The Study Of Phone-Space Reconstruction Prediction Method In Geostatistical Data Analysis

Chao, the nature of a broad irregular movement, is a behaviour generated by deterministic non-linear dynamic system. Most data in geological analysis are non-linear, chaotic and fractal. Prediction in geological science is of vital importance. Nonlinear time ( space ) sequence is an important part in geological data analysis. The traditional method of geostatistics is the theoretical foundation of our studies on data analysis. However, although the traditional geostatistical methods are still very effective, the non-linear characteristics of most geological data make them insufficient. In recent years, phase-space reconstruction technology based chaotic and fractal theory develops rapidly, which has become an effective means for non-linear sequence prediction.This paper comprehensively discusses the prediction method based on the phase-space reconstruction.We discuss the method of chaotic time series, phase-space reconstruction as well as Kriging method of the traditional geostatistics, and improve weighted local forecasting methods by combining the advantages of these two methods. We also put forward the new local prediction method to determine a weighted coefficient with the combination of Kriging method, and based on the principles of minimum variance estimation error, provide how to determine the set point with the combination of Kriging weighted local prediction method. This paper studied the basic theory of chaos forecast and based on the theory of phase-space reconstruction nonlinear chaotic sequence prediction, conducted in-depth discussions. There is a detailed discussion on the parameter-determining methods, such as embedding dimension, time-delay, the Lyapunov exponent parameters. At the same time this paper discussed the current epidemic basis methods of chaotic sequence prediction which based on the theory of phase-space reconstruction. This paper studied the theory of geostatistic Kriging method and its application, discussed the existing problems, analyzed their strengths and the problems on the defect of solving nonlinear problems, and did a corresponding improvement concerning the existing problems.Concerning with local prediction method, the choice of adjacent point is the first-encountered problem and it must be solved. The determination of the adjacent point decided the choice of prediction methods and played a decisive role in prediction accuracy. By now, the method of adjacent point determination usually bases on the inverse distance method and correlation. This paper discussed these methods and put forward a new set of asynchronous multi-point method to fix the adjacent point.Local forecast believes that the state of next step is determined by the state of current moment, from all the history and current status of adjacent moment to the next phase of change from fitting. This approach, taking into account only the current moment state to the next phase of change, does not take into account the impact of the moment before the current moment. This paper put forward asynchronous two local forecast methods combining with asynchronous multi-point set method of adjacent point.Error analysis on current method of phase-space reconstruction prediction was discussed in this paper and estimation error analysis with the combination of Kriging method was given. Combining multiple interpolation theory, the inherent mechanism of phase-space reconstruction forecasting methods and error were analyzed.This paper analyzed the method of phase-space reconstruction prediction from the perspective of multi-angle approximation, and put forward a novel phase-space reconstruction prediction method of multi-angle approximation.In the actual log identification data, the improved methods were applied. The effect of numerical examples comparison test is good.