Prediction Of Temperature Change In Protected Area Based On Artificial Neural Network

The ever-changing global climate will trigger extreme weather and climate events,which will affect human survival and social development.With the rapid development of science and technology in the modern era,people's requirements for meteorological prediction accuracy have increased.Solutions have been proposed by means of deep learning and data mining to deal with massive meteorological data and improve forecasting accuracy,which has pushed the meteorological modernization industry to a new level.In this paper,BP neural network prediction model is established based on temperature time series.Based on the analysis of time series modeling prediction and neural network modeling prediction at home and abroad,an adaptive BP neural network model is proposed to solve the problems of local optimum,slow convergence speed and difficult determination of network structure of standard BP neural network prediction model.The main contents of this paper are as follows:(1)BP algorithm optimization.The improved methods of the standard BP algorithm mainly include learning rate adaptive method,increasing momentum item method,quasi-Newton method and LM method.In this paper,based on the learning rate adaptive method,an adaptive BP algorithm is used to establish the temperature prediction model.The algorithm dynamically adjusts the weights between different neurons,changes the learning rate,realizes the weight difference adjustment,and shows the adaptiveness of the learning rate.(2)BP model input optimization.According to the multi-dimensionality of temperature series,the stepwise regression analysis method is used to reduce the dimension of the sample to ensure the optimal input sample data.For the temperature sample to be periodic,the introduction of the monthly period variable in the input sample can further improve the fitting accuracy of the BP model.(3)Application analysis of adaptive BP model.The temperature data from January 1,2008 to December 31,2017 in Dalinuoer Nature Reserve was selected to establish a prediction model,and the three variables of daily average temperature,minimum temperature and maximum temperature were predicted and analyzed.Compared with the BP model established by the adaptive BP model and the learning rate adaptive method,quasi-Newton method and LM method,the adaptive BP model has better learning efficiency and prediction accuracy.Compared with the prediction accuracy of the time series prediction ARIMA model,the fitting degree is higher.The adaptive BP prediction model has achieved good results in temperature prediction in the Dalinuoer Nature Reserve,which helps to analyze the climate change of the protected area and take timely measures to protect its ecosystem.