Study On Water Quality Simulation And Prediction Of Lihe River Section In Luanhe-Tianjin Water Diversion Project

Water quality simulation and prediction is the key foundation of water environmental management and water pollution controlling. The mechanism water quality models take account of the factors that have impact on the change of water quality, so the results are usually satisfactory. But they often require a great deal of background information and always complex. However, water quality models without considering mechanism often acquire satisfactory results because it was built on aiming at specific water quality system by using statistical method or other mathematical methods.According to the actual condition of Lihe River section in Luanhe-Tianjin Water Diversion Project, this thesis builds up a mechanism water quality model and a non-mechanism water quality model, i.e. the two-dimensional vertical mechanism water quality model from the outlet of tunnel to Guo He-qiao and the non-liner time series of BP neural network water quality model at Qian Mao-zhuang cross section.The two-dimensional vertical mechanism water quality model is built by using the continuity equation, the momentum equation, the free water surface equation, the equation of state and water quality transport equation. These equations are solved by the finite difference method. This model can simulate the hydrodynamic effect and the actual state of water quality. The model must need to be verified and its parameters must need to be identified, when the model's prediction expected error is accepted eventually. The comparison of the computational results with field data of hydrodynamic and water quality shows that this model can simulate and predict the water quality of the study area well.The non-liner time series of BP neural network water quality model is a black-box model. It is based on the theory of artificial neural network. The model is established by the non-liner time series neural network forecasting method. BP neural network model, the L-M algorithm and the trial and error method is used to train and test the network. When the model's prediction expected error is accepted, the model can simulate and predict the water quality.Compared the computational results of these two models, we can see that the non-liner time series of BP neural network water quality model is more accurate than the two-dimensional vertical mechanism water quality model at this research area, when the same reaserch data is used.