| This research subject is sub-project"Research of water quality and quantity scheduling scheme"of"Surface Water Environment Quality Standard"of national major water project"Nansi Lake degraded wetland ecological restoration and water quality improvement technology and demonstration topics"(No. 2009ZX07210-009).This research taked Xinxue River constructed wetland of Nansi Lake as the area for studying. The purpose was to establish scheduling scheme of water quantity and quality under the restrained conditions of III standard of surface water quality . First, the main influent and effluent water quality and quantity conditions of the constructed wetland were monitored and analyzed the seasonal changes of wetland. On the basis of experimental data, constructed the wetland water quality models based on BP neural network in different seasons, and predicted the modelthe largest influent loads under the restrained conditions of the III standard of Surface Water Quality. With the software of Origin, established the influent water quality and quantity scheduling nonlinear functional relations. These constraints of functions were the prediction results of BP neural network. With these constraints, solved and confirmed optimal influent loads, set up water quality and quantity scheduling scheme of Xinxue river.As for the innovation of this paper, with the prediction result of the wetland water quality models based on BP neural network, using nonlinear programming techniques, established the influent water quality and quantity scheduling functional relations. Through the solving of figures, confirmed optimal influent loads. The main findings were as follows:(1) Based on experimental and data collection in Xinxue river construction wetland, confirmed influent and effluent water quality and quantity conditions, and provided datas base for the establishment of wetland water quality model. According to the actual running conditions of wetland and experimental datas, analyzed and confirmed that the pollutant removal effect had obviously changed on the seasonal variation. In view of influent water quality changes were more obvious, so used the curve of variation, and divided into three periods: February to May, June to September, October to December. The forecasting period partition of model based on these periods.(2) Established simulation and prediction model of BP neural network between the influent water quantity,water quality and effluent water quality of constructed wetland in different seasons. In the process of model building, through carrying out a series of optimal design and circular trail calculations to the network, determined the network topology structure and network learning parameters. After the model was established, trained and tested the model through using test samples. Test results showed that the relative prediction errors of the neural network model water quantity, CODCr, ammonia, TN, TP from February until May were 0.27%, 2.93%, 1.74%, 8.87% and 1.4%, respectively; from June until September they were 5.43%, 15.51%, 2.63% and 15.01%,respectively; from October until December they were 2.11%, 1.67%, 26.02%, 20.79% and 25.55%, respectively. According to the decision rules of the model prediction, their relative prediction errors of the the model established were under allowed band, the network training was successful, and its performance could meet the requirements of practical application.(3) The applications of the constructed wetland prediction model. Xinxue River constructed wetland covers an area of about 1.33 km2, After prediction and analysis, on the premise of assuring that the effluent of Xinxue River wetland met the III standard of Surface Water Quality, the largest predictive value of wetland model from February until May about influent water quantity, CODCr, ammonia, TN, TP should be less than 8556.4 m3/d, 26.37 mg/l, 1.81mg/l, 10.51 mg/l, 0.16mg/l, respectively; the largest predictive value of wetland model from June until September, they should be less than 31747.9 m3/d, 45.37 mg/l, 1.93mg/l, 6.18mg/l, 0.09mg/l, respectively; the largest predictive value of wetland model from October until December, they should be less than 11069.2 m3/d, 29.78mg/l, 1.46mg/l, 0.7mg/l, 0.17mg/l, respectively.(4) According to the test results, established the influent water quality and quantity scheduling functional relations of Xinxue River constructed wetland. In the process of function establishing, through repeatedly attempting a variety of theoretical methods, selected trinomial fitting method, with the software of Origin, established the function formulas between water quantity and quality, and made the functions curve. (5) Using the prediction results of the constructed wetland prediction model as the function constraint conditions, through the solving of figures, confirmed the optimal influent water quality and quantity conditions of constructed wetland in each season. from February until May, The influent water quantity of the wetland should be less than 8556.4m3/d, the range of CODCr ,ammonia,TN were (25.47mg/l, 26.37mg/l), (0.11mg/l, 1.0mg/l), and (10.28mg/l, 10.51mg/l), the range of TP was 0.16; from June until September, the influent water quantity of the wetland should be less than 31747.9m3/d, the range of CODCr was (26mg/l, 32.36mg/l), or (37.15mg/l, 45.37mg/l), the range of ammonia, TN, TP were (0.48mg/l,1.78mg/l), (5.15mg/l, 6.18mg/l), (0.07 mg/l, 0.09mg/l), respectively; from October until December, the influent water quantity, ammonia, TN of the wetland should be less than 11069.2m3/d, 0.75 mg/l, 8.61 mg/l respectively, the range of CODCr was (24.55mg/l,26.91mg/l), the range TP of was (0.10mg/l ,0.12mg/l), or (0.16mg/l ,0.17mg/l). |
PDF Full Text Request