Development And Application Of Modeling Approach For Forecasting Of Water Environment

Along with the increasing problems of water environment, a comprehensive, system or quantitatively research on the simulation, forecasting, supervision of water quality become concerned focus in solving various problems of water environment. It has important practical significance that development and application of mathematical model of water environment. By this method, one can not only carry on the appraisal, predict the development trend of water environment and carry out how to control the pollution source, but also which may provide the decision-making basis for the environmental supervising management.This paper firstly to combine the grey model with the control program using MATLAB2011b, which can carry out error feedback to original model, in order to obtain the higher accuracy of prediction model. Secondly, the prediction model is applied to the case study of Yaan section of the Qingyijiang River Basin. The accuracy grade distribution of the water quality index can be determined based on the simulation and evaluation of two mainstreams and six tributaries of sections (DO, pH, CODMn, NH3-N) from2006to2011. And the Grey Forecasting Model are established based on observed time series, which could reveal the statistical properties of dynamic data, and predict future value from2012to2018, based on past observations. Lastly, y(t)min and Tmax can be determined.In addition, autoregressive moving average model and BP artificial neural network model are established based on observed time series. Meanwhile, three models were evaluated separately based on four kids of evaluation indexes, MSE, MAPE, MSPE and Theil IC. The results indicated that the three models are better, and the BP model fitting precision is highest, but its practicability and controllability are poor. Owing to the BP model can only be used based on the original code of the operation and the network training process is not stable. The ARMA prediction model must meet the requirements of control parameters along with white noise sequence, which is not suitable for long-term prediction model and dynamic instability of sample. The fitting precision of the GM (1,1) model equated with ARMA model, which is a simple, practical prediction model compared with other prediction methods.