UV Disinfection Knowledge Base [Project #3117]

Transcription

1 UV Disinfection Knowledge Base [Project #3117] ORDER NUMBER: 3117 DATE AVAILABLE: August 2012 PRINCIPAL INVESTIGATORS: Harold Wright, David Gaithuma, Mark Heath, Chris Schulz, Travis Bogan, Alexander Cabaj, Alois Schmalweiser, Marcia Schmelzer, and Janet Finegan-Kelly OBJECTIVES: The goal of this project was to develop a knowledge base on the practice of drinking water UV disinfection. BACKGROUND: The practice of drinking water UV disinfection has evolved considerably over the last 10 years in terms of regulations, science, and technology. Accordingly, utility personnel, engineers, and state regulators have questions who is installing UV disinfection, what are the design criteria, how efficient is UV dose monitoring and dose pacing, how reliable are UV system components, how much labor is involved in maintaining UV systems, and what are the lessons learned from UV implementation. APPROACH: The project approach involved the following tasks: Identify issues and questions with drinking water UV disinfection from participating utilities, regulators, and consultants. Collect and analyze UV system data through surveys of participating utilities and field evaluations of installed systems. Conduct an evaluation of mercury release with the breakage on low-pressure highoutput (LPHO) and medium-pressure (MP) lamps and develop engineering approaches for mitigating mercury release. RESULTS/CONCLUSIONS: Survey data collected during the spring of 2008 indicates that 161 utilities in Canada and 148 in the United States have installed or are implementing drinking water UV disinfection at plant flows greater than 0.5 mgd. Seventy-eight percent of UV systems were installed for Cryptosporidium and/or Giardia inactivation, 12% for virus inactivation based on a UV dose of 40 mj/cm 2, and 9% for heterotophs, total coliform, or bacteria inactivation. Twenty-four percent of the UV systems were treating groundwaters while 76% were treating surface waters. Design flowrates ranged from 0.03 mgd to 2,200 mgd and the total design capacity was 6.1 bgd. The design UVTs range from 70 to 98% with a median value of 90%. The histogram of design UVT shows distinct peaks at 85, 90, and 95% UVT suggesting that utilities are either guessing at their design UVT values or rounding off the measured UVT to one of these values. The limited

2 data shows lamp aging and fouling factors used for UV system sizing ranging from 60 to 90%. Fifty-two percent of UV systems treated the combined filter effluent while 14% treated individual filter effluents in the filter gallery. The MP system headloss ranged from 0.8 to 30 inches with a median value of 5.9 inches. LPHO system headloss ranged from 6 to 30 inches with a median value of 15 inches. Seventy-three percent of systems used MP lamps and 27% used either LPHO or amalgam LPHO UV lamps. The nominal power of LPHO and MP lamps ranged from 240 to 427 W and 2.4 to 21.6 kw, respectively. The average ratio of amalgam to MP lamps was 10 to 1. MP UV systems were typically equipped with automatic wipers while LPHO systems used automatic mechanical or physchem wipers, offline acid cleaning, or manual cleaning. Offline acid cleaning chemicals included phosphoric or citric acid. Utilities indicated that 25% of LPHO systems and 53% of MP system required additional manual cleaning. UV systems used either DVGW or ÖNORM-compliant or proprietary UV sensors. ÖNORM sensors were used more with LPHO systems while DVGW sensors were used more with MP systems. Only 69% of utilities stated that they used reference UV sensors to check duty UV sensors, and a majority of utilities reported having only one reference UV sensor. Forty-eight percent of the utilities used flowmeters on each reactor train while 40% used a single flowmeter on the combined flow. Sixty-three percent of utilities reported using an on-line UVT monitor, and only 52% report conducting UVT monitor checks. The data indicates a need for improved QA/QC with operating UV systems. Forty and 25% of U.S. and Canadian systems have offspec requirements. Reporting with U.S. systems includes UV dose and flowrate (78%), UV sensor readings and check data (55%), UVT and check data (44%), and off spec performance (33%). In contrast, Canadian systems report UV dose and flowrate data (90%), UVT data (60%), UV sensor data (10%), UV sensor and UVT monitor check data (5%), and none report off spec performance. The differences between U.S. and Canadian systems likely reflect the influence of the LT2ESWTR and UVDGM, which is greater with U.S. systems. The number of reactor trains with LPHO and MP systems ranged from 1 to 56 and from 1 to 15 trains, respectively. Surprisingly, 18 and 34% of LPHO and MP systems, respectively, used only one train (i.e., no redundant reactor). All LPHO systems were installed with one reactor per train whereas 5% of MP systems had more than one reactor per train, either to deliver higher UV dose or meet NWRI UV Guidelines. Sixty-two percent of UV systems with more than one train used modulating flow control valves and the remainder used passive flowsplit. Most LPHO systems were installed with horizontal reactor alignment while 79% of MP systems were installed with horizontal alignment and 21% with vertical alignment. Twenty-five and 35% of LPHO and MP systems, respectively, conducted a UV pilot study.

3 The labor hours per month as 10 th and 90 th percentiles ranged from 1 to 20 hours per month with LPHO systems and 1 to 10 hours per month with MP systems. Fifty percent of LPHO systems and 41% of MP systems observed sleeve and UV sensor port window fouling. The median ratio of spare lamps, sleeves, ballasts, and UV sensors was 12:2:3:1 with LPHO systems and 4:2:1:1 with MP systems. While utilities have observed issues with UV disinfection, overall they report that their UV systems are effective and simple to use and maintain, with UV component performance exceeding warranties and vendors providing good service. UV system performance was also evaluated with eight installed UV systems. Many of those systems did not have access to validation reports and did not understand their UV dose-monitoring algorithm. One system used an unvalidated algorithm. Over dosing was common (e.g., factor of 2 to 3), due to use of unnecessary safety factors, inefficient implementation of UV sensor setpoint monitoring approach, or limitations on reactor turndown. The combined lamp aging and fouling (CAF) index, calculated as the ratio of the observed UV sensor readings to those predicted using the UV sensor equation, was used to quantify lamp aging and fouling. Data showed lamp aging well within design criteria. Fouling was site specific with some locations experiencing little if any fouling, even with MP lamps, while others show significant fouling over time. At locations where fouling was observed, the automatic mechanical and physical-chemical wipers used with MP lamps kept the sleeves and UV sensor port windows clean. However, internal sleeve fouling was an issue with MP systems. Significant fouling was observed with LPHO systems using offline acid cleaning. It appears operators may not be motivated to clean the reactor if the PLC indicates the reactor is delivering the required UV dose even though the fouling significantly impacts O&M costs. The site visits showed significant issues with the accuracy and calibration of on-line UVT monitors (i.e., UVT errors > 2%). The impact of UVT monitor errors on UV dose monitoring can be determined using the UV dose-monitoring algorithm given in the reactor s validation report. If the impact is unacceptable, the UVT measured by a lab spectrophotometer can be entered into the PLC and used for monitoring. UV sensors used with MP systems showed more variability compared to those used by LPHO systems. Field calibration of MP UV sensors, by comparison to reference UV sensors, reduces that variability. Checking wet UV sensors (i.e., sensors in direct contact with the water) was difficult and time consuming because the reactor had to be drained. These UV sensors can be retrofitted using dry UV sensors that use a UV sensor port. Mercury release following a lamp breakage event was evaluated using a pilot reactor and showed that mercury transport following a lamp break depends on lamp type and operation. During a lamp break, the vapor phase mercury dissolves into solution and is carried downstream of the reactor whereas liquid and amalgam mercury settles to the bottom of the reactor. The mass of vapor phase mercury with an operating LP and LPHO

4 lamps is orders of magnitude less than an operating MP lamp because the former operate at much lower temperatures. A mercury mitigation plan should include prevention, detection of lamp breaks, modeling of mercury release and transport, capture and containment, sampling, treatment and disposal of contaminated water, and cleanup and re-commissioning of the UV reactor. Resonant sleeve vibration is a cause of lamp breaks not mentioned in the UVDGM. The mercury concentrations in piping and basins downstream of the reactor following a lamp break can be predicted using advective-dispersion equations, CFD-modeling, or residence time distributions obtained from tracer studies. The models predict that dispersion of mercury following a LP or LPHO lamp break will reduce mercury concentrations to orders of magnitude below the MCL. Dispersion following a MP lamp break may only decrease concentrations below the MCL following long transmission lines, basins, or reservoirs. If dilution is not sufficient, the mercury release may be contained using downstream valves or diversion, the location of which should account for the response time for detecting the breakage event and valve closure times. Valve closure times may be limited by water hammer, which can cause water pressures to exceed the bursting pressure of the sleeves. A response plan to a lamp breakage event should describe sample locations, frequency of sampling, treatment and disposal of contaminated waters, and cleanup of liquid or amalgam mercury and quartz shards mercury remaining within the reactor. Sampling locations downstream of the reactor should be selected based on the concentration profiles predicted using transport models. Samples should be analyzed using EPA Methods 1631E and 245.7, with a detection limit of 5 ng/l or less. Mercury contaminated water contained within the reactor may be treated by passing the water through sulfur impregnated activated carbon. Many states specify a maximum concentration of 12 ng/l mercury for the protection of aquatic life, and watershed specific TMDLs may specify below 1 ng/l. RECOMMENDATIONS: The project team and the participating utilities identified numerous recommendations for utilities planning to implement UV disinfection. Utilities implementing UV disinfection should consider the UV dose requirements specified within the 2006 UVDGM based on optimized validation test microbes as an alternate to an MS2 dose of 40 mj/cm 2. Utilities should collect a robust data set on UVT to define design criteria. Lamp aging and fouling factors used for design should also be clearly defined by the design team. UV design recommendations include providing a redundant reactor for maintenance, using high flow reactors to minimize the number of reactor trains, using large drain lines with large reactors, balancing building capital with requirements for upstream and downstream straight pipe lengths, using automatic motorized isolation valves, sending offspec water to the plant headworks, equipping MP systems with wipers even with waters with low iron levels, providing dedicated space for spare parts and maintenance tasks, and having operators and plant personnel participate in all phases of design and

5 construction. Off-site validation is recommended over on-site validation due to challenges obtaining required flows and UVTs, and water disposal. Utilities should have clear documentation describing the UV dose-monitoring algorithm used by their UV systems and confirm the algorithm programmed into the PLC matches that provided in the validation report. Utilities should evaluate and improve the efficiency of UV dose monitoring and control to reduce over dosing and associated O&M costs. Utilities need to improve UV system operational QA/QC, including UV sensor and UVT monitor checks. Wet UV sensors are not recommended and reference UV sensors should use electronics independent of the reactor. Criteria for UVT monitor accuracy should be based on the impact of that error on UV dose monitoring. UV system operators should calculate the CAF index for their reactors on a weekly basis and use the results to optimize UV system operation and maintenance such as manual cleaning or lamp replacement. Operators should inspect sleeves for internal fouling. Utilities report that UV disinfection requires regular maintenance it is not a low maintenance technology. Utilities report not planning for enough operational staff, and recommend having a dedicated maintenance technician. Overall, the project identified a need for improved training with UV system operators. Utilities should implement a mercury response plan that accounts for the transport of mercury expected with LPHO and MP lamps. The plan should prevent resonant sleeve vibration and address detection of the lamp and sleeve breaks, response time of valves and water hammer, sampling locations and method detection limits, and discharge limits and the need to treat mercury contaminated water. MULTIMEDIA: The project report includes a Microsoft Access database containing data collected on drinking water UV systems with participating utilities.