Research On Artificial Neural Networks For Water Quality Forecasting And Realization On The MATLAB

Human's living and development will be faced with shortage of water resource and deterioration of water quality. In recent years, forecasting of water quality has become one of the point of most scholar. However, the nondeterminancy and the highly non-linearity of the water environment pollution processing, come to a decision that the water quality model is complex and difficult to be solved, make the traditional methods and tools seem to be lack the ability to do.Artificial neural network (ANN) plays a leading role in the sciences for complex non-linear phenomena and artificial intelligence. Researches on its application in the forecasting of water quality are still in preliminary stage in the world. On the basis of analysis of present situation of the researches in water quality forecasting with ANN, the writer try to explore on this field. The structure and algorithm of the BP network are studied in the paper, and the BP network which is ameliorated by the L-M algorithm is trained with the data.This paper take Hei river as an example, analyzing the water environment, pollution source and their influence to water pollution, choosing 7 items monitor data of the two cross section of Hei river constitute to the sample and carrying on the analytical processing to the sample data. A 3 layers BP network that includes one hidden layer has been built up, compile program and realize the model of the netwok on the platform of MATLAB. Use the discharged data to training the BP network and the forecasting of water quality will use the network which has been well trained. Compare the forecasting result with the monitor data, the result is found objective and reasonable.Studies of cases have proved that the BP neural network which is ameliorated by the L-M algorithm has good prospects for validity and application, and the accuracy of this ANN water quality forecasting model has been enhanced more greatly than the traditional method. This research demonstrates that this method can foresee the pollution condition of the water environment quickly and accurately, proposes a new way to provide reference to the prevention and treatment work of the water environment.