Research On Chaos Characteristics And Forecasting Of Thunderstorm Time Series

The sequence of thunderstorm is a complex nonlinear time series, study of its variation is the premise and foundation to conduct disaster prevention and mitigation planning and make decision. For quite some time, many scholars applied traditional deterministic methods or stochastic methods, or combined of both methods to describe the change process of thunderstorm to reveal the variation of thunderstorms. In this paper, take Nanjing as the object of study to research the variation of thunderstorm. The first of all, research the long range dependence of thunderstorm time series; Then apply the chaos theory in the study of the variation of thunderstorm to qualitatively and quantitatively analyze the chaotic characteristics of thunderstorm time series, and combine the chaos theory and RBF-NN to forecast it. The main findings are as follows:(1) Through use detrended fluctuation analysis method in month thunderstorm sequence of five stations in Nanjing for 53 years to reveal the long range dependence and inherent laws of thunderstorm. The results show that:Thunderstorm time series in Nanjing has obvious long range dependence and timescales nature. The general trend of future thunderstorm sequence of Nanjing Station increase weakly, and the others decrease. The power law relationship of every station in Nanjing changed significantly before and after the change point what also means the slope changes, which has two to three significant scale-invariant region, reflecting complex physical mechanisms of the thunderstorm systems in Nanjing;(2) Phase space reconstruction of thunderstorm time series. The phase space reconfiguration is completed about the average month thunderstorm time series of Nanjing. The results show that:Respectively, time delay τ calculated by Mutual Information method and the best recent embedding dimension m calculated by False Nearest Neighbors method of thunderstorm time series about Nanjing is 3 and 6, then reconstruct the phase space of thunderstorm time series, its attractor trajectory has complex fractal structure, qualitatively reveal chaotic characteristics of thunderstorm system.(3) Chaotic characteristic indentification of thunderstorm time series. Quantitatively analyze the chaotic characteristic of the average month thunderstorm time series by calculating correlation dimension, Kolmogorov entropy and Lyapunov exponents of time series of thunderstorm days about Nanjing. The study shows:The correlation dimension D is 2.3238, Kolmogorov entropy is 0.0462 and Lyapunov exponents is 0.0012 of thunderstorm time series, The results of the three feature quantities are greater than non-zero integers, which quantitatively reveal chaotic characteristics of thunderstorm system, what means that thunderstorm time series is chaotic time series with the chaotic characteristics, the system exist chaotic mechanism intrinsically.(4) Chaotic forecasting of thunderstorm time series. For non-linear characteristics of thunderstorms time series, build a RBF-NN chaotic prediction model to predict thunderstorm time series in Nanjing. The results show that: RBF-NN model can be used to predict thunderstorm time series, the prediction has higher precision that can well reflect the inherent variation of thunderstorm time series, and the prediction accuracy of thunderstorm time series after the phase space reconfiguration is completed is significantly higher than The prediction accuracy of the original thunderstorm time series. The results confirmed that the model can truly and accurately reflect the changes in the overall trend of thunderstorm, which provides a new method to judge the nonlinear nature of time series at the same time.