Simulation and Prediction of Rural Sewage Treatment by Biological Aerated Filter Using LMBP Artificial Neural Network

摘要

Treatment efficiency of refractory rural sewage could be hardly predicted, resulting in a poor operation performance and an effluent threat to the receiving water body. This study fabricates a Levenberg-Marquardt back propagation artificial neural network (LMBP-ANN) model for the efficiency prediction of rural sewage treatment. The rural wastewater was treated by four different aerated biological filter (BAF), which were filled with different filter materials, which was zeolite, modified zeolite, ceramsite. The automatic control system was constituted of the programmable logic controller and pH, DO, ORP on-line water quality indicators sensors to enhance the management level. The hydraulic retention time was set as 18 h. During 90-day operation, the effluent of all the four filters reached the first class type A of discharge standard of pollutants for municipal wastewater treatment plant (GB18918-2002). The spline interpolation, cross-validation, and weight decay were used to develop the LMBP-ANN model. The model provided accurate estimation and satisfactory prediction of the effluent qualities i.e. chemical oxygen demand (COD), ammonia (NH_4~+-N), and total nitrogen (TN), using on-line water qualities i.e. pH, DO, and ORP. The correlation coefficients of the model were all up to 0.95 in all the four scenarios. The results show that the LMBP-ANN model can be used as a predictive tool for the complex and nonlinear analysis of rural wastewater treatment processes.