**1. Introduction**

The oil sector continues to be the main primary energy source, contributing with about 35% of global fuel consumption in 2017 [1]. Meeting the demand for oil products, mainly in the industry and fuel sectors, requires large amounts of water, mostly for thermal exchange. This scenario is mostly observed in the countries with the largest volume of oil manufactured goods production, such as Brazil, which occupies the 9th position in the world rank [1]. However, due to Brazil's local reality—of irregular water density and, therefore, scarcity of this natural resource in large urban centers at certain times of the year, several regulatory agencies have increased their restrictions on water collection and effluent disposal, in order to mitigate the depletion of the national water network. This scenario has stimulated recent research aimed at increasing the efficiency of effluent treatments and, consequently, decrease damage to the extraction and disposal sites of this resource.

Water reuse is an approach that has been gaining ground in this area, consisting in treating part, or even, when possible, the whole effluent and subsequently destining it for the supply of industrial or

domestic needs. This practice can equate water demand problems. Nevertheless, water reuse can also lead to environmental impacts as energy resource consumption (fluid heating and pumping), from the synthesis of process inputs and waste generation due to the water reconditioning.

Current literature reports varied contributions on water reuse practices. When the investigations addressed the environmental domain, approaches were once again diversified. Chang et al. [2] evaluated energy use and greenhouse gas (GHG) emissions for urban water reuse systems in Korea, while Hendrickson et al. [3] evaluated the same parameters for an operating Living Machine (LM) wetland treatment system, which recycles wastewater in an office building. The study also assessed the performance of the local utility's centralized wastewater treatment plant, which was found to be significantly more efficient than the LM. The life-cycle approach was adopted for cases in which the specification of the environmental variable required a systemic scope [4–11]. Morera et al. [4] applied the Water Footprint to assess the consumption of water resources in wastewater treatment plants (WWTP). Their findings indicated that the choice for WWTP systems leads to a significant decrease in grey resources but, on the other hand, also generates a small blue water footprint in comparison to no-treatment scenarios. Cornejo et al. [5] carried out a comprehensive literature review of carbon footprint (CF) reports from water reuse and desalination systems. The study recognized general CF trends associated with the technologies and recommended improvements to this method based on limitations, challenges and knowledge gaps identified throughout the analysis.

The life cycle assessment (LCA) itself was applied to identify impacts resulting from control strategies in wastewater treatment plants [6], verify the environmental performance of large [7] and small scale [8] WWTP, and compare decentralized wastewater treatment alternatives for non-potable urban reuse [9]. LCA was also used to subsidize environmental and economic assessments performed to determine optimal water reuse in a residential complex [10] and develop an eco-efficiency analysis (EEA) framework for the evaluation of treatment systems proposed for greywater recycling in domestic buildings for non-potable uses [11].

Baresel et al. [12] quantified the environmental effects of using the energy grids from different countries (USA, Spain, and Sweden), as well as the option of applying sludge resulting from effluent treatments as fertilizer within water reuse systems for agriculture and industry. Their findings indicate that altering external treatment aspects can impact process performance, compared to seeking out improvements for a given technology.

O'Connor et al. [13] evaluated 14 different process arrangements for the treatment of a pulp and paper mill effluent by the commutative use of six operations: flotation, clarification, use of activated sludge, an upflow anaerobic sludge blanket (UASB) reactor, ultrafiltration, and reverse osmosis (RO). The study assessed climate change (CC), freshwater ecotoxicity (FEC), eutrophication (Eut) and water recovery impacts. The results demonstrate the significant influence of the solid waste originating from the treatment and the use of electricity in CC impacts, as well as the higher RO efficiency for the quality of the reuse water compared to the ultrafiltration technology.

On the other hand, no optimal scenario for all evaluated categories has been determined. Pintilie et al. [14] tested the feasibility of reuse practices in the industrial sector regarding effluents recovered by a treatment plant in Spain, concluding that electricity consumption displays the greatest weight concerning the environmental impacts and that water reuse can be an adequate alternative for non-potable uses, such as applications in the industrial sector.

To the best of our knowledge, however, no studies in the literature are available regarding LCA application in the evaluation of technologies applied for water reuse in the closed looping process itself. This study contributes to the theme, verifying the environmental performance of different scenarios conceived to conditioning the treated effluent from an oil refinery located in Brazil, so that it can be reused in processes within the facility itself. To pursue this aim, attributional LCA was applied according to a 'cradle-to-gate' approach.

Apart from identifying which reuse treatment bottlenecks should be analyzed in further detail in order to reduce environmental impacts, such an analysis can subsidize information to water resources managements in situations in which water collection is prohibited by legal regulations.

Environmental performance has the potential to become a management criterion under reuse practices. In addition to its close correlation with economic aspects, the adoption of this approach can support optimization actions in procedural arrangements, and for technology selection.

#### **2. Materials and Methods**

The method applied in this study encompasses six steps: (i) specification of the effluent quality upstream from the industrial wastewater treatment plants (IWTP) in the refinery, and the choice of a process use for which recovered water is intended for; (ii) definition of water recovery strategies and setting the analysis scenarios; (iii) description of the recovery systems in terms of their technological approach and operational conditions, as well as resource consumption and emissions; (iv) designing of mathematical models to represent each system, from the data and information obtained in the previous step; (v) application of the LCA technique to establish an environmental diagnosis for each scenario concerning primary energy demand and global warming; and (vi) perform a critical review of the obtained results.

#### *2.1. Effluent Specification and Destination of Recovered Water*

The refinery effluent is submitted to a conventional treatment at the IWTP. The primary step consisting of an API oil–water separator and a floating filter removes suspended solids, oils, and greases. The API separator is an equipment used to separate gross amounts of oil and/or suspended solids from the effluents of oil refineries. Its design is based on the specific gravity difference between the oil and the wastewater, which is smaller than that one between suspended solids and water. Thus, the suspended solids will settle to the bottom of the separator, the oil will rise to top, and the wastewater will occupy the middle layer, being recovered separately from the other components [15]. Dissolved solids that assign organic load to the effluent are treated during the secondary step, comprising aeration ponds (complete and facultative) and biodisks.

Finally, in the tertiary step, contaminants that result in color and conductivity, as well as nutrients, metals, non-biodegradable compounds, and volatile suspended solids, are removed from the liquid stream by passing through a clarifier and an activated carbon filter. Table 1 shows the characteristics of a typical effluent from an IWTP. The IWTP output effluent characteristics can vary significantly depending on the type of oil processed, the refinery configuration, and the operating procedures used in the treatment [16]. On the other hand, such information is not easily accessible. Because of this, Table 1 shows concentration ranges of contaminants present in that stream. The limits of each interval were established from data of Gripp [17], Moreira [18], and Pantoja [19] for refineries of the same technological concept but at different periods.

The effluent presents salt characteristics due to the earlier IWTP treatment, where suspended solid, greases, oils and organic loads have been removed. As the scope of this study begins precisely at this point, the environmental IWTP performance was disregarded.

The process adopted to determine the purpose of the reclaimed water took into account two criteria: (i) volume consumed by productive sector or activity; and (ii) water quality restrictions for each type of use. The worldwide reference for specific water consumption in refining processes ranges from 0.7–1.2 m3 water/ m3 processed crude oil [20,21]. Regarding Brazilian plants, this demand is distributed as follows: 46% replace losses occurring in cooling towers; 26% serve the steam production in boilers; 9.0% are allocated to fire-fighting systems; and the remaining (~19%) is absorbed in chemical preparation (dilution), cleaning, and for human consumption [22].

The boiler feedwater requires the most restrictive quality standards in terms of salt, organic matter, and dissolved gas concentrations. The cooling water makeup predisposes intermediate quality levels; in such cases limits are established in order to regulate (or prevent) scale development, corrosion, and

slime and algae formation. In general, water used for firefighting does not require any treatment [23]. Considering these arguments, in addition to water recovery system performance regarding the quality of the final product and the effectiveness with which such thresholds can be met, the feed from one of the refinery's cooling towers was established for the final use of the desalinated water.

#### *2.2. Setting the Analysis Scenarios and Description of the Recovery Systems*

The effluent quality at the outlet of the tertiary IWTP step (Table 1) obligates a complementary removal of metals (Ca2+, Ba2+, and Na+), chlorides (Cl∓), and carbonates (CO3 <sup>2</sup>−) to reconditioning this flow as makeup water in the cooling tower. In this regard, six water recovery arrangements were defined. RO, evaporation (EV) and crystallization (CR) operations are common to all strategies. In contrast, these vary in how the effluent is pre-treated before entering the RO and due to the technology adopted in the EV process.


**Table 1.** Characteristic values of indicators in effluent. IWTP: industrial wastewater treatment plants.

<sup>1</sup> TDS: total dissolved solids.

Based on precipitation principles, pre-treatment aims to remove ions that may reduce, or even compromise, RO performance. The definition of the chemical agents applied in this stage took into account their removal potential, through mechanisms that affect ion solubility in aqueous solutions. Regarding the EV technology, two alternatives were investigated: (i) multi-effect distillation, and (ii) steam recompression. Each recovery strategy resulted in an analysis scenario, coded as S1 to S6. Table 2 presents some specificities for each scenario regarding technological conditions and the origin of the energy supply.

S1 was defined as the baseline scenario and, therefore, no pretreatment technique was charged for this situation. Due to the high ion concentrations in the effluent, a risk for encrustations in the osmosis membrane is noted, making it necessary to add anti-fouling reagents to the liquid effluent at the entrance of the RO module. In S1, this situation was addressed by dosing EDTA, often applied in such cases. However, EDTA excesses can cause side effects, i.e., biofouling, harmful in extreme circuit closure situations, such as the Zero Liquid Discharge regime [24].

In S2, pre-treatment occurred by desupersaturation through the addition of BaSO4 seeds (BaDs). This strategy aims to remove barium sulphate itself through its accumulation in the presence of the crystals formed by this salt [25]. BaSO4 is commonly found at supersaturation concentrations

in industrial effluents (particularly in oil refineries), and is one of the main fouling agents in RO plants [26].


**Table 2.** Technical characteristics and energy sources for each scenario.

<sup>1</sup> BR grid: Brazilian electricity matrix; <sup>2</sup> NG: natural gas; <sup>3</sup> WH: waste heat.

The coprecipitation method (CPT) by alkalization to pH = 11 applying Ca(OH)2 and CaCO3 seed addition [24] was adopted in S3 for pre-treatment. This allows for the precipitation of not only a large part of the calcium carbonate in solution, but SiO2 as well, which is also a limiting component in RO, due to its decreasing solubility with increasing acidity levels (pH). CPT may also lead to the removal of metal ions—Ba2+, Sr2+, Fe2+, Mn2+, and Cd2+—by incorporation into the precipitating CaCO3 via isomorphic adsorption, absorption or substitution [27]. The supernatant arising from the CPT stage should be acidified to pH = 8.0 to prevent scales mainly in the RO system. In S3 and S5, this occurs via the addition of HCl (1.0 M) and in S6, with HNO3 (1.0 M).

S4 assessed the situation in which the thermal energy consumed by EV and CR results from waste heat. In this case, part of the heat released to the atmosphere from the boiler chimneys, or even by the cooling tower, is reused by the desalination unit. This option is economically attractive as it reduces heat generation costs and waives the use of RO—and, therefore, of also any pre-treatment method—by the system [28].

S5 examined a substitutive arrangement of S3, in which the energy demand of EV is supplied exclusively by electricity. This scenario assumes a technological change in the evaporation stage, from which the multi-effect distillation, adopted regularly by scenarios S1–S4 and S6, is replaced by steam recompression [29]. This technology applies steam produced in the evaporator itself in order to lead the evaporation phenomenon at that equipment. For this purpose, a compressor driven by an electric engine is used to raise the functions of state (temperature and pressure) of the steam. In addition to evaporating some of the water existing in the saline solution, the vapor stream is also used to increase the temperature of effluent fed in the evaporator [30].

Since electrolysis, a technology conventionally adopted to obtain HCl, requires high electricity consumption [31], a variant of S3—coded S6—was also investigated, in which the acidification of the supernatant leaving CPT was carried out by the addition of HNO3, as described previously. The use of nitric acid becomes viable because the ion NO3 − is, with a few exceptions, quite soluble in water. Therefore, its presence in the effluent does not revert to the risk of fouling in the subsequent treatment stages.

Finally, it should be noted that effluent pretreatment by BaDs increases the water recovery rate by 2.8% at the RO stage (*ηS*2) in comparison to that observed in the baseline scenario. In cases where the action occurs via CPT (S3, S5 and S6) the efficiency gain is even more expressive, surpassing *ηS*<sup>1</sup> in about 13%.

#### *2.3. Mathematical Model Design*

The Hydranautics® 'IMS Design' computational tool was used to model the RO unit. The design of this stage considered an average limit flow of *Q* = 20 Lmh, maximum limiting polarization concentration (*βmax* = 1.18), and a standard high saline rejection membrane CPA7 MAX. It was also assumed that RO pumps achieved average yields *η* = 87%. Except for S5, all the assessed scenarios adopted the multi-effect distillation technology for the EV stage. In such cases, a heat consumption of 230 MJ/m<sup>3</sup> of effluent to be treated was required to place the steam under process conditions, and 9.00 MJ/m<sup>3</sup> of electricity were employed in pumping operations [31]. In S5, evaporation occurs by steam recompression, therefore expending only electricity. A unit that operates with this technology achieves a consumption of 115 MJ/m3 [28]. Regarding the CR stage, the effluent flow fed to the desalination system (250 m3/h) also justifies the use of steam recompression [32]. The energy demand related to the crystallization technology is of 238 MJ/m3 [29]; this performance was evenly considered for modeling all scenarios.

The waste heat that meets the thermal demand of S4 was modeled as an elementary flow. In methodological terms, this decision represents the elimination of all the environmental burdens associated with this flow (and the impacts arising from it) generated by anthropic interventions prior to its use by the desalination system. This is supported by the hypothesis that heat, one of the crude oil refining residues, is revalued in S4, when it becomes a process fluid. For EV, the energy decrease assumed in other scenarios (9.00 MJ/m3) was also adopted for S4. In contrast, the modeling of CR stage admitted that fluid pumping would consume about 576 MJ/t of obtained salt. This value was estimated by Jongema [33] from a thermodynamic approach based on an exergy analysis.

The precipitation estimations were performed using an OLI Systems Inc. simulator®, which provides the approximate compositions of the formed salts. For S1, this calculation included an additional rate of 4.0 ppm EDTA. In S2, it was assumed that the exchange of BaSO4 seed is performed once a year (C ~ 10 g/L). Moreover, a consumption of 79.2 kJ/m<sup>3</sup> for agitation in the precipitation stage and a monthly cleaning of the equipment with HCl solution (pH = 3.0) were also considered.

S3 includes an electricity consumption of 317 kJ/m<sup>3</sup> from shaking in the coprecipitator, the insertion of 12.5 g/L CaCO3 seeds (annual renewal) and the acidification of the supernatant obtained from the precipitation stage (addition of HCl, 1.0 M) to achieve pH = 8.0, in order to prevent alkaline incrustations.

The transports considered in the models were the piping that leads the recovered water from the desalination plant to the cooling tower (L ~ 200 m) and the salt from the precipitation and crystallization stages to a landfill located 20 km from the refinery. Finally, the Brazilian energy matrix (BR grid) was the exclusive source of electricity supply from the desalination system in all analyzed scenarios.

#### *2.4. Life-Cycle Modeling*

#### 2.4.1. Scope Definition

The environmental diagnoses were performed by attributional LCA, under a 'cradle-to-gate' approach and in line with the guidelines provided by ISO 14044 standard [34]. A reference flow (RF) of 'add 1.0 m<sup>3</sup> of reused water to a cooling tower within quality requirements that allow the proper functioning of this equipment' was defined to carry out the analyses.

Figure 1 displays an overview of the set of elements, highlighting all the interconnections among the refining process, the IWTP, and the water recovery system. The product system that represents scenarios S1–S4 and S6 is contained in the dark gray rectangle. Figure 2 details the same arrangement for S5. Modeling of the Product System was based on primary and secondary data, that express local conditions. No multifunctionality was found.

**Figure 1.** General view of effluent recovery systems (S1–S4 and S6) for reuse in a cooling tower.

**Figure 2.** Detail of the desalination system for scenario S5.

The life cycle impact assessment was carried out in two levels. First, the energy consumption of the process in terms of primary energy demand (PED) was quantified by the cumulative energy demand (CED) method—v1.10 [35], which addresses the contributions of the different energy sources, both renewable (biomass: RB; wind: RWD; water: RWA) and non-renewable (fossil: NRF; nuclear: NRN; biomass: NRB). In the second level, the environmental effect of global warming (GW) was calculated by the ReCiPe 2016 Midpoint (H) method—v1.13 [36].

#### 2.4.2. Life Cycle Inventory (LCI)

The effluent concentrations at the input of the process train (Table 1) are primary data obtained from a refinery. In contrast, electric energy and natural gas (NG) supplies were modeled from secondary data. The electricity generation (BR grid) was edited for Brazilian 2015 conditions [37]. Hydropower still remains the most expressive source of energy supply (64% of the BR grid) and biomass (8.0%) was expressed by sugarcane bagasse. For coal (4.0%), a mixed supply model (national and Colombian) was created considering mining operation and the distances of the Brazilian mines to the main thermoelectric plants [38]. For the natural gas from the BR grid (13%) and for heat supply, a model obtained from the Ecoinvent database was tailored to Brazilian conditions considering local offshore extraction, activities involved in refining raw natural gas and transport of final product.

In Brazil, inland transport occurs, mainly, by road. The diesel consumed in these actions was customized from the LCI 'Crude oil, in refinery/US' available at the USLCI database, [39] once again considering the procedural and technological requirements practiced in the country. The inputs and auxiliary materials, such as EDTA, barite and calcium hydroxide were edited from the Ecoinvent datasets entitled 'EDTA, ethylenediaminetetraacetic acid, at plant/RER U' [40], 'barite, at plant/RER' [41] and 'lime, hydrate, loose weight at plant/RER' [42] to more closely reproduce the national reality.

The same approach was applied to HCl production (adjusted from the Ecoinvent: 'Hydrochloric acid, 30% in H2O, at plant/RER U') [31], HNO3 (whose LCI 'nitric acid, 50% in H2O, at plant/RER U' existing in Ecoinvent was adapted) [31], and water (converted from Ecoinvent: 'Tap water, at user/U') [31] used for dilution of concentrated acids to 1.0 M. As in the case of diesel, the original energy inputs—electricity, NG, the diesel itself and other petroleum derivatives—were replaced by inventories created specifically for this study.
