The Regional Forest Fire Research Based On BP Neural Network

In 2013, the 8th national forest inventory conducted by national statistics departments shows that there are 208 million hectares of forest area with 21.63% of forest coverage rate. Total stumpages are 16.433 billion cubic meters with forest reserves 15.137 billion cubic meters. The natural forest covers an area of 122 million hectares with accumulation of 12.296 billion cubic meters; 69 million hectares plantation area, volume 2.483 billion cubic meters. Compared with the international forest resources, forest resources in our country are relatively scarce. Forests, which are called the lungs of the earth, are the oxygen manufacturing sites which are important to our survival space, while at the same time have a filter effect for the dust in the air and play an important role in defending sandstorm. Forests are necessary precious natural resources both for the development of the national economy and environment protection. However, the forest fires make these resources cause great harm. Therefore, prevention of forest fires and put forward effective measures are urgent affairs.Prediction and prevention of forest fires have great significance for the sustainable development of the economic environment and the social environment.Artificial neural network is an effective method for the prediction and prevention of forest fires.Through putting the climatic factors which might cause forest fire as the input of neural simulation and output of fire situation, weight of factors affecting fires and connection between these factors can be effectively analyzed. Artificial neural networks can optimize their network structure and work out the desired function computing architecture through a large number of statistics simulation practice and simulation operations. This approach has more advantages over logical reasoning calculus on massive data processing and analysis.This thesis first provides an overview of the current situation of forest fires,then describes the artificial neural network and BP neural network architecture, features and algorithms and gives a brief analysis of the relevant climatic factors affecting forest fires correlation. In this study, the historical data climate change and forest fires of Guilin in Guangxi Province and Guangzhou City in Guangdong Province is collected as the research basis, and an improved particle swarm optimization algorithm was introduced to optimize the performance of the BP neural network.Compared with the traditional BP algorithm and particle swarm optimization, the improved particle swarm optimization algorithm can not only quickly converge to the neural network learning objectives, but also improve the accuracy of the prediction model.