Coal, Emissions

Boiler Feedwater Treatment: Improvements in Continuous Electrodeionization

Issue 9 and Volume 107.

By: Jonathan Wood and Joseph Gifford, USFilter Corporation

The conventional means of purifying boiler feedwater has been to use chemically regenerated ion-exchange deionization. Over the past decade, the power industry has increasingly relied on the use of reverse osmosis (RO) as a roughing demineralizer to remove the bulk of the mineral, organic and particulate contaminants, and reduce the chemical consumption of the ion exchange system. More recently, improvements in continuous electrodeionization (CEDI) technology have caused a movement toward chemical-free deionization systems, as RO/CEDI has become more cost competitive with conventional ion-exchange technology.

Recent advances in electrodeionization module construction have led to further cost reductions both at the module and system level. For instance, incorporation of an ion exchange resin filler in the concentrate compartments has allowed for elimination of both the concentrate recirculation pump and brine injection system that is often used to increase water recovery and reduce module resistance.

CEDI Module Design

Continuous electrodeionization was first commercialized in 1987.1 For the first ten years, nearly all commercial CEDI devices were plate-and-frame designs, and used “thin cell” product water compartments (about 2.5 mm between ion exchange membranes) with a mixed-bed ion exchange resin filler. In recent years, a variety of new designs have emerged, including different module configurations (spiral wound), thicker product cells (8-9 mm inter-membrane spacing), and different resin configurations (clustered bed, layered bed, separate bed).

For example, USFilter’s CDI-LX is a thick cell, layered bed, plate-and-frame technology.2,3,4 The developers put much thought into the mechanical reliability of the device, particularly the ability to handle higher operating pressure and temperature as well as pressure and flow fluctuations. The engineered plastics used for the spacers are polysulfone or polyphenylsulfone – chosen due to their extremely high yield strengths and their ability to maintain mechanical properties at high temperatures.

Click here to enlarge image

Figure 1 is an exploded view of the CEDI module, shown without the ion exchange resin, which would be filled into the open area in the center of the spacers. The four holes labeled “flow ports” feed the water into the product and reject compartments in a co-current, downflow configuration. The smaller holes around the perimeter of the spacer are where the tie bars pass through to compress the module and hold it together. For protection against electrical short-circuiting, the tie bars are encased in plastic sleeves, with an elastomeric seal between the outside of the sleeve and the electrode block (not shown).

The thick dilute spacer affords the ability to incorporate grooves for O-ring seals. There is an O-ring around the perimeter of the resin bed as well as an O-ring around every flow port. This yields a double seal between the different flow ports and the resin beds, thus isolating purified product from concentrated reject water. There is also a second perimeter O-ring to provide the same double seal to the outside of the module, eliminating external leakage.

It is well known that ion exchange resins swell as they are converted from the salt form to the regenerated form. In fact, some anion resins can expand 20 percent or more. Testing has shown that in a layered bed this swelling can impart an internal pressure equivalent to about 100 psi of hydrostatic pressure. The module was designed to handle 100 psi maximum feedwater pressure. Therefore the total internal pressure acting on the module assembly is 200 psi. In that respect, the spacers, without any additional supporting hardware, were designed to withstand 200 psi with a safety factor greater than two. In an assembled module, the safety factor is actually higher due to the additional strength contributed by the tie bars and endplates.

Another point of note regarding this CEDI technology is the design of the concentrate cells. In most CEDI technologies, the concentrate is some type of gasketed screen. Therefore, the amount of salt in the concentrate streams controls the overall electrical resistance of the module. Many suppliers incorporate concentrate recirculation and/or salt injection to increase the conductivity of the concentrate and reduce the electrical resistance of the module. To lower the module resistance without resorting to such measures, ion exchange resin is included in the concentrate and electrode cells, as well as in the dilute cells, rendering the resistance independent of the concentrate water conductivity.

RO/CEDI Process Considerations

Since its introduction in 1987, CEDI has gradually evolved into a polishing demineralization process that is almost always employed downstream of a reverse osmosis system. There are several reasons for this evolution. First, CEDI devices are susceptible to hardness scaling, organic fouling, and physical plugging by particulates and colloids. Second, the CEDI product water quality is somewhat dependent on the feedwater quality. While some CEDI devices may be able to produce “two-bed quality” product water directly from a softened feedwater, most power plant applications now require “mixed-bed quality” water, which would not be produced by CEDI alone.

Click here to enlarge image

Using RO pretreatment ahead of the CEDI reduces total dissolved solids (TDS) to a level that allows the CEDI device to meet the feedwater quality requirements of a high-pressure boiler (Table 1). In addition, the RO removes organics that could foul the ion exchange resins in the CEDI modules, and removes particulates that could clog the narrow flow channels in the resin compartments (spacers) or the resin bed itself.

There are some issues regarding the use of RO/CEDI systems that are sometimes overlooked, to the detriment of system performance and reliability. One such issue is hardness, since most CEDI modules can tolerate only about 1 ppm total hardness. This is often surprising to those familiar with saturation indices such as the Langelier Saturation Index (LSI), since in most cases the LSI of the bulk CEDI reject stream will be negative. However, the water splitting that is necessary for electrochemical regeneration of the ion exchange resin can also lead to localized pH shifts, creating regions (such as near the ion exchange membrane surface) where the scaling potential is greater than in the bulk solution.

Many systems will use ion exchange softening to protect both the RO and CEDI systems from scaling, but in some cases this is undesirable. While the RO system alone may be able to produce steady-state permeate hardness of less than 1 ppm, most boiler makeup water systems operate in “start/stop” mode, only operating when the deionized water storage tank is calling for water. When an RO system starts up from a standby condition, the initial slug of RO permeate can be worse than the RO feedwater (since the concentration gradient causes some salt to continue to diffuse through the RO membrane after the permeate flow stops), and the first few minutes of RO permeate may not meet the CEDI system feedwater specifications. This phenomenon is illustrated for permeate conductivity in Figure 2, but the results are similar for most ionic constituents in the feedwater. In this particular instance, it took about two minutes for the RO permeate to approach steady state. Even though the volume is relatively small, it is important to flush this water to drain rather than send it to the CEDI system, especially since most RO/CEDI systems do not have a buffer tank between the RO and CEDI. An alternative approach is to flush the RO system with demineralized water before it shuts off.

Click here to enlarge image

The other main issue affecting CEDI operation is the presence of CO2 in the raw feed water. Any CO2 in the gaseous form will not be rejected by the RO membrane, and will impart an ionic “load” on the downstream CEDI system. For many CEDI devices, a concentration of 10 ppm CO2 in the CEDI feed may be enough to prevent the system from meeting the boiler makeup water specifications. Therefore, many RO/CEDI systems will employ a separate means of CO2 removal. The most common methods are forced draft degasification, membrane degasification, or pH adjustment before the RO.

Since the reject stream from a CEDI system typically contains anywhere from one-fifth to one-half the salt concentration of the raw water, recycling the CEDI reject to the inlet of the RO is often desired. This may be possible, but it is necessary to consider that while the CEDI reject may be lower in salt, it typically has ten times the CO2 of the RO feedwater. In the absence of a degasification or pH adjustment step, recycling this water could result in as high as a three-fold increase in the CEDI feed CO2 concentration, and could have a significant impact on the CEDI product water quality. In most cases this would be impractical without a CO2 removal step as part of the process train.

CEDI systems now often use multiple smaller stacks in parallel to attain high product flow rates. There are two possible design approaches for the DC power supply: a single large unit, which supplies the same voltage to all the CEDI modules, or multiple separate rectifiers. New devices are available that are low enough in cost so that one can be supplied to operate each module individually. If one rectifier fails, that particular rectifier and module can be taken out of service until a new one can be installed. Having individual rectifiers also offers some degree of flexibility in monitoring, control and optimization of the DC power applied to each module. The power supply is operated in “constant current” mode, where the DC voltage is adjusted as necessary (in response to changes such as feed water temperature) to keep the DC amperage stable.


New developments in CEDI module construction have improved both physical integrity and module reliability while simultaneously enabling process simplification such as elimination of concentrate recirculation and elimination of salt injection into the concentrate stream. However, reliable long-term operation of an RO/CEDI system requires careful attention to process design, and in particular hardness and CO2. With good module and system design, it is possible to design deionized water systems based on RO/CEDI that will consistently meet the makeup water quality requirements of high-pressure boilers without the use of hazardous chemicals and without creating regenerant waste.


1. Ganzi, G.C., Y. Egozy, A.J. Giuffrida, and A.D. Jha, “High Purity Water by Electrodeionization: Performance of the Ionpure ™ Continuous Deionization System,” Ultrapure Water, Vol. 4, No. 3, pp. 43-50 (1987).

2. DiMascio, F. and G. Ganzi, Electrodeionization apparatus and method, U.S. Patent No. 5,858,191.

3. Ganzi, G. et al., Electrodeionization apparatus and method, U.S. Patent No. 5, 868,915.

4. DiMascio, F. et al., Electrodeionization apparatus and method, U.S. Patent No. 6,284,124.


Jonathan Wood is Director of Applications R&D for USFilter in Lowell, Mass. He has more than 25 years of experience in water purification, encompassing research, product development, product management and technical service. Wood is currently involved primarily in applications development for electrodeionization and reverse osmosis equipment. He holds a B.S. in chemical engineering from Worcester Polytechnic Institute and an M.S. in environmental engineering from Northeastern University.

Joseph Gifford is a Senior Development Engineer in Applications R&D at USFilter, where he has spent the last seven years working primarily with the development and application of continuous electrodeionization. Gifford holds a B.S. in chemical engineering from Worcester Polytechnic Institute and an M.S. in chemical engineering from the University of Massachusetts.

Membrane Filtration Making Inroads

By Brad Buecker, Contributing Editor

Many power plants that came on-line in the 1960s and 1970s relied on pretreatment schemes consisting of raw water clarification followed by multimedia filtration. Final polishing often consisted of activated carbon filtration followed by ion exchange demineralization. Reverse osmosis (RO) has since become very popular as an intermediate step between pretreatment and final polishing. RO systems require feedwater with very low suspended solids to minimize fouling of the membranes. This issue and others are pushing the drive towards pretreatment alternatives to clarification and multimedia filtration.

Membrane filtration techniques – microfiltration (MF), ultrafiltration (UF) and nanofiltration (NF) – are gaining acceptance in many industries. MF and UF are most popular for the power industry. Antonia J. von Gottberg of Ionics discussed the basics of these systems in a presentation at the 23rd Electric Utility Chemistry Workshop (EUCW) in May in Champaign, Ill. Each of the three techniques filters water through membranes containing extremely small pores that become progressively smaller in the order of MF, UF, and NF. Unlike RO, however, the pore sizes in MF and UF membranes are large enough that they do not remove dissolved solids, only suspended solids. (NF can remove divalent cations, e.g., calcium and magnesium.) Pressure drop is much lower through MF and UF membranes than their RO counterparts.

While most RO membranes are fabricated in a spiral-wound configuration, a hollow fiber design is more common for MF and UF applications. In these systems, thousands of small membrane fibers are bundled together into an element. The number of elements can then be selected to achieve the desired flow rate. Either inside-out or outside-in water flow patterns through the membranes are possible. MF and UF systems can reduce raw water turbidity to less than 0.1 NTU, making them strong potential candidates as pretreatment devices for gray water sources such as secondary sewage treatment plant effluent. Because MF and UF membranes consist of hollow fiber systems that can accumulate solids, plugging is possible, so periodic backwashing is required. Anyone contemplating installation of a MF or UF system should thoroughly discuss the application with a qualified vendor. Besides Ionics, other reputable vendors include Pall, US Filter, and Zenon Environmental Systems.

Condenser performance monitoring is still a very important issue. Even utilities with on-line condenser monitoring systems may encounter trouble if the data is not tracked closely or chemical feed systems malfunction. Tony Selby, president of Water Technology Consultants, presented a paper at EUCW in which he pointed out that a one-inch (Hg) rise in condenser absolute pressure (also known as backpressure) due to tube fouling or excess air in-leakage can cost up to $4800 per day in efficiency losses in a 500 MW unit. (This figure compares closely with thermodynamic studies from my department.) Condenser performance may be fine in the spring, perhaps after a mechanical or chemical tube cleaning, only to drop off in the summer just as power costs reach a maximum. For plants that do not have on-line condenser performance monitoring capabilities, straightforward computer programs are available that will allow plant personnel to reasonably calculate condenser performance, or more critically, changes in condenser performance.

A critical aspect of such programs is that the data output is readily available to warn plant technicians if biocide or scale-inhibiting treatments are not performing effectively. The programs can also be very effective at detecting sudden increases in air in-leakage. On a related issue, Pete Kutzora of We Energies and George Lucina of Structural Integrity Associates offered a paper that outlined performance of a biofouling monitor known by its trademark name of Bi°GEORGE. This instrument, simple in design, uses electrochemical principles to detect the onset of microbiological fouling before it begins to appear in condenser tubes.

With regard to air quality control chemistry, Rob Moser of Codan Development presented an interesting paper on the injection of simple salts to lower sulfur trioxide (SO3) in plant flue gas emissions. SO3, of course, raises the acid dew point of flue gas and is the primary culprit in boiler backend corrosion. SO3 is also of concern to the EPA, as it is a main contributor to aerosol formation. These issues are becoming even more important because selective catalytic reduction systems used extensively for NOx control fractionally catalyze SO2 to SO3. Preliminary results indicate that the straightforward process outlined in Moser’s presentation will remove almost all of the SO3 in flue gas. This in turn allows for lower exit gas temperatures and better unit efficiency. There are some concerns about the effects on precipitator performance, as SO3 enhances particulate removal, but according to Moser, the lower flue gas exit temperatures offset this negative impact on precipitator operation.