Prediction Of Zhengzhou City Precipitation Based On Chaotic Time Series

The global climate change leads to the occurrence of natural disasters is becoming more and more frequent.The prediction and prevention of natural disasters has become a subject of great concern.Rainstorm and flood are the most common natural disasters,precipitation is an important influencing factor of rainstorm and flood disasters.Accurate precipitation prediction can reduce unnecessary property losses and casualties,which has very important guiding significance for disaster prevention and mitigation.Rainfall is one of the main sources of surface fresh water,which is comprehensively affected by geographical location,atmospheric circulation and weather system conditions.It is the most basic link of water cycle process.The quantity of rainfall is collectively referred to as precipitation.Since precipitation is affected by various factors and its distribution is inequality and unstable in time and space,the series of precipitation shows the characteristics of nonlinearity and randomness.Therefore,introducing chaos theory to analyze it can better reflect the internal law of precipitation change.In this thesis,based on the monthly precipitation data of Zhengzhou from 1979 to 2020,the chaotic characteristics of the precipitation time series were analyzed.The method of machine learning combined with Variational Mode Decomposition was used to predict the precipitation time series,so as to provide a certain reference for preventing drought and flood disasters.Firstly,the chaotic characteristics of monthly precipitation time series were identified,the time delay τ and embedding dimension d of the parameters required for phase space reconstruction were calculated according to the C-C method.The obtain parameters were applied to Wolf method to get the maximum Lyapunov index ?λ﹥0.The chaotic characteristics of monthly precipitation time series were known.Secondly,the Support Vector Regression model is used to predict the precipitation time series with chaotic characteristics.Compared with the prediction accuracy of Random Forest,RBF neural network model and BP neural network model,it is found that the prediction accuracy of Support Vector Regression model is better than the two models.Thirdly,the method of Variational Mode Decomposition was introduced to decompose the chaotic precipitation time series into several relatively stable components,predict each component and sum the results to get the final prediction result.The prediction results of Empirical Mode Decomposition and Ensemble Empirical Mode Decomposition were compared,it was found that the prediction accuracy of the established VMD-SVR model was the highest.Finally,the parameters of VMD-SVR model were optimized to find a suitable prediction model,which was applied to the prediction of precipitation time series.