基于人工神经网络对燃煤电厂汞排放的预测-热能工程专业论文.docxVIP

II II Abstract* As the largest source of atmospheric mercury emission, mercury emitted from coal-fired power plants has been identified as a hazardous to both human health and environment. And the relative research concerned with the speciation and control of mercury in coal-fired power plants is currently an active topic. The programmers of BP neural networks were implemented on MATLAB. Because the speed of BP networks convergence is too slow and enters local minimum point easily, the article provides advanced BP algorithm through adjusting study rate and momentum to improve the performance of the forecast model. Six explanatory variables (i.e., mercury concentration, heating value of coal, chlorine content, sulfur content, ash content and temperature) were selected based on availability of collected data and known mercury correlations. These inputs were used to develop the NN model. The result shows that actual output is similar to the actual figures. For a better understanding about the coal combustion Hg emission in China, a field measurement was conducted on a 200MW pulverized coal fired boiler. As halides being proved to be mercury transformation participators, the U.S. EPA recommended Ontario Hydro Method (OHM) and EPA Method 26A were adopted to determine the speciation of Hg and halides in post combustion flue gases, respectively. Results indicated that, as the flue gas cooling down, the percent of oxidized mercury in total gas phase mercury (Hg2+(g)/HgT(g)) increased from 41% to about 74% across the electrostatic precipitator (ESP) outlet. Oxidized mercury (Hg2+(g)) was more apt to be absorbed onto the fly ash. The main halides measured in flue gas were HF and HCl，while the concentration of Cl2 and HBr was extremely low and none Br2 was detected in flue gas. Analysis indicated that HCl, HF and SO2 showed none linearly promotion on Hg oxidation, while NO displayed no obvious correlations. *This project is supported by the National Youth Science