摘要
在实际污水处理厂运行过程中,其最终出水水质会受多种因素影响制约,而基于生物反应机理的活性污泥数学模型(ASM)并未将这些生物反应以外的因素考虑在内,由此带来一些不足.对此,本文提出可通过基于数据挖掘技术的黑箱模型对污水厂处理效果进行模拟预测.结合具体实际分析,提出可将BP神经网络与马尔可夫链组合应用于污水处理脱氮效果预测中.首先,通过BP神经网络模型对北京某大型污水处理厂实际进出水数据和工艺参数进行粗略拟合;其次,利用马尔可夫链对拟合结果及误差进行状态划分以进一步提高预测精确度;最后,运用基于BP神经网络与马尔可夫链的组合模型预测分析了该厂的实际出水水质.试验结果表明,BP神经网络适用于污水处理脱氮过程的拟合计算,而通过与马尔可夫链组合,可以提高模拟预测的精度和可靠性.
The effluent quality of a wastewater treatment plant（ WWTP） is restricted by many factors. However,the activated sludge mathematical model（ ASM）,which is only based on biological reaction mechanism,does not involve the factors other than biological reactions. Therefore,It will bring inaccuracy in real forecast. To solve this problem,the author put forward a＂black box model＂which is based on data mining technology to simulate and predict a WWTP operation. First,the actual WWTP data and process parameters were fitted by using neural network model. Second,the markov chain was applied to fit the results and error state division to further improve the prediction accuracy. Finally,BP neural network and Markov chain were combined to analyze and forecast a large WWTP nitrogen removal efficiency. The results show that the BP neural network can be used to simulate the WWTP nitrogen removal process,and the accuracy and reliability of the simulation can be improved when combined with the Markov chain.
出处
《环境科学学报》
CAS
CSCD
北大核心
2016年第2期576-581,共6页
Acta Scientiae Circumstantiae
基金
国家水体污染控制与治理科技重大专项(No.2014ZX07201-001)~~
