Pilot-scale study on seeded precipitation assisted nanofiltration for flue-gas desulfurization wastewater softening Open Access

As a conventional fossil energy, coal plays an important role in China, in which more than half of power plants are coal-fired power plants (China Statistical Yearbook 2018). The coal-fired power plant is the main place where sulfur dioxide is produced during coal combustion,which must be eliminated before being released into the atmosphere (Liu et al. 2021). For this reason, the wet limestone-gypsum flue gas desulfurization (FGD) process is introduced in most large and medium-sized coal-fired power plants in China (Koralegedara et al. 2019). FGD wastewater is the drainage of the FGD absorption tower, having the characteristics of corrosivity, acidity, considerable chloride ions, high hardness and many kinds of heavy metal contamination (Shuangchen et al. 2016). The traditional treatment process of FGD wastewater includes neutralization, chemical precipitation and coagulation (NPC) units. Since the NPC process cannot remove the main salt ions in FGD wastewater, the wastewater cannot be recycled and reused (Dou et al. 2017). In general, the NPC effluent is used for ash field spraying or hydraulic slug flashing, which lead to further environmental pollution (Han et al. 2020).

Due to severe water scarcity there are increasingly strict regulations in many parts of the world, such as the Water Pollution Control Action Plan in China (State Council-2015–17), which demands coal-fired power plants carry out a zero liquid discharge (ZLD) process for FGD wastewater (Xiong & Wei 2017; Sun et al. 2021). Typically, the ZLD process for FGD wastewater consists of three modules, a pretreatment module by chemical softening and/or nanofiltration (NF), a concentration module by membrane system, and a curing module by evaporative crystallization (Schantz et al. 2018; Xin et al. 2020). Chemical precipitation is the most widely-used pretreatment method, in which calcium hydroxide and sodium carbonate are always used to remove calcium and magnesium ions. Here, a high concentration of calcium and magnesium ions exists in the FGD wastewater. A large amount of sodium carbonate is needed for hardness removal in the chemical softening process, so the pretreatment cost of FGD wastewater is very high (Conidi et al. 2018; Jia & Wang 2018).

NF has been widely used for trapping divalent and hypervalent ions in the ZLD process. Compared with water softening by chemical precipitation, water softening by NF membrane can reduce chemical and energy consumption, decrease operating cost, and increase operational safety (Fang et al. 2013; Shi et al. 2020). However, there are still lots of divalent ions in NF brine, which are mostly removed by chemical softening or seeded precipitation (Sanciolo et al. 2012; Liu et al. 2016; Sanciolo & Gray 2017).

The seeded precipitation process is an effective method to treat supersaturated brine for achieving high water recovery, and it is usually used in combination with a membrane system. In the process, a seeded crystallizer is usually needed for low solubility salts to crystallize with the help of enough seed crystal (Choi et al. 2021). Feed water with scale inhibitor flows into the membrane unit, where the feed water is concentrated to be supersaturated with low solubility salts (Pervov 2015). Then, the concentrated feed water flows into the seeded crystallizer, in which the degree of supersaturation is decreased. The desupersaturated effluent from the seeded crystallizer is recycled back to the original membrane module, or concentrated by another membrane module. Since there is no need to add a chemical agent in the membrane concentrate for removing hardness, seeded precipitation can be proved to be a cost-effective process in the future.

In the 1980s, the seeded precipitation was first applied in tubular RO systems to remove hardness in mine water saturated with calcium sulphate (Harries 1985). Then, the seeded precipitation process was used in association with electrodialysis (ED) or electrodialysis reversal (EDR) in order to decrease the disposed concentrate volume; the results showed that the process recovery was above 98% (Korngold et al. 2005; Oren et al. 2010). In recent years, the seeded precipitation process has been designed to combine with the NF system to remove hardness in the NF concentrate, and it is more cost effective compared with seeded precipitation used in association with ED and EDR. In the previous work of our research team, a laboratory-scale seeded precipitation assisted RO system was designed and tested with the wastewater from coal chemical industry. The results showed the laboratory system achieved an overall water recovery of 90.9% without obvious indication of fouling, and substantial calcium sulfate dihydrate crystallization was confirmed in the ATC unit, which indicated the feasibility and scalability of seeded precipitation assisted RO system (Xiong et al. 2017).

Since there is lots of hardness in FGD wastewater, it is greatly suitable for application of the seeded precipitation process. Up to now, there have been no articles about seeded precipitation process designed for FGD wastewater and the chemical softening cost of FGD wastewater is high in China. The objective of this work is to use a seeded precipitation assisted NF system to remove the hardness in FGD wastewater, which is supposed to decrease the cost of chemical softening.

The pilot-scale test process and devices of the seeded precipitation assisted NF system are shown in Figure 1.

The pilot system for this test was located in a coal-fired power plant in Shandong province with a capacity of 1 m³/h, and there were four main units: the lime softening unit, the ambient temperature crystallizer (ATC) unit, the ultrafiltration (UF) unit and the NF unit. In the lime softening unit, there were a reaction tank, a flocculating tank and a settler. One water tank was set between two units with a volume of 2 m³, and there were four water tanks in total. The ATC was the core unit used to carry out the seeded precipitation process, which was made of carbon steel with anticorrosive coating on the surface. Two membrane elements (FUF-8040) were used in the UF unit from Hangzhou Creflux Membrane Company. Two membrane vessels were installed in parallel in the NF unit, and there were four elements DK4040F30 in each membrane vessel from Suez Environment. In addition, two filter presses were set in the pilot system, the filter areas of which were 6 m².

Three main dosing systems were included in this system, which were calcium hydroxide dosing system, sodium sulfate dosing system and scale inhibitor dosing system. A few pumps except the three main pumps in Figure 1 as well as other auxiliaries were also employed. A prescreened chemical was fed into the NF brine, which was used as deactivation agent for the residual scale inhibitor.

Flowmeters and pressure gauges were employed to measure the flux and pressure in the whole process. Liquid level indicators were introduced into all tanks to detect the liquid level. Two pH on-line analyzers were installed in the lime softening unit and tank 2 to measure the corresponding pH. The on-line density meter from E + H was introduced into ATC to measure the density of calcium sulfate dihydrate suspension, and the density value could be converted to the mass fraction of suspended solid.

The characterization of raw water, effluent from each unit, and salt from the ATC unit were implemented by sampling. Cations were analyzed with the EDTA titration method or inductively-coupled plasma optical emission spectrometry (ICP), and anions were measured using ion chromatography (IC) in the water samples. All water samples were filtered, and diluted 100 ∼ 200 times with deionized water before being analysed by ICP and IC, the analysis range of which was 0.1 mg/L ∼ 100 mg/L. The ICP (SPECTRO ARCOS) was under the following operational conditions: radiofrequency power of 1,400 W, plasma gas flow rate of 12.0 L/min, auxiliary gas flow rate of 1 L/min and nebulizer gas flow rate of 1.0 L/min. The IC separation was conducted using a Thermo Intergrion system, samples were injected to the separation column with the elute solution and separated according to the different relative affinity of the ion exchange resin. The determination of anion used potassium hydroxide as elute solution, and samples were separated by a chromatographic column and a guard column. The anionic suppressor was used with suppressor current of 150 mA. The elute concentration was 30 mmol/L with flow rate of 1.0 mL/min.

The method of testing the purity of calcium sulfate dihydrate (gypsum) was based on JC-T 2074-2011 Flue Gas Desulfurization Gypsum. Malvern2000 analyzer from Malvern Instruments Company was employed to measure particle size distribution of the seeds and calcium sulfate dihydrate in our laboratory, in which the saturated calcium sulfate solution was selected as the dispersant. In the long term operation, the electricity consumption was recorded by the electricity meter and the chemical consumption was recorded manually.

Before the pilot test, typical ionic compositions of the FGD wastewater were analyzed, and are shown in Table 1.

The pilot system ran 168 h based on the Code for Fossil Power Construction Project from The Unit Commissioning to Completed Acceptance in China.

As can be seen from Figure 1, the FGD wastewater was fed into the lime softening unit with a flux of 1 m³/h. Calcium hydroxide was added in this unit to adjust the pH about 11.2 to remove magnesium ions in the raw water. After flocculation and clarification, the liquid supernatant entered tank 1, and small amount of sludge was discharged where the main ingredient was magnesium hydroxide and calcium sulfate dihydrate.

The wastewater was fed into ATC unit by pump 1 from tank 1, where a small amount of sodium sulfate was dosed to remove calcium ions. When the ATC unit was firstly filled with wastewater, calcium sulfate seeds were introduced with 5% loading by weight. The seeds were fully suspended by a mixer, providing enough surface area for efficient crystallization. While the ATC was under operation, the mass fraction of the seed would be increased. Some calcium sulfate dihydrate seeds should be released to the filter press unit to maintain the mass fraction of the seeds at 5% in the ATC. The ATC effluent flowed into tank 2 after being clarified by a settler in the ATC unit.

The ATC effluent was injected to the UF unit to eliminate suspended solids in the wastewater, and met the feed water quality requirement of the NF membrane. Water recovery of the UF unit was set as 90%, the concentrate of which was returned to tank 2. Due to the dosing of calcium hydroxide in lime softening unit, the wastewater was alkaline in tank 2. Thus, hydrochloric acid was added into tank 2 to regulate the wastewater to be neutral. For the UF membranes, backwash with UF permeated water was employed for 60 s and air wash was employed for 30 s every 30 min.

The NF membranes were fed by pump3 with the UF permeate in tank 3. In order to prevent the NF membrane from scaling, scale inhibitor was injected into the feed water of NF with a dosage of 9 ppm. The NF permeate flowed into tank 4, and the NF concentrate was recycled to tank 1 and pumped to the ATC for desupersaturation. The prescreened chemical with a dosage of 5 ppm was added into the NF concentrate. The prescreened chemical acted as an inactivation agent to deactivate the residual scale inhibitors, eliminating the bad effect on crystallization in the ATC. So calcium and sulfate ions in the NF concentrate spontaneously crystallized with the aid of seeds, and calcium sulfate dihydrate was generated.

Moreover, the NF permeate in this pilot test can be introduced into the concentration module to get high quality water that can be reused. Then the concentrate from the concentration module can be fed into the curing module to obtain high purity sodium chloride. Since the technologies of these two modules are relatively mature, they are not the research content of this paper.

The main objective of this pilot system was to remove hardness in the wastewater, so magnesium ions and calcium ions were selected as the main research subjects. Figures 2 and 3 show magnesium and calcium ion concentration in the main flows. It can be seen from the figures that the system has a great removal effect on hardness.

Both the concentrations of magnesium ions and silicon were enough low in the lime softening effluent to have little fouling effect on the downstream membrane system, reduce the frequency of chemical cleaning, and prolong the service life of the membrane.

As shown in Figure 2, the concentration of magnesium ions was fluctuating in the lime softening effluent. In fact, the fluctuation amplitude was small, within the 23 mg/L range. Maximum fluctuation amplitude of the magnesium ion concentration was about 1,040 mg/L in the raw water, which may be due to the quality of limestone and makeup water in the FGD absorption tower. Although the concentration of the magnesium ion fluctuated in the raw wastewater, the concentration of magnesium ions was not affected in the lime softening effluent.

It can be seen from Figure 3 that the ATC unit has a good removal effect on calcium ions. The calcium ion concentration in the ATC effluent flow was the key parameter to ATC, which was tested every four hours. The calcium ion concentration in other flows was measured every day. As shown in Figure 3, calcium ion concentration was increased after lime softening, and was about 2,655 mg/L in the ATC inlet flow. This is due to the reaction of calcium hydroxide with magnesium ions, which produced magnesium hydroxide and calcium ions. However, the calcium ion concentration was significantly decreased in the ATC effluent flow, at 800–1,000 mg/L, indicating the effectiveness of ATC in getting rid of calcium ions. The decrease of calcium ions can also illustrate the effectiveness of the deactivation agent injected in the NF concentration. In addition, there was a small fluctuations of the calcium concentration in the ATC effluent, which might be due to fluctuations of hardness and sulfate ion concentrations in the raw water. The fluctuation of calcium concentration in the ATC effluent was also attributed to the solubility of calcium sulfate dihydrate. The solubility of calcium sulfate dihydrate was affected by other ion concentrations, in the wastewater, especially the chloride ion, which was negatively correlated with the solubility of calcium sulfate dihydrate.

In Figures 2 and 3, the concentrations of calcium and magnesium ions were greatly reduced in the NF permeate compared with that in the raw water, especially the magnesium ion concentration, most of which was less than 1 mg/L. The concentration of calcium ion was less than 80 mg/L. It can be calculated from the data in the figure that the total hardness is less than 204.2 mg/L in NF permeate, so the hardness removal rate of this system is more than 98.9%. Table 2 shows the final compositions of the FGD wastewater after treated by the pilot system. Compared with the raw FGD wastewater in Table 1, the effluent of this system mainly contained sodium ions and chloride ions, which can form sodium chloride in subsequent curing module. There were few other ions in the system effluent due to the high hardness removal rate of this system, which can ensure the high purity of the sodium chloride salt.

The concentration of sulfate ion in NF permeate was also recorded since it could affect the purity of the subsequent crystalline salts. Figure 4 shows the change of sulfate ion in the in the NF permeate. It can be seen from the figure that the concentration of sulfate ion was less than 80 mg/L, indicating high removal rate of sulfate ion in this system, which was more than 99%. Almost all sulfate ions were concentrated in the NF brine and recycled back to the ATC for crystallization, which reduced the dosage of sodium sulfate. Moreover, a small fluctuation of the sulfate ion concentration occurs in NF permeate, which might be attributed to the feed quality of the NF unit.

Particle size and purity of the crystallization salts were measured and analyzed in the ATC unit. The particle size has an important influence on clarification and crystallization. The saturated calcium sulfate solution was selected as the dispersant during particle size measurement, because the crystallization salt was calcium sulfate dihydrate, which would remain stable and insoluble in solution. A Hydro MU2000A dispersion unit was used to measure the samples, and the measurement range of particle size was 0.02 μm ∼ 2,000 μm. The pump and agitator speed were set to 2,500 and 850 rpm, respectively. The measured value consists of D10, D50 and D90, and the value of D50 was chosen to monitor the change trend of particle size, which was more representative. Since the crystallization of calcium sulfate dihydrate was carried out at ambient temperature, other impurity ions were not concentrated in the wastewater during crystallization, which ensured the high purity of the crystalline salt.

The average particle size for the crystallization salts generated in the ATC during the pilot test are shown in Figure 5. It can be seen from the figure that the crystallization salt size increased as the pilot test went on, and the average particle size increased from 20 μm to 75 μm, indicating that new calcium sulfate dihydrate was formed on seed crystal. However, the growth rate of crystallization salt decreased gradually as the slope of the curve got smaller, which might be due to the increased surface area of the crystallization salt. Otherwise, both the stirring intensity and seed circulating pump can reduce the growth rate of crystallization salt. According to the phenomenon in the test, crystallization salt of 75 μm was easy for solid-liquid separation and it also had enough surface area for crystallization.

According to data statistics, the output of calcium sulfate dihydrate generated was about 11.8 kg/h in the ATC. After being analysed in our laboratory, the calcium sulfate dihydrate had a high purity of 95.8%, and could be reused as gypsum building material or cement retarder.

Figure 6 shows changes of the TMP and permeate flux in the UF unit during continuous operation. As is shown in Figure 6, the TMP was 0.02 MPa ∼ 0.05 MPa and the permeate flux was about 2.20 m³/h ∼ 2.30 m³/h. During the operation, these two parameters remained relatively stable, illustrating that membrane fouling did not occur in the UF unit. It was interesting to note that there were slight fluctuations in the TMP and permeate flow, which was caused by the variations in SS content and the backwashing. After the UF membrane was backwashed, the permeate flux significantly increased and the operating pressure decreased. In addition, the water samples of UF permeate were taken and tested, and the results showed that the turbidity of the UF permeate was less than 0.2 NTU, which met the feed stream requirement of the NF unit.

It can be seen from the figure that the permeate flux of UF was much greater than the raw water treatment capacity, which was caused by the internal circulation in this system. The feed stream flux of the UF unit was about 2.50 m³/h, which consisted of the ATC effluent, UF concentrate, and the UF backwashing wastewater. The ATC effluent flux was about 1.92 m³/h, including the lime softening effluent (0.98 m³/h) and NF concentrate (0.96 m³/h). Continuous flow flux of the UF concentrate and backwashing wastewater was 0.20 m³/h ∼ 0.30 m³/h. The function of internal circulation was to realize the crystallization of calcium sulfate dihydrate and prevent calcium sulfate from scaling.

The feed stream flux of the NF unit was 1.92 m³/h. During the operation of the NF unit, the water recovery had been maintained at 50% by adjusting the operating pressure. Therefore, the flow rate of NF brine and permeate were 0.96 m³/h and corresponding membrane flux was about 13.2 L/m²/h. The water recovery of the NF system was fixed and not high in order to prevent the NF membrane from scaling and maintain the water balance of the whole system. If the water recovery of NF was increased, the flux of internal circulation would be decreased in the system, but calcium and sulfate ions would be improved simultaneously in the NF brine, which would lead to the high probability of calcium sulfate scale in the NF membrane. Hence, the water recovery of NF was set at 50% according to our laboratory research. Even so, the water recovery rate of the whole seeded precipitation assisted NF system could reach 96%, and the 4% water loss was the water contained in sludge and gypsum. The water recovery rate of this system refers to the ratio of NF permeate flux to raw wastewater flux.

Figure 7 shows the operating pressure and pressure drop of the NF membrane. The operating pressure includes membrane inlet and outlet pressure, and the pressure drop is the difference between membrane inlet and outlet pressure. As can be seen from the figure, the membrane inlet pressure was 0.90 ∼ 1.10 MPa and the membrane outlet pressure was 0.70 ∼ 0.90 MPa. The pressure drop kept quite stable, at 0.20 MPa. It is interesting to see that the inlet and outlet pressures were not high, which was attributed to the low water recovery rate of the NF unit. Obviously, the inlet and outlet pressure showed insignificant fluctuations, which might be due to the change of sulfate ion concentration in the feed water of the NF unit. Though the inlet and outlet pressure fluctuated, the inlet and outlet pressure were not increased and the pressure drop remained unchanged, indicating that no scaling occurred on the NF membrane surface.

Calcium hydroxide and sodium sulfate were the main chemical agent used in this pilot system. The total dosage of calcium hydroxide and sodium sulfate during the pilot test were recorded manually. The consumption of calcium hydroxide and sodium sulfate were got by using total dose divided by time, which were 13.56 kg/m³ for calcium hydroxide and 8.60 kg/m³ for sodium sulfate. The price of calcium hydroxide and sodium sulfate is about 600 RMB/ton, so the softening cost of this pilot system is 13.30 RMB/m³. The cost of other chemical agents was about 0.61 RMB/m³, which mainly consist of the scale inhibitors, flocculant, and hydrochloric acid. Thus, the cost of all chemical agents used in this system was about 13.91 RMB/m³.

When the FGD wastewater in this paper was softened by calcium hydroxide and sodium carbonate, or sodium hydroxide and sodium carbonate, the corresponding cost of the chemical agent is shown in Table 3, which was got by theoretical calculation.

It can be seen from the table that the softening cost of this pilot system is much less than the two softening processes above, illustrating that the seeded precipitation assisted NF process for FDG wastewater is a greatly cost-effective process. The low softening cost of this system is due to the following: 1) the calcium ions were removed by sodium sulfate instead of sodium carbonate, and the market price of sodium sulfate is much lower than that of sodium carbonate; 2) there is no need to add chemical agent in the NF brine since the seeded precipitation process was adopted to deal with the NF brine; 3) The NF brine contains a lot of sulfate ions, which reduced the dosage of sodium sulfate.

From the data of electricity meter in the test, the electricity consumption is 9.07 kWh/h. Since the raw water flow is 1 m³/h, the electricity consumption is 9.07 kWh/m³. The electricity price is 0.3737 RMB/kWh, so the electricity cost of the pilot system is 3.39 RMB/m³. Generally, the electricity cost of the concentration module by membrane is 2 ∼ 4 RMB/m³, and the electricity cost of the curing module by evaporative crystallization is 2 ∼ 3 RMB/m³. Therefore, the electricity cost of the complete ZLD process would be 7.39 ∼ 10.39 RMB/m³ using the seeded precipitation assisted NF process to remove hardness in FGD wastewater.

By summarizing the data above, it can be concluded that the whole cost of the ZLD process using the seeded precipitation assisted NF system for the FGD wastewater would be 21.30 ∼ 24.30 RMB/m³ if the cost of the concentration module and curing module were included, since little chemical agent is needed for the concentration module and curing module. Moreover, high purity gypsum and sodium chloride would be produced during the ZLD process, which could be sold as by-product and reduce the production of solid waste. Thus, the seeded precipitation assisted NF process is a cost-effective and attractive pretreatment module for treatment of FGD wastewater to remove hardness and recover salt.

The seeded precipitation assisted NF system was firstly introduced into the ZLD process of FGD wastewater to reduce the cost of chemical softening, and a corresponding pilot system was designed, built and run for 168 h in a coal-fired power plant. The concentration of magnesium ion and silicon was decreased to less than 24 mg/L and 0.15 mg/L through adjusting the reaction pH to 11.2 with calcium hydroxide. High purity gypsum was got from the ATC unit, of which the purity was 95.8%. The pressure and permeate flux of the UF and NF unit kept stable, which revealed that no scaling occurred in the units during the 168 h test. Full process mass balance was carried out in the test, in which the water recovery and hardness removal rate of this system reached 96% and 98.9%, respectively. By carrying out economic analysis, the softening cost of this system was about 13.30 RMB/m³, which was much less than that softened by sodium carbonate. The results showed that the seeded precipitation assisted NF process is technically feasible and economically viable for softening of the FGD wastewater.

This study is supported by CHN ENERGY group with the project number of KJ9300000562.

All relevant data are included in the paper or its Supplementary Information.