Frequency Selecting Technology For HF Wireless Communication System Based On Machine Learning And Chaotic Theory

Due to the unique advantages of long-distance,non-relay,low-cost and flexible-deployment,HF communication plays an extremely important role in the fields of military communication,disaster relief,global broadcasting services,etc.At present,The worldwide research interests focus on developing the next generation HF communications,where one of the key features have been recognized as intelligent in the future.In order to meet the long-term spectrum planning and short-term frequency optimization for smart HF communication,the technology architecture for selecting usable frequency is optimized.And the statistical machine learning and chaotic adaptive predicting method were introduced to develop the long-term prediction and the short-term forecast model of the usable frequency for HF communication.Based on the above-mentioned model,the regional refined long-term predictions and real-time short-term forecast results of usable frequency can be obtained.The main contributions of this thesis are as follows:1.As the basis of the long-term prediction of the usable frequency for HF communication,a model based on statistical machine learning method is proposed to improve the accuracy of predicting the monthly median ionospheric critical frequency of the F2 layer（identified as f_oF₂）,which is one of the key parameters for predicting usable frequencies for HF communication.The annual dynamic variation map of the proposed model is achieved by the two solar activity parameters of the 10.7-cm solar radio flux and sunspot number.And the geomagnetic dip latitude and its modified value are first together chosen as features of the geographical spatial variation for reconstructing spatial dynamic variation map.The proposed model can provide higher prediction accuracy for f_oF₂ over Asia.Compared with the international reference ionosphere（IRI）model with CCIR and URSI coefficients（identified as IRI-CCIR and IRI-URSI）,the root-mean-square error of the proposed model is reduced by 0.27MHz and 0.23MHz respectively,and the accuracy is improved by 2.90%and 1.85%respectively.2.As to the long-term prediction method of the usable frequency for HF communication,an enhanced model is proposed to provide fine granularity and higher prediction accuracy for the maximum usable frequency（MUF）,the optimum working frequency（OWF）and the highest probable frequency（HPF）over Asia.First of all,the refined mapping model of MUF propagation factor at a distance of 3000 km of the F2layer（identified as M（3000）F2）is reconstructed by using statistical machine learning method.And then the new mapping models of conversion factors of OWF-MUF and HPF-MUF are proposed by using the fine-grained solar activity parameters and coupling with two geomagnetic activity parameters.Compared with ITU recommended model,the root-mean-square errors of MUF,OWF and HPF are reduced by 1.18MHz1.64MHz and 1.06MHz respectively,and the accuracies are improved by 10.89%,15.47%and 9.10%respectively.3.To achieve short-term forecasting the usable frequency for HF communication,a chaos-based adaptive forecasting model of f_oF₂ is first proposed as an important basis.The proposed model is based on the Volterra series adaptive filtering method,which is introduced in the ionospheric field for the first time.And it can one-hour-ahead forecast f_oF₂ with high accuracy by using a small training dataset of 27 days（one solar rotation period）.Compared with IRI-CCIR and IRI-URSI model,the root-mean-square error of the proposed model is reduced by 1.66MHz and 1.59MHz respectively,and the accuracy is improved by 31.38%and 29.97%respectively.4.As to short-term forecast of the usable frequency for HF communication,a chaos-based adaptive forecasting model of M（3000）F2 is first proposed,and a spatial interpolation method based on geomagnetic coordinates is first proposed to interpolate these characteristic parameters for HF communication.In the end,a short-term dynamic forecast model of MUF is proposed by referring to the above research achievement.Compared with ITU long-term prediction model,the root-mean-square error of the proposed model is reduced by 1.87MHz,which is corresponding to the improved accuracy of 11.30%.