Modeling And Prediction Of Time Series For S-band Spectrum Use In Tiantong-1 Satellite Downlink

Satellite communication gradually plays an increasingly important role in various fields of people's lives,with its advantages of large coverage,long communication distance,and large communication capacity.Satellite spectrum resources are beginning to show a shortage,in consequence of the stimulus of satellite communication demand and the gradual increase in the types of satellite communication service applications,however,it has been found through actual measurement that the fundamental reason for the shortage of spectrum resources is the low spectrum utilization rate.Based on the research experience of the terrestrial mobile communication networks,cognitive radio technology can be used to solve the problem of spectrum utilization,but the premise of the implementation of this technology is to find spectrum gaps.In order to discover and use these spectrum gaps,not only the advanced spectrum monitoring equipment is needed but also the theoretical knowledge of the spectrum occupancy model of the target frequency band is needed.Therefore,a comprehensive and detailed understanding of the spectrum occupancy model in the satellite communication system can provide a basis for solving the contradiction between the shortage of spectrum resources and the low spectrum utilization rate in the satellite communication network.It can be said that the development of cognitive radio has largely benefited from the availability of practical and accurate spectrum occupancy models.In this paper,it gives a research work of modeling and prediction of time series for S-band spectrum use in satellite downlink,and all these research are based on the actual measured data of the spectrum used by Tiantong-1 satellite transponder.In the aspect of spectrum occupation model,the spectrum occupancy model proposed in many literatures so far can describe and reproduce the statistical characteristics of the occupied time series.For example,the busy/idle-period lengths of traditional terrestrial mobile communication networks can be fitted by generalized Pareto distribution,exponential distribution,etc.However,in some complex scenarios such as satellite link spectrum occupancy,the traditional parameter estimation distribution cannot give the goodness of fit.For the reason that,this paper proposes to use the kernel density estimation method to estimate the probability density function of the spectrum occupancy time series.The conclusion shows that,compared with the generalized Pareto distribution,exponential distribution used in the traditional ground network,kernel density estimation can study the distribution characteristics by starting from the data sample itself,so that it can more accurately describe the statistical characteristics of the occupied time series of the S-band used in the satellite downlink.In the aspect of spectrum occupancy prediction,aiming at the defect that classic ARIMA model prediction cannot capture the nonlinear relationship in the data,this paper proposes to use fuzzy neural network to predict the time series of spectrum occupancy model.The conclusion shows that the fuzzy neural network prediction is more accurate than ARIMA model prediction,because it has the ability of learning and self-adaptive,and can capture the non-linear relationship in the time series data.