A WATER POLLUTION PREDICTION MODEL BASED ON IMPROVED LSTM NETWORK IN THE IOT ENVIRONMENTAL MONITORING SYSTEM

With the development of industrialization and urbanization of human society, the problem of water quality and environmental safety has become increasingly prominent. Water pollution prediction is the most important part of water quality protection. Its accurate prediction can provide more reliable support for water resource management and application. In order to further improve the accuracy of water pollution prediction, a water pollution prediction model based on Internet of things and improved long and short-term memory neural network was proposed. In this paper, the water quality parameter collection terminal is designed based on embedded software, and the missing value filling algorithm is adopted to supplement the two core parameters of missing value and time label of water quality sample. Finally, the long and short-term memory neural network is improved by using gating cycle unit module. Based on the time sequence of water quality parameters, prediction models of different water pollution parameters are constructed. Finally, feasibility and validity of the model are verified by using Taihu Lake online water quality monitoring data as samples. The experimental results show that proposed water pollution prediction model has better adaptability. The root means square error of dissolved oxygen, ammonia nitrogen and turbidity are 0.0871, 0.0932 and 0.515 respectively, which is more accurate than traditional SVM model and BPNN model, and has important theoretical feasibility and engineering applicability.