A Study On Application Of BP Neural Networks Based On GA And LM Algorithm In Urban AIR Quality Frediction

With the word "haze" appearing frequently as a hot topic in our country and international in recent years, people realized that the blind pursuit of rapid industrial development, although temporarily brought the rapid progress of society, but also had to pay a heavy price, in which Urban air pollution is one of the most painful price. Sulfur dioxide, carbon monoxide, the respirable particulate matter and other air pollution index gradually coming into sight, has also become an important reference for the weather forecast and people’s travelling. At present, China has put the environmental pollution problem as one of the basic national policy task, carried out large-scale environmental governance. How to effectively use urban air quality monitoring data for forecasting and analysis, not only to provide a favorable basis for people to travel, but also to provide an effective help to further take appropriate measures to control pollution.The city Jinan of Shandong Province nowadays has 15 sites using to monitor pollutants in urban air quality data, After about ten years of accumulation, already have begun to take shape data can be an important source of information for us on urban air quality prediction. How to use the vast amounts of data as accurate prediction of urban air quality, it is our main problem in this study. Considering about the neural network widely using in social research for those years and the BP neural network processing unique non-linear problem of capacity, the paper choose to use the most widely used BP neural network as model and use the air quality testing data of Jinan Springs Plaza site as samples with considering barometric pressure, temperature, humidity, wind speed and direction and other limiting factors of the city, according to the current ambient air quality standard GB3095-1996 to the urban air quality forecasting problems to study. Aimed at the shortcomings of BP neural networks, such as the slow convergence, easy to fall into local minimum value,we put forward the corresponding GA and LM optimization algorithm to improve it. The final data show that it is feasible to use the optimized BP neural network in urban air quality prediction.