Based On The Time Series, Neural Network, Grey And Combination Forecasting Of Ionospheric Tec Forecast Research

Ionospheric TEC is an important characteristic parameters of the ionosphere, the temporal and spatial variations of the wireless communication, radar, navigation system, especially for the satellite communication system has a great influence on. At present, the existing prediction methods in various areas play an important role in the ionospheric prediction technology, many scholars have done a lot of work. In this paper, according to the2010Delingha ionospheric data, the use of appropriate data processing method, China Delingha region ionosphere technology forecast method research.First the ionospheric TEC forecast data processing and interpolation method were studied. On the ionospheric TEC raw data, using diagonal technical data, according to the GPS receiver site density using different data processing methods, the station vertical time sequence of TEC series. On the ionospheric TEC data, receiver or other reasons caused by the lack of value, use the autocorrelation method for interpolation of time, achieve higher accuracy of interpolation, interpolation error and missing value length, error to meet the need.Study on ionospheric TEC time series prediction is studied. There are a lot of time sequence model of non-stationary time series, first difference to eliminate the trend, meet the stable condition after modeling, namely ARIMA model, this paper finally selected as ARIMA model on TEC value prediction. Through the SAS software programming, using the inverse correlation coefficient, regression coefficient, AIC, SBC index parameters to determine the model parameters, finally get the model with11months prior to the data of December TEC value prediction. The experimental results can meet the basic requirements in advanceStudy on ionospheric TEC neural network prediction is studied. Artificial neural network is of biological neural system imitation. With the development of the neural network model has been developed, there are numerous model, in which BP neural network model is the most commonly used models, reflects the essence of neural network. This paper adopts BP neural network model of the Delingha December TEC values were predicted. And ARIMA model prediction results are comparedStudy on ionospheric TEC grey model prediction is studied. Gray GM (1,1) model in many data quantity is less predictive of good performance, through a variety of aspects of the study, find out by using GM (1,1) model of the vast amounts of data in the TEC value prediction methods, and the above two methods were compared, three kinds of prediction methods each has advantages and disadvantages, but the results meet the basic requirements.The above three kinds of prediction methods have their own advantages, but also have the defects. It is better to be different models are combined together, learn widely from others’strong points, using each method of useful information, so as to achieve better prediction results, improve the precision of the prediction, the prediction method is combination forecasting. This paper uses the linear combination forecasting and nonlinear combination forecasting, the above three methods were optimized, and the results are compared. The results show that combined forecasting method the result is better than the above three methods improved.Eventually each independent method forecasting result has achieved the desired results, the error is less than3in the proportion of more than65%, less than Sin85%above. After combination forecast method of optimization results are much better than the independent prediction method, especially the nonlinear combination forecasting method, the optimization results of error is less than3in70%, less than5in97%above.