Water-guzzler turns water producer

By Rob Heijboer and Frank de Vos, KEMA

A power plant or waste incinerator that doesn’t need a supply of expensive potable water could soon be a reality, thanks to a membrane system that recovers clean water from flue gas. As well as providing water, the system could cut an average plant’s annual energy bill significantly.

Most people who see white clouds billowing from the stack of a power plant or waste incinerator think that what they’re looking at is pollution. Few people realize that these clouds contain harmless water. So, for PR reasons alone, the removal of water from flue gas is a worthwhile goal. However, the drivers for introducing water recovery are primarily economic.

Breathable raincoat
The most obvious way to recover water from flue gas is to cool the flue gas until the water vapor condenses. Such an approach is highly energy-consuming, however. It also results in the release of harmful substances such as SO2 and NOX in the water requiring even more energy to be used and even more money to be spent in order to remove these pollutants. Several years ago, KEMA and its energy sector partners started a project with the aim of recovering water from flue gas in the most energy-efficient way possible. The idea of a process based on membrane technology was appealing, because it offered the possibility of combining extraction and purification in a single step. Interested parties wanted a membrane through which liquid water couldn’t pass, but water vapor could – essentially, a breathable raincoat.

Highly selective

Working with partners such as the European Membrane Institute at the University of Twente in the Netherlands, and with support from the Dutch government, a new type of membrane capable of separating large quantities of water was developed. The membrane is in effect a sophisticated version of those used in breathable raincoats like GORE-TEX (figure 1).

Figure 1 Working principle of GORE-TEX

The GORE-TEX material is permeable for water vapor but rejects liquid water. The driving force is the water vapor pressure close to the human skin which is higher than the water vapor pressure on the outside of the clothing. The material used in the textile industry isn’t suitable, however. Because at a typical flue gas temperature of 60 degrees Celsius, it isn’t selective enough in relation to nitrogen – which makes up to more than 70 percent of a combustion plant’s flue gases. An alternative synthetic membrane has therefore been developed, which remains selective at the appropriate temperature. Based on the fact that flat membranes have an uneconomical surface to volume ratio the choice has been made for the use of hollow fiber membranes.

Figure 2. Idealistic scheme of water recovery

The membrane modules are placed directly after the FGD. The recovered water vapor (permeate) is transported to the condenser and condensates together with the steam from the steam turbine. The surplus of condensate compensates for steam loss and boiler blow down and is fed to the boiler. Due to the water recovery, reheating of the flue gas can be omitted.

Proof of principle

Before any development can be made in the direction of the above-depicted ideal closed cycle, the working principle of the membranes has to be proofed. To carry out this proof a curtain shaped module is made and placed in a side stream of the flue gas from a coal fired power plant in a cylindrical membrane chamber (figure 3) .

Figure 3. Proof of principle

The flue gas is fed to the shell side of the fibers. The permeation of the water vapor through the membrane is a pressure driven process. The water vapor is transported to a condenser system by applying a vacuum at the tube side of the fibers; any non condensables are expelled by the vacuum system.

Thermal and chemical stability

After successfully carrying out the proof of principle, thermal and chemical stability of the membrane material has been assessed as well. After 5300 hours under flue gas conditions the main part of the membrane material is still defect free and complies with the specification of having a low nitrogen flux. Because nitrogen is the main compound in flue gas, this compound is chosen as the target compound for this assessment. The water flux of the membrane is very important because it determines the surface area needed to recover an acceptable amount of water. During the first experiments the water flux was between 0.6 and 0.2 L/m2/h. This had to be improved because a rather low water flux requires a rather high surface area. Implementing large surface areas in an existing power plant is quite a challenge. One in particular is the maximum allowable pressure drop in the flue gas duct. This pressure drop will be limited to 10 mbar over the total of membranes. Besides, large surface areas put pressure on the economy of the process.

To keep the recovery process feasible a higher flux is necessary and based on modeling and experiments a curtain shaped module has been built with hollow fiber membranes with an increased internal diameter. The module was placed in a side stream of the flue gas directly after the FGD demisters (figure 4). During the experiments a flux improvement could be achieved with a factor of 7 and a permeate quality of 20 µS/cm.

Figure 4. Curtain shaped module

It can be concluded that SO2 is one of the main compounds affecting the quality of the recovered water. Comparing the measured conductivity to that of fresh surface water one has to keep in mind that acidic compounds cause the conductivity in the recovered water. This means that for comparison this conductivity has to be divided by a factor of 3-3.5 resulting in a value of 6-7 µS/cm.


Interestingly, the system’s energy benefit doesn’t derive directly from the water recovery process itself. It’s after the water’s been extracted that the savings can be made. Before flue gas is released into the atmosphere, it is often necessary to reheat, in order to prevent condensation in the stack, which can cause enormous damage. If 20 percent of the water is removed from the flue gas leaving a 400 MW coal-fired plant, savings could be made up to more than a million euros ($1.4 million USD).

Little wear and tear

Initial tests with membranes having an effective surface of one square meter installed at a coal-fired power plant and at a waste incineration plant have shown that the system performs as intended and isn’t unduly prone to wear. At the power plant, the membrane was able to collect 1.5 liters of water an hour per square meter effective surface. That was twice what was originally expected. At the waste incinerator plant, the collection rate was even better: four liters an hour. The increase in efficiency was due to the higher flue gas temperature and water content.

The quality of the water was found to be close to that of industrial demineralized water; however, the electricity sector needs water that is even purer, so before the recovered water can be used for energy production, the small quantities of ionic compounds/ acids it contains need to be removed.


At present, KEMA is busy scaling up the water recovery system. A pilot project is being organized at an existing power plant which will involve five hundred square meters of membrane surface arranged in modules fitted side-by-side and in series inside a flue gas duct. If the pilot is a success, commercialization will follow within one to two years.


Within the energy and waste processing industries, there is considerable interest in the concept. In the Netherlands and other European countries, the energy savings are the system’s main selling point. However, in countries such as Australia and South Africa, it is the prospect of recovering such a large amount of good-quality water that really appeals. In such countries, water is a valuable commodity. So a power plant that is water self-supportive would be a real winner.