Predictive Analysis Of Rainfall And Runoff Time Series Based On Chaos Theory In Hulun Basin

In recent years, rainfall and runoff decrease sharply of Hulunhu Basin because of influenced by many factors such as the global climate change and human activity, its lead to vegetation degradation, grassland atrophy and soil desertification. In view of the problem, used of modern new theory and new methods in this paper, including the chaotic theory and the analysis method, LS-SVM model, RBF neural network model and ARIMA model, used them to predict analysis of monthly rainfall and monthly runoff of Wuerxun River and Kelulun River in the Hulunhu Basin.Application of phase space reconstruction, used it to phase space reconstruction of monthly rainfall and monthly runoff time series of Wuerxun River and Kelulun River, used the autocorrelation function method to determine delay time, used saturation correlation dimension to determine embedding dimension, and used saturation correlation dimension and Lyapunov exponent method to carry on the chaos identification, the results showed, they have chaotic characteristics of monthly rainfall and monthly runoff time series of Wuerxun River and Kelulun River.Application of LS-SVM model and RBF neural network model respectively make model of monthly rainfall and monthly runoff chaotic time series of Wuerxun River and Kelulun River, and analyzed the prediction results of LS-SVM model and RBF neural network model. The results showed, the two models have their own advantages and disadvantages, the relative error percentage variation of prediction results is relatively stable of LS-SVM model, the prediction accuracy is very good. RBF neural network model in plentiful water period prediction accuracy is very high, but in dry period, the prediction accuracy is relatively low, the results of prediction relative error percentage fluctuation.ARIMA model is application to predication and analysis of monthly rainfall and monthly runoff time series of Wuerxun River and Kelulun River. And comparative analysis to the predict results of ARIMA model, LS-SVM model and RBF neural network model. The results showed, the qualified rate of LS-SVM model is higher than ARIMA model and RBF neural network model, and the qualified rate of ARIMA model is higher than RBF neural network model. Generally speaking, the relative error percentage of LS-SVM model is less than ARIMA model and RBF neural network model, and the relative error percentage of ARIMA model is less than RBF neural network model.In short, application of the new theory, new method in this paper, multiple pathways research to rainfall and runoff variation characteristics of Hulunhu basin, for further study hydrologic characteristic of Hulunhu basin provides reference.