The Research Of Aviation Dangerous Weather Forecast For Fog&haze Based On BP Neural Network

The weather phenomenon as fog and haze is common in autumn and winter. Along with rapid development of economy and society, the energy demand grows continuously, and the ability of environmental self-purification has reduced, the frequency of fog and haze with scope of the problem to be appearing more and more popular in present China. It has produced increasingly large impacts with negative effects in the social life. In fact, fog and haze are two kinds of weather phenomenon, yet they are closely related to each other in their generation and dissipation that make it possible to forecast together. In the air transport sector, low visibility from fog-haze is the main factor leading to flight delays, even safety accidents. To improve the accuracy of their forecasting has a positive that can not only bring down the security accident rate, but also can be rationally adapted the flight, dispatch procedure in advance. It has very active means for the society and economy.The BP neural network is the representative in the artificial neural network(ANN), it have characteristics of high precision and nonlinear which has been widely used in pattern recognition, control system etc. It also can be used to design for elaborate weather forecasting that is difficult using conventional methods. As the most extensive application of the ANN model, BP neural network also has very good performance in forecasting of meteorological elements. Therefore, it is feasible for BP neural network to forecast for dangerous fog and haze, and be applied in the aviation meteorological guarantee business. The contents of this dissertation mainly include the following aspects.(1) A data-collecting with screening and correction, datasets required for modeling work to fog-haze forecast that are not limited to manual observation and automatic weather station, but also include completing daily documentation related to fog and haze.(2) After analyzing development of the numerical weather prediction actuality and the application of Neural Networks on forecasting, and for the problems in aviation weather safeguard that the present research investigates on traditional experienced and linear systems which cannot solve it, to design of the fog-haze forecast model based on BP neural network.(3) In order to improve the output precision of fog-haze forecast model based on BP networks, we used Genetic algorithm with decimal float-Coding to optimize connecting value and threshold neurons that could increase the efficiency of the BP neural network, and refined the design steps of genetic algorithm optimization which was realized with MATLAB software system. Airport environment and the principle of generation and dissipation would be determined prediction factor for model, and it is suitable for single forecast model to fog-haze forecasting.(4) The computer simulation modeling in output precision, stability and fitting with using MATLAB software, would be carried on to further confirm the conclusion obtained from these analysis that the forecasting precision of FGA-BP model. In the end, we can draw conclusion that this model have certain practical value and worth popularizing.