The Research Of Optimal Estimation Method Based On Time Series Prediction

Modeling and prediction of time series is one of the most active research topics in academic research and practical engineering application, which is widely used in automatic control, aviation, aerospace, marine, communications, industrial production and so on. In this paper, modeling and forecasting of time series and optimal estimation algorithm are discussed, the optimal estimation method and its improved algorithms are applied to time series forecasting. Time series model parameters are optimized by optimal estimation methods so that the accuracy of time series prediction is improved.In this paper, with linear time series models in mind, time series model parameters updating algorithm through the updated Kalman filter is studied based on the discussing of recursive parameter estimation method of Kalman filter, including the extended Kalman filter (EKF) time series model parameter updating method, the unscented Kalman filter (UKF) time series model parameter updating method, and simulation shows that accuracy of time series prediction is improved.In this paper, the parameter of linear model of time series is estimated through particle filter technology. First, the parameters of the preliminary time series model will be used as particles under a certain degree of disturbance, as a result, particle set is formed, observations undertake a state transition according to the parameters of the model, particle weight is obtained based on the deviation between the true value and observations, the estimated state transition model parameters are obtained. And the real time ARMA model parameters updating algorithm through the improved particle filter are discussed, including the time series model parameter updating method of the extended Kalman particle filter (EPF), the time series model parameter updating method of the unscented Kalman particle filter (UPF), the time series model parameter updating method of the kernel particle filter (KPF). The feasibility of optimal estimation algorithm optimizing time series forecasting model parameters is analyzed by experiment.