T-S模糊神经网络在厌氧反应器预测中的应用  

Application of T-S neural network in prediction for anaerobic bioreactors

作　　者：曹刚[1] 李明玉[1] 王君[1] 房献宝[1] 

机构地区：[1]暨南大学环境工程系,广州510632

出　　处：《环境工程学报》2007年第11期119-123,共5页Chinese Journal of Environmental Engineering

基　　金：广东省自然科学基金重点资助项目(7117909);暨南大学引进人才科研启动项目(51204018)

摘　　要：3个厌氧反应器运行稳定后,用三氯甲烷和2、4-二硝基酚作为毒物负荷对它们进行了冲击试验。利用负荷冲击试验所得的数据集建立了T-S模糊神经网络,并用其预测了反应器的容积产气率、挥发性脂肪酸和CH4体积含量。研究结果表明,基于某一反应器建立的T-S模糊神经网络可以很好地预测毒物负荷冲击下该反应器的容积产气率、挥发性脂肪酸和CH4变化规律,实测值与预测值的相关系数均>0.850;但是基于某一反应器建立的模糊神经网络用来预测其他反应器时,其预测能力较差,预测值和实测值的相关系数基本上<0.500。After the three bioreactors became steady-state, they were shocked by the chloroform and 2, 4- dinitrophenol. The T-S fuzzy neural networks were created based on the database collected from the anaerobic system shock, and were used to predict the biogas production rate, volatile fatty acid and CH4 of the bioreactors. The results showed that the fuzzy neural network based on a bioreactor can perfectly predict performance of the bioreator, its correlation coefficients of observed and simulated values were above 0. 850 for both training data set and testing data set. But the fuzzy neural network based on a bioreator could not predict well the other bioreactor, the values of correlation coefficients of observed and simulated were below 0. 500 for the biogas production rate, volatile fatty acid and CH4.

关 键 词：模糊神经网络 毒物负荷 厌氧反应器 预测 

分 类 号：X703.1[环境科学与工程—环境工程]