Problems Solving With Granular Computing And Deep Learning For Water Quality Prediction

The water of life feeds everything of our planet, and it is particular significant for this world. However, with the improvement of people's life quality in the 21 st century, the water situation becomes more and more serious. In addition, the water resources is facing serious pollution.In order to control the floods, produce electricity, protect and make full use of water resources, the Three Gorges Dam which is the largest hub of water resources and hydropower engineering of the world has been built in China. To make sure the quality of water and environment in the Three Gorges region and improve the management of the Three Gorges Dam, Chongqing Institute of Green and Intelligent Technology, Chinese Academy of Sciences participated in the major national science and technology special project, the Three Gorges water environment perception system and operation platform. The content of this article based on this project and use the advantage of granular computing for water quality prediction.Water quality data has complex nonlinear characteristics, and granular computing can simulate the solving process of human brain. It can use a few simple problems to replace the intricate problems. Then, the uncertain and nonlinear complex problems can be solved easily by granular computing. The masteral thesis will combine the granular computing and water quality prediction, study the mechanism and non-mechanism water quality prediction, specific research content is as follows.(1) The mechanism water quality prediction model fully consider the various factors which influence the water quality. Among them, the convection diffusion equation can be used to describe the diffusion process of pollutants. It is great theoretical and practical significant that quickly and accurately get the pollutants concentration distribution. This article will combine the reconstruction method of the solution between the granular layers which is a part of granular computing and finite element method to construct a new algorithm for solving nonlinear convection diffusion equation, and analyze theoretically the astringency of this method and the time complexity. At the same time it also prove the effectiveness of this method using mathematical examples. This method will make the nonlinear problem which originally need to be solved directly on fine particle layer(target granularity) converted into first solving on the coarse grain layer, then deducing the solution of the fine particle layer with coarse grain. This method leaves all the nonlinear problems in coarse grain layer which is easy to solve, and the fine particle layer only make the linear calculation. It is not only to eliminate the numerical oscillation, but also reduce the complexity of the problem without sacrificing accuracy, at the same time it also improve the solution efficiency.(2) Based on the water quality forecast of deep learning. Although mechanism water quality model can fully understand the water quality change process, it need a lot of data and information to determine the model parameters, and it is only valid for a specific waters. Deep learning technology has good data nonlinear approximation ability and self-learning ability and generalization ability, and it can running simulation to predict the complex nonlinear phenomena. This article attempts to use the water quality prediction model based on deep learning to predict the data of sewage treatment plant which has complex nonlinear characteristics, and uses the particle swarm optimization algorithm. It can dynamically optimize the number of neural unit of hidden layer in prediction model and learning rate, and improve the convergence speed and generalization ability of prediction model, and make the prediction results more scientific and accurate, and also provide a new approach to solve water quality forecast.