Reverse Osmosis Membrane Combined with Ultrasonic Cleaning for Flue Gas Desulfurization Wastewater Treatment

: Flue gas desulfurization (FGD) wastewater treatment is currently of interest, as stringent standards have been released in order to limit the pollution emissions from the energy industry, and concerns about water scarcity are also increasing. Reverse osmosis (RO) membrane is a promising alternative for highly efﬁcient FGD wastewater treatment. However, membrane fouling strongly limits its application. This study developed a suitable treatment system by combining RO membrane with ultrasonic cleaning. The introduction of low-frequency and high-intensity ultrasonic cleaning improved the cleaning efﬁciency of membrane fouling, as the permeate ﬂux recovered 49% of the reduced value within 10 min of cleaning. The lifespan of the membrane was also extended, as the time of permeate ﬂux declined to the same level, increasing from 2 h to 4 h after ultrasonic cleaning. The efﬂuent of the system could meet the standard of desulfurization wastewater treatment. The treatment system is feasible for FGD wastewater treatment at a laboratory scale. These ﬁndings proved that the combination of RO membrane and ultrasonic cleaning could be applied to FGD wastewater treatment. The study provided an efﬁcient, cost-saving, and convenient way to develop the FGD wastewater treatment system.


Introduction
Coal dominates the energy structure of China, and about 63.22% of electricity is generated by coal-fired power plants [1]. The consumption of coal directly increases the emissions of dust, sulfur dioxide, and nitrogen, causing severe atmospheric pollution. From 2011 to 2014, several national standards and plans were released to limit the pollution emissions from the energy industry, and flue gas desulfurization (FGD) systems have become mandatory for ultra-clean flue gas treatment in coal-fired power plants [2]. The most widely applied treatment method is limestone gypsum wet flue gas desulfurization, which accounted for about 69% and 80% of market shares worldwide and in China [3]. This process generates gypsum saturated FGD wastewater, which needs further treatment to avoid environmental pollution. The content of total dissolved solids in FGD wastewater is relatively high, including large amounts of SO 2− 4 , SO 2− 3 and Cl − . Therefore, treatment for FGD wastewater is necessary to improve water quality and reduce environmental concerns [4].
Technologies for FGD wastewater treatment include surface impoundments, chemical precipitation, biological treatments (anoxic/anaerobic), evaporation systems, constructed

Materials
All chemicals were of analytical reagent grade. The Na2SO4 solution was freshly prepared with deionized water.

Design of Laboratory-Scale Membrane System with Ultrasonic Cleaning
The experimental system combined a commercial reverse osmosis membrane with an ultrasonic cleaning apparatus. The membranes were commercial household reverse osmosis (RO) membranes (HP18122-50, Vontron Membrane Technology CO., Ltd., Guiyang, China), with a membrane area of 0.18 m 2 . The ultrasonic cleaning apparatus consisted of six ultrasound transducers (KMDcsb, Shenzhen, China) and a resonance cell. The transducers were connected to an ultrasound generator (KMDcsb, KMD-K1, Shenzhen China) to precisely adjust the frequency and intensity of ultrasound. The feed for the pilot system was Na2SO4 solutions with specific concentrations for operating conditions optimization experiments and actual flue-gas desulfurization (FGD) wastewater for cleaning performance experiments. The feed was induced to the RO membrane via a high-pressure pump. A schematic of the system is shown in Figure 1a and photograph is shown in Figure  1b.

Optimization Experiments for RO Membrane Operation
Five parameters (feed flow rate, operating pressure, feed concentration, temperature and pH) were evaluated separately to obtain the optimal operating conditions for the RO membrane. In each experiment, four factors were maintained constant to evaluate the effect of the remaining factor on RO membrane performance. The membranes were operated for 10 h under different operating conditions. Permeate flux and rejection were calculated according to the following equations to evaluate the performance of the RO membrane process: where J = permeate flux, Qp = permeate flow rate, S = membrane surface area, R = rejection, Cp = ions concentration in permeate flow, and Cf = ions concentration in feed flow.

Optimization Experiments for RO Membrane Operation
Five parameters (feed flow rate, operating pressure, feed concentration, temperature, and pH) were evaluated separately to obtain the optimal operating conditions for the RO membrane. In each experiment, four factors were maintained constant to evaluate the effect of the remaining factor on RO membrane performance. The membranes were operated for 10 h under different operating conditions. Permeate flux and rejection were calculated according to the following equations to evaluate the performance of the RO membrane process: where J = permeate flux, Q p = permeate flow rate, S = membrane surface area, R = rejection, C p = ions concentration in permeate flow, and C f = ions concentration in feed flow.

Optimization Experiments for Ultrasound Generation Conditions
The optimal generation conditions for ultrasound were also examined to achieve better cleaning effects. As in the above experiments, frequency and intensity were adjust separately by the ultrasound generator to evaluate the influence of the two parameters on ultrasonic cleaning. The membranes were also operated for 10 h under optimal operating conditions and then cleaned for 10 min under different ultrasound generation conditions. Permeate flux and rejection after cleaning were also calculated according to the above equations to evaluate the cleaning effect.

Ultrasonic Cleaning Performance of RO Membrane after Actual FGD Wastewater Treatment
After determining the optimal conditions of membrane operation and ultrasound generation, the laboratory-scale membrane system was fed with actual FGD wastewater to evaluate the cleaning effect of ultrasonic cleaning. As the actual ion concentration was too high for the commercial household RO membrane treatment, the actual FGD wastewater was diluted 10 times to avoid membrane fouling. The system was operated for 10 h and cleaned for 10 min. The permeate flow was sampled and measured every 15 min in the first 2 h and then measured every 30 min. The ions concentrations in feed wastewater were measured before membrane filtration. Moreover, the concentrations of the ions were measured before ultrasonic cleaning and after cleaning.

Analytical Methods for Evaluation Parameters
Grab samples of permeate were collected and analyzed for inorganic constituents three times during each experiment. According to National Standard, the analyses were performed (GB/T 5750.5-2006, GB/T 15453-2008, GB/T 15452-2009, HJ 535-2009). The ion concentrations were analyzed via UV spectrophotometer. In optimization experiments for RO membrane operation and ultrasound generation conditions, sulfate was the only anion of interest. In the experiment of ultrasonic cleaning effect of RO membrane after actual FGD wastewater treatment, the anions of interest were calcium, magnesium, and ammonium and the cations of interest were sulfate and chloride.

Effects of Operating Conditions on Membrane Performance
Five parameters (feed flow rate, operating pressure, feed concentration, temperature, and pH) were studied to optimize the operating conditions, as they influence the wastewater treatment effect of the reverse osmosis (RO) membrane [34]. Feed flow rate, operating pressure, temperature, and pH are the essential parameters that directly determine the overall operating conditions. The optimization experiment on feed concentration was to testify the concentration range of wastewater that the system can handle. The experiment results (permeate flux and rejection) could also testify the feasibility of using the RO membrane for FGD wastewater.
Permeate flux and rejection versus feed flow rate are shown in Figure 2. Both permeate flux and rejection increased with feed flow rate and reached maximum (36.19 L·m −2 ·h −1 and 96.15%, respectively) at a feed flow rate of 35.00 L·h −1 , representing the best performance of wastewater treatment. The improvement of membrane performance was related to the mitigation of membrane fouling. Higher feed flow rate increased the flow velocity and provided a better mass transfer coefficient [35]. The improved flow velocity and mass transfer coefficient further reduced the ion concentration near the membrane surface and diminished the concentration polarization and membrane fouling.
High operating pressure increased the permeate flux and rejection ( Figure 3). The increment of permeate flux was expected, as RO membranes are pressure-driven membranes. An increase in the rejection with increasing pressure was achieved until a constant and steady rejection of ca. 96.10% at 0.35 MPa was obtained. The trend of rejection variation was similar to the experimental results of Gur-Reznik Shirra [36]. The variation of rejection may be attributed to the path change of the high content of salts with the permeate, which decreases the salt rejection efficiency [37]. The maximum permeate flux and rejection were obtained at 0.50 MPa (37.25 L·m −2 ·h −1 and 96.16%, respectively). In this study, 0.50 MPa was applied to the membrane system for higher experimental efficiency, while 0.35 MPa could be applied for cost-saving purpose. High operating pressure increased the permeate flux and rejection ( Figure 3). The increment of permeate flux was expected, as RO membranes are pressure-driven membranes. An increase in the rejection with increasing pressure was achieved until a constant and steady rejection of ca. 96.10% at 0.35 MPa was obtained. The trend of rejection variation was similar to the experimental results of Gur-Reznik Shirra [36]. The variation of rejection may be attributed to the path change of the high content of salts with the permeate, which decreases the salt rejection efficiency [37]. The maximum permeate flux and rejection were obtained at 0.50 MPa (37.25 L·m −2 ·h −1 and 96.16%, respectively). In this study, 0.50 MPa was applied to the membrane system for higher experimental efficiency, while 0.35 MPa could be applied for cost-saving purpose.   The high concentration feed resulted in reducing the wastewater treatment effect ( Figure 4). Compared with other experiments under the same operating conditions, both the value of permeate flux and rejection declined to the minimum (30.15 L·m −2 ·h −1 and 85.44%) under high concentration. The highly concentrated feed contained more polluting particles at fixed feed flow. These particles will deposit on the surface or the pores of the membrane, resulting in membrane scaling and the decrease of permeate flux and rejection [34]. The feed concentration suitable for RO membrane treatment was 1000.00 to 3500.00 mg·L −1 in this study, as the rejection value remained at ca. 95.00%. The high concentration feed resulted in reducing the wastewater treatment effect ( Figure 4). Compared with other experiments under the same operating conditions, both the value of permeate flux and rejection declined to the minimum (30.15 L·m −2 ·h −1 and 85.44%) under high concentration. The highly concentrated feed contained more polluting particles at fixed feed flow. These particles will deposit on the surface or the pores of the membrane, resulting in membrane scaling and the decrease of permeate flux and rejection [34]. The feed concentration suitable for RO membrane treatment was 1000.00 to 3500.00 mg·L −1 in this study, as the rejection value remained at ca. 95.00%. Since this study evaluates the feasibility of simplifying desulfurization wastewater treatment by combining RO membrane treatment with ultrasonic cleaning, the laboratoryscale experimental system was built with commercial RO membranes to save on experimental costs. The results above indicate that the commercial RO membranes showed inadequacy to high concentration (above 3500.00 mg·L −1 ) wastewater. The actual FGD wastewater (25,000.00 mg·L −1 ) was diluted 10 times to avoid poor membrane treatment effects under high feed concentrations. In future research, this deficiency could be made up when applying the industrial RO membrane to the system.
Feed temperature had the opposite effects on permeate flux and rejection ( Figure 5). The permeate flux increased linearly with temperature, while the rejection declined slightly at high temperatures. The rising trend of permeate flux is similar to previous studies, with a 60% increase in the permeate flux when the feed temperature increased from 20 to 40 • C [38]. The high temperature increased the pore size of the membrane because of thermal expansion, allowing more water to pass through. With the increment of temperature, the solubility of the solute also increased, and a higher diffusion rate of solute through the membrane is possible, causing the decrease of rejection at high temperature [39]. According to the experimental results, the optimal temperature for membrane operation was 25.00 • C (maximum rejection value of 96.17%). Considering the heating cost, the operating temperature for the system was 18.00 • C. 20 to 40 °C [38]. The high temperature increased the pore size of the membrane because of thermal expansion, allowing more water to pass through. With the increment of temperature, the solubility of the solute also increased, and a higher diffusion rate of solute through the membrane is possible, causing the decrease of rejection at high temperature [39]. According to the experimental results, the optimal temperature for membrane operation was 25.00 °C (maximum rejection value of 96.17%). Considering the heating cost, the operating temperature for the system was 18.00 °C. The effect of feed pH on membrane performance was negligible, as the rejection value remained at about 96.00% ( Figure 6). However, the variation trend of permeate flux and rejection is similar to Hoang's experiment [40]. Alkaline wastewater may cause membrane structure deformation or damage, leading to permeate flux decline at high pH values. The impact of feed pH on RO membrane structure may require further studies. Since the pH of actual FGD wastewater is often acidic, the RO membrane can treat it without membrane The effect of feed pH on membrane performance was negligible, as the rejection value remained at about 96.00% ( Figure 6). However, the variation trend of permeate flux and rejection is similar to Hoang's experiment [40]. Alkaline wastewater may cause membrane structure deformation or damage, leading to permeate flux decline at high pH values. The impact of feed pH on RO membrane structure may require further studies. Since the pH of actual FGD wastewater is often acidic, the RO membrane can treat it without membrane damage. In this study, the optimum operating condition of pH was 7 because of the maximum value of permeate flux and rejection (35.97 L·m −2 ·h −1 and 96.14%).
To determinate optimal membrane operating conditions, permeate flux and rejection are the key factors. According to the experimental results above, the optimal conditions (feed flow rate, 35.00 L·h −1 ; operating pressure, 0.50 MPa; feed concentration, 2500.00 mg·L −1 ; feed temperature, 18.00 • C; feed pH, 7) for the RO membrane operation in this study were fixed. The permeate flux and rejection results also proved that the RO membrane was capable of FGD wastewater treatment.

Effect of Frequency and Intensity on Membrane Cleaning
Since RO membranes are vulnerable and have complex and composite structures, cleaning conditions are vital to membrane lifetime extension. Operational factors of ultrasound generation can influence the effectiveness of ultrasonic treatment [32]. This study examined the effect of frequency and intensity on the ultrasonic cleaning effect. All membranes ran for 10 h using 2500.00 mg·L −1 Na 2 SO 4 under optimal operating conditions (feed flow rate: 35.00 L·h −1 , operating pressure: 0.50 MPa, temperature: 18.00 • C, pH: 7). The time of ultrasonic cleaning was 10 min.
Both permeate flux and rejection reduced with increasing ultrasound frequency, indicating the cleaning efficiency decreased under high frequencies (Figure 7). This result is consistent with previously published literature, in which high-frequency treatment led to low permeate flux [41][42][43][44]. The findings further support that lower ultrasound frequencies have higher cleaning efficiencies [45]. The foulant on the membrane surface was removed by collapses of microbubbles generated by ultrasound [46]. Under high frequencies, the rarefaction and compression cycles of cavitation bubbles formation were reduced, lowering the lifetime and the size of the microbubbles before collapse [47]. The energy contained in microbubbles was also reduced. The diminished energy cannot generate powerful collapses for pollutant cleaning, resulting in a poor membrane cleaning effect. The optimal frequency for ultrasonic cleaning is 25 kHz in this study, as the permeate flux and rejection reached the maximum (45.26 L·m −2 ·h −1 and 97.37%).  To determinate optimal membrane operating conditions, permeate flux and rejection are the key factors. According to the experimental results above, the optimal conditions (feed flow rate, 35.00 L·h −1 ; operating pressure, 0.50 MPa; feed concentration, 2500.00 mg·L −1 ; feed temperature, 18.00 °C; feed pH, 7) for the RO membrane operation in this study were fixed. The permeate flux and rejection results also proved that the RO membrane was capable of FGD wastewater treatment.

Effect of Frequency and Intensity on Membrane Cleaning
Since RO membranes are vulnerable and have complex and composite structures, cleaning conditions are vital to membrane lifetime extension. Operational factors of ultrasound generation can influence the effectiveness of ultrasonic treatment [32]. This study examined the effect of frequency and intensity on the ultrasonic cleaning effect. All membranes ran for 10 h using 2500.00 mg·L −1 Na2SO4 under optimal operating conditions (feed flow rate: 35.00 L·h −1 , operating pressure: 0.50 MPa, temperature: 18.00 °C, pH: 7). The time of ultrasonic cleaning was 10 min.
Both permeate flux and rejection reduced with increasing ultrasound frequency, indicating the cleaning efficiency decreased under high frequencies (Figure 7). This result is consistent with previously published literature, in which high-frequency treatment led to low permeate flux [41][42][43][44]. The findings further support that lower ultrasound frequencies have higher cleaning efficiencies [45]. The foulant on the membrane surface was removed by collapses of microbubbles generated by ultrasound [46]. Under high frequencies, the rarefaction and compression cycles of cavitation bubbles formation were reduced, lowering the lifetime and the size of the microbubbles before collapse [47]. The energy contained in microbubbles was also reduced. The diminished energy cannot generate powerful collapses for pollutant cleaning, resulting in a poor membrane cleaning effect. The optimal frequency for ultrasonic cleaning is 25 kHz in this study, as the permeate flux and rejection reached the maximum (45.26 L·m −2 ·h −1 and 97.37%). By increasing the intensity of ultrasound, the permeate flux and rejection after cleaning reached the maximum (45.26 L·m −2 ·h −1 and 97.37%), indicating that the cleaning effect improved (Figure 8). In accordance with the present results, previous studies have demonstrated that higher intensity will provide better cleaning effect and greater perme- By increasing the intensity of ultrasound, the permeate flux and rejection after cleaning reached the maximum (45.26 L·m −2 ·h −1 and 97.37%), indicating that the cleaning effect improved (Figure 8). In accordance with the present results, previous studies have demonstrated that higher intensity will provide better cleaning effect and greater permeate flux, and the permeate flux increased nearly linearly with intensity [41][42][43]48]. Contrary to the influence of frequency on the cleaning effect, an increment of intensity will increase the acoustic pressure amplitude and generate more powerful ultrasonic bubbles [47]. More pollutants on the membrane surface can be removed under high intensities. The optimal intensity for RO membrane cleaning was 3.10 W·cm −2 in this study. According to the experimental results above, the optimal conditions (frequency, 25 kHz; intensity, 3.10 W·cm −2 ) for ultrasound generation in this study were also fixed. In summary, for ultrasonic cleaning of the membrane module, the optimal conditions for ultrasound generation should be low frequency and high intensity. This could be guiding advice when applying the ultrasonic cleaning technology to other membrane technologies.

Treatment of Actual FGD Wastewater and Ultrasonic Cleaning of the Fouled Membrane
After determining the optimal conditions for membrane operation and ultrasound generation, the system was fed with actual FGD wastewater (diluted 10 times) to evaluate the treatment effect of RO membrane combined with ultrasonic cleaning. The flux data during the experimental process were examined and presented in Figure 9. After 6 h of filtration, the flux decreased from 47.78 to 34.55 L·m −2 ·h −1 . Then the flux increased to 41.07 L·m −2 ·h −1 after 10 min of ultrasonic cleaning. The permeate flux recovered for 49% of the reduced value within 10 min of cleaning, suggesting that ultrasonic cleaning can efficiently remove pollutants on the membrane surface. Although the recovery rate of permeate flux (49%) was only about half of the physical cleaning (ca. 93%), the cleaning time (10 min) was only one-third of physical cleaning (30 min) [49]. The ultrasonic cleaning was more efficient than physical cleaning without consuming a large amount of clean water. The recovered flux was also higher than the chemical cleaning of sugar-fouled reverse osmosis membrane with some chemical agents (i.e., 1% HNO3, 1% NaOH, 1% HCl, 1% CTAB) [50]. After another 4 h of filtration, the flux declined to 36.40 L·m −2 ·h −1 . The decline rate of flux slowed down, as the flux declined to 36.41 L·m −2 ·h −1 within only 2 h in the first 6 hours of filtration. This result indicated that ultrasonic cleaning extended the lifetime of According to the experimental results above, the optimal conditions (frequency, 25 kHz; intensity, 3.10 W·cm −2 ) for ultrasound generation in this study were also fixed. In summary, for ultrasonic cleaning of the membrane module, the optimal conditions for ultrasound generation should be low frequency and high intensity. This could be guiding advice when applying the ultrasonic cleaning technology to other membrane technologies.

Treatment of Actual FGD Wastewater and Ultrasonic Cleaning of the Fouled Membrane
After determining the optimal conditions for membrane operation and ultrasound generation, the system was fed with actual FGD wastewater (diluted 10 times) to evaluate the treatment effect of RO membrane combined with ultrasonic cleaning. The flux data during the experimental process were examined and presented in Figure 9. After 6 h of filtration, the flux decreased from 47.78 to 34.55 L·m −2 ·h −1 . Then the flux increased to 41.07 L·m −2 ·h −1 after 10 min of ultrasonic cleaning. The permeate flux recovered for 49% of the reduced value within 10 min of cleaning, suggesting that ultrasonic cleaning can efficiently remove pollutants on the membrane surface. Although the recovery rate of permeate flux (49%) was only about half of the physical cleaning (ca. 93%), the cleaning time (10 min) was only one-third of physical cleaning (30 min) [49]. The ultrasonic cleaning was more efficient than physical cleaning without consuming a large amount of clean water. The recovered flux was also higher than the chemical cleaning of sugar-fouled reverse osmosis membrane with some chemical agents (i.e., 1% HNO 3 , 1% NaOH, 1% HCl, 1% CTAB) [50]. After another 4 h of filtration, the flux declined to 36.40 L·m −2 ·h −1 . The decline rate of flux slowed down, as the flux declined to 36.41 L·m −2 ·h −1 within only 2 h in the first 6 hours of filtration. This result indicated that ultrasonic cleaning extended the lifetime of the membrane. The cleaning may loosen the foulant layer on the membrane surface and diminish the speed of layer formation. Table 1 shows the variation of permeate concentration and rejection before and after ultrasonic cleaning. Rejections increased by 2-3% after cleaning, suggesting that the ultrasonic cleaning slightly improved the treatment effect. As the rejection value was at a high level (ca. 95.00-97.00%) after cleaning, ultrasonic cleaning was beneficial to maintain wastewater treatment performance. The effluents also met the standard of desulfurization wastewater treatment during the whole experimental process. These results proved that ultrasonic cleaning is an efficient method for fouled RO membrane. Furthermore, combining RO membrane treatment with ultrasonic cleaning can extend the lifetime of the membrane and improve the wastewater treatment effect.
Water 2022, 14, x FOR PEER REVIEW 11 of 14 cleaning was beneficial to maintain wastewater treatment performance. The effluents also met the standard of desulfurization wastewater treatment during the whole experimental process. These results proved that ultrasonic cleaning is an efficient method for fouled RO membrane. Furthermore, combining RO membrane treatment with ultrasonic cleaning can extend the lifetime of the membrane and improve the wastewater treatment effect.

Conclusions
In this work, flue gas desulfurization (FGD) wastewater was treated by a laboratoryscale treatment system with the combination of reverse osmosis (RO) membrane and ultrasonic cleaning. The optimization experiments on operating parameters proved that the commercial RO membrane was capable of FGD wastewater treatment. The investigation of optimal ultrasound generation conditions has shown that ultrasound generated under low frequency and high intensity had the best cleaning effect on the fouled membrane. These findings provide the optimal conditions for RO membrane operation (feed flow rate, 35.00 L·h −1 ; operating pressure, 0.50 MPa; feed concentration, 2500.00 mg·L −1 ; feed temperature, 18.00 • C; feed pH, 7) and ultrasound generation (frequency, 25 kHz; intensity, 3.10 W·cm −2 ). The feasibility of the system was also tested.
The introduction of ultrasonic cleaning improved the cleaning efficiency of membrane fouling. The permeate flux recovered 49% of the reduced value within 10 min of cleaning, which is more efficient and cost-saving than physical cleaning and some chemical cleaning methods. Furthermore, the time for permeate flux decline after cleaning had doubled (from 2 h to 4 h), suggesting that the ultrasonic cleaning can delay the formation of membrane fouling and extend the lifespan of the RO membrane. The treatment effects were maintained as the rejection values were high (95.00-97.00%). The effluent of the system also met the standard of FGD wastewater treatment.
Overall, the study indicated that the treatment system is capable of FGD wastewater treatment at a laboratory scale. The combination of RO membrane and ultrasonic cleaning sheds new light on building a system for FGD wastewater treatment in a more efficient, cost-saving, and convenient way. Further research should focus on applying this system at a factory-scale experiment by changing the commercial membrane to industrial membrane.