Extracting resources from wastewater
Wastewater has been described nowadays as both a resource and a problem by several authors. Since 2013, publications from Water Environment Federation (WEF) started using the term Water Resource Recovery Facility (WRRF) instead of WWTP, in order “to better focus on the products and benefits of treatment rather than the waste coming into such facilities” (WE&T, 2013).
In the past, the main idea behind wastewater treatment was the accomplishment of permitted reject limits. Nowadays, the goals are moving towards the maximization of recovery of valuable resources although water quality is maintained and robustness in process is achieved.
When conducting a balance across the potential energy that might be recovered from wastewater, Shizas and Bagley (2004) estimated that wastewater contains 9.3 times the energy required to treat it. However, this value might vary between 3.6 (when considering recoverable energy with recovery from organics and nutrients) and 13 (when including heat recovery) times higher than the energy required for treating it (Gude, 2015).
When considering nutrients, the most important part comes from urine. For instance, as discussed by Tidaker et al. (2007), the urine from Swedish population contains approximately 36 kilotons of nitrogen and 3.3 kilotons of phosphorus while 170 kilotons of nitrogen and 15 kilotons of phosphorus were consumed from mineral fertilizers in Sweden in 2001. Therefore, if nitrogen, phosphorus and even potassium in urine were utilised in substitution of synthetic fertilizers, the industrial production of those might be decreased and the discharge of those nutrients in water bodies would also be reduced (Johansson et al., 2000).
However, extracting resources from wastewater is not new. Some alternatives such as the production of fit-for-purpose water, the biosolids used in lands and the energy generation from anaerobic digestion are already common in WWTPs. Nowadays, new processes are starting to be studied in order to produce/recover specific compounds from wastewater such as precipitated struvite to be used as a slow-release fertilizer, the production of biologically-deriving polyhydroxyalkanoates from sludge and the use of digester gas to produce methanol or ammonia (WERF, 2011).
Also, the use of biosolids and the wastewater itself in farmlands is gaining more acceptances by population. While in developing countries the wastewater reuse is driven among others by the limited capacity to treat wastewater and the lack of other acceptable water sources for agriculture, in developed countries water reuse and recycling are increasingly seen as a means to respond to physical water scarcity and water reallocations from agriculture to other uses. Also, nowadays, all the strict environmental standards lead to land application of wastewater and sludge to be unavoidable and economically feasible (Drechsel et al., 2010).
Anaerobic digestion nowadays is a currently valorisation process in several WWTPs. In this process, as the redox potential is low, microbial fermentation takes place and converts organic material to biogas (mainly CH4 and CO2) that can thus be used to produce energy in cogeneration systems. Another consequence of anaerobic digestion is the mineralisation of biodegradable organic compounds that leads to high concentration of NH4+ and PO43- in solution and the stabilisation of the sludge (van Lier et al., 2008). Several studies have been conducted in order to improve biogas production aiming to invert the energy balance and lead WWTPs to be in energy autarky (Schaubroeck et al., 2015; Aichinger et al., 2015). Considering that biogas production is fully dependent of volatile suspended solids (VSS) entering the digester that will be degraded, one possibility of increasing biogas production is forcing the input of more organic matter into the digester. This can be done by recovering the organic matter entering the WWTP by an enhanced primary clarification (with an addition of chemicals to achieve better flocculation), previously to its degradation in the water line by the microbial oxidation. Another advantage of recovering organic matter at this point is that less bacterial growth to treat carbon will take place in the aerated basins and thus less energy will be required to aerate the activated sludge (Flores-Alsina et al., 2014). More recent research have been also developed for high rate activated sludge process (HRAS) – Jimenez et al., 2015 – working at low SRT and also for partial nitritation and anammox process on the mainstream (Laureni et al., 2016). If the coupling of the two previously mentioned approaches succeeded, this innovative plant would allow minimising the aerobic degradation of organic matter and producing more fermentable sludge for AD (Xu et al., 2015).
One of the promising opportunities in the phosphorus recovery is the struvite precipitation from wastewaters. Struvite (ammonium magnesium phosphate hexahydrate – MgNH4PO4 · 6 H2O) is a slow release fertilizer that might substitute conventional industrial fertilizers. The recovery of struvite from wastewater would allow not only a decrease in mineral fertilizers production and depletion of natural phosphorus resources but also a decrease in rejected phosphorus in water bodies together with the necessary production of crops to respond to population growth. However, in order to be feasible and present real advantages, struvite precipitation should be conducted in nutrient-rich wastewaters containing mainly phosphate and ammonium. Several studies have been conducted for the struvite precipitation using, among others, digester supernatant, landfill leachate and urine. However, as discussed by Maurer et al. (2006) without addition of phosphate, only 3% of nitrogen in urine might be eliminated in struvite precipitation (with an efficient precipitation of 98% of phosphorus).
As discussed previously, several recovery opportunities are available and well applied in WWTPs. However, currently, one of the bottlenecks to fully benefit from resource recovery in wastewater is that, in municipal wastewater, nutrient-rich streams (e.g.: urine and faeces) should not be diluted (e.g.: by greywater) and thus, source-separation should be encouraged. At the same time, the choice of centralized or decentralized systems is not evident: Among others, the management approach depends on the area (urban or rural), the size and density of the population, the development level, the technical feasibility, the treatment quality required, the education and culture of the population, being necessary to evaluate case-by-case (Corcoran et al., 2010; Libralato et al., 2012).
Evaluating benefits: Life Cycle Assessment (LCA)
Even when having data to evaluate conventional and new technologies, one could claim that depending on the methodology used to assess the process, results would be different. There is thus a need for a standardized methodology to evaluate the whole process and its requirements. Currently, different assessment tools are available to evaluate the sustainability of systems such exergy analysis, economic analysis and Life Cycle Assessment (LCA) (Balkema et al., 2002).
Aiming to analyse environmental footprint of WWTPs, several studies have been suggesting the use of LCA. It is a methodological framework for assessing the environmental impacts attributable to the life cycle of a process or a product. The idea of using LCA in WWTPs is to account for all background process (e.g. energy consumption, chemicals production and transport utilization) besides the foreground process which has his own environmental emissions.
According to ISO 14044 (2006), LCA is defined as the “compilation and evaluation of the inputs, outputs and potential environmental impacts of a product system throughout its life cycle”. Considering this, when analysing conventional and alternative WWTPs for long-term environmental sustainability, effluent quality discharge impact has to be considered (end-of-pipe approach) but also all the processes associated to the main treatment such as the sludge treatment and disposal, energy consumption and production of ancillary materials (background processes).
Also, the four main steps that are recommended by the ISO 14044 (2006) are to be followed: i.) the goal and scope definition; ii.) the inventory definition; iii) the impact assessment phase and iv.) the interpretation.
The first step, the goal and scope definition, allows the description of the system in terms of the system boundaries, function and functional unit (FU), and allocation methods. It is important as the correct definition of the FU allows the latter comparison between alternatives. The second step, the Life Cycle Inventory (LCI) is the compilation of all estimated consumption of resources from the environment and emitted substances to the environment during the process/product life cycle. By the end of this step, an inventory of the system is obtained based on a well-defined functional unit. Following, the third step, Life Cycle Impact Assessment (LCIA) provides the correlation between emitted substances and indicators of impacts on the environment. Finally, the last step, the life cycle interpretation occurs naturally as when conducting the LCA one wants to take decisions after comparing options (Rebitze et al., 2004) and identifying the hot spots of the system.
In the field of water and wastewater treatment, LCA has already been used in several studies to evaluate the environmental performances of proposed technologies. Different LCA applications have been published so far (Corominas et al., 2013a) for different WWTPs configurations as well as for sewage sludge management technologies (Yoshida et al., 2013) and for the full urban water system (Loubet et al., 2014). However, as results are usually obtained by site collected data, it can neither be used to automatically analyse general trends nor to process optimization. Therefore, the consideration of an LCA together with WWTP modelling and simulation tools is be a powerful approach to allow the modification of operational and design parameter when aiming to conceive more sustainable systems.
However, it has to be highlight that, in order to be realistic and to provide a fair comparison against studied scenarios, LCA have to be conducted considering the appropriate boundaries and the proper allocation methodology. For instance, when analysing WRRFs, the conventional WWTP Life Cycle Assessment has to be adapted to account for all avoided impacts generated by the production of by-products. Similarly, the new functions of the system have to be added as it will not deliver the same functional unit anymore.
Another important feature of the LCA that allows it to be applied to the quantification of impacts and the evaluation of scenarios in WWTPs considers the main emissions of the plant. Even if nowadays, life cycle impact assessment methodologies are not capable of integrating all substances leaving a WWTP, such as personal care products and medicaments residues, the main burdens of the implementation of a wastewater treatment (for instance, the reject of non-treated pollutants, the greenhouse gases emitted and the high consumption of energy and chemicals) can be correctly quantified and evaluated by the currently available impact categories (for instance, marine and freshwater eutrophication, human toxicity, climate change and resources depletion).
Improving systems: Multi-objective optimization
When a novel technology is able to be evaluated considering sustainability but also technical and economic aspects, an optimization can be conducted. Additionally, it may consider not only one optimal functioning point; it is possible to compromise between all trade-offs without any preliminary judgement.
As described by Deb (2011), multi-objective optimization (MOO) consists of optimizing more than one objective simultaneously. As opposed to the single objective optimization, the multi-objective optimization minimizes all usually conflicting objective functions simultaneously without using expressions of weight between objectives. Hence, a set of solutions, called Pareto-optimal solutions, is usually obtained by the end of a multi-objective optimization. The concept of domination is generally used in the context of multi-objective optimization to discriminate between solutions and to locate the globally non-dominated (minimum) solutions, while maintaining the diversity in a given Pareto-front (Deb, 2001). A further processing step, the decision making process, will be therefore required in order to comprise between the trade-offs and to find one optimal functioning point, when needed.
Additionally, as conflicting objectives are most often involved, none of the optimal solutions found can be improved without worsening at least one of the other objectives (Hakanen et al., 2011) and thus, solutions cannot be easily sorted only on their objective value.
There exist different ways to solve a multi-objective optimization problem among which evolutionary algorithms (EA), widely known due to their robustness. According to EAs, the optimization is carried out by using a population of solutions, usually created randomly and therefore robustness is ensured independently from the quality of initialization. Thereafter, the algorithm provides a generation-based (iterative) operation updating the current population to create new populations based on genetic operations such as genetic selection, crossover, mutation and migration. This generation-based operation is pursued until one or more pre-specified termination criteria are met (Deb, 2011).
From the application point of view, the optimization of WWTP design and operation has been applied since 1990s. However, commonly, the optimization strategies, described in the literature, are most often aggregation-based, which is to say that the optimization is conducted by aggregating several objectives into a unique objective function through weight factors representing the importance of each objective (Hakanen et al., 2011; 2013).
However, as described by Hreiz et al. (2015) in a review of ASPs optimization, several objectives have not the same units (and sometimes also contradictory) and cannot be instinctively combined in order to form a single objective function.
From a practical point of view, even if highly non-linear processes are present in a WWTP due, for instance, to the rigorous consideration of biochemical reactions, it is important to avoid the aggregation of objectives and preserve the intrinsically multi-objective structure of optimization problem. For instance, the quality of treated wastewater and the operational costs are two conflicting objectives as, reaching low organic matter, ammonium or phosphate concentrations in effluent leads to high consumption of energy and chemicals in the plant. Also, operating a plant with short SRT risks the stability of nitrification and may produce an excess of sludge; however, suspended solids quality and BOD is ensured (Hakanen et al., 2013) and thus only the operator expertise might compromise between different conflicting outcomes that are sometimes non quantifiable or highly dependent.
Multi-objective optimization is recognized to be more suitable to deal with WWTPs due to the conflicting nature of objectives taken into account. Additionally, the use of derivative-free algorithms such as EAs is favoured to avoid uncertainties due to the numerical approximation of gradients in highly-linear systems, and to ensure the robustness of the algorithm. Nevertheless, regarding the EAs, the main drawback is the considerable numerical budget required. However, the full WWTP case study is an example of an expensive optimization problem, where the resolution without an efficient optimization tool would be practically impossible.
The integrated DM-LCA methodology
As mentioned above, the integration of the dynamic modelling approach and LCA tools is a prerequisite when trying to analyse the total environmental footprint of a WWTP system.
The DM-LCA approach developed here used three different platforms, interconnected as shown in Figure II.1. WWTP scenarios were simulated with BioWin® v220.127.116.116, a Windows-based wastewater treatment process simulator that includes biological, chemical, and physical processes (Envirosim, 2014). The interface between WWTP dynamic modelling and LCA calculations were performed through Python™ scripts.
To achieve the study objectives, model parameters were fixed initially and dynamic influent data was provided to the simulator (Fig. 1 data flow 1). Dynamic simulations were also designed to reach effluent quality limits (e.g. 10 g.m-3 of total N, 1 g.m-3 of total P, 35 g.m-3 of total suspended solids, 100 g.m-3 of total COD and 4 g.m-3 of ammonium ion). As a result of the dynamic simulation, process inventories (Fig. 1 data flow 2) were generated with their own inputs and outputs. After the dynamic simulations, Python™ scripts (Fig. 1 data flow 3) integrated the results over the simulation time. All parameter values and examples of calculations can be found in the Supplementary Information document (SI, Section 1).
The results were then converted to an Umberto®-compatible input file for foreground and background processes. Python™ scripts also performed complementary calculations, based on the literature (e.g. calculation of cogeneration and energy requirements) and adjusted assignments between the output flows resulting from the BioWin® simulation and Umberto® input flows (Fig. 1 data flows 4, 5, 6 and 7).
LCA calculations were then performed with Umberto® (Fig. 1 data flow 8) using the Ecoinvent database (Fig. 1 data flows 9 and 12). This last step completed the LCI by adding the contribution of background processes to WWTP ones (inventory details in section 2.4.2), and calculated the LCI (Fig. 1 data flow 10) then the environmental impacts (Fig. 1 data flow 11). Three main types of results were generated: effluent concentrations and quality violations (Fig. 1 data flow A), energy parameters and nutrient recovery (Fig. 1 data flow B), and environmental impact results (Fig. 1 data flow C).
Life cycle inventory
The inventory took account of all flow types proposed by the reference study included in the Ecoinvent 2.2. database (Doka, 2009) and also some others judged sensible in our case, such as a cogeneration system with electricity and heat production, external carbon source addition and production, fertilizer production from urine and its utilization.
For the direct gas emissions of carbon, it is important to emphasize that, in order to be consistent with IPCC (2006) guidelines, all organic carbon in sewage was considered to be biogenic. However, to achieve complete denitrification, it was necessary to use methanol (produced from natural gas) in some scenarios. Thus, there was a percentage of produced carbon dioxide (CO2) that originated from a fossil source. Emissions of N2O from WWTPs are considered to be 0.5% of ammonia nitrified flows in dynamic conditions (Czepiel et al., 1995). The volume and composition of offgas were calculated (from Biowin® software) based on gas/liquid transfer models. Calculations were based on transfer coefficients and concentration gradients with atmosphere. For anoxic reactors the volume of gas emitted was mainly related to the dinitrogen produced by denitrification which was calculated with the transfer surface. Heavy metal concentrations are not taken into account by Biowin as they are considered to be inert for biological processes. Their input concentrations in WWTPs were therefore taken from Doka (2009) and Henze and Ledin (2001) and allocated to effluent and sludge in specific quantities, using their specific transfer coefficients proposed by the same authors.
The amounts of chemicals consumed (FeCl3 for P precipitation (coagulant), methanol, MgO, NaOH) that would consequently need to be produced in background processes were calculated according to simulation demands. The total amount of FeCl3 required was calculated considering it to be used for both P precipitation and biogas purification (to avoid H2S formation). Grit removal in pre-treatment was also included in LCA, considering 31g of grit to be present in 1 m3 of raw sewage (50% as plastics and 50% as paper to be disposed of in a municipal waste incinerator; Doka, 2009).
The electricity consumption was calculated by taking the sum of all electricity requirements (aeration of AER and NITRITATION tanks and THK, mechanical mixing of ANOX, ANAMMOX and AD tanks, pumping of main lines – influent input, dosing of chemicals, sludge outputs, recirculation lines, and effluent output – scrapping and dewatering unit) and subtracting the electricity produced in the cogeneration unit. Electricity that was produced was consumed by the WWTP itself and, when more electricity was produced than consumed, it was considered to be injected into the network and the avoidance of electricity production was calculated. Heat production was calculated in a similar manner to electricity and it was used to heat sludge to be digested and to compensate for heat losses by AD walls. Process sludge was firstly preheated in countercurrent flow against digested sludge and the heat transfer was complemented by heat generated in the cogeneration unit (HE2). Details on the energy balance are given in SI Section 3.
Table of contents :
Chapter I. General Introduction
I.1. Wastewater as a pollution stream
I.2. Wastewater as a resource recovery opportunity
I.2.1. Extracting resources from wastewater
I.2.2. Urine source separation
I.3. Quantifying benefits: Wastewater treatment modelling and simulation
I.3.1. Obtaining data: Influent generation
I.4. Evaluating benefits: Life Cycle Assessment (LCA)
I.5. Improving systems: Multi-objective optimization
I.6. Research objectives and tasks
I.6.1. Thesis outline
Chapter II. Coupling Dynamic Modelling and LCA
Entitled of the paper: Evaluation of new alternatives in wastewater treatment plants based on Dynamic Modelling and Life Cycle Assessment (DM-LCA)
II.2. Materials and methods
II.2.1. The integrated DM-LCA methodology
II.2.2. Plant layout and scenarios
II.2.3.1. Goal & scope
II.2.3.2. Life cycle inventory
II.3. Results and discussion
II.3.1. Reference scenario
II.3.2. Results of alternative scenarios: nutrient recovery, efficiency and energy consumption
II.3.3. Results of alternative scenarios: environmental impacts
II.3.3.1. Endpoint impacts
II.3.3.2. Midpoint results
Chapter III. Influent Generator
Entitled of the paper: A dynamic influent generator to account for alternative wastewater management: the case of urine source separation
III.2. General overview
III.3. Flow generation
III.4. Pollutants generation
III.4.1. General aspects
III.4.2. Composite variables
III.4.3. Fractionation into state variables
III.5. Noise addition
III.6. Example of simulations obtained with the generated influent
III.7.1. Average results
III.7.2. Daily and weekly profiles
III.8. Use of the generated influent for process simulation
III.8.1. Comparison between different models
III.8.2. Effect of urine retention levels
Chapter IV. Feasibility of Multi-Objective Optimization
Entitled of the paper: Feasibility of rigorous multi-objective optimization of wastewater management and treatment plants
IV.2. Materials and methods
IV.2.1. Dynamic Modelling (DM) approach
IV.2.2. Life Cycle Assessment (LCA) approach
IV.2.3. Efficient Multi-Objective Optimization (EMOO)
IV.2.3.1. Problem formulation: objective functions, constraints and range of decision variables
IV.2.3.2. Expensive optimization algorithm and general settings
IV.2.4. Integrated framework: DM-LCA-EMOO
IV.3. Results and discussion
IV.3.1. WWTP optimization and objectives’ dependencies
IV.3.2. Drivers for an optimal treatment
IV.3.3. Problem formulation and computational feasibility of multi-objective optimization
Chapter V. Case Studies on Multi-Objective Optimization
Entitled of the paper: LCA based multi-objective optimization of conventional and innovative wastewater management and treatment scenarios
V.2. Materials and methods
V.2.1. Dynamic Modelling – Life Cycle Assessment – Efficient Multi-Objective Optimization (DM-LCA-EMOO) coupling approach
V.2.1.1. Dynamic modelling (DM)
V.2.1.2. Life Cycle Assessment (LCA)
V.2.1.3. Efficient Multi-Objective Optimization (EMOO)
V.2.2. WWTP scenarios (conventional vs. innovative)
V.2.3. Optimization problem formulation
V.2.3.1. Objective function
V.2.3.2. Decision variables and constraints handling
V.2.3.3. Recall on optimization problem formulation
V.3. Results and discussion
V.3.1. Reference scenario
V.3.1.1. General results
V.3.1.2. Analysis on decision variables
V.3.1.3. Steady state versus dynamic modelling approach
V.3.1.4. Total Endpoint versus Midpoint Global Warming Potential
V.3.2. Alternative scenario (ANA)
V.3.3. Consequences on energy autarky
V.3.4. Benefits achieved from REF and ANA scenarios
Chapter VI. Conclusions and perspectives