Research On Temporal-spatial Correlation Of The Near-surface Wind Field Based On The Fractal Theory

Gas diffusion is governed by the near surface wind.Since the near surface wind field is influenced by several factors such as air pressure,temperature,and complex topography,wind signals exhibit strong nonstationary and uncertainty,which further brought great difficulties to the estimation of poison gas diffusion path and the localization of leakage source.Dynamic mechanism and temporal-spatial evolution analysis of the near-surface wind will contribute to a better understanding of gas diffusion behavior,and will help to build effective warning and emergency responding system for gas leakage accident.Hence,we investigate the temporalspatial correlation of high frequency wind signal,and thoroughly analyze the complex oscillatory mode in wind speed time series.The main studying contents of this thesis are as follows:Firstly,aiming at problem that the common wind angle transient in wind direction time series for the 2D ultrasonic anemometer may have adverse effects on the further fluctuation analysis,we study the causes of this problem,and then introduce a correction algorithm based on the magnitude/sign decomposition.For the problems that the 3D ultrasonic anemometer can only output the three dimensional wind speed components,while the instantaneous wind speed and wind direction data should be calculated indirectly,we discuss the synthesis algorithm of instantaneous wind speed and the calculation methods of wind angle.Secondly,the temporal correlation behavior of 3D wind signals is studied by using Hurst exponent estimators and the multifractal detrended fluctuation analysis.Result shows that the high frequency wind speed time series are not only long range power-law correlated,but also exhibit multifractal behavior.However,the horizontal component and vertical component of 3D wind speed signal have different multifractal structure.The multifractality in horizontal component is due to the long-range correlations rather than the probability density,while both long-range correlations and the probability density affect the multifractality in vertical wind speed signal.When the sampling frequency is higher and lower than 5Hz,there exist two different power law relations between the sampling frequency and the singularity value at the peak of multifractal spectrums.Thirdly,the spatial correlation of 2D wind signals is studied by using the conventional cross-correlation analysis methods and the DCCA(Detrended Crosscorrelation Analysis)cross-correlation coefficient.Results show that the 2D wind speed time series measured at different locations are indeed power law crosscorrelated,and the cross-correlation levels are varying both with temporal and spatial scales.The cross-correlation coefficients calculated by using the conventional cross-correlation analysis methods can only reveal the spatial variations between wind signals,while fail to uncover the temporal variations.The DCCA method is a qualitative approach to analyze the complex cross-correlation structure,but it does not quantify the level of cross-correlations.The DCCA cross-correlation coefficient can accurately quantify the temporal-spatial variations of the cross-correlations between nonstationary wind speed time series.Fourthly,in order to further investigate the valuable information contained in the multiscale fluctuation structures of wind signal,we propose an oscillation mode recognition method based on the empirical mode decomposition and time series reconstruction.Our method can identify the role of each intrinsic mode function by monitoring the reconstructive process.Experimental results based on measured wind data indicate that the method can recognize the secular variation in long term wind speed time series,and it can also determine the roles of each mode in the fluctuation structure of wind signal.The results agree well with the physical reality.