Applications And Research On Atmospheric Quality Evaluation And Forecast By The Genetic Algorithms And BP Neural Network

In recent years, the speed of the genetic algorithm and BP neural network development is very fast. There are many weaknesses in the traditional genetic algorithm. It can’t maintain the ability of continuous global optimization search and cannot ensure the superiority of the species and the convergence speed. And then the number of individual diversity of the groups will gradually become less in the process of arithmetic operations. As it has a poor ability of the local search, it cannot avoid the phenomenon of prematurity, and thus it cannot ensure the computing efficiency of the algorithm.By the study of genetic algorithm (GA) and deep understanding of the genetic algorithm theory, I have improved its structure and training methods. I added the elite individual migration operator, the local search operator, to the point of crossover operator in the traditional genetic algorithm to improve the training process of the genetic algorithm. And I had made a new improved hybrid accelerating genetic algorithm (IHAGA), and use it in the air quality assessment model, and proposed a universal air quality assessment method for a variety of atmospheric pollutants. And achieved a good results from the experimentsArtificial neural networks show some of its own shortcomings and problems in the actual prediction and application gradually along with the artificial neural network in the continuous development of environmental quality and atmospheric sciences and other fields and in-depth. Such as How to determine the initial weights and the structure model of the neural network-the number of hidden nodes and the number of hidden layer of the artificial neural networks, etc. And it is difficult to determine the determined learning parameter and momentum factor of the network. These are all important conditions to ensure the correct operation of the neural network, but through repeated practical training in order to be determined. And which in some practical applications, artificial neural networks are often over-fitting, so that serious affected the entire neural network generalization capability, which greatly limits the practical application of neural networksBy the study of genetic algorithm (GA) and artificial neural network method, I optimization and improved the BP neural network by using the improved genetic algorithm. Optimize the network structure of neural network and weights and threshold at the same time. And this method can make up for the lack of a single BP neural network. And then use the IHAGA-BP neural network to predict the pollutants in the atmosphere. Instance of the prediction results show that the improved BP neural network by IHAGA has a certain application prospects in the field of air pollution forecast future.