GNSS Tropospheric Delay Interpolation Based On RBF Neural Network Model

With the development of GNSS in different fields,users have higher requirements on the accuracy of the system.How to quickly locate and improve accuracy is the focus and development trend of current GNSS research.The tropospheric delay is highly susceptible to atmospheric influences and is closely related to the angle of satellite signal's path.When the altitude angle is 90 degrees,the error reaches 2-3m,but when the altitude angle is 5 degrees,the maximum error is 25 m.Since the correction of tropospheric errors is very complicated,this has become an important factor affecting the accuracy of GNSS.Improving the accuracy of tropospheric error correction has great significance in the application of positioning and measurement.At present,RBF neural network has been applied in many fields.It can approximate any function with arbitrary precision,and has high learning speed.In order to improve the accuracy of tropospheric delay interpolation,a GNSS tropospheric delay interpolation model based on RBF neural network is established.The model only needs to input the latitude and longitude and elevation of the survey site,which brings convenience in practical applications.In the process of constructing the RBF neural network model,firstly,the tropospheric delay is solved by using the GAMIT software according to the coordinate data of 10 CORS base stations in the Anhui power system.The coordinates and tropospheric delays of 6 CORS base stations in the Anhui power system are used as modeling data.Four CORS stations tropospheric delays are used as test data.The appropriate neural network learning algorithm is selected to determine the network structure parameters,and the model is built.Secondly,the output value of the model tropospheric delay is compared with the tropospheric delay calculated by the GAMIT software,analyzing the error to verify the reliability of the model.Experimental results show that the tropospheric interpolation accuracy of the test data can be up to mm,and the maximum error is about 2mm.The prediction of tropospheric delay with RBF neural network model has high accuracy.