Application Of Kalman Filter Based On Evolutionary Algorithm In Time Series Analysis

Kalman filter is a common used method in time-series analysis. It is a computer-based real-time algorithm. The object is random signals. We use system noise and observation noise of statistical characteristics of measurements as the filter input, the value of estimate as filter output. The filter between the input and output are time update and observation update algorithm linked to estimate the needs of all the signals according to the system equation and observation equation. Kalman filter algorithm has the advantage of using state-space description of the system which is recursive form. With small amount of data storage, not only it can handle stationary stochastic processes, but also can handle multi-dimensional and non-stationary random process. However, the traditional Kalman filter is built on the model of precision and the random signal of interference of statistical characteristics is known. For a practical system, the model is uncertain or the statistical characteristics of interference signal are not entirely known. These uncertainties made traditional Kalman filter lost optimality.Evolutionary Computation is a parallel problem solver which uses ideas and gets inspirations from natural evolutionary process. It uses simple coding techniques to express complex structure, through a group code for simple operation and the survival of the fittest genetic natural choice to guide the study and determine the Search direction. Evolutionary algorithm with the traditional algorithm is very different, but the main difference lies in evolutionary computation is intelligent and parallel. Due to its intrinsic parallelism and some intelligent properties such as self-organizing, adaptation and self-learning, evolutionary computation has been applied successfully to problems where heuristic solutions are not available or generally lead to unsatisfactory result. In recent years, the interest in evolutionary computation is growing dramatically. It has been considered that simulated evolution may be the most promising way to develop machine intelligence.Currently, scholars focus on adaptive Kalman filtering techniques with intelligent information integration. It is a new development direction of the integrated navigation algorithm. It designs Kalman filter with fuzzy logic, neural network technology and information fusion technology. In this paper, according to different reasons of Kalman filter error, we use intelligent technology to take control of the calculation of the corresponding, without loss of the accuracy of the original filter estimation. Improve the system to various disturbances adaptability, inhibit Kalman filter divergence. Conventional Kalman filter needs to know process noise matrix Q and measurement noise matrix R accurately which is difficult to engineering applications. In order to improve the performance of filter, we put forward adaptive Kalman filter based on evolutionary computation. According to the residual information of the filter, using evolutionary algorithm to adjusted Q and R values, improving the accuracy of the filter.Simulation examples show that the Kalman filter based on evolutionary computation method is effective and feasible. Under the same conditions, it better than the traditional Kalman filter. It restrains the Kalman filter divergence effectively. It has practical significance and broad prospects in engineering applications.