Trend Analysis And Prediction Algorithm Of Satellite In Orbit Mutation Status

Because the satellite is a great strategic significance for the country, and the production cost of satellite is very expensive. The health of the satellite becomes an important topic of the experts’ research recently. Also, because the satellite is disturbed by various factors, the prediction of the satellite mutation status becomes especially significant.Accordingly, this also makes the prediction is even more difficult.In this thesis, based on satellite mutation data, I give two directions to study the prediction algorithm. The first is based on the trend prediction method. In this method,I use wavelet algorithm to distinguish the salellite’s data to the high frequency and low frequency components. Then I use the ARMA model prediction model and the SVR prediction model to predict each component. At last I get The effective prediction results.The second is based on the similarity prediction method. In this method, I propose fuzzy clustering method to calculate similar sequences. Then I can predict the satellite mutation points by similarity matching.First, in the trend prediction algorithm, according to the characteristics of the satellite’s data, this thesis uses the wavelet algorithm, SVR prediction model, ARMA Prediction Model. Because the satellite signal has a lot of high-frequency jitter, this thesis uses wavelet algorithm to distinguish the salellite’s data to the high frequency and low frequency components. And then I use SVR prediction model to predict the low-frequency component and ARMA forecasting model to predict the high-frequency component. Finally, the two prediction component reconstruct the final prediction by using wavelet reconstruction.Second, in the similarity prediction algorithm, I propose using fuzzy clustering methodthis into the prediction of the similarity matching. Fuzzy clustering algorithm is used to classify the similar historical data area. Then put the predicteing data and the history database doing similarity matching. According to the parameters and features,to predict the final result. The results of this prediction method is much better than the rezult of commonly used prediction method.Finally, if the satellite mutation happens in a short time, this thesis proposes a comprehensive prediction algorithm to solve the problem of insufficient data samples.The prediction algorithm of satellite mutation state in this thesis, can predict the accurate time when satellite mutation happenes, which could solve the problems in the satellite health management work properly. At the same time, in terms of similarity prediction, the fuzzy clustering this thesis proposes, provides a new way to predict complex signal.