Full-scale experience with the membrane bioreactor-reverse osmosis water reclamation process

The combination of conventional activated sludge followed by ultrafiltration and RO is widely accepted to reclaim water for reuse from municipal wastewater. It has been further hypothesized that combining activated sludge and ultrafiltration in the MBR process might prove more cost-effective and provide operational and performance advantages when coupled with RO, as compared to the more widely demonstrated activated sludge, ultrafiltration, and RO municipal wastewater reclamation process train (Lozier & Fernandez 2001; Comerton et al. 2005; Qin et al. 2006; Freeman et al. 2011; Moreno et al. 2013; Farias et al. 2014a, 2014b). More stable biological treatment is expected due to the retention of biomass that occurs in an MBR, due to the relatively long solids retention time (SRT). However, circumstances have not resulted in the construction of many full-scale municipal wastewater reclamation facilities using MBR followed by RO. Consequently, full-scale experience from the GWF can provide valuable insight into the operational and performance characteristics of this water reclamation process train.

The GWF domestic MBR began operation in late 2009, with stable operation since late 2010. The RO system has been operational since early 2012. A detailed evaluation of plant performance was completed in 2012 (Gippsland Water Factory 2012), and a subsequent evaluation of the domestic MBR was completed in 2014 (CH2M HILL 2014). This paper presents the results of those evaluations, along with additional operational and performance results through 2014 so that they may be compared to the operational and performance results from other water reclamation facilities.

Table 1 summarizes the major facilities that comprise the GWF domestic treatment train. While the design average capacity of the facility is 16,000 m³/day, served by four membrane cells, the peak wet weather capacity is approximately 40,000 m³/day. Treatment of the peak wet weather flow can be accommodated in the MBR by use of the eight industrial membrane cells for domestic treatment duty during peak wet weather periods (membranes only, not the industrial bioreactor), effectively increasing the available membrane area by a factor of three (i.e., from 4 cells to 12 cells). Facilities are available to store industrial wastewater during such events, allowing the use of the industrial membranes for peak wet weather service. Twelve Memcor Memjet submerged hollow-fiber ultrafiltration membrane racks were initially installed in each of the 12 membrane cells. However, for a variety of operational and commercial reasons the 12 Memjet racks in each of the four dedicated domestic modules were replaced with 15 Mempulse racks.

The Memcor Mempulse racking system is a newer format for the Evoqua (formerly Siemens) submerged UF membrane modules in which agitation air rises through the submerged fiber bundles in sporadic large bubbles or pulses. This agitation method produces a more beneficial boundary layer clearing effect at the membrane surfaces than the older Memjet system, in which a continuous stream of air was injected into the bottom of the membrane bundles. A key additional feature of the Mempulse agitation format is that is uses less agitation air than the Memjet system, and thus is less costly to operate.

These changes were completed by early 2011. The Memjet agitation format remained in the eight industrial membrane cells, although extra membrane racks were added to some cells and progressive conversion of all cells to the Mempulse format is underway.

The RO facility is sized to produce 8,000 m³/day of product water on a yearly average basis, based on 75% recovery of the domestic filtrate feed water. Based on prior experiences of the design team, an availability of 93.6% was assumed, making the required daily production capacity 8,550 m³/day. A balance tank for raw domestic wastewater is provided to capture diurnal peak flows during dry weather conditions so that the RO plant can continue operating during daily lower flow periods.

Each bioreactor consists of three passes and is configured with three zones to provide biological nitrogen and phosphorus removal (Daigger et al. 2013, 2015). Including the aerated volume in the submerged membrane cells, it is configured as a four-stage Bardenpho facility consisting of initial mixed zone, main aerated zone, second anoxic zone, and final aerated zone in the aerated submerged membrane cells. The initial mixed zone is 28% of the bioreactor volume, the main aerated zone 48%, and the second anoxic zone 24%. The aerated cells in the membrane tanks add a further 4% volume to the system. Recycle from the submerged membranes is directed to the main aerobic zone rather than the initial mixed zone because of the elevated dissolved oxygen (DO) concentrations it contains. Mixed liquor (ML) recirculation from the main aerobic to the initial mixed zone at a rate of four times the design average flow is also provided. Process modeling during design indicated that biological phosphorus removal would occur, even though a dedicated anaerobic zone was not provided. Ferric chloride feed capability was also provided as a back-up, although it has not been used as sufficient phosphorus removal has been achieved as predicted. This performance is described further below.

Much of the data presented were obtained through routine operation of the full-scale GWF using standard sampling and certified analytical procedures. Details of these procedures have been documented elsewhere (GWF 2013; Daigger et al. 2013), and interested parties are referred to these documents for further details.

The domestic treatment train, except for the RO facility (referred to here as the domestic train), was fully operational with the revised submerged ultrafiltration membrane racks by late 2010. In contrast, the RO facility did not become fully functional until early 2012. The domestic train performed well (as described below), and only modest efforts were made to optimize its performance through 2012, including the period included in the overall detailed plant evaluation. Further efforts were made to improve the performance of the domestic MBR in 2014, as described below. The performance of the RO facility was characterized during much of 2012. It was run periodically during 2013 and 2014 as it was not needed as a water supply by GW during this time period due to relaxation of the previous drought conditions. The performance of the domestic train and the RO facility are summarized below.

Domestic train influent flows and constituent loadings for the intensive evaluation period of December 2010 through October 2012 are compared in Table 2 with the design values and demonstrate that the domestic train was essentially loaded to design average values through this period. The average influent flow was marginally lower than the design value, while constituent loadings exceeded the design values modestly for most parameters. Thus, this period is appropriate for assessing the capability of the plant under full design load. Table 3 summarizes domestic MBR influent (primary effluent) flows and constituent loadings for the same period, compared to the design values, further confirming that the domestic train MBR was loaded to its design values. Flows and constituent loadings were similar in 2013 and 2014, indicating again that the domestic train MBR was consistently loaded to its design values.

A further issue is associated with the pressure decay rate (PDR). The Victorian water quality regulator required that the GWF demonstrate a log removal value for virus (LRV) of six when operating in the reclamation mode. This was partially accomplished in the initial design of the GWF by including a requirement to maintain the domestic MBR membrane PDR less than 7 kPa/min (GWF 2010). Experience demonstrated, however, that the frequency of membrane repair (pinning) required to maintain this level of integrity was neither practical nor necessary. With the full-scale plant in operation, it was demonstrated that the required level of overall reclamation treatment train LRV could be demonstrated by monitoring total organic carbon (TOC) removal across the downstream RO system, thereby gaining two logs of virus removal by the overall treatment system. This relieved some of the treatment performance validation requirement from the ultrafiltration system and the PDR requirement was relaxed to 70 kPa/min, which, although still an onerous performance requirement, permits a practicable level of membrane maintenance activity. Table 5 summarizes LRV values provided by the facility with this change.

Energy requirements for the submerged membranes were initially on the order of 1 kWh/m³ for operation with the Memjet system. However, the energy consumed declined to just under 0.4 kWh/m³ as the Mempulse system became fully operational and the overall process was optimized. Although the Mempulse system does require less agitation blower energy, it should be noted that much of the energy saving reported here, although not all of it, is thought to be attributed to the optimization efforts.

The RO facility became operational in early 2010. However, addressing filtration integrity issues associated with the MBR was generally the focus during 2010 and 2011, and this led to infrequent RO operation through this period. Improved filtration integrity, and the resulting decrease in turbidity, was achieved by late 2011, allowing more consistent operation of the RO system for process proving and optimization purposes during 2012. The RO system was operated only intermittently during 2013 and 2014 as reclaimed water was not needed by GW to meet the overall water supply needs of its service area. Consequently, this analysis focuses largely on operation during 2012.

A detailed summary of RO influent quality is presented in Table 6 for the period of January through October 2012, compared to the specified design influent quality. Both 50th and 90th percentile values are presented. Influent values exceeding the specified design values are indicated in bold. The principal issues indicated by these data are iron and manganese (because of potential oxidation and fouling of the RO membranes), and nitrogen and phosphate species (because of the stringent reclaimed water discharge standards and potential for precipitation of calcium phosphate within the second stage of the RO system). Actual total dissolved solids (TDS) values were significantly less than the design value. Even considering that operating temperatures were generally lower than the 90th percentile design value of 26 °C, the head on the RO feed pumps was more than sufficient to achieve the specified flux and recovery at the observed TDS values.

Table 7 summarizes 50th and 90th percentile flux and recovery values for the two RO trains for 2012, indicating that they were generally operated at reasonable flux and recovery values. Analysis of sparingly soluble salts indicated that calcium carbonate, calcium phosphate, barium sulfate, and silica were supersaturated in the RO concentration at 75% recovery for both the 50th and 90th percentile feed water concentrations, but this was controlled by anti-scalant addition (GWF 2012). Analysis of feed water data indicated some concern related to iron precipitation. An analysis of normalized product flow, normalized differential pressure, and normalized salt passage for the entire period of operation (2010 through 2012) indicated little evidence of fouling or increased salt passage (GWF 2012), although it was decided to clean the membranes with both sodium hydroxide and citric acid in mid-2011 to facilitate commissioning of the RO cleaning systems. Biofouling of the RO membranes is controlled by continuous dosing of chloramines which acts to suppress biological growth within the RO system.

A total of 49 effluent quality parameters are specified for the RO product water, but not all must be routinely monitored because many are expected to be consistently below the specified values as long as design influent values are not exceeded and membrane integrity is maintained as demonstrated by compliance with critical control points. Table 8 summarizes performance for the parameters routinely monitored and generally demonstrates routine compliance with the required performance.

Energy use for RO averaged 0.73 kWh/m³ of product water in 2012. The total energy usage for both MBR and RO treatment was calculated including the bioreactors, MBR membranes, and the RO system. It averaged 3.04 kWh/m³ of product water for the same period.

Performance results for the MBR-RO water reclamation train at the GWF demonstrate the robustness and resilience of this process combination. In spite of significant membrane filtration system integrity issues experienced with the domestic MBR during its initial operating phase, reliable performance of the MBR-RO train was achieved when sufficient wastewater volume was available to allow for routine operation. Consistently good quality reclaimed water was produced, with occasional deviations from the desired 90th percentile values. Effluent nutrient (TN and TP) values were consistently below the very stringent limits specified, even though consistent performance by the MBR has not yet been achieved. Although the ability to add ferric chloride to the MBR for further TP control was provided, overall performance indicated that this was not needed and has not been practiced. While membrane integrity issues adversely impacted operation and performance initially, these issues are now considered resolved.

Operating experience with the RO system demonstrated the need to continuously maintain the analytical systems (instruments, SCADA, and controls) which support it. Intermittent operation of the RO system sometimes led to difficulties with these components of the system when it was started up after a period of inactivity. Operational procedures have been developed by plant staff to more routinely verify the readiness status of these system components, and they are now routinely returned to service with little difficulty. Comprehensive management of process assets for operational readiness should be a key component of the operational plan for any RO system, not only one following MBR. The DoH requirement to demonstrate membrane integrity via online TOC measurement adds an additional instrumentation requirement not typically required for RO systems utilized for removal of TDS or specific inorganic constituents. Likewise, RO systems are capable of removing a wide range of constituents to low levels. This does not mean that all constituents should be monitored as such a practice leads to excessive analytical costs and adds little value as the tendency is to not make use of the data collected. Routine operation demonstrating that RO membrane integrity is maintained (including online monitoring of both conductivity and TOC removal, as implemented at the GWF) can be supplemented by occasional confirmation of effluent quality.

In spite of initial operational and performance issues, the GWF has operated successfully and met its performance requirements. Water reclamation is practiced only when required to supplement the regional water supply during periods of drought. Operation, to date, has allowed GW to gain a full understanding of the operational procedures required to achieve the intended capacity and performance. Consequently, the facility is fully available when needed as a drought-proof supplemental water supply. MBR membrane integrity issues were unexpected but have been successfully dealt with, and it is understood that the lessons derived from this experience have been applied elsewhere by the membrane supplier. GW owns and operates other wastewater treatment facilities which use activated sludge processes with clarifiers, and consequently has a basis for evaluation of the decision to use the MBR process at the GWF rather than a more conventional activated sludge process followed by tertiary membranes and RO. GW is fully satisfied with selection of the MBR process for the GWF, and on the basis of several years’ operational experience at GWF, it is reasonable to conclude that the combination of MBR and RO is a viable means of producing reclaimed water from municipal wastewater at a very high quality standard.