**3. Results and Discussion**

Table 3 describes the environmental PED and GW impacts of the systems included in the S1–S6 scenarios, generated for the supply of 1.0 m<sup>3</sup> of recovered water for a cooling tower at the refinery. S4 displayed the lowest impacts of the entire series for both categories, mainly due to the use of waste heat as a thermal energy source for the system. On the other hand, S1 (baseline scenario) displayed the worst performance indices in the same analyzed dimensions, due to the absence of effluent pretreatment admitted for RO. This condition overloads the osmosis system, which then presents reduced water recovery rates (Table 1), and causes all the elements of the arrangement to require greater heat and electricity amounts in order to generate water within the quality standards required by the cooling tower. The adoption of pre-treatment methods improved the PED performance of all scenarios. The use of BaDs (S2) led to a 13% reduction in impact with respect to S1, whereas the gains were even more expressive with CPT, reaching 55%, 66%, and 54% attenuation, respectively for S3, S5 and S6.


**Table 3.** Impacts related to primary energy demand (PED) (by subcategory and total impact) and global warming (GW) for scenarios S1–S6.

The technological change applied to EV also brought benefits to the category, since the accumulated PED for S5 was 24% lower than for S3. On the other hand, the use of HNO3 (S6) instead of HCl (S3) did not generate significant energy behavior variations.

This exchange in positions is justified because S1–S3 and S6 present finite demands for thermal energy in EV (Table 4), which are supplied by natural gas, while S4 uses more heat during the EV process than the other scenarios (230 MJ/RF). However, due to waste heat, this utility does not contribute to PED impacts. S5 does not make use of thermal flows during this stage of the process. On the other hand, S4 and S5 present the highest electric consumption rates among the evaluated scenarios (13.1 and 10.4 MJ/RF). The BR grid, predominantly hydroelectric energy, meets these demands [37], with RW as the impact agent.


**Table 4.** Total energy consumption per stage of the system for each scenario (in MJ/RF). RO: reverse osmosis; EV: evaporation; CR: crystallization.

As depicted in Table 4, natural gas consumption for heat generation in S1/EV constitutes the main NRF contribution source, and thus, PED for this scenario. The effluent pretreatment raises electrical S2/RO consumption in 19% in relation to S1. In contrast, the use of BaDs makes *ηS*<sup>2</sup> > *ηS*<sup>1</sup> (Table 2), leading to a 16% decrease in thermal (heating) and electric (pumping) EV demands, relative to evaporation in S1. The combination of these effects is favorable for S2, since, as it uses lower amounts of natural gas, displays NRF contributions about 15% lower than those observed for S1. Since the contributions to the other subcategories are equivalent for both scenarios (Table 3), *PEDS*<sup>2</sup> overcame *PEDS*<sup>1</sup> in a little over 13%.

S3 follows similar trends as those observed for S2. In this case, however, the fact that *ηS*<sup>3</sup> is higher than *ηS*<sup>1</sup> due to CPT, leads to higher electric consumption in RO compared to that expended by S1, while also reducing EV energy demands even more intensely. Thus, natural gas consumption was dampened to sufficient levels so that the total *NRFS*<sup>3</sup> was lower than *NRFS*<sup>1</sup> in 64%, which immediately resulted in a 55% decrease in *PEDS*<sup>3</sup> compared to the baseline scenario. However, a closer examination of S2 and S3 RW results or indicates an 11% increase in this source of impact due to the replacement of BaDs by CPT. This disparity can be explained by the environmental loads associated with the HCl manufacture.

Electrolysis is an energy intensive technology and the most widely used methods for synthesis of Cl2. The manufacture of Cl2 from sea water consumes from 414 to 500 kJ/kg Cl2 of electricity [31]. Therefore, the manufacture cycle of HCl displays a contribution of 757 kJ/RF, corresponding to 14% of the entire RW value for S3.

The waste heat option reduced *PEDS*<sup>4</sup> concerning impacts generated on the account of the electrical demand of the system by the BR grid. Therefore, it was already expected that RW impacts would represent the largest share (54%) of the total. Regarding the process, the individual electricity expenditure to pump the effluent to EV (6.23 MJ/RF) is highlighted.

The change in evaporation technology implemented in S5—from multiple-effect to steam recompression—dismissed the use of natural gas at this stage of the system, but increased electricity consumption. Although 22% of the BR grid is composed of non-renewable sources [37], the effect of the discontinuation of the use of natural gas in EV on *NRFS*<sup>5</sup> performance prevailed over the inputs that the subcategory received due to the increased electricity use. Moreover, steam recompression caused RW to become the dominant *PEDS*<sup>5</sup> precursor, with 49% contribution.

Substitution of HCl by HNO3 introduced in S6 had a positive effect. By inserting electrolysis in the HCl production process, the system's accumulated electricity consumption was reduced *RWS*<sup>6</sup> < *RWS*3. On the other hand, HNO3 synthesis has as its essential raw material ammonia obtained through natural gas steam reforming (0.63 m<sup>3</sup> NG/kg NH3). In addition, as both the NH3 and HNO3 processes are endothermic, another portion of natural gas will be added to the system to meet those needs. The reconciliation between these elements explains the fact that *NRFS*<sup>6</sup> supplanted *NRFS*<sup>3</sup> at 1.08 MJ/RF, in spite of the decreased global electric consumption observed in S6.

Finally, a collective analysis of the findings described above indicates an intrinsic correlation between actions carried out for the process and its performance in terms of PED. With respect to scenarios S1–S3 and S6, which differ only in the applied pretreatment technology, the increased *ηSi*

(Table 2) is reverted to mitigation of the overall PED impacts (Table 3). EDTA (S1) substitution by BaDs (S2), and of BaDs (S2) by CPT (S3 and S6), led to an accentuated water recovery efficiency in RO. This trend is based on the (high) efficacy achieved by CPT in the effluent-dissolved salt removal compared to other possibilities.

The increased *ηSi* led to improve power consumption in RO (Table 4), due to the additional resistance that the solute formed by the pre-treatment imposes to the solvent flow (water in use by the cooling tower) through the membrane. However, the concentrate flows transferred to EV were reduced. This effect led to higher PED benefits than the deleterious effects caused by increases in the boiling temperature of the solutions, resulting in decreased heat consumption of the multi-effect distillation (Table 4). A similar phenomenon was noted for the concentrate streams routed to CR, which, due to the low flows, provided decreased electrical demands for this stage.

The heat production by natural gas burning resulted in more significant PED contributions than the electricity generation from the BR grid. Therefore, the gain in *ηSi*, provided from the greater effectiveness of the pretreatment is reverted to contribution retention for the category in all the arrangements that depend on such utilities to operate. S5's success can be explained by the same reasoning. Although it displays higher electrical consumption compared to the other scenarios (4.90 MJ/RF), the synergy established between S5 water recovery rate (*ηS*<sup>5</sup> = 98.5%), and the exemption of the use of heat in EV, compensates this disadvantage.

The situation described by S4 is quite specific. Due to waste heat, the desalination process is limited to the EV-CR set, and even so, this scenario achieved the best PED result among the other options. It should be noted, however, that, in addition to relying on other refinery sectors, the operation of this arrangement requires special care to serve the purposes for which it is intended. This condition makes it not recommended for regular use. S4 can, therefore, be characterized as the lowest PED impact level to be achieved by the system concerning its base technologies.

S4 also presented the lowest GW performance index among the assessed options, again due to the use of waste heat. Under these circumstances, electricity generation from the BR grid becomes the main source of impact for the category, accounting for 98% of the contributions for this scenario.

S3 prevails over the other scenarios that make use of natural gas for heat generation, followed by S6 and S2, with S1 being the most impactful alternative of the set. In addition, none of these options surpassed S5 performance, which applied electricity only during the steam recompression operation. This finding corroborates Pintilie et al. [14], who state that GW impacts are more sensitive to the intensity oscillation of thermal demands than to the electrical requirements of the treatment systems.

In scenarios with lower water recovery rates in RO (S1 and S2), CO2 emissions of fossil origin (CO2,f) derived from the combustion of natural gas for heat production in EV were the main source of impact. According to Cornejo et al. [5], this is not only due to the use of fossil sources for the thermal energy supply of the stage, but also due to the fact that EV is naturally more energy intensive than RO. In S1 and S2, the CO2,f losses to air represented 78% and 75% of their total impacts as GW. The disparity between the results for these scenarios is due to the thermal demand fluctuations caused by the use of BaDs, which decreased from 1.03 m3/RF in S1 to 0.87 m3/RF in S2.

The reduction of natural gas consumption caused by the CPT led to significant benefits for S3 concerning GW. However, even if dampened in comparison to S1 and S2 performances, CO2,f emissions originating from the heat generation still accounted for 41% of the *GWS*3. Another relevant focus of this scenario's impact is on GHG emissions from electricity generation. Responsible for 38% of the *GWS*3, these contributions originate (once again) from CO2,f, summed to CH4 losses, which occur in the life cycle of the natural gas that feeds thermoelectric plants. Impact precursors comprise the releases of dinitrogen oxide (N2O) and CO2 from land transformation (CO2,LT) from bioelectricity generation, a source that represents 8.0% of the BR grid [37]. In addition to intervening in S3 to meet RO, EV, and CR energy requirements, the BR grid acts indirectly on the system by participating in the manufacturing stages of the osmosis membrane (147 kJ/m2) and Ca(OH)2 (22.9 kJ/kg), as well as, mainly, in the HCl production chain. As mentioned previously, this action focuses on the electrolysis

of sodium chloride, which, due to its energy intensive character contributes to about 11% of the total impact of the category. Finally, the use of Ca(OH)2 in CPT also brought a significant contribution to *GWS*3, because of the regular technology applied to obtain quicklime (CaO)—thermal decomposition of limestone, a material that contains calcium carbonate (CaCO3)—in a lime kiln. The calcination of CaCO3 releases 909 gCO2/kg CaO [42], corresponding to an impact of 146 gCO2eq/RF.

The option of the use of steam compression for water evaporation was promising for GW as effects of increases in electric EV demands, also in this category, were compensated by the suppression of natural gas burning and, therefore, of the GHG emissions that this operation would entail. Thus, the impact observed by S5 was about 18% lower than that achieved by S3. In any case, the increased electricity consumption during the evaporation stage transformed the BR grid into the main source of *GWS*5, accumulating 74% of the total impact.

Unlike observed for PED, the exchange of HCl for HNO3 led to worse GW performance in S6 compared to S3. This is due to the N2O emissions generated during acid synthesis (the Ostwald process), in which anhydrous NH3 is oxidized to HNO3 by metal catalysts and stringent temperature and pressure conditions. The dinitrogen oxide ends up being formed because a small portion of the NH3 is partially oxidized [31]. The high impact factor of N2O for GW (298 kgCO2,eq/kg, [43]), led to these losses (8.40 g/kg HNO3) representing 22% of the *GWS*<sup>6</sup> impact.

Due to the similarities between the primary energy demand and global warming precursors, the correspondence diagnosed for PED between type of pretreatment, *ηSi*, and the cumulative performance of each system also remains valid concerning GW impacts.

#### **4. Conclusions**

This study evaluated the reuse of a saline effluent from an oil refinery to supply a cooling tower inside the chemical plant, in a closing loop movement. An arrangement consisting of reverse osmosis (RO), evaporation (EV), and crystallization (CR) was defined for water recovery. Six scenarios were assessed, mainly observing pre-treatment options (desupersaturation or coprecipitation), and different approaches for energy supplying during EV. These are: (i) the use of waste heat to replace the natural gas burning fumes to meet the thermal demand of the system, and (ii) the application of steam recompression in substitution of multi-effect distillation.

The estimation of the environmental effects of the treatments was carried out by attributional LCA, according to a scope from 'cradle-to-gate', for the primary energy demand (PED) and global warming (GW) impact categories.

The scenario that makes use of waste heat as a source of thermal energy for EV (S4) presented the lowest impacts indices among the analyzed possibilities (PED: 11.9 MJ/RF; and GW: 720 gCO2,eq/RF). However, the vulnerability of this arrangement, because its operation is subordinate to the operation of other refinery sectors, makes the alternative not recommended. Thus, S5, that applies coprecipitation as a pretreatment technique for RO-fed effluent and adopts steam recompression to meet EV energy demands appears as the lowest impact scenario of the series (PED: 17.2 MJ/RF; and GW: 1.24 kgCO2,eq/RF). This occurs because the BR grid provides lower contributions in terms of PED and GW than the natural gas life cycle for heat-generating purposes.

A comprehensive analysis of the research findings identified an intrinsic correlation between water recovery rates in RO and overall PED and GW impacts in the arrangements that use heat and electricity for their operation.

Despite the performance variability observed among the scenarios, the benefits identified by the analysis indicate that water desalination is an environmentally efficient alternative in the reduction of water consumption and effluent discharge. These results can still be improved by the application of less aggressive compounds in the environment to remove salts during the pre-treatment phase and the reuse of residual energy sources.

**Author Contributions:** F.M.R. designed the evaporator and crystallizer models; M.M.S. provided technical information for crystallizer modeling; H.S. designed the RO simulations and the LCA assessment; L.K. and H.S. analyzed results, carried out the discussion, wrote the paper and reviewed the text.

**Funding:** This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior—Brazil (CAPES) Finance Code 001.

**Conflicts of Interest:** The authors declare no conflict of interest.
