Review of technologies for mercury removal from flue gas from cement production processes

摘要

Mercury is a pollutant of concern and mercury emissions from cement plants are under environmental regulation. After coal-fired power plants, mercury emissions from cement and mineral production are the second largest anthropogenic sources. Compared to fuels, cement raw materials are the major sources of mercury in the cement kiln flue gas. Cement plants are quite different from power plants and waste incinerators regarding the flue gas composition, temperature, residence time, and material circulation. Cement kiln systems have some inherent ability to retain mercury in the solid materials due to the adsorption of mercury on the solids in the cold zone. However, recirculation of the kiln dust to the kiln will cause release of the captured mercury. The mercury chemistry in cement kiln systems is complicated and knowledge obtained from power plants and incinerators cannot be directly applied in cement kilns. Among the mercury control technologies, sorbent injection upstream of a particulate control device has shown the most promise. Due to material recirculation, and high moisture level in the cement kiln flue gas the application of sorbent injection to cement plants will be more challenging. The sorbent injection system should be installed downstream of the main kiln filter and upstream of a new added polishing fabric filter to avoid the cement kiln dust recycling and disposal issues. To reduce the sorbent cost and possible disposal expense, non-carbon based sorbents that could be added to cement or regenerated in-situ are desired and should be developed. Various mathematical models have been developed to simulate mercury removal in fixed-bed reactors and by sorbent injection upstream of a fabric filter. The fabric filter adsorption models use the adsorption isotherms coupled with diffusion in the sorbent particle and the parameters are obtained by fitting the model to experimental data. Verification of the models by full-scale or pilot-scale data is very limited.