Performance Evaluation And Modeling Optimization Of Low-strength Complex Wastewater Treatment System

In most of low-strength complex wastewater treatment systems of the industrialestates chosen the aerobic biological processes as the main process, and lots ofproblems are arising, such as poor effluent water quality, system works in unstablestate, sewage load lower than the design requirements, high energy consumption andhigh cost of wastewater treatment. Avoiding the impact of the uncertainties onwastewater treatment plants and re-rating wastewater treatment plants for greatercapacity without significant facility improvement become the most frequent practiceof the wastewater treatment field in recent years.A wastewater treatment was chosen as the object in this research, itsconventional influent and effluent water qualities were monitored for one year.Mathematical statistics methods were introduced to the study which found that anideal pollutants removal efficiency was observed in the LSCWWTs during theexperimental period, the average COD, BOD₅, TN, NH₃-N, TP and SS removal rateswere90.6%,95.2%,71.9%,94.1%,90.3%and95.4%, respectively. The resultshowed that the stability of the LSCWWTs about the effluent COD, BOD₅, SS andTN were perfect （the stability assessment grade A）, the stability of effluent TP wasacceptable （the stability assessment grade B）, and the effluent NH₃-N had a poorstability and high dispersion （stability assessment grade D）. The results of thereliability research demonstrated that all of the reliability probabilities which theeffluent water quantity reached the GB18918-2002Level1B standard were greaterthan98%, and only the COD, TN and NH₃-N reliability probabilities which reachedthe GB18918-2002Level1A standard were greater than80%, others were less than80%.GA-BPNN model was not only built for further evaluating and analysing waterquality, also for optimizing the wastewater treatment system. Besides, the predictionof the effluent water quantities showed that the GA-BPNN had a well performance oneffluent quality simulation, the average relative error values between the GA-BPNNsimulation value and the measured value of COD, BOD₅, TN, NH₃-N, TP and SSwere4.58%,4.92%,4.96%,4.04%,4.14%and4.81%, respectively. The process model of the whole LSCWWTs was built based on ASM2D whichto simulate the A²/O process combined with the subsidence model to simulate thesecondary settling tank as well as the empirical model to simulate the primary settlingtank. Perfect results of effluent water quantity simulation were achieved after thepreliminary analog, parameter sensitivity analysis and parameter adjustment ofASM2D model, the relative errors between simulated and measured A²/O effluentCOD, TN, TP and NH₃-N were0.18%,0.24%,0.06%and1.00%, respectively. Inaddtion, the subsidence model also has an ideal simulation of the distribution of SS inthe secondary settling tank.The weight coefficient transformation method was employed to solve theequations which describe the optimum operating conditions of the LSCWWTs. Thencoupled the optimal solution value which obtained by solving the intelligent modelwith the process model found that the process models simulated results were almostthe same with the optimal solution values, the average errors between process modelsimulated COD, TN, NH₃-N, TP and GA-BPNN calculated values were4.56%,1.55%,-1.92%and6.39%, respectively. The simulation of secondary sedimentation tankeffluent SS by process model was8.6954mg/L, it’s very close to the GA-BPNNoptimal value of9.2561mg/L. The results showed that the intelligent model can bewell combined with process model and they can apply to improve the performance ofthe LSCWWTs and optimize the LSCWWTs’s operation.