Water quality prediction method based on LSTM neural network

Abstract

Water quality prediction has more practical significance not only for the management of water resources but also for the prevention of water pollution. It's a time series prediction problem which the traditional neural network isn't suitable. A new water quality prediction method based on long and short term memory neural network (LSTM NN) for water quality prediction is proposed in this paper. Firstly, a prediction model based on LSTM NN is established. Secondly, as the training data, the data set of water quality indicators in Taihu Lake which measured monthly from 2000 to 2006 years is used for training model. Thirdly, to improve the predictive accuracy of the model, a series of simulations and parameters selection are carried out. Finally, the proposed method is compared with two methods: one is based on back propagation neural network, the other is based on online sequential extreme learning machine. The results show that the method is more accurate and more generalized.