Research On Tropospheric Delay Correction Model Without Meteorological Parameters

Tropospheric delay is one of the main factors that affect the accuracy of GNSS positioning.GNSS satellite signal through the atmosphere,by the impact of neutral atmospheric refraction will produce the phenomenon of delay and path bending,resulting in GNSS signal propagation delay.At present,the method of correcting the tropospheric delay is the model correction method,and the model correction method is divided into meteorological model and non-meteorological parameter model.The measured meteorological model needs to obtain meteorological parameters measured at the station,which will be limited in the practical engineering application.The accuracy of the tropospheric delay correction is lower than that of the measured meteorological model,but the meteorological parameter model is more convenient.In this essay,a new method for improving the non-meteorological parameter model based on BP neural network is proposed.The main contents and conclusions of this essay are as follows:(1)Based on the high-precision tropospheric delay data provided by the IGS station,the temporal and spatial variations of the tropospheric delay in China are analyzed in detail.Tropospheric delay increases with latitude decreases,the eastern coastal areas of tropospheric delay is higher than the western inland regions.(2)Based on the EGNOS model and the cosine function model,BP neural network is used to compensate the error of EGNOS model,and a new IEGNOS fusion model is proposed.In 5 IGS stations of China,the IEGNOS model is better than the EGNOS model and the cosine function model in both the mean absolute value and the average root-mean-square error.The average root-mean-square error of the EGNOS model on the five IGS stations is ± 5.5cm,while the average root-mean-square error of the IEGNOS model is ± 2.9cm.Compared with the traditional EGNOS model,the accuracy of IEGNOS model is improved by 47%.(3)Research on process and error of PWV inversion by zenith wet delay.The water vapor conversion coefficient calculated by the sounding data is regarded as the true value,and the least square method is used to establish the Emardson water vapor conversion coefficient calculation model,i.e.IEmardson model.The average root-mean-square error of the Emardson model in the four stations is ± 0.00495,while the average root,mean-square error of the IEmardson model is ± 0.00112,compared with the Emardson model,its accuracy is improved by about 77%.(4)The IEGNOS model was used to invert the PWV in the Kunming area and compared with the PWV obtained by sounding data.The average deviation error of the IEGNOS model is 1.38mm and the average root-mean-square error is ± 3.58mm,which is basically consistent with the trend of PWV obtained by sounding data.In the absence of measured meteorological parameters,IEGNOS model inversion PWV has a high degree of credibility.