Research On The Processed Method Of Dynamic Data On The Ground Of Modern Time Series Analysis Method

This dissertation mainly focuses on the processed method of dynamic measurement data, including the solution of adaptive factor, the solution of colored noises and weighted measurement fusion and so on.The main works and contributions are summarized as follows:1. There are three processed methods of dynamic measurement data, included Wiener filter, Kalman filter and modern time series analysis method.First the basic idea and applicability of these methods are discussed systematically. Then the relations of these methods are proved in this article.2. In the processed of adaptive sequential adjustment, a method to measure the range of the adaptive factor is brought up on the ground of the variance component estimation principle. We can use mean square error as criterion which judged the results good or bad, and then an approach of calculating the adaptive factor by decompounding spectrum and the formula of the way is deduced. It is shown that the value of the adaptive factor not only makes the mean square error least but also the results subsequently by the way have the actual meanings.3. On the base of analyzing colored noises thoroughly which influence the estimate results of parameter. A new approach called random model series-express of colored noises and compensation is presented by polynomial-quotient and the formula of the way is deduced. Because this method control the influence of colored state noises by polynomial-quotient so it is provided with expansibility which can be used not only the case colored state noises model is AR but also ARMA or MA.4. When the observation noises are colored, how to identify the outliers is studied. On the ground of the routine methods which identifying outliers, a new method is presented by random model series-express of colored noises, the variance of dummy white observation noises is calculated, then use the variance modify innovation series orthogonal property and present outliers identification function. Simulation calculations demonstrate that the method can identify outliers effectively, reduce error rate, and improve the precise of results.5. In the process of information fusion, weighted measurement fusion is discussed. when the colored observation noises model are AR, observation-expand filter could eliminate the influence of colored observation noises and the optimal weighted measurement fusion formula is obtained. On the base of analyzing observation-expand filter, a new approach is presented in the article by random model series-express of colored noises, the variance of sensors dummy white observation noises is calculated and the weighted measurement fusion filter is constructed. The simulation results show that the method is provided with expansibility which can deal the observation noises model is ARMA.6. The application of the new approach of colored noises in the real-time GPS dynamic data in this artical, utilize pseudo-range of C/A code and use Kalman filter to calculate the results of dynamic point position, the process included the set up of system state equation, the linearization of observation equation,the select of original state value.