Research on a global periodic climate prediction method based on improved BP neural network

Climate prediction has a wide social demand, but due to the complexity of climate system formation and the uniqueness of Chinas terrain, accurate prediction is extremely complex and difficult. Therefore, this article proposes a global periodic climate prediction method based on an improved BP neural network. Firstly, based on historical meteorological data, establish a climate model and summarize and analyze the characteristics of the patterns; Then, based on the BP neural network, considering the relationship between adjacent days, rolling sampling of historical data is carried out according to seasonal and meteorological factors in the sampling interval, establishing a multi-state transition probability matrix, and constructing an annual time series forecasting model; Finally, taking precipitation prediction as an example, an example was used to simulate the annual precipitation situation. The calculation results have verified the effectiveness of the proposed method, indicating that it can simulate climate conditions well and provide guidance for climate governance and engineering construction.