By Steve Blankinship, Associate Editor
Several technologies will probably be needed to remove mercury from coal-plant stack emissions as mandated by new mercury emission control legislation. One of the most promising mercury removal approaches is the injection of a sorbent, such as powdered activated carbon (PAC), to make it much more controllable. All leaching tests performed indicate that with such treatment, mercury will not be re-released from landfills.
Injecting a sorbent such as PAC into the flue gas represents one of the simplest and most mature approaches to controlling mercury emissions from coal-fired boilers. Sorbent injection captures both elemental and oxidized mercury and is effective on both bituminous and sub-bituminous coals, so it is applicable to a large number of units. It requires capital equipment of less than $1 million and can be installed without a plant outage.
There is a good chance that computational fluid dynamics (CFD) modeling of ductwork could be included in future commercial bids for sorbent injection to optimize the efficiency of mercury removal and minimize sorbent expenses, according to Dr. Jens Madsen, senior consulting engineer for Fluent International.
“The main value of the CFD modeling is that it allows for inexpensive testing and comparison of different injection grid arrangements,” says Madsen. “Doing this at the actual plant can be difficult and expensive. With CFD you can identify a better injection system, which will typically allow getting the desired mercury capture with a smaller sorbent feed rate. This directly addresses the main cost component of this mercury control technology, namely the sorbent cost.”
But modeling mercury capture through sorbent injection is a challenging task that has only recently been accomplished successfully. Mercury capture is simulated using a process that concisely describes the equations of motion of the sorbent particles. The sorbent particle trajectories are calculated by integrating the relevant forces on the particles, including drag and gravity, while gaseous mercury concentration in the flue gas is solved separately using a set of transport equations. These two calculations are coupled together for mass and momentum exchange. The amount of mercury adsorption is calculated based on the sorbent trajectories, which determine the sorbent’s exposure to the mercury vapors.
“Mercury adsorption involves three steps,” says Dr. Michael Durham, president of ADA-ES, whose company has been developing and testing one of the major sorbent injection mercury control technologies. “First, the gas phase mercury adheres to the external surface of the sorbent. Second, the mercury undergoes diffusion mass transfer to the interior of the sorbent particle where, third, it is physically/chemically absorbed.”
Durham explains that during the adsorption process, the mercury field is depleted, and this reduction in mercury is accounted for during the next round of gas phase calculations. “The trajectories are recomputed, and the new mercury concentration field is again coupled to the new trajectories,” he says. “This process is repeated until convergence is reached.”
With the support and cooperation of the U.S. Department of Energy’s National Energy Technology Laboratory (DOE/NETL), ADA-ES recently worked with consultants at Fluent, Inc. to simulate field tests of sorbent injection at New England Power Company’s Brayton Point Power Plant in Somerset, Mass., where activated carbon sorbent was injected using a set of eight lances upstream of the second of two electrostatic precipitators (ESPs). Fluent consultants created a computational model of the ductwork and injection lances. The simulation results showed that the flue gas flow was poorly distributed at the sorbent injection plane, and that a small region of reverse flow occurred, a result of the flow pattern at the exit of the first ESP. The results also illustrated that the flow was predominantly in the lower half of the duct, and affected by some upstream turning vanes.
The field tests at Brayton Point showed that, due to their higher inertia, larger sorbent particles travel in bands across the lower part of the duct. The smaller particles showed a more diffuse motion, with some sorbent getting caught in the reverse flow. The average residence time of the sorbent particles was estimated at 0.45 seconds based on plug-flow conditions. However, (CFD) simulations showed that as a result of the skewed velocity distribution, the actual mean sorbent residence time was only on the range of 0.25 seconds. In spite of this, ADA-ES measured excellent mercury capture efficiencies of up to 90 percent based upon CFD simulations.
The CFD simulations revealed that this success was promoted by uniform sorbent dispersion because of a very high degree of turbulent mixing in the injection duct. The strong mixing means that just 12 feet downstream of the injection grid, more than 80 percent of the duct cross section is sufficiently covered with sorbent. Two different injection lances – a multi-nozzle and a single nozzle design – were evaluated at Brayton Point. The multi-nozzle lance is a slender cylinder with a capped-off end. Eight holes (four pairs) of holes drilled perpendicular to the cylinder axis of revolution served as the nozzles. Sorbent and a small amount of carrier gas are fed through the inside of the cylinder to these nozzle openings.
The tests showed an insignificant difference in performance between the single and multi-nozzle designs, which was surprising since the designers of the nozzles expected the multi-nozzle design to provide superior performance. The CFD simulation explained that the gas flow is almost evenly distributed around the eight holes of the multi-nozzle lance, while the vast majority (>90 percent) of sorbent exits from the lowest set of holes. The simulation showed that this is because the bigger particles have too much momentum to make the curve required to exit at the other nozzle openings. The fine particles were much more evenly partitioned among the different nozzles
“Results of the Brayton Point test simulations were consistent with the results of the tests themselves,” says Madsen. “The simulations clearly demonstrated the value of CFD as a diagnostic tool. They were performed in a fraction of the time and cost required for the physical tests yet provided far more diagnostic information, such as the distribution of mercury and sorbent at each point in the computational domain.”
Madsen notes that a typical 300 MW power plant will require between $1 and $2 million of sorbent per year and CFD simulation can be expected to reduce that amount about 20 percent. “Cost reductions of this magnitude will substantially reduce the costs of complying with the new mercury reduction regulations,” he says.