By: Ted Prato, Ionics, Inc., Jeff Purvis, Duke Energy Hinds, LLC, and Brian Ahern, Maine Independence Station
There are many different process designs available for producing the high-purity water needed at modern power plants. Although not typical, initially successful system performance may deteriorate for a wide variety of reasons, including fouling of ion-exchange media, deterioration of membrane performance, and feed changes. This article profiles two combined-cycle power plants, owned by Duke Energy, where the original water treatment systems were not able to meet required specifications. Replacing the original equipment with current water treatment technology quickly led to effective production of in-spec water.
The Maine Independence Station in Veazie, Maine, is a two-on-one 520 MW gas-fired combined-cycle facility that sells power in the Northeast Power Coordinating Council. The Hinds Energy Facility in Jackson, Mississippi, is a nearly identical two-on-one 520 MW gas-fired combined-cycle plant that sells power into the Southeastern Reliability Council and surrounding regions. Maine Independence Station entered commercial operation in May 2000 and Hinds came on-line in May 2001.
The Maine Independence plant uses surface water for feed to the water treatment system, while Hinds Energy uses well water. Table 1 shows average feed analysis at the two plants and includes typical power plant treated water specifications. A total organic carbon (TOC) specification is not always required. Specifications for aqueous species like chloride, sodium and sulfate may vary from site to site.
These analyses present a moderate water treatment challenge compared to more aggressive high salinity waters or wastewater feeds like secondary sewage. Feed conductivity is fairly low at both sites, and species like silica and calcium are of moderate concentration. In the absence of significant changes in feed water quality, an adequately designed water treatment system should be able to provide good quality product over a long period of time.
At both power plants, the original water treatment vendor installed a packaged water treatment system sized to the flow requirements. These systems included a number of unit operations designed to cover a range of water treatment challenges. At Maine, the process train included two greensand filters, a heat exchanger, two cartridge filters, three-stage reverse osmosis (RO), electrodeionization (EDI), and backup ion exchange (IX). At Hinds, the process train consisted of multi-media filtration, reverse osmosis, EDI, and backup ion exchange.
The purpose of the manganese greensand filter at Maine Independence was to oxidize and remove iron and particles from the feed, trapping the particles in the media. Greensand product fed the RO heat exchanger and was heated to a constant 73 F for optimization of the RO system. The RO system produced 37 gpm of RO permeate feed for the EDI system.
The purpose of the multi-media pretreatment at Hinds Energy was to filter out particles through a series of layers of differently sized media including gravel, sand, garnet, and anthracite. Often a polymeric coagulant was dosed to increase filtering efficiency.
Both water treatment schemes relied on EDI technology as a critical process step. Using RO permeate as feed, the EDI process converts aqueous neutral or weakly-ionized species to ionized species that can then be easily removed by a DC electrical field.1
EDI systems are all based on similar components. A combination of semi-permeable membranes, flow-directing spacers and a diluting compartment filled with mixed-bed ion-exchange resin comprises the heart of the system. The EDI diluting compartment contains a strong ion-removal region near the inlet and a polarized region adjacent to the spacer outlet.2 Chemical conversions occur in high or low pH micro-regions on either anionic or cationic ion-exchange resin beads, creating ionized species for easy removal. An example is the conversion of dissolved CO2 gas to bicarbonate ion by addition of a hydroxyl ion.
The combination of the electrodialysis function with the polarized region of the EDI diluting compartment enables very high percentage removal of strongly ionized, weakly ionized and neutral aqueous species. The EDI systems at Maine and Hinds consisted of three process modules, each with a nominal flow rate of 12.5 gpm, close to the system design flow of 34 gpm. Nominal product quality was 10 megohm-cm with less than 10 ppb silica, and less than 3 ppb of sodium and chloride.
The water treatment trains at Maine and Hinds worked well initially, but product quality declined and the plant failed to meet product quality or flow requirements. Various factors associated with the original equipment contributed to the poor performance:
- The manganese-coated greensand filter may not have been required given the feed iron concentration of 10 ppb or less.
- Greensand filters can act as a particulate filter, similar to a media filter. However, greensand filters typically reduce turbidity (composed of very fine suspended particles) by only 50-80%. The higher turbidity is passed along to (and through) the 5-micron cartridge filters to the RO system. RO inlet silt density index (SDI) numbers are not known, but are believed to have exceeded the rule-of-thumb threshold of SDI 3.0. (Silt density index is a method of determining fine particulate loading by comparing a small volume feed flow through a filter paper initially and after 15 minutes. The ratio of initial flow through the filter paper compared to final flow gives a good indication of membrane fouling potential from particle loading.)
- The three-stage (single vessel each stage) RO system did not improve flow and pressure sufficiently with chemical cleanings, probably due to loading with particulates and membrane surface fouling.
- The EDI systems, which were supplied by the original vendor and did not use Ionics technology, could not deliver product that met the silica spec of 10 ppb. Factors could include membrane type (heterogeneous vs. homogeneous), flow path length, and residence time. The original process EDI used a heterogeneous membrane, which means it is manufactured by pressing ion exchange materials together with a polymer matrix or support. The polymer matrix takes up space and blocks some regions of the membranes. As a result, a heterogeneous membrane has less ion exchange capacity and is less efficient than a homogeneous membrane, which is composed mainly of pure ion exchange material cast on a thin sheet of fabric. A heterogeneous membrane may foul more easily and is harder to clean, thus performance tends to fall off after a period of time. Flow path length and residence time are also important because the feed flow must remain within the ion-exchange resin-filled diluting compartment long enough for the ionization chemistries to take place. The original EDI process equipment had a short path length and short residence time.
Maine Independence Station also experienced some difficulty in cleaning the three-stage RO system, possibly caused by inadequate RO pretreatment by the greensand filter, heat exchanger and cartridge filter. Table 2 shows EDI unit performance at Maine Independence, at start-up and after six months of operation (as an indicator of general plant performance). Detailed analytical results were not available.
Table 2 reveals a significant decline in EDI product quality over the first several months of operation. Product quality could not be recovered with multiple membrane cleanings and system adjustments. The power plants were forced to bring in temporary trailer-mounted equipment to supply spec quality water, and plant management ultimately decided to upgrade the water treatment system.
Initial operating data for Hinds Energy could not be recovered, but the overall performance followed a similar trend, and plant management also decided to upgrade the water treatment system.
Both plants replaced the original equipment with a process train consisting of cartridge filtration, two-stage ultrafiltration, two-stage reverse osmosis, electrodeionization, and backup ion exchange:
- Cartridge Filters — 5 micron, for removal of feed particles.
- Ultrafiltration — Spiral-wound elements (8-inch x 40-inch) are installed in a 7:5 array (seven membranes in first-stage vessels, five membranes in second-stage vessels) with a high recycle rate. These UF membranes have a nominal molecular weight cut-off (MWCO) of 100,000. MWCO is a means of projecting membrane performance by measuring the average molecular weight of organic molecules that cannot pass through the membrane. Membranes are designed and manufactured to have different molecular weight cut-offs for different applications.
- Reverse Osmosis — Two-stage design in a 4:2 array, with six membranes per vessel and 8-inch x 40-inch elements. Anti-scalant and sodium bisulfite (SBS) injection are used for membrane protection. Anti-scalant inhibits the formation of damaging crystals, and SBS reacts with residual oxidants such as free chlorine to prevent damage to oxidant-sensitive membrane surface chemistry.
- EDI — This is composed of a two-unit system, each unit at a flow rate of 50 gpm.
- Ion Exchange Backup – Each process plant has a set of 3.6 ft3 bottles filled with mixed-bed resin for extra protection of product water.
- Post-Retrofit Performance
Figure 1 provides an indication of the product quality and performance of the new process train at Maine Independence Station for the past 18 months. Although the feed temperature cycled between 5 C and 22 C (41 F and 72 F), EDI product resistivity was maintained between 17 and 18 megohm-cm.
Overall feed quality to the system did not vary much, consistently measuring under 200 uS. The EDI system was not challenged by the RO product silica feed of 30-90 ppb, and removal easily met plant specifications of less than 10 ppb (Table 3).
The ion exchange system that follows the EDI consists of two separate sets of ion exchange bottles, the primary and the polish. Typical product quality ranges from 18.0 to 18.3 megohm-cm. Silica levels after the polish bottles (by Hach 5000 Analyzer) measure less than 1 ppb consistently.
The process train at Hinds Energy has demonstrated similar performance. Again, EDI performance is shown as an indicator of overall plant performance (Fig. 2). Final product resistivity of the plant ranged between 18.0 and 18.3 megohm-cm.
Table 4 shows EDI performance for percent removal of conductivity (micro-Siemens) and silica (ppb), as well as sodium, chloride and sulfate. Product silica is presented as <2 ppb, indicating detection threshold. The true concentration of product silica may be significantly less than 2 ppb.
1 Allison, Robert, P., “The Continuous Electrodeionization Process,” American Desalting Association 1996 Biennial Conference, Monterey, CA, August 4-8, 1996.
2 Hernon, B.P., Zanapalidou, H., Prato, T., Zhang, L., “Removal of Weakly-Ionized Species by EDI,” 59th International Water Conference, October 19-22, 1998, Pittsburgh, PA.
Ted Prato is a Senior Process Engineer for the Equipment Business Group at Ionics Inc. He can be reached at [email protected]
Jeff Purvis is a plant engineer at the Hinds Energy Facility, which is owned by Duke Energy. He can be reached at [email protected]
Brian Ahern is the Operations & Maintenance Superintendent at the Maine Independence Station, owned by Duke Energy. He can be reached at [email protected]