The Study On Predictive Models About Drinking Wellhead Water Quality Parameter In Some District Of Chongqing By Wavelet Neural Network

Objectives: Discuss the feasibility about using wavelet neural network to predict drinking wellhead water quality parameter. To provide the methodology reference for the drinking wellhead water quality parameter prediction and control as well as for evaluation of the drinking wellhead water quality changing tendency.Methods: 1. we explain the basic principle of wavelet analysis, artificial neural networks, as well as wavelet neural network. 2. Apply BP neutral network (BPNN) and wavelet neural network (WNN) to forecast the monthly average concentration of potassium permanganate index from 2001 to 2005 in Yuzhong District of Chongqing, one drinking wellhead water quality parameter. 3. The predictive results of two models were compared by some related statistical indexes.Results: apply RMSE and MAPE to evaluate the forecasting results. And it indicated that the predictive precision of WNN prediction model about research data was superior to that of BPNN prediction model.Conclusions: 1.the research data have a greater part of random component by wavelet analysis. So it's difficult to predict the research data by the traditional artificial neural network methods and the predictive results was not so good, because it's poor in generalization ability. 2. As for research data, the WNN predictive model was strong in simulation ability for function and in generalization ability. And the predictive performance was good. 3. it's a practical and beneficial exploration for applying wavelet neural network method to predict and control the drinking wellhead water quality parameter.