Study On Adaptability Of Establishing Forecast Model For Water Quality By Artificial Neural Network

In many years, forecast model for water quality is one of the water environment lessons concerned of by all countries scholars.The United States scientist Streeter established with the Phelps the first headstream one-demention model for water quality in 1925 in the world and it was the basis of mathematics model for water quality . Afterward every kind of mathematics model develops continuously.These models rises very important function in the environmental protection of water. However, the mathematics model application suffered very general situation limit as identify the problem of its parameter .In recent years, along with artificial neural network(briefly named ANN) theories and its application technical increasingly perfect, its application is wider and wider.ANN is becoming the most attractive growth point in every course. Now,there are only some papers simply applying ANN on forecast model for water quality but few paper further study model structure,hiding cell number, train epoch,data process and learn rate.Whether the ANN adapt to establish forecast model for water quality or not is lesson to deserve to study .The thesis made the first step study on adaptability of establishing forecast model for water quality. According to two parameters as train epoch and test error, the thesis contrast structure of three-layer network and that of four-layer network through trials. Study show that the ANN is adapted to establish the forecast model for water quality and ANN's unstability.ANN's structure , hiding cell number, learn rate and transfunction are researched.The conclution can provide beneficial aid for the same problem. Finally, some questions and experience are given in order to do well in the future...