A pilot-scale investigation of ozonation and advanced oxidation processes at Choa Chu Kang Waterworks

Public utilities in tropical areas face the challenges of seasonal occurrences, variation in taste and odor compounds, and/or other emerging organic contaminants in their source waters. That leads to the application of treatment processes other than the conventional ones (Agus et al. 2011; Xu et al. 2011). Advanced oxidation processes (AOPs) are one of the recommended approaches. The key is to produce highly reactive hydroxyl radicals (OḢ) that can react rapidly with most organic compounds (von Gunten 2003a, 2003b). To study the feasibility of full-scale AOP application, Xylem and PUB have jointly conducted this comprehensive pilot study to obtain first-hand data on treating actual effluent from water treatment plant, which helps to evaluate the options. The test results will help decision makers upgrading plant design or emergency measurements to handle potential organic compound issues in the future.

The pilot plant (Wang et al. 2012) was manufactured and supplied by Xylem Water Solutions Herford GmbH, Germany (Figure 1). It is designed for a water flow of 5–20 m³/h and hydraulic retention times (HRTs) of up to 10 minutes. Ozone, UV, and hydrogen peroxide dosages are all adjustable at a predetermined water flow rate. In total, 19 organic compounds were spiked into the AOP inlet water for this study. They comprised two taste and odor (T&O) compounds (i.e. 2-methylisoborneol (MIB) and geosmin), four artificial sweeteners (e.g. sucralose), four endocrine disrupting chemicals (EDCs) (e.g. E1, E2, and EE2), eight pharmaceuticals and personal care products (PPCPs) (e.g. diclofenac and ibuprofen), and one industrial solvent (i.e. 1,4-dioxane) as listed in Table 1. During the pilot testing, the selected compounds were mixed in cocktail solution and spiked into water at the same time. Grabbed water samples were analyzed at the PUB laboratories (Zhang et al. 2006; Hu et al. 2008; Wu et al. 2008; He et al. 2012).

Table 2 is a summary of the inlet water characteristics during the pilot tests. Depending on the operating mode at the waterworks, the inlet water for each test trial was either sand filtered, or a mixture of sand filtered water and MF membrane permeate.

The test data shows that 10 compounds, including six PPCPs including carbamazepine and diclofenac, and all four EDCs can be effectively degraded by dosing with 2 mg/L ozone and 5 minutes contact time. For the other nine compounds including two T&O, two PPCPs (Figure 2 left), four artificial sweeteners (Figure 2 right), and the industrial solvent, higher ozone dosing rates or ozone-based AOP are necessary to achieve a removal rate as high as 90%.

Figure 2 shows some of the treatment results, with a fixed, 5-minute HRT. Under ozone-based AOP conditions, the generation of OḢ radicals was promoted and the removal efficiencies of the compounds increased significantly in comparison to those achieved with the use of ozone alone.

Figure 3 shows some of the treatment results with HRTs between 0.4 and 10 minutes. It clearly shows that in the ozone-based AOP case, faster and more aggressive reactions took place. Overall, 3 mg/L ozone-based AOP with only 2.8 minutes HRT achieved an equivalent or higher removal efficiency than 5 mg/L ozone only with 10 minutes HRT.

Figure 4 shows some of the treatment results with UV and UV-based AOP. It is clear that, UV-based AOP is much more effective in removing the organic compounds studied than UV alone. To achieve removal efficiencies comparable to ozone/ozone-based AOP, the UV and hydrogen peroxide dosages required were above 8,000 J/m² and 10 mg/L, respectively.

To get a first idea about the energy consumption required to remove some of the chemical species studied, we calculated the electrical energy per log removal (E_EO) for removing 90% (i.e. 1-LOG) of representative compounds, i.e. MIB, geosmin, and 1,4-dioxane.

‘3 mg/L ozone + 0.8 mg/L H2O2’ and 5 mg/L ozone only were chosen for the E_EO calculation for the ozone-based processes; and a UV dose rate of 8,742 J/m² with H2O2 concentrations of 5 and 10 mg/L, respectively, was used for the E_EO calculation for the UV-based processes. The calculations are based solely on the power consumptions of the ozone generator and UV lamps in the pilot system. The E_EO results indicate that ozone-based AOP is a more energy efficient solution for the oxidation of organic compounds in this specific water matrix – it should be noted that the vertical scales in Figures 5(a) and 5(b) differ by a factor of 10.

For full-scale ozone AOP plant design, depending on the scope of the components, e.g. types of feed gas, ozone generator, cooling system, injection system, H2O2 dosing station, etc., the full-scale E_EO value could be up to twice that of the pilot plant. For a full-scale UV AOP plant design, it is recommended that a computational fluid dynamics calculation is carried out in order to achieve a more energy efficient design than that of the pilot UV reactor. In general, even though the differences between the full- and pilot-scale systems were not taken into account, an ozone-based AOP will probably be more energy efficient than UV AOP in full-scale applications.

For the specific water matrix studied, ozone and ozone-based AOP, with their relatively low energy consumptions and hydrogen peroxide utilization rates, are superior to UV-based AOP in the degradation of the compounds selected.

Throughout the tests, no bromate formation was detected in the treated water. Thus, for this specific water source, both ozone and ozone-based AOP have been shown to be appropriate. The only question is determining the treatment goal in order to select between ozone only and ozone-based AOP processes in real applications. If organic contaminant removal is the treatment goal, an ozone-based AOP will definitely be superior as compared to the use of ozone only. If the conventional ozone disinfection goal has higher priority than the removal of organic contaminants, the ozone only operation might be superior to an ozone-based AOP. In this study, an effort has been made to find a balance between the two philosophies: the combination of appropriate ozone and low hydrogen peroxide usage with certain shortened reaction times could be a suitable option for utilities. This idea could be further supported by a follow-up disinfection effect study under optimized ozone-based AOP conditions.

As the optimized ozone and hydrogen peroxide dosages for the specific water matrix were obtained in this pilot-scale test, a preliminary cost estimate could be made for a waterworks intending to install a full-scale ozone or ozone-based AOP application. Table 3 shows the preliminary CAPEX and OPEX estimations at a water flow of 7,600 m³/h, which is also the water flow of one treatment line at Choa Chu Kang Waterworks.

Figure 6 illustrates the calculation results from Table 3 to enable comparison between the options. The economic advantage of ozone-based AOP is further proved in terms of achieving the same T&O removal goal.

The following recommendations can be made on the basis of the results of the pilot study:

• MIB is recommended for use as an indicator compound to set the treatment goal(s) for this specific water source under tropical climate conditions. This study has shown that 90% (i.e. 1-LOG) can be a suitable and realistic goal for treatment processes.

• For a waterworks like Choa Chu Kang Waterworks, if there is an existing ozone contactor with more than 10 minutes contact time, ozone only with a dosage of 5 mg/L shall be efficient to control MIB, or in other words to control T&O. However, with the addition of hydrogen peroxide, the OPEX on site can be reduced.

• For an upgrading design of new waterworks without existing ozone contactor, ozone-based AOP design can be a good option. In this case, the designed ozone dosage can be reduced to 3 mg/L and the designed contact time can be reduced to 5 minutes or even shorter.

We thank Xylem Water Solutions and PUB, Singapore's National Water Agency for financial and professional contributions to this study.

Thanks to Mr Yongjun Xiao, Ms Xiaoqing Qian, Dr Jingming Wu, Mr Ruikang Hu, Dr Gao Pingping from labs of PUB for quantitative and sensory analysis, and important advice on the analysis of all water samples.

Special thanks to Mr Robin Wong for setting up this research collaboration and Ms Adeline Choo for local support from Xylem Singapore.