The Research On Predicting The Water Quality Of Lake With Grey Elman Neural Model

The prediction of the status of lake water is about deducing and estimating the development trend of the quality and quantity of water resources in a region scientifically.The prediction is used to achieve the purpose of guiding people to use and develop water resources more reasonably and effectively and protect the ecological environment of lakes during the development of economy and society.The main task is to predict the future trend of lake water resources in future,and to provide basis for the use and management of water resources.Taking the characteristics of the water quality data itself into account,the grey GM(1,1)model and the traditional grey Elman neural network model are firstly made to predict the water quality.Grey GM(1,1)requires less original data,and nonlinear fitting ability of Elman neural network is strong.Combination grey with Elman neural network complement each other.The combination offsets the poor data fitting ability of grey GM(1,1)in greater volatility data and the poor data fitting ability of Elman neural network in small sample.Lake water quality index dissolved oxygen(DO),chemical oxygen demand(COD),NH3-N week average from November 24 2014 to January 11 2015 are taken as the original data in Xiao Guan Yi of The Erhai Lake to predict DO,COD,NH3-N concentration from January 19 to Feburary 1.So the precision of two models can be tested.Through a series of research and analysis,the final results show that: Variance of the NH3-N original data set is minimum.When predicting the original data the prediction accuracy of traditional gray neural network is poorer than grey GM(1,1)'s.The volatility of DO original data is moderate and variance is about 0.14.When predicting the original data the prediction accuracy of traditional gray neural network is almost the same with grey GM(1,1)'s.The volatility of COD original data is large and variance is about 0.19.When predicting the original data the prediction accuracy of traditional gray neural network is better than grey GM(1,1)'s.Therefore,we need to find a new improved model to obtain more accurate results in any different situations.Considering systematic the defects of the traditional grey Elman neural network,the improved grey Elman neural network model is established.The model retains the advantages of the traditional model,and changes the structure of the combination of traditional way to make it more optimized.Lake water quality index dissolved oxygen,COD,NH3-N week average from November 24 2014 to January 11 2015 are taken as the original data in Xiao Guan Yi of The Erhai Lake to predict DO,COD,NH3-N concentration from January 19 to Feburary 1.So we can validate the precision of the improved grey Elman neural network model.Compared with the traditional model,the accuracy of the prediction results is greatly improved,which shows that the improved model is feasible.