Study On Internal Recirculation Flow Intelligent Control Of A/O Wastewater Treatment System Based On Fuzzy Neural Network

With the increasingly serious water pollution, sewage treatment has become increasingly urgent and arduous task. In order to achieve sustainable development strategy, China must vigorously develop the cause of the sewage treatment, sewage treatment process, activated sludge is a more advanced technology, in which A/O nitrogen removal process is better, but the activity Wastewater treatment sludge by the water quality, water, equipment, and many other factors, resulting in wastewater treatment with multi-variable, nonlinear, large delay, uncertainty and complexity of features, simple manual operation can't achieve stability and good the water effect. Conventional sewage treatment process control automation level is relatively backward, the control result is not satisfactory, in view of this, the paper concentrates on the study of fuzzy neural network algorithm to achieve the A/O ratio of wastewater treatment system, intelligent control of reflux, the following results:Under the existing conditions in the laboratory experiments,reached the correlation between TOC and COD_Cr of inflow: y=1.8646x-19.385, R=0.9956; the correlation between TOC and COD_Cr of effluent: y=1.7338x+1.0675 , R=0.9952; which had a good correlation between the two;Under laboratory conditions to determine A/O wastewater treatment process, completed MCGS PLC configuration software configuration and program design, in this based on the automatic control of wastewater treatment systems, and automatic control system in this test to obtain a large number of sample data. In-depth analysis of the existing modeling methods, based on characteristics of wastewater treatment and analysis of fuzzy neural network, using sample data obtained by Anfis network was A/O anoxic nitrate prediction model, using MATLAB software to create the prediction model simulation analysis, simulation results show that the prediction model has good learning ability and generalization ability of absolute relative error of training data range is 0⁰.003%, the relative error of the absolute value of the test data range of 0 to 0.06%, indicating network generalization is acceptable;Through the establishment of the Takagi-Sugeno model and forecast simple BP network prediction model of performance comparison, the former training data and test data, the average absolute error rate of MAPE was 1.186×10^-3%,0.263%, while the latter's training data and test data, the average absolute error rate of MAPE was 1.847%, 1.855%. By comparing the performance of the two known, Takagi-Sugeno model prediction MAPE mean absolute error rate far less than the simple BP network prediction model, indicating that the former is more suitable for predicting A/O biological value of nitrate nitrogen pool; Mamdani fuzzy rules based on neural network control model, using oxygen tank at the end of the change in nitrate concentration and the oxygen tank at the end of the rate of change of nitrate concentration as a control model input variables, the reflux ratio correction as a control model output, for regulating the return flow. According to practical experience, the definition of the two input control model contains seven linguistic values, respectively, the model for the 2-14-49-49-1 structure, with the test sample data for training the network, under control before and after the pieces of the model parameters combined hybrid algorithm to complete the network structure and parameter identification, control model to determine the specific parameters;Advanced development kit in MCGS VB program generated fuzzy neural algorithm is prepared, in accordance with the interface function specifications MCGS embedded MCGS in the algorithm, the realization of paper-making wastewater treatment system intelligent control. Oxygen tank from the end of the experiment the optimal concentration of nitrate is 2.45 mg/L, the concentration of the oxygen tank at the end of the expected concentration of nitrate set 2.45 mg/L, the results show that under different influent loading, lack of pool at the end of the nitrate concentration of oxygen in the 2.45 mg·L^-1 near the fluctuations, fluctuations in the range of 2.14 mg·L^-1 2.68 mg·L^-1, show that the algorithm based on fuzzy neural network intelligent control system that can effectively achieve A/O wastewater treatment system, intelligent control.The reasearch provided an effective way to achive autocontrol for wastewater biochemical treatment system and provided guidance for the further study of intelligent control in the field of wastewater treatment and the popularization of wastewater treatment project with intelligent control.