| Biogas is a kind of green renewable resources. As a reserve energy, biogas gets more and more extensive attention due to the growing shortages of energy nowadays. But hydrogen sulfide in biogas is not only harmful to human and environment, but also has strong corrosion action to the pipelines, metal instruments and equipments. The concentration of hydrogen sulfide can not exceed 20mg/m3 in biogas if methane energy is utilized according to China national standards. Therefore, the research on biogas desulfuration plays an important role in biogas recovery and utilization. With high efficiency, low pollution and low energy consumption, the biological desulfurization has become present research focus. In this paper, the biotrickling filter was used to take the biogas desulfuration experiments and the optimum operation parameters were determined. Based on this BP (back-propagation) neural network model was established, the optimal structure and parameters of the model were determined, and then model simulation, prediction and control application were conducted. These researches were expected to provide technical support and parameter guidance to application promotion of biogas desulfurization process.Haydite was finally chosen as the filler for the biotrickling filter through experiments. The biotrickling filter filled with haydites was used to carry out the hydrogen sulfide removal experiments, and the optimum operation parameters of the reactor were determined. Inlet H2S concentration should be controlled within 800mg/m3, 650mg/m3, 400mg/m3, 300mg/m3 when the inlet flow rate was respectively at 150L/h, 200L/h, 250L/h, 300L/h. The optimal inlet flow rate was 350L/h, 300L/h, 200L/h, 200L/h, 100L/h, 50L/h, while the inlet H2S concentration was respectively at 100mg/m3, 300mg/m3, 500mg/m3, 700mg/m3, 900mg/m3, 1100mg/m3, with the corresponding gas retention time of 7.07s, 8.24s, 12.37s, 12.37s, 24.73s, 49.46s. The best circulating liquid spraying flux was 1L/h, 2L/h, 3L/h, 4L/h, 6L/h, respectively at the inlet H2S concentration of 300mg/m3, 500mg/m3, 700mg/m3, 900mg/m3, 1100mg/m3. The maximum inlet volume load of this reactor was 130g/(m3·h). The operating temperature was controled at room temperature. In these above conditions, the outlet H2S concentrations of the reactor were lower than 13.1mg/m3 and the H2S removal efficiencies were over 98%.By the experiments of biotrickling filter removing hydrogen sulfide from biogas under four different biogas flows of 0.05m3/h, 0.10m3/h, 0.15m3/h, 0.20m3/h, it was indicated that the H2S removal efficiency was high at every flow in the reactor, and 0.15m3/h was chosen as the actual operation flow. By investigating the pH value and pressure drop in the reactor, it was found that the pH value varied between 2.0 and 7.0 during operation, without the need for adjustment, and the reactor could achieve a good desulfurization effect. During the film-forming period, the pressure drop in the reactor increased with time, and finally stabilized. The reactor pressure drop was proportional to increase with the inlet flow rate whlie stable operation. Throughout the trial process, the reactor pressure drop maintained at a relatively low leve, so there was no need to stop the tower for backwashing.The experiment on the results of bacteria respectively utilized four kinds of sulfur source proved that Na2S2O3 was the best, sulphur was the worst, Na2SO3 was better than Na2S. Through the study on the concentration change rule of various sulfides, it was inferred that the sulfide conversion process under microbial action was S2-â†'S0â†'SO32â†'SO42-.BP neural network model was established on the process of biotrickling filter removing hydrogen sulfide, and the influence of model parameters was investigated to determine the optimum ones. With the optimal parameters the model was trained and simulated, as a result the mean square error was 0.00341329, the linear regression correlation coefficient of model simulation values and the corresponding experimental values was 0.9965. Using the model to predict the experiment results, the mean absolute error of model predictions and the corresponding experimental values was 2.092033mg/m3, the average relative error was 4.215287%, and the correlation coefficient was 0.9988. It showed that the experiment results could be reasonably predicted by this model when the test conditions were in the scope of the model establishment conditions. Inlet H2S concentration control curve and gas retention time control curve were obtained by using the BP neural network model, which could respectively realize the control of inlet H2S concentration under different gas retention time conditions and the control of gas retention time under different inlet H2S concentration conditions to achieve different treatment requirements, including H2S did not disclose and outlet H2S concentrations did not exceed the environmental requirements of 20mg/m3. |
PDF Full Text Request