Research On Precipitation Prediction In Huanghuai Area

Precipitation is the condensation of water vapor formed in the atmosphere and falling to the ground,which is commonly seen in life in the form of rain,snow,hail,etc.Precipitation as a direct or indirect supplemental source of a variety of water bodies on land has an important impact in agricultural production,soil and water resource loss,etc.Precipitation is affected by a variety of meteorological factors in time and space with strong uncertainty,and the annual precipitation-related meteorological drought,heavy rainfall,flooding,mudslides and other disasters in China cause losses of up to 100 billion yuan.The Huanghuai region is located at the intersection of the south and north of China,where there are many plains and is an important food production area in China.At the same time,the precipitation situation in the Huanghuai region varies greatly from year to year due to its geographical location,and is one of the areas seriously affected by droughts and floods in China.The article uses wavelet analysis and neural network methods to predict precipitation time series data from 1951 to 2020 at observation sites in the Huanghuai region.The main contents of the study are as follows:firstly,the precipitation index is constructed,and the standardized precipitation index(SPI)at different monthly scales is generated using the original monthly precipitation data as the basic research data and the regional precipitation condition is evaluated.In view of the non-stationary and non-linear characteristics of meteorological data,the article enhances the learning ability and local analysis ability of the data through neural network models and wavelet analysis methods,which help to promote the improvement of the prediction ability of regional precipitation compared with traditional statistical models.The article constructs four single neural network models and compares the multi-cycle prediction results of different kinds of neural network models to compare and analyze the prediction effect of single neural network models.The article makes use of the excellent characteristics of wavelet transform tool in signal local analysis,constructs a hybrid model of wavelet analysis and neural network,generates subseries of different frequencies of precipitation index by wavelet decomposition and reconstructs each subseries for prediction,uses multiple hybrid models for multi-round prediction of regional precipitation,and compares the prediction results of hybrid models with those of single models.It is found that the neural network model has excellent learning ability for non-stationary nonlinear data,and achieves better results in precipitation prediction,and the results are very stable after multiple rounds of prediction.For the hybrid model of wavelet analysis and neural network,due to the excellent performance of wavelet transform in local analysis of data,the hybrid model achieves better results than the single neural network model in precipitation prediction in the Huanghuai region,which greatly improves the prediction accuracy of the model and provides an important reference for improving the regional precipitation prediction capability and promoting the research of meteorological data.