hybrid separate executable file. OMC symbolically manipulates the equations to differential-algebraic equations (DAE). Component model equations are reduce the non-linear system, resulting in code that allows for more encapsulated in a reusable format that allows an instance of the efficient execution. The procedure of building separate executable files component to be used instead of the original equations. A system model is very different from the typical MINLP superstructure based on a single is assembled by dragging, dropping, and connecting different compo­ complete model in which plant design configurations are achieved by nents and subsystem models. There are many commercial modeling and discrete integer variables that activate/deactivate connections/flows simulation environments based on the open Modelica language such as between stages or process units. The typical MINLP formulation involves 3 Ž. Južnič-Zonta et al. Journal of Cleaner Production 367 (2022) 132873 Fig. 1. SP-DSS framework developed for a plant superstructure with resource recovery units. Table 1 List of process models available in SP-DSS and table indicating which processes can be applied at each stage of the superstructure. Name Description of the modelled process ST1 Dynamic fine-screen and post-processing of cellulosic sludge ST2a Polyurethane-based anaerobic digestion bio-filter ST2b Mainstream Short-Cut Enhanced Phosphorus and PHA Recovery (SCEPPHAR) ST3 Tertiary hybrid ion exchange for N and P recovery ST4a SCENA and ordinary digestion ST4b SCENA and CAMBI-enhanced digestion ST5 SCEPPHAR side-stream process A2O Anaerobic, anoxic, oxic EBPR process AD Anaerobic digestion ADF Anaerobic digestion with fermentation DF Drum filter unit to remove all the effluent TSS HRAS High rate activated sludge PS Primary sludge clarifier Empty PS A2O HRAS AD ADF DF ST1 ST2a ST2b ST3 ST4a ST4b ST5 Stage1 ✓ ✓ ✓ Stage2 ✓ ✓ Stage3 ✓ ✓ ✓ ✓ Stage4 ✓ ✓ ✓ Stage5 ✓ ✓ ✓ Stage6 ✓ ✓ ✓ ✓ compiling a very large superstructure dynamic model once, where many state variables are not used, which unnecessarily slows down simulation Where TCOD, TKN, NOx, TSS, TNP and PO3− 4 are the effluent mass flows time. On the contrary, the approach used results in shorter simulation (kg/day) of total COD, total Kjeldahl nitrogen, nitrogen oxides, total time because no dummy states are created (i.e., the numerical Jacobian suspended solids, total non-reactive phosphorus and phosphate phos­ matrix of the DAE solver is smaller, there is less data storage overhead phorus. The weights applied to each component allow calculating an …) and the equation reduction routines are more efficient. In addition, overall pollution index. Similarly, the Influent Quality Index (IQI) is having separate executables for each configuration facilitates parallel defined to assess the pollution at the inflow. computation. The accumulated effluent violation for a given pollutant (AEVi) is For the simulation of each configuration, design parameter values calculated with Equation (2) based on the effluent concentration (Ci) (volumes, flow rates …) based on reference books (Tchobanoglous et al., and the relative legislation limit (Clim i ). 2014) are used. During the optimization phase (Step 6), the design pa­ ∫T [ ] rameters can be refined for the most promising cases. 1 ( {( ) } ) kg AEVi = if Ci > Cilim Ci − Cilim ⋅Q else {0} dt (2) T d Step4: Objective values estimation AEVIi terms are aggregated within a weighted sum (wi) to get an The calculation of the criteria related to effluent quality, economics, overall AEV index (Equation (3)). and environmental impact is performed in the Python environment by ∑ [ kg Pollution ] running the executable files and processing the results. The EQI function AEV = AEVi ⋅wi (3) d used was similar to Flores-Alsina et al. (2011) (Equation (1)). i ∫T [ ] 1 ( ) kg Pollution EQI = 1 TCOD + 30 TKN + 10 NOx + 10 TSS + 100 TNP + 100 PO3−4 dt (1) T d 4 Ž. Južnič-Zonta et al. Journal of Cleaner Production 367 (2022) 132873 The cost analysis is performed using NPV analysis (Pasqual et al., relative proximity of the alternatives to the ideal solution is estimated, it 2013) (Equation (4)), where CFn is the cash flow for year n, T is life of the is possible to rank and sort all alternatives. It should be noted that the project in years and r is the interest (or discount) rate. final ranking depends on the initial design parameters assigned by handbook guides and expert knowledge, and the consideration of a ∑ T NPV = CF0 + CFn (1 + r)− n [€ ] (4) constant influent, so the classification could change if an additional n=1 optimization loop is performed as explained in the following Step6. The project life is set to 25 years and the interest rate to 4%, as Step6: Design parameter optimization recommended by the guidelines of the European Commission for cost- benefit analysis in the water supply/sanitation sector (European Com­ This step can be applied to all configurations or only to the top- mission, 2014). The NPV estimation provides an economic comparison graded in Step 5. The design parameters are refined by maximizing the framework to show the relative costs of each plant design. An alternative NPV and limiting the AEV and other design criteria. Although the most measure to NPV is the internal rate of return (IRR), calculated by solving important parameters to optimize are the reactor volumes, the optimi­ the non-linear equation (5), which is obtained from equation (4) setting zation problem can be extended to the controllers (e.g., set-points and the NPV to zero. tuning) (Guerrero et al., 2012; Rojas et al., 2011). Optimization methods ∑ T that are specially designed for expensive, discontinuous and constrained 0 = CF0 + CFn (1 + IRR)− n [€ ] (5) objective functions should be considered since configurations with n=1 discrete-continuous model structure (for example technologies ST2b, Capital expenditures (CAPEX) and operational expenditures (OPEX) ST4a/b and ST5 described in Table 1) could have long simulation times. were calculated as detailed in section SI-1. The incomes from WWTP products were derived from the net production of electricity (from Step7: Uncertainty analysis cogeneration with produced biogas), nutrient recovery and PHA pro­ duction. The wastewater tariff was not included as income since it can The model uncertainty input is composed of two main sources: (i) a vary considerably depending on the location. Taxes deductions on in­ stochastic process in time produced by a WWTP inflow generator and (ii) comes were not accounted for (i.e., pre-tax-NPV). Moreover, no grant or a set of uncertain bio-chemical model parameters. Similarly to the project co-financing was considered. Monte-Carlo (MC) approach used by Benedetti et al. (2013), the un­ The environmental impact assessment was calculated based on certainty input is fed to perform the uncertainty analysis of each WWTP estimated GHG emissions. Direct emissions account for biological NO, configuration with the design parameters that need to be optimized N2O and CH4 emissions from wastewater treatment, while indirect (Step6) or let fixed at their initial values (Step3). After each MC simu­ emissions account for CO2 equivalents produced from electricity, con­ lation TOPSIS is recalculated and the robustness of the classification is sumables, durables (construction of the plant units) and credits. Credits checked. are negative contributions to indirect GHG emissions, since they are given for recovering resources from the wastewater (i.e., electricity, 2.3. WWTP models struvite, PHA, etc.) that avoid fossil CO2 production. Biological N2O emissions are quantified following the model described by Massara et al. The different stages of the superstructure contain one or more (2018) with the total N2O emission factor (EFN2O) that considers both treatment and/or resource recovery units that need to be described by the stripped N2O (N2OST) and the N2O in the effluent (N2OEF) in relation appropriate models. To facilitate the task of modelling complex pro­ to the total nitrogen mass flow in the influent (NIN): cesses, an automatic code generation routine was designed, developed, N2OST + N2OEF and implemented in Python to translate a model described in MS-Excel EFN2O = 100⋅ (6) sheets into Modelica code and to check the stoichiometry. The following NIN sections briefly describe the models implemented. The maturity and reliability of a WWTP configuration was assessed with the SRL, which was estimated averaging the technology readiness 2.3.1. ST2b: Short-Cut Enhanced Phosphorus and PHA Recovery levels (TRL) as defined by the ISO 16290:2013 standard of all the stages. (SCEPPHAR) When a stage was empty, its TRL was not included in the average. A scheme of the operation of the SCEPPHAR process (Larriba et al., Finally, the plant land area (Area) (Equation (7)) was estimated as 2020) is presented in Fig. S2. It is a two-sludge system based on four the sum of all surface areas of the process stages (Astages), access roads units: two sequential batch reactors (SBRs), one focused on heterotro­ (Aroads) (Equation (8)) and an assumed 50% idle area which depends on phic processes (HET-SBR) and another for autotrophic processes the relative process unit design parameters. (AUT-SBR), an interchange vessel (INT) and a precipitation reactor for ( ) [ ] Area = Astages + Aroads ⋅2 m2 (7) P-recovery as struvite (PRE). The system works with nitrogen removal via nitrite (nitritation in AUT-SBR and denitritation in the anoxic phase [ ] Aroads = 0.32⋅Astages m2 (8) of HET-SBR). The anaerobic purge of HET-SBR allows to obtain sludge containing PHA. The ASM2d + N2O model (Massara et al., 2018) was used to describe Step5: Design configuration sorting this system, but including some extensions to allow a wider range of operational conditions to be simulated. The changes implemented were: All configurations were ranked applying the technique of preference (i) effect of temperature on kinetics; (ii) inhibition of ammonia and ni­ ordering by similarity to the ideal solution (TOPSIS) (Tzeng and Huang, trite on nitrifiers; (iii) decrease of reaction rate under unmixed condi­ 2011), which was introduced by Hwang and Yoon (1981). This method tions in SBR configurations; and (iv) saturation of PHA accumulation for selects a decision alternative based on the shortest distance to the ideal polyphosphate-accumulating organisms (PAO) to a maximum ratio of 2 solution and the farthest distance to the negative-ideal solution in geo­ gCOD-PHA/gCOD-PAO to avoid unrealistic PHA accumulation in some metric terms (e.g., Euclidean distance). Moreover, each criterion (e.g., scenarios. The settling phase was modelled as a static solid/liquid sep­ NPV, GHG emissions, EQI) should monotonically increase or decrease aration process, while P precipitation and struvite formation were utility to easily define the ideal and negative-ideal solutions. Criteria modelled by solving the chemical equilibrium Mg-NH4-PO4. The volume normalization is necessary when they have different units and weights exchange ratio (VER) of the SBRs was set to 80%, assuming good are assigned by the decision maker to express preference. Once the biomass settling characteristics. The control system required 5 Ž. Južnič-Zonta et al. Journal of Cleaner Production 367 (2022) 132873 proportional-integral (PI) controllers for dissolved oxygen (DO), alka­ calibration and validation step, the absolute objective values are only linity and acetate dosing and a phase control programmed with Graf­ relative. The SP-DSS performs the final ranking by including a reference chart (Fig. S3 and Section SI-2 for a more detailed explanation). A buffer case represented by a conventional WWTP configuration (primary tank was included to connect the continuous wastewater inflow to the settling, A2O and anaerobic digestion, i.e. PS-A2O-AD). Target values SBR process. In a full-scale WWTP, this buffer tank would not be for EQI, AEV, NPV and GHG are calculated in relation to this reference necessary because the inflow would be distributed in parallel ST2b lines. configuration, and finally ranked by TOPSIS and made available to the However, it was preferred to scale only one treatment line to maintain decision-maker. Step6 and Step7 of SP-DSS framework (optimization of the parsimony of the model and reduce the complexity of the simulation. design parameter and uncertainty analysis) are not implemented in this case study. 2.3.2. AD: Ordinary anaerobic digestion The raw wastewater inflow and characteristics are fixed to the time The anaerobic digestion (AD) process (Fig. S4) is based on a con­ invariant average values shown in Table 2. The assumed medium ventional WWTP sludge digestion. The kinetic model used is based on strength raw wastewater characteristics imply an IQI of 76658 kg Poll/ the ADM1coP, a modification of the conventional ADM1 model (Bat­ day and a population equivalent (PE) of 91000 PE. Water, ground and stone et al., 2002). The interface model between the ASM2d + N2O and air temperature are also considered constant: 20, 15 and 25 ◦ C. The ADM1coP models, based on the work of Solon et al. (2017), integrates digested sludge is sent to an incineration facility rather than to agri­ co-substrate flows, fractionation of particulate COD and instantaneous culture. Stage4 includes a drum filter (DF) unit to remove all the effluent conversion of N-compounds, including those associated to N2O forma­ TSS, to avoid the interference of the settling models used in each case. tion. Finally, the centrifuge processing the primary sludge is adjusted to The parameters for ASM2d-N2O were set to default values except for the produce a 5% TSS concentrate with a COD removal efficiency of 90%. following cases: (i) overall rates of lysis and decay for biomass were decreased in order to avoid excessive accumulation of inert materials 2.3.3. A2O: Anaerobic/anoxic/oxic EBPR process (bH = 0.12 d− 1, bPAO = bPP = bPHA = 0.06 d− 1); (ii) AOB affinity constant The process presented in Fig. S5 is a typical A2O configuration for HNO2 was increased (KHNO2_AOB = 0.6 gN/m3) to better describe the applied in benchmarking studies (Guerrero et al., 2013; Meneses et al., emissions of N2O in accordance to experimental evidence from the ST2b 2015) and represents an appropriate reference for SCEPPHAR. The total pilot-plant (Larriba et al., 2020). The parameters of the ADM1 model reaction volume is divided into an anaerobic zone (10%, 2 reactors), an relative to the side-stream digester are maintained at their default anoxic zone (15%, 2 reactors) and an aerobic zone (75%, 3 reactors). values, where only the pH was set to a fixed value of 7 since it is assumed The purge flowrate was manipulated to control the solids concentration that enough buffer capacity is provided for this operation conditions. in the last reactor around 3000 mg/L. The secondary settler was Table 3 summarises the three plant configurations used for this case modelled as a simple liquid/sludge separator. study, indicating the units selected for each SP-DSS stage. The main assumptions for each configuration are given as follows: 2.3.4. PS: Primary sludge clarifier The primary clarifier is modelled as a solid/liquid separator (Fig. S6) 1. PS-A2O-AD. Raw wastewater goes to a primary settler unit (PS) after with separation efficiencies for COD and TSS depending on surface which the clarified wastewater is treated in an A2O system with overflow rate (SOR), influent TSS and wastewater temperature (Chris­ 17300 m3 of total volume distributed as detailed in section 2.3.3. For toulas, 1998). The operating conditions are set at HRT = 2.5 h and SOR maintenance and back-up reasons, there are two PS units: one in = 40 m3/m2/d. If the SOR exceeds the maximum allowed, a by-pass operation and the other in stand-by. The return sludge and internal valve in front of the clarifier is opened. A sludge buffer tank is recycle factors are 1 and 4. DO in the aerobic tanks is set to 2 g/m3 included to smooth the feed to the digestion unit. This buffer tank can be with PI-controllers manipulating the airflow of the blowers. considered as a sludge storage tank which is sometimes present in a real 2. PS-ST2b-AD. Raw wastewater goes to a PS unit after which the WWTP or as a model for the sludge blanket inside the primary clarifier. clarified wastewater is treated in the mainstream SCEPPHAR unit (ST2b). The total volume of HET-SBR and AUT-SBR are each of 10920 m3. ST2b working cycle is 8 h and only one treatment line 2.4. Case study definition treating the average inflow is used. Two buffer tanks (V = 21840 m3) are installed to account for continuous inflow dynamics found in a This section presents an example of the SP-DSS application, focusing real situation, one for feeding and the second as an effluent equalizer. on assessing the feasibility to transform the WWTP of Manresa (Cata­ The purge of waste sludge is constant and set in relation to the SRT: lonia, Spain) into a WRRF using ST2b. Although the SP-DSS is based on 10 and 12 days for the HET-SBR and AUT-SBR. Concentration of TSS dynamic models to consider the variability of the input in the evaluation in the clarified effluent from HET-SBR and AUT-SBR are assumed 25 of alternatives, this first presentation of the SP-DSS focuses on a com­ and 10 g/m3 to reproduce what was observed in the pilot-plant parison of two steady-state alternatives to demonstrate its multi-criteria (Larriba et al., 2020). DO setpoint for the aerobic phase of both re­ evaluation capability. Further work will compare how the evaluation of actors is set to 2 g/m3. Struvite recovery is done by dosing MgCl2 and alternatives changes when the design is carried out with constant or air-stripping to increase pH in R3-PRE reactor. PHA recovery from variable influent. waste sludge is enabled only if the ratio PHA/TSS is higher than 20%; Considering that the whole procedure does not include a model otherwise, the waste sludge from HET-SBR is directed to the AD. Methanol dosing is provided in HET-SBR during the anoxic phase to Table 2 Raw wastewater inflow and characteristics for the Manresa WWTP case study. Table 3 Parameter Units Value Selected process units in each stage of the plant superstructure for the three Qliq m3/d 21840 scenarios evaluated. TCOD gCOD/m3 592 PS-A2O-AD PS-ST2b-AD ST2b-AD SCOD gCOD/m3 178 TKN gN/m3 54 Stage1 PS PS - NH4 gN/m3 42 Stage2 - - - TN gN/m3 56 Stage3 A2O ST2b ST2b PO4 gP/m3 4 Stage4 DF DF DF TP gP/m3 7 Stage5 AD AD AD TSS g/m3 289 Stage6 - - - 6 Ž. Južnič-Zonta et al. Journal of Cleaner Production 367 (2022) 132873 guarantee the denitrification of nitrite and nitrate. Recovered stru­ weight was low because it was assumed that legal requirements were vite is used as fertilizer in agriculture. satisfied: any effort to improve even further the effluent quality may be 3. ST2b-AD. The only difference with PS-ST2b-AD is the lack of PS, i.e., secondary. In relative terms, GHG emissions improvements were the raw wastewater is fed directly to the mainstream SCEPPHAR unit weighted slightly lower than the improvements in effluent quality. Note (ST2b). PS-ST2b-AD configuration adds the PS unit to ensure good that there was no legal constraint so far on GHG emissions and numeric clarification of SBRs and lower operation risks since ST2b operates estimations or field measurements are still affected by considerable with a high VER of 0.8, instead of typical SBR systems working with a uncertainties. Very low weights were set for technology maturity and VER of 0.15–0.35. Theoretically, the optimal performance for ST2b space requirements since it was assumed that for the current case study would be without a PS unit since this last removes valuable COD space was not an issue (its cost was already accounted in CAPEX) and needed for PHA accumulation and increases costs and WWTP area. that most operation problems relative to technology maturity could be However, a long stable operation with PS-ST2b-AD was maintained solved. In relation to the economic assessment, it should be noted that during the ST2b pilot-plant operation (Larriba et al., 2020), while the NPV estimation reported in Table 4 does not account for any only a shorter period was tested with ST2b-AD. Because of this, it is wastewater tariff income, but only incomes from direct product pro­ assumed that ST2b-AD is not yet proven to be as robust as duction (i.e., electricity, PHA, etc.). On the other hand, the tariff shown PS-ST2b-AD. In any case, ST2b-AD is a useful configuration to study in Table 4 represents the amount of money needed to achieve a neutral as it gives an idea of possible improvements in the performance of NPV after 25 years with an interest rate of 4%. ST2b. In relation to the TOPSIS sorting (see Table 4), the best configuration is the baseline PS-A2O-AD (shortly A2O), followed by ST2b-AD and PS- For the economic assessment, struvite indirect benefits from the ST2b-AD. Running a simple robustness analysis over the decision avoided maintenance of the piping system were not considered. For land criteria weights, with the constraint that their sum is equal to one (see costs, a value of 80 €/m2 was assumed, but the range could be very large, Table 5), it can be concluded that the PS-A2O-AD first rank position is from agricultural land prices of 1.5 €/m2 (Eurostat 2017, Catalunya) to very robust and that only ST2b-AD is able to achieve the best rank, but 170 €/m2 for land suitable for building (Tsagarakis et al., 2003). In this only for a short range of weights combinations. study, it is not assumed the real sludge disposal costs to agriculture of One of the reasons for the A2O success is its well-balanced objective Manresa WWTP (96 €/ton), but an European average value of 317 €/ton values. It has the best wastewater treatment performance (EQI of 2.34 g (Foladori et al., 2010) is set to reproduce what is generally observed in Poll/m3), the lowest investment cost (CAPEX of 0.130 €/m3), GHG WWTPs around Europe (i.e. EU sludge disposal costs are 40–60% of emissions (0.14 kgCO2/m3) and plant area needs (0.24 m2/PE). More­ OPEX). Then, it was assumed that digestate was incinerated in a facility over, A2O achieved the highest technology maturity level (i.e., highest 20 km away instead of disposed in agriculture. In consequence, no GHG SRL). On the other hand, ST2b configurations performed better in terms emission credits relative to fertilization were applied here. of OPEX (0.397 €/m3 for ST2b-AD) and benefits (5.260 ¢€/m3 for PS- The resulting PWM configurations were simulated for 300 days to ST2b-AD). Note that ST2b had almost a twofold area increase if remove the influence of transient conditions and the objective values compared with A2O because of its buffer tanks. However, if buffer tanks were estimated for the last 50 days. were replaced by a parallel multi-line configuration its area would be only 4% larger than that of A2O (results not shown). 3. Results and discussion The effect of removing the PS from PS-ST2b-AD is relevant since it improves the tariff by 7% (see Table S1). The main reason for this This section reports and compares the results obtained by evaluating economic feasibility improvement in relation to PS-ST2b-AD was a each of the criteria for all the alternatives in the case study. Subse­ lower sludge production (see Section 3.3), mainly due to the lower quently, the cost estimates obtained through the LCCA study are shown methanol requirements for denitrification, which decreased sludge and the ability of the SP-DSS to generate different levels of reporting production. This fact implied that ST2b-AD OPEX was 9% lower than PS- results is demonstrated. Finally, the planned improvements to the soft­ ST2b-AD and 23% lower than A2O. Decreased sludge production ware are discussed. implied a loss of 8% in benefits from biogas if compared with PS-ST2b- AD. However, this loss was by far compensated by the improvement in 3.1. Comparison of decision criteria values OPEX savings. There was even an improvement in the treatment The simulations relative to the environmental and economic Table 5 assessment are reported in Table 4 and Table S1. The configurations Robustness analysis of the TOPSIS ranking. The ranges represent the weights of were sorted with the TOPSIS method, where EQI, Tariff, GHG, SRL and the decision criteria (step size of 0.05) for which a given plant configuration is Area were taken as decision criteria with respective weights 0.25, 0.6, best ranked after sorting (i.e., rank equal to 1). Note that for any combination of 0.15, 0.05 and 0.05. These weights were selected by the authors to try to the weighs, PS-ST2b-AD is never the best option and is therefore not present in the table. reflect the point of view of a public water service body that needs to build a new WRRF in line with the circular economy policy. These EQI Tariff GHG SRL Area weights can be easily modified to consider different criteria. The highest PS-A2O-AD 0–1 0–0.90 0–1 0–1 0–1 weight was assigned to the economic factor since its estimation was ST2b-AD 0–0.10 0.75–1 0–0.10 0–0.20 0–0.05 reliable and traditionally relevant to decision makers. Effluent quality Table 4 Summary of the results of the environmental and economic assessment of different plant configurations. The configurations were ranked using the TOPSIS method, where EQI, Tariff, GHG, SRL and Area were taken as decision criteria. PS-A2O-AD (underlined) is the reference configuration. TOPSIS order Configuration EQI CAPEXa OPEX Benefits NPV Tariff SRL Area GHG EFN2O 3 3 3 3 3 3 2 3 g Poll/m €/m €/m ¢€/m €/m €/m - m /PE kgCO2/m % 1 PS-A2O-AD 2.34 0.130 0.515 4.946 − 0.421 0.674 9.00 0.24 0.14 0.23 3 PS-ST2b-AD 3.74 0.201 0.438 5.260 − 0.442 0.706 8.25 0.46 0.18 0.95 2 ST2b-AD 3.30 0.191 0.397 4.820 − 0.409 0.654 8.00 0.44 0.19 1.09 a Normalized over 25 years of plant life. 7 Ž. Južnič-Zonta et al. Journal of Cleaner Production 367 (2022) 132873 efficiency of wastewater: ST2b-AD had a 12% lower EQI than PS-ST2b- AD because of a slightly better nitrogen removal. Note that CAPEX of ST2b-AD was not much improved (5% lower than PS-ST2b-AD) since the PS unit investment cost was not relevant. Finally, overall GHG emissions only slightly increased in relation to PS-ST2b-AD (5%) due to the improved nitrogen oxidation and in consequence increased N2O emissions. 3.2. LCCA methodology validation In order to validate the cost estimation methodology used, the CAPEX of the A2O configuration was estimated with a statistical cor­ relation provided in the “Compliance Costs of the Urban Wastewater Fig. 2. Pollutant removal efficiencies for the Manresa WWTP case study. The Treatment Directive” (MEDG and MMS, 2010) from the Water Infor­ configurations are ranked with TOPSIS, where EQI, Tariff, GHG, SRL and Area mation System for Europe (WISE). For the case study of Manresa WWTP are taken as decision criteria (from left to right, best to worst). PS-A2O-AD (in inflow, the estimated CAPEX for a WWTP with N and P removal without bold) is the reference configuration. a sludge treatment line is 235.9 €/PE relative to 2008 for the Denmark price level of 146 (Eurostat). If adjusted to the EU28 price level (109) imposes TCOD, TSS, TN and TP efficiencies to be higher than 75, 90, 80 and updated to 2019 within CEPCI (i.e., factor 1.21) the adjusted CAPEX and 80%. In most of the cases, the legal effluent requirements are is 213.8 €/PE, which is 33% lower than the A2O estimate from the guaranteed and improved. In this case-study, no further optimization of SP-DSS (i.e., 284.7 €/PE). The difference could be explained by the fact design parameters is attempted. If further fine tuning were to be made, that the WISE-correlation accounts only for cost of the mainstream unit effluent quality could be improved at the expense of operation or con­ processes. struction costs. For example, despite TN removal percentage was ach­ The current WWTP of Manresa, which is a N and P removal system ieved, nitrite and nitrate slightly exceed the legal requirements (see (modified Ludzack-Ettinger process, MLE) with chemical precipitation Fig. S7). To improve nitrite and nitrate removal efficiency in all con­ and sludge treatment line had an investment cost of 10 M€ and was built figurations, an additional post-denitrification process could be added. in 1985; if adjusted to 2019 (i.e., CEPCI factor 1.99), its CAPEX would be The SP-DSS reporting tool also builds interactive liquid phase mass 218.4 €/PE, 30% lower than the estimate provided by the SP-DSS. In this flow diagrams of COD, N and P for all configurations under study (see case, the difference could be explained by the different configuration Fig. S8 in S1 and SI S2). Only liquid-phase mass flows are represented in (MLE vs. A2O) and the accuracy of the methodology, which would be these Fig. S8 to keep the diagram as simple as possible but, if needed, the around ±50%. These results partially validate the procedure of CAPEX gas phase mass flows could easily be derived from the mass balance estimation of the SP-DSS. Unfortunately, to our knowledge, OPEX equations and presented in similar figures. cannot be compared to other estimates, as no correlations are available, Life Cycle Costing Analysis (LCCA) is broken down in Fig. 3 (for or they are simple relations based on CAPEX estimates. Finally, the additional information see Fig. S9, Table S2 and Table S3). To compare estimated tariff of 0.65 €/m3 for the A2O system has a correct order of CAPEX with benefits and OPEX, CAPEX is normalized over 25-year life magnitude since the Spanish average is 0.73 €/m3 with a range of of the plant in terms of total wastewater treated. The benefits of struvite 0.37–1.34 €/m3. recovery from ST2b are negligible, as quantities and market prices are This result shows that the SP-DSS could be used to accomplish pre- low (150 €/ton). It is interesting to note that PS-ST2b-AD and ST2b-AD feasibility assessments of WWTP projects that require an accuracy of have 17% and 28% less digestate production than A2O, which would ±30–50%. Moreover, since the same cost-estimation methodology is intuitively lead to lower biogas production. However, the gross energy used, it allows to obtain a reasonable benchmarking ranking among the production benefit of PS-ST2b-AD equals A2O performance and ST2b- multiple plant configurations evaluated. AD only has a 9% lower production than A2O. This is because, in the case of ST2b, the excess sludge has a higher PHA content, which results 3.3. SP-DSS reporting levels in a higher biogas yield when compared to a conventional waste sludge (Chan et al., 2020). An important difference between PS-ST2b-AD and Besides the main decision criteria values, the SP-DSS provides a great ST2b-AD is that more COD is available for the PHA accumulation pro­ amount of extra detailed information that could give a better insight of a cess, increasing the weight ratio of PHA to TSS from 16% to 19%, which given WWTP design. Here, an overview of its reporting capability is raises the biogas production yield. presented. The SP-DSS provides three reporting levels: RL0 as the main In terms of CAPEX, PS-ST2b-AD has an expensive pumping system objective values for decision making (the values presented in Table 4), (almost 300% higher than A2O) but a lower cost of solid/liquid sepa­ RL1 with aggregated Key Performance Indicators (KPIs) at plant-wide ration units, since no secondary settlers are needed (44% lower than level and RL2 with aggregated KPIs at process stage level. At RL2, all A2O). In relation to the cost of tanks and mixers, ST2b costs are 250% the KPIs are aggregated from single process units that build-up a given and 300% higher compared to A2O. In relation to the investment costs of SP-DSS stage. On the other hand, RL1 KPIs are aggregated from all the the aeration system (i.e. blowers and diffusers) ST2b is approximately stage related KPIs of RL2. The values of the RL0 criteria are used as input five times more expensive than A2O due to the discontinuous operation data for the TOPSIS ranking method, which provides the decision on the of the SBR and the stripping system (pH control) in relation to struvite best WWTP configuration. However, the SP-DSS reporting enables the recovery. These costs could be reduced if the stripping system in the PRE decision maker to access finer scales of information to support the final reactor was replaced by chemical dosing for pH control and if the time to decision or to understand better the pros and cons of a given reach the DO set-point at the beginning of the aeration phase in AUT- configuration. SBR was relaxed. As expected, in all configurations the OPEX was mainly due to sludge 3.3.1. Aggregated KPIs at plant-wide level (RL1) disposal (44–47% of total OPEX) and maintenance (30–33% of total Fig. 2 provides insight on the effluent quality achieved by each OPEX). The lower sludge disposal costs of ST2b in relation to A2O (17% scenario in terms of daily basis removal efficiency and effluent N and P for PS-ST2b-AD) are the main driver of its economic feasibility. If concentrations. Daily averaged removal efficiencies are close or above maintenance costs are considered, ST2b shows a slight reduction the legal requirement of the Council Directive 91/271/EEC, which 8 Ž. Južnič-Zonta et al. Journal of Cleaner Production 367 (2022) 132873 Fig. 3. LCCA results for the Manresa WWTP case study. The configurations are ranked with TOPSIS (from left to right, best to worst). PS-A2O-AD (in bold) is the reference configuration. “Benefits” are the incomes from selling energy (i.e., electricity) and re­ sources recovered from wastewater, “Investments” are related to purchased equipment (sum gives CAPEX) and “Operation” costs are related to mainte­ nance, energy, consumables and digestate disposal (sum gives OPEX). Black points are the Net Present Values for 25-year plant life and discount rate of 4%. compared to A2O (6% for PS-ST2b-AD). This positive result is explained efficient use of methanol. PS-ST2b-AD is the only net producer of elec­ by the assumption that maintenance and replacement costs are usage- tricity, which implies 391% lower indirect electricity emissions than dependent for all equipment operating discontinuously (e.g. blowers A2O (see Fig. 4). and pumps). In the case of the A2O configuration, the contribution to OPEX of 3.3.2. Aggregated LCCA-KPIs at process stage level (RL2) energy (8%), consumables (8%) and “other” (7%) terms was limited (see The process stage level of LCCA reporting enables to focus on each Table S3). Approximately the same is true for both ST2b configurations stage of a given WWTP design configuration. Fig. 5 represents the and any improvement on these terms will only lead to a slight overall benefits, CAPEX and OPEX of each process stage for the three improvement of the plant NPV. However, it is interesting to note that PS- scenarios. Benefits are dominated by the incomes generated from the ST2b-AD has lower energy needs (see Table S2) than A2O (14%), most anaerobic digester (Stage5). Struvite produced in ST2b in Stage3 has only likely due to the implementation of the N removal process via nitrite. a limited contribution to benefits, since in this case study it was assumed Moreover, both ST2b configurations have considerably lower consum­ that undesired struvite precipitation in tubing does not affect mainte­ ables costs than A2O (51%) thanks to a lower methanol addition for the nance costs. denitrification process. If the CAPEX of each process stage is considered (Fig. 5 - middle), From the LCA perspective (see Fig. 4 and Table S4, Table S5 in SI), ST2b has a much higher investment cost of Stage3 compared to A2O. On ST2b has a fourfold higher direct N2O emissions than the baseline A2O the other hand, since the sludge production of ST2b is lower than in configuration because it operates with via-nitrite nitrogen removal, A2O, the cost of Stage5 (sludge treatment) is significantly lower. The which is reported to increase N2O emissions (Massara et al., 2018). All lower production of sludge impacts favorably on Stage5 investment configurations are self-sufficient in terms of thermal energy consump­ costs, but also on the operation and maintenance efforts (see Fig. 5 - tion (i.e., AD heating) and consequently produce no direct CO2 emis­ bottom) in comparison to the baseline A2O configuration. sions because no external natural gas is consumed. Finally, the SP-DSS reporting tool generates detailed LCCA stage In terms of indirect GHG emissions, PS-ST2b-AD and ST2b-AD have level assessment diagrams of each design configuration (see Fig. S10 to 55% and 61% lower consumables emission than A2O due to their Fig. S12 in SI). 3.4. Future software development Future development efforts will focus on tuning the wastewater inflow generator module (Step2 of the framework) to assess the eco­ nomic and environmental feasibility of the WRRF under dynamic inflow conditions. At the moment, the most relevant design parameter values are taken from reference handbooks or expert knowledge. This sub- optimal practice will be improved by implementing surrogate-based global optimization methods that are suitable for expensive function evaluations (Step6 of the framework). The task to parallelize model simulations (i.e. objective function evaluations) is straightforward, since the build-up of WWTP configurations is automatic and the relative Fig. 4. LCA results for the Manresa WWTP case study. The configurations are model simulations are independent. Parallel computing of models will sorted with TOPSIS (from left to right, best to worst). PS-A2O-AD (bold) is the allow the uncertainty analysis (Step7 of the framework) and will boost reference configuration. the global optimization step. Finally, since models of new wastewater 9 Ž. Južnič-Zonta et al. Journal of Cleaner Production 367 (2022) 132873 Fig. 5. Benefits, CAPEX and OPEX of each process stage for each configuration. The configurations are ranked with TOPSIS (from left to right, best to worst). treatment and resource recovery technologies could be easily integrated CRediT authorship contribution statement into the SP-DSS superstructure, its architecture will be extended ac­ cording to stakeholders’ needs for pre-design of new WRRFs or retro­ Živko Južnič-Zonta: Conceptualization, Methodology, Software, fitting old ones. Validation, Formal analysis, Investigation, Data curation, Writing – original draft, Visualization. Albert Guisasola: Conceptualization, 4. Conclusions Validation, Writing – review & editing. Juan Antonio Baeza: Concep­ tualization, Validation, Resources, Writing – review & editing, Super­ In this work the Smart-Plant Decision Support System (SP-DSS) was vision, Project administration, Funding acquisition. presented, a tool built to advise potential stakeholders on the effects of implementing resource recovery technologies for their specific waste­ Declaration of competing interest water treatment problem. This SP-DSS decreases the complexity of technology benchmarking, while allowing the estimation of economic, The authors declare that they have no known competing financial effluent quality and environmental impact multi-criteria values, in order interests or personal relationships that could have appeared to influence to align the decision-making process with the holistic vision of the cir­ the work reported in this paper. cular economy. The following strengths emerge from the framework developed: Acknowledgements • The use of open-source software makes the SP-DSS attractive for This work was supported by European Union’s Horizon 2020 future collaborative research and development activity; research and innovation programme under grant agreement No.690323, • Python scientific and data visualization libraries greatly facilitate the project SMART-Plant (Scale-up of low-carbon footprint material recov­ benchmark analysis and results reporting; ery techniques in existing wastewater treatment plants). We would like • Modelica component-oriented modelling allows to automatically to thank Borja Solís for the fruitful discussion on mass balance model build WWTP configurations from a common superstructure; check and ASM2d/ADM1 model interface implementation. • Dynamic bio-process models and control systems are both integrated into a plant-wide model and there is no need for static-model ap­ Appendix A. Supplementary data proaches that may lead to filter out promising WWTP configurations; • Simulation of the many WWTP configurations is easy to scale within Supplementary data to this article can be found online at https://doi. distributed computing, since models can run as independent org/10.1016/j.jclepro.2022.132873. instances; References As a case study of SP-DSS application the SCEPPHAR process was benchmarked against a conventional A2O process under steady-state Arnell, M., Rahmberg, M., Oliveira, F., Jeppsson, U., 2017. Multi-objective performance inflow conditions. Future work will expand the economic and environ­ assessment of wastewater treatment plants combining plant-wide process models mental assessment to all the SMART-Plant technologies. and life cycle assessment. J. Water Clim. Chang. 8, 715–729. https://doi.org/ 10.2166/wcc.2017.179. Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S.V., Pavlostathis, S.G., Rozzi, a, Sanders, W.T.M., Siegrist, H., Vavilin, V. a, 2002. The IWA anaerobic digestion model No 1 (ADM1). Water Sci. Technol. 45, 65–73. https://doi.org/10.2166/ wst.2002.0292. 10 Ž. Južnič-Zonta et al. Journal of Cleaner Production 367 (2022) 132873 Benedetti, L., Belia, E., Cierkens, K., Flameling, T., De Baets, B., Nopens, I., Weijers, S., Hwang, C.-L., Yoon, K., 1981. Multiple Attribute Decision Making, Lecture Notes in 2013. The incorporation of variability and uncertainty evaluations in WWTP design Economics and Mathematical Systems. Springer Berlin Heidelberg, Berlin, by means of stochastic dynamic modeling: the case of the Eindhoven WWTP Heidelberg. https://doi.org/10.1007/978-3-642-48318-9. upgrade. Water Sci. Technol. 67, 1841–1850. https://doi.org/10.2166/ Jiménez-Benítez, A., Ferrer, F.J., Greses, S., Ruiz-Martínez, A., Fatone, F., Eusebi, A.L., wst.2013.064. Mondéjar, N., Ferrer, J., Seco, A., 2020. AnMBR, reclaimed water and fertigation: Bozkurt, H., Quaglia, A., Gernaey, K.V., Sin, G., 2015. A mathematical programming two case studies in Italy and Spain to assess economic and technological feasibility framework for early stage design of wastewater treatment plants. Environ. Model. and CO2 emissions within the EU Innovation Deal initiative. J. Clean. Prod. 270 Software 64, 164–176. https://doi.org/10.1016/j.envsoft.2014.11.023. https://doi.org/10.1016/j.jclepro.2020.122398. Bozkurt, H., van Loosdrecht, M.C.M., Gernaey, K.V., Sin, G., 2016. Optimal WWTP Jimenez, J., Miller, M., Bott, C., Murthy, S., De Clippeleir, H., Wett, B., 2015. High-rate process selection for treatment of domestic wastewater – a realistic full-scale activated sludge system for carbon management – evaluation of crucial process retrofitting study. Chem. Eng. J. 286, 447–458. https://doi.org/10.1016/j. mechanisms and design parameters. Water Res. 87, 476–482. https://doi.org/ cej.2015.10.088. 10.1016/j.watres.2015.07.032. Castillo, A., Cheali, P., Gómez, V., Comas, J., Poch, M., Sin, G., 2016a. An integrated Johnsson, C., Årzén, K.-E., 1999. Grafchart and grafcet: a comparison between two knowledge-based and optimization tool for the sustainable selection of wastewater graphical languages aimed for sequential control applications. IFAC Proc. 32, 19–24. treatment process concepts. Environ. Model. Software 84, 177–192. https://doi.org/ https://doi.org/10.1016/S1474-6670(17)56006-1. 10.1016/j.envsoft.2016.06.019. Larriba, O., Rovira-Cal, E., Juznic-Zonta, Z., Guisasola, A., Baeza, J.A., 2020. Evaluation Castillo, A., Porro, J., Garrido-Baserba, M., Rosso, D., Renzi, D., Fatone, F., Gómez, V., of the integration of P recovery, polyhydroxyalkanoate production and short cut Comas, J., Poch, M., 2016b. Validation of a decision support tool for wastewater nitrogen removal in a mainstream wastewater treatment process. Water Res. 172, treatment selection. J. Environ. Manag. 184, 409–418. https://doi.org/10.1016/j. 115474 https://doi.org/10.1016/j.watres.2020.115474. jenvman.2016.09.087. Lorini, L., Martinelli, A., Capuani, G., Frison, N., Reis, M., Sommer Ferreira, B., Castillo, A., Vall, P., Garrido-Baserba, M., Comas, J., Poch, M., 2017. Selection of Villano, M., Majone, M., Valentino, F., 2021. Characterization of industrial (food, drink and milk sector) wastewater treatment technologies: a multi- polyhydroxyalkanoates produced at pilot scale from different organic wastes. Front. criteria assessment. J. Clean. Prod. 143, 180–190. https://doi.org/10.1016/j. Bioeng. Biotechnol. 9, 1–13. https://doi.org/10.3389/fbioe.2021.628719. jclepro.2016.12.132. Machado, V.C., Lafuente, J., Baeza, J.A., 2020. Systematic comparison framework for Chan, C., Guisasola, A., Baeza, J.A., 2020. Correlating the biochemical methane potential selecting the best retrofitting alternative for an existing water resource recovery of bio-P sludge with its polyhydroxyalkanoate content. J. Clean. Prod. 242, 118495 facility. Water Environ. Res. 92, 2072–2085. https://doi.org/10.1002/wer.1368. https://doi.org/10.1016/j.jclepro.2019.118495. Mannina, G., Cosenza, A., Rebouças, T.F., 2020. A plant-wide modelling comparison Christoulas, D., 1998. An empirical model for primary sedimentation of sewage. Environ. between membrane bioreactors and conventional activated sludge. Bioresour. Int. 24, 925–934. https://doi.org/10.1016/S0160-4120(98)00076-2. Technol. 297, 122401 https://doi.org/10.1016/j.biortech.2019.122401. Comas, J., Alemany, J., Poch, M., Torrens, A., Salgot, M., Bou, J., 2004. Development of a Massara, T.M., Solís, B., Guisasola, A., Katsou, E., Baeza, J.A., 2018. Development of an knowledge-based decision support system for identifying adequate wastewater ASM2d-N2O model to describe nitrous oxide emissions in municipal WWTPs under treatment for small communities. Water Sci. Technol. 48, 393–400. https://doi.org/ dynamic conditions. Chem. Eng. J. 335, 185–196. https://doi.org/10.1016/j. 10.2166/wst.2004.0887. cej.2017.10.119. Conca, V., da Ros, C., Valentino, F., Eusebi, A.L., Frison, N., Fatone, F., 2020. Long-term MEDG, M.M.S., 2010. Compliance Costs of the Urban Wastewater Treatment Directive, validation of polyhydroxyalkanoates production potential from the sidestream of pp. 28–30. municipal wastewater treatment plant at pilot scale. Chem. Eng. J. 390, 124627 Meneses, M., Concepción, H., Vrecko, D., Vilanova, R., 2015. Life Cycle Assessment as an https://doi.org/10.1016/j.cej.2020.124627. environmental evaluation tool for control strategies in wastewater treatment plants. Crutchik, D., Frison, N., Eusebi, A.L., Fatone, F., 2018. Biorefinery of cellulosic primary J. Clean. Prod. 107, 653–661. https://doi.org/10.1016/j.jclepro.2015.05.057. sludge towards targeted Short Chain Fatty Acids, phosphorus and methane recovery. Mussati, M., Gernaey, K.V., Gani, R., Jørgensen, S., 2002. Computer aided model analysis Water Res. 136, 112–119. https://doi.org/10.1016/j.watres.2018.02.047. and dynamic simulation of a wastewater treatment plant. Clean Technol. Environ. Da Ros, C., Conca, V., Eusebi, A.L., Frison, N., Fatone, F., 2020. Sieving of municipal Policy 4, 100–114. https://doi.org/10.1007/s10098-002-0153-z. wastewater and recovery of bio-based volatile fatty acids at pilot scale. Water Res. Nopens, I., Benedetti, L., Jeppsson, U., Pons, M.N., Alex, J., Copp, J.B., Gernaey, K.V., 174, 115633 https://doi.org/10.1016/j.watres.2020.115633. Rosen, C., Steyer, J.P., Vanrolleghem, P.A., 2010. Benchmark Simulation Model No de Faria, A.B.B., Ahmadi, A., Tiruta-Barna, L., Spérandio, M., 2016. Feasibility of 2: finalisation of plant layout and default control strategy. Water Sci. Technol. 62, rigorous multi-objective optimization of wastewater management and treatment 1967–1974. https://doi.org/10.2166/wst.2010.044. plants. Chem. Eng. Res. Des. 115, 394–406. https://doi.org/10.1016/j. OSMC, 2022. OpenModelica (open source Modelica Consortium) [WWW Document]. cherd.2016.09.005. www.openmodelica.org (Last access 07/07/22). Dragan, J.M., Zubov, A., Sin, G., 2017. Methodology for Plantwide Design and Palmieri, S., Cipolletta, G., Pastore, C., Giosuè, C., Akyol, Ç., Eusebi, A.L., Frison, N., Optimization of Wastewater Treatment Plants, pp. 859–864. https://doi.org/ Tittarelli, F., Fatone, F., 2019. Pilot scale cellulose recovery from sewage sludge and 10.1016/B978-0-444-63965-3.50145-8. reuse in building and construction material. Waste Manag. 100, 208–218. https:// European Commission, 2014. Guide to Cost-Benefit Analysis of Investment Projects: doi.org/10.1016/j.wasman.2019.09.015. Economic Appraisal Tool for Cohesion Policy 2014-2020. Publications Office of the Pasqual, J., Padilla, E., Jadotte, E., 2013. Technical note: equivalence of different European Union. https://doi.org/10.2776/97516. profitability criteria with the net present value. Int. J. Prod. Econ. 142, 205–210. Fernández-Arévalo, T., Lizarralde, I., Fdz-Polanco, F., Pérez-Elvira, S.I., Garrido, J.M., https://doi.org/10.1016/j.ijpe.2012.11.007. Puig, S., Poch, M., Grau, P., Ayesa, E., 2017. Quantitative assessment of energy and Rigopoulos, S., Linke, P., 2002. Systematic development of optimal activated sludge resource recovery in wastewater treatment plants based on plant-wide simulations. process designs. Comput. Chem. Eng. 26, 585–597. https://doi.org/10.1016/S0098- Water Res. 118, 272–288. https://doi.org/10.1016/j.watres.2017.04.001. 1354(01)00782-7. Flores-Alsina, X., Corominas, L., Snip, L., Vanrolleghem, P. a, 2011. Including Rojas, J.D., Baeza, J.A., Vilanova, R., 2011. Effect of the controller tuning on the greenhouse gas emissions during benchmarking of wastewater treatment plant performance of the BSM1 using a data driven approach. In: Watermatex 2011. 8th control strategies. Water Res. 45, 4700–4710. https://doi.org/10.1016/j. IWA Symposium on Systems Analysis and Integrated Assessment. International watres.2011.04.040. Water Association, London (UK), pp. 785–792. Foladori, P., Andreottola, G., Ziglio, G., 2010. Sludge Reduction Technologies in Seco, A., Ruano, M.V., Ruiz-Martinez, A., Robles, A., Barat, R., Serralta, J., Ferrer, J., Wastewater Treatment Plants. IWA Publishing. https://doi.org/10.2166/ 2020. Plant-wide modelling in wastewater treatment: showcasing experiences using 9781780401706. the Biological Nutrient Removal Model. Water Sci. Technol. 81, 1700–1714. https:// Garrido-Baserba, M., Reif, R., Hernández, F., Poch, M., 2012. Implementation of a doi.org/10.2166/wst.2020.056. knowledge-based methodology in a decision support system for the design of Solís, B., Guisasola, A., Flores-Alsina, X., Jeppsson, U., Baeza, J.A., 2022. A plant-wide suitable wastewater treatment process flow diagrams. J. Environ. Manag. 112, model describing GHG emissions and nutrient recovery options for water resource 384–391. https://doi.org/10.1016/j.jenvman.2012.08.013. recovery facilities. Water Res. 215, 118223 https://doi.org/10.1016/j. Grau, P., de Gracia, M., Vanrolleghem, P.A., Ayesa, E., 2007. A new plant-wide watres.2022.118223. modelling methodology for WWTPs. Water Res. 41, 4357–4372. https://doi.org/ Solon, K., Flores-Alsina, X., Kazadi Mbamba, C., Ikumi, D., Volcke, E.I.P., 10.1016/j.watres.2007.06.019. Vaneeckhaute, C., Ekama, G.A., Vanrolleghem, P.A., Batstone, D.J., Gernaey, K.V., Guerrero, J., Flores-Alsina, X., Guisasola, A., Baeza, J.A., Gernaey, K.V., 2013. Effect of Jeppsson, U., 2017. Plant-wide modelling of phosphorus transformations in nitrite, limited reactive settler and plant design configuration on the predicted wastewater treatment systems: impacts of control and operational strategies. Water performance of simultaneous C/N/P removal WWTPs. Bioresour. Technol. 136, Res. 113, 97–110. https://doi.org/10.1016/j.watres.2017.02.007. 680–688. https://doi.org/10.1016/j.biortech.2013.03.021. Tchobanoglous, G., Stensel, H.D., Tsuchihashi, R., Burton, F., Abu-Orf, M., Bowden, G., Guerrero, J., Guisasola, A., Comas, J., Rodríguez-Roda, I., Baeza, J.A., 2012. Multi- Pfrang, W., 2014. Metcalf & Eddy. Wastewater Engineering. Treatment and Resource criteria selection of optimum WWTP control setpoints based on microbiology-related Recovery, fifth ed. McGray-Hill Education. failures, effluent quality and operating costs. Chem. Eng. J. 188, 23–29. https://doi. Torregrossa, D., Hernández-Sancho, F., Hansen, J., Cornelissen, A., Popov, T., Schutz, G., org/10.1016/j.cej.2012.01.115. 2017. Energy saving in wastewater treatment plants: a plant-generic cooperative Guida, S., Potter, C., Jefferson, B., Soares, A., 2020. Preparation and evaluation of decision support system. J. Clean. Prod. 167, 601–609. https://doi.org/10.1016/j. zeolites for ammonium removal from municipal wastewater through ion exchange jclepro.2017.08.181. process. Sci. Rep. 10, 1–11. https://doi.org/10.1038/s41598-020-69348-6. Tsagarakis, K.P., Mara, D.D., Angelakis, A.N., 2003. Application of cost criteria for Hvala, N., Vrečko, D., Bordon, C., 2018. Plant-wide modelling for assessment and selection of municipal wastewater treatment systems. Water. Air. Soil Pollut. 142, optimization of upgraded full-scale wastewater treatment plant performance. Water 187–210. https://doi.org/10.1023/A:1022032232487. Pract. Technol. 13, 566–582. https://doi.org/10.2166/wpt.2018.070. Tzeng, G.H., Huang, J.J., 2011. Multiple Attribute Decision Making: Methods and Applications, Multiple Attribute Decision Making: Methods and Applications, 11 Ž. Južnič-Zonta et al. Journal of Cleaner Production 367 (2022) 132873 Lecture Notes in Economics and Mathematical Systems, vol. 186. Springer-Verlag Zaborowska, E., Czerwionka, K., Mąkinia, J., 2021. Integrated plant-wide modelling for Berlin Heidelberg. evaluation of the energy balance and greenhouse gas footprint in large wastewater Vaneeckhaute, C., Claeys, F.H.A., Tack, F.M.G., Meers, E., Belia, E., Vanrolleghem, P.A., treatment plants. Appl. Energy 282, 116126. https://doi.org/10.1016/j. 2018. Development, implementation, and validation of a generic nutrient recovery apenergy.2020.116126. model (NRM) library. Environ. Model. Software 99, 170–209. https://doi.org/ Zarei, M., 2020. Wastewater resources management for energy recovery from circular 10.1016/j.envsoft.2017.09.002. economy perspective. Water-Energy Nexus 3, 170–185. https://doi.org/10.1016/j. Vasilaki, V., Conca, V., Frison, N., Eusebi, A.L., Fatone, F., Katsou, E., 2020. A knowledge wen.2020.11.001. discovery framework to predict the N2O emissions in the wastewater sector. Water Zhou, Y., Stanchev, P., Katsou, E., Awad, S., Fan, M., 2019. A circular economy use of Res. 178, 115799 https://doi.org/10.1016/j.watres.2020.115799. recovered sludge cellulose in wood plastic composite production: recycling and eco- Xue, W., Zeng, Q., Lin, S., Zan, F., Hao, T., Lin, Y., van Loosdrecht, M.C.M., Chen, G., efficiency assessment. Waste Manag. 99, 42–48. https://doi.org/10.1016/j. 2019. Recovery of high-value and scarce resources from biological wastewater wasman.2019.08.037. treatment: sulfated polysaccharides. Water Res. 163, 114889 https://doi.org/ 10.1016/j.watres.2019.114889. 12