Research On Yunnan Regional Drought Prediction Method Based On NGM(1,1,k) Model

China is a country with severe water shortages,and drought is the main natural disaster facing China.In recent years,with climate change,economic development and population expansion,water shortage has become more and more serious.More frequent and serious drought has become a constraint to social development,agricultural production and resident life in China,which seriously threatens human life.Yunnan is located in the Yunnan-Kweichow Plateau,due to the influence of the plateau monsoon,the change rate of precipitation is very large,and it is easy to cause surface drought.In recent years,frequent drought has seriously restricted the local social and economic development,also,it has a negative impact on the quality of life of residents and the level of agricultural development.Therefore,it is very significant to study the occurrence law of drought in Yunnan and forecast the drought in the future.Based on the meteorological data of Yunnan since 1961,this article studies the occurrence and degree of drought,and rainfall prediction methods.The main contents are as follows:(1)Studies and analyses the Meteorological Drought in Yunnan,based on Precipitation Anomaly in percentage(PA),Z index and Standardized Precipitation Index(SPI),The results showed that PA had low sensitivity to the drought in Yunnan,the occurrence rate and degree of drought was not consistent with the actual situation,which could not correctly reflect the characteristics and laws of drought in the area.SPI and Z index are sensitive to drought in Yunnan,which can accurately describe the drought in this area and can be used for local drought analysis.Moreover,the drought incidence of Z index and SPI is about the same.Therefore,Z index and SPI should be used in the analysis of drought in Yunnan.(2)Based on the geometric meaning of Non-homogenous discrete exponential grey mode(1,1,k),this article proposes a background value optimization method.Through the verification of homogeneous,approximate non-homogeneous,strictly nonhomogeneous,descending concave,descending convex,ascending concave and ascending convex data,it is shown that the model deduced in this article achieves the highest accuracy in all seven types of data and achieves the method innovation.(3)Improves the calculation method of topology prediction.Compared with the traditional topological prediction method,the verification of 2012-2016's spring rainfall forecast in Yunnan area shows that the average accuracy is improved by 77.88%.(4)Taking NGM(1,1,k)model as the core,combining with wavelet decomposition,a series prediction optimization scheme is constructed.The results show that the prediction accuracy of wavelet-NGM(1,1,k)model is very high,and the sequence prediction method of NGM(1,1,k)model is optimized.