Analysis And Prediction Of Lake Water Quality Change Trend Based On LSTM And BP Neural Network

The situation of my country’s water resources is becoming increasingly severe.As an important component of water resources,lakes play an important role in my country’s ecological balance.However,with the development of science and technology,the water quality of lakes is increasingly polluted by industry,agriculture and daily life,and the management of the water environment is urgent.Water quality evaluation and water quality trend analysis are important content of water quality improvement work,which can provide scientific theoretical basis for establishing scientific water quality management and control programs.In addition,as an extension of water quality evaluation,water quality prediction can provide scientific guidance for water pollution prevention and control,and it has a pivotal position in water environmental protection.Taihu Lake is the second largest freshwater lake in my country.It is the foundation of life,economy,and society,and it has an important mission.The thesis took the water quality change trend of the Taihu Lake Basin as the research object,and carried out the following research:(1)Use single factor index method,comprehensive index method and other water quality evaluation methods to evaluate the water quality of lakes.The results of the two evaluation methods show that the water quality of Taihu Lake from 2005 to 2015 was classified as Class V water and heavily polluted water.(2)Spearman rank correlation coefficient method and seasonal Kendall test method are used to analyze the trend of water quality changes in the Taihu Lake Basin.The evaluation results of the two methods are roughly the same.From 2005 to 2015,the ammonia nitrogen,chemical oxygen demand,total phosphorus and total nitrogen of the Taihu monitoring station mostly showed a downward trend,and the dissolved oxygen showed no obvious change trend.(3)Establish a water quality prediction model based on long and short-term memory network and error back propagation neural network to predict water quality,use the 2005-2013 data set as the training set,and the 2014-2015 data set as the test set.And set up four sets of comparative experiments: long and short-term memory network,error back propagation neural network,random forest and gated cycle unit water quality prediction model to evaluate the water quality prediction model.It is found that the long-short-term memory network shows good performance in predicting the concentration of dissolved oxygen and total phosphorus in Taihu Lake.(4)Combine the constructed water quality prediction model with the particle swarm algorithm to find the optimal parameter combination of the water quality prediction model,and compare and evaluate it with the model before optimization.It is found that the particle swarm has the highest improvement in the prediction accuracy of dissolved oxygen for the water quality prediction model based on the error back propagation neural network and the long short-term memory network.The study adopted the Spearman rank correlation coefficient method and the seasonal Kendall test method to prove that the water quality of Taihu Lake showed a trend of improvement from 2005 to 2015.The seasonal Kendall test method obtained more accurate results and smaller fineness;the water quality prediction results of Taihu Lake showed that,The long and short-term memory network water quality prediction model has higher accuracy in predicting the water quality of Taihu Lake,especially in the prediction of dissolved oxygen and total nitrogen;particle swarms can shorten the optimization time of the optimal parameter combination of the prediction model,and The accuracy of the model in the prediction of dissolved oxygen in Taihu Lake has been improved the most.