Design deviations to blame at Tampa
The operational difficulties encountered with the pretreatment system for the Tampa Bay Desalination Plant, following its completion in January 2003, threw doubt into the minds of many who were about to embark on seawater reverse osmosis (SWRO). However, as since shown by SWRO successes elsewhere, Tampa's pretreatment was a unique case. Brett Boyd and Dominic Janssen of Parkson Corporation explain why.In December 2000, a new company assumed the construction management responsibilities for the Tampa Bay SWRO desalination project in Florida, USA, after one of the original proposers filed for bankruptcy. The pilot pretreatment system was based on a continuously backwashed upflow dual-sand filtration (CBUDSF) technology which had been successfully completed at Tampa using a two-stage DynaSand® Filter.
Based on the positive results from the pilot test, which indicated favourable economics compared with other available options, the construction managers ordered components for 128 deep-bed and 128 standard-bed model DSF-50 DynaSand® Filter modules from Parkson Corporation.
This pretreatment system was designed to produce a filtrate with a 4.0 SDI or better at design conditions. The Tampa SWRO pretreatment system with the two-stage DynaSand® Filter as its centrepiece was proven capable of meeting the required filtrate quality despite numerous design deficiencies, many of which were the result of design modifications outside the original pilot study.
Some of the main RO pretreatment areas that Parkson believed were not properly addressed at the Tampa installation include:
- Screening of feed
- Mixing of flocculant
- Polymer use
- Filter cell feed flow distribution
- Materials of construction
- Loading rate of stage 2
- Bed turnover of stage 2
- Process control
There are two feed water sources for the Tampa Bay Desalination Plant:
- Prechlorinated cooling water from the power plant
- Cooler, unchlorinated water pumped directly from Tampa Bay.
The rampant growth of Asian green mussels and other biota on the inner walls of this second feed pipeline adds an unnecessary challenge to the filters. The ability to chlorinate at the head of this secondary intake pipe when it is in use and the ability to drain the pipe when it is not in use would greatly reduce this problematic biogrowth.
Screening of feed
Both the power plant cooling water and the supplemental feed-water supply, which is pulled directly from the bay, pass through 1.25cm mesh screens prior to entry to the power plant. However, fish chunks up to 30cm long and other foreign objects routinely enter the feed to the plant from not one, but both sources of desalination feed water. This phenomenon has the following possible explanations:
Primary feed source
Tampa Electric Company (TECO) screens its intake before using it as cooling water. These screenings are then dumped back into the discharge line downstream of the power plant before being returned to the bay.
The plant's intake is unfortunately located downstream from this screenings dumping location. These screenings are composed of an assortment of aquatic plant and animal matter, and are quite voluminous at times.
This ineffective SWRO feed intake location creates the conditions for completely unscreened influent entering the desalination plant. The desalination plant intake is also close enough to the discharge point, that large fish are capable of swimming up and getting sucked in by the pumps, particularly during periods of high tide.
Supplemental feed source
The supplemental feed water supply, which also passes through the same 1.25cm mesh screens as the primary feed-water source, also delivers a large quantity of problematic aquatic plant and animal matter. This occurs due to the fact that after passing though the 1.25 cm mesh screens, the water enters a sizable wet well.
This wet well serves as a breeding tank for an assortment of aquatic species, including large fish. When this supplemental Desalination feed water source is used, the intake pumps take in the water and marine life from this wet well and deliver them to the desalination pretreatment system.
Evidence of this was gathered by the use of makeshift feed screens installed in mid-2003 ahead of the pretreatment system. It can be inferred that countless large objects entered the pretreatment system, and ended up in the filter beds.
Normal DynaSand Filter operation requires that the entire media bed including its contents pass through the 3.8cm diameter airlifts several time as day. Presence of large objects such as the fish explains why airlifts at this facility repeatedly lost prime, resulting in bed plugging, channeling, and poor performance.
Mixing of flocculant
The original pretreatment system design called for mechanical mixing to provide an even dispersion of flocculant throughout the feed water, however, the flocculant-mixing step was never installed. As a result of this design change, the bulk of the dense flocculant immediately settled to the bottom of the 6m-deep feed channel prior to the filters.
On occasion, a sudden increase in feed flow would then stir this settled flocculant, creating a massive overdose of flocculant in the feed to the DynaSand Filters. As a result, the flocculant was either greatly under dosed or overdosed in the feed to the DynaSand Filters.
Similar problems were experienced as a function of excess amounts of polymer being fed to the filters in the early months of the plant's operation. The presence of sticky gelatinous masses on the tops of the media beds in late 2003 indicates that previous polymer feed was either inadequately mixed into the feed and/or polymer dosage was in excess of what was required for flocculating the feed solids.
Such events of excess polymer dosage undoubtedly rendered large portions of the media beds useless through solidification, thereby reducing the effective cross-sectional area of the filters. This solidification could be broken up through air lancing of the media bed, however, the 10cm-thick concrete covers at this facility prevent reasonable access to use the air-lancing method.
Filter cell feed flow distribution
Proper and even flow distribution between all filter cells is critical to ensure optimum performance. At design flow, the surface velocities present in the influent channels at the Tampa plant are substantial.
In contrast, the bottom half of the 6m-deep channel is virtually stagnant. This influent channel flow pattern leads to the first filter cell receiving up to 50% greater flow compared to that of the last cell. Such maldistribution is detrimental to optimum filter performance.
Materials of construction
The majority of internal filter components were specified as PVC, fibreglass, or AL6XN alloy. However, the materials of a number of components were not so carefully and properly selected.
For example, the original design and specifications have called for a hard-pipe connection between the pipelines evacuating the filter backwash (reject) and the backwash troughs of the individual filter cells. Rather than following the original design, flexible rubber connections were installed instead.
The hose clamps for the rubber Fernco boots, used to connect the reject evacuation pipes, were specified as 304 SS stainless steel. These clamps failed within a month of their installation as a result of direct and continuous exposure to seawater.
Furthermore, the filter-cell concrete walls were not properly protected from the seawater with the coating specified in the original design and within months of operation the walls began loosing material into the media beds. Some of the module dividing-walls, made of concrete grout poured between glass-fibre cones, were short as much as 8cm of grout in less than a year of operation.
Loading rate of stage 2
According to the engineering design for the Tampa desalination plant, the pretreatment system was to handle 2.12m3/s so that the RO membrane system could produce 1.1m3/s of permeate (based on an RO retentate/permeate ratio of 50/50). However, upon start-up, the pretreatment system was expected to operate at even higher loading rates.
Both first and second stage filters were required to perform at a design-flow surface loading rate of 14.7m/h and higher. For the stage 1 scalping filters, which contain 1.1mm effective size media, this loading rate is conservative and proper. However, for the stage 2 polishing filters with the 0.5mm effective size media, this 14.7m/h is outside of the recommended operating range. The hydraulic loading rate for stage 2 filters should not exceed 12.3m/h at design conditions.
Bed turnover of stage 2
The performance observed from a two-stage filtration system is greatly improved when the media bed turnover rate for Stage 2 is significantly less than the normal rate of 25-50cm/h used for Stage 1. The adoption of a lower turnover rate in Stage 2 at the Tampa plant would contribute to a better filtrate quality, however, in light of all of the other challenges presented at this plant, lower Stage 2 turnover rates were not possible as the beds were already overloaded with feed solids.
Given the magnitude of this plant and the importance of high quality filtrate, it would seem prudent to have employed tighter control loops for all aspects of the process pretreatment. This real-time responsiveness would greatly assist in protecting the downstream cartridge filters and RO membranes from potential upsets.
Deviations from the original project design and specifications, along with construction shortcuts at the Tampa Bay Desalination facility have detracted from the benefits of two-stage filtration as a pretreatment option for SWRO systems. However, the information contained in this paper has gone over in detail the numerous deficiencies in the design at Tampa, suggesting that no filter system could have performed properly under the circumstances.
With a properly designed system, such as the one being tested at the SWRO desalination pilot facility in Carlsbad, California, it becomes apparent that a state-of-the-art two-stage filtration system, such as Parkson Corporation.s DynaSand D2", does produce a filtrate quality that can be considered equivalent to that of a low-pressure membrane system. This performance, coupled with the economic benefits associated with simpler two-stage filtration technology, makes the DynaSand D2" an attractive option for reliable pretreatment for an SWRO treatment facility.
This is an updated extract from a longer paper about continuous backwash upflow sand filters presented at the International Desalination Association's World Congress in September 2005.