Enhanced Beidou precise time service method based on receiver clock model

Abstract

The invention relates to the technical field of radio orientation, in particular to an enhanced Beidou precise time service method based on a receiver clock model, which comprises the following steps: acquiring multi-source data, and preprocessing the multi-source data; establishing a Beidou receiver clock model, optimizing model parameters through a least square method, outputting a first clock drift value and a first coefficient, and predicting a second clock drift value and a second coefficient by using a machine learning model; then, a Bayesian fusion model is utilized to be combined with the Beidou receiver clock model and the machine learning model, a clock drift predicted value is output, and a time migration predicted value is calculated; calculating a carrier-to-noise ratio loss value, an autocorrelation function distortion value and a phase change value, generating a signal quality index, and calculating an error compensation value by using a weight distribution mechanism; and finally, correcting the clock skew according to the error compensation value. Through combination of multiple models, the precision of clock drift prediction is improved, and the clock skew is effectively corrected, so that the precision of mountainous area time service is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of radio direction finding, and specifically to an enhanced Beidou precise timing method based on a receiver clock model. Background Art

[0002] As a global satellite navigation system independently developed in China, the Beidou navigation system is widely used in communication, timing, and positioning. Especially in scenarios with limited communication conditions such as remote mountainous areas and complex terrains, the Beidou satellite system has important strategic value. However, the terrain in mountainous areas is complex, often with mountain blockages and multipath interferences, which affect the signal reliability of the Beidou satellite system, reduce the timing accuracy or even cause interruptions. Moreover, the electromagnetic environment in mountainous areas is complex, with weak signal coverage for a long time, which will increase the positioning and timing errors, bringing challenges to key applications such as emergency communication and disaster monitoring.

[0003] Existing Beidou satellite timing technologies can provide high-precision timing services, with advantages such as low cost, wide coverage, and good continuity. Especially in open areas and conventional scenarios, they show good stability and reliability. However, their adaptability in mountainous area scenarios is still insufficient. First, the modeling of the local clock drift of the receiver clock by these technologies is relatively simple, and may not be able to cope with the dynamic changes of clock errors in the complex mountain environment, resulting in the difficulty of maintaining stable timing accuracy. Second, the observation errors caused by multipath interferences and blocked signals are relatively large, and these error correction models are mainly based on fixed parameters, lacking the ability of real-time dynamic adjustment and unable to effectively respond to the changes in the mountain environment. Third, the mountain communication scenario has high requirements for the real-time performance of timing devices, but most of these systems rely on offline processing and post-processing of data, and it is difficult to meet the requirements of precise Beidou satellite timing in mountain real-time communication. Therefore, the accuracy of Beidou satellite timing needs to be further improved.

[0004] Therefore, an enhanced Beidou precise timing method based on a receiver clock model is proposed. Summary of the Invention

[0005] The object of the present invention is to provide an enhanced Beidou precise timing method based on a receiver clock model. By acquiring multi-source data and preprocessing it; establishing a Beidou receiver clock model, optimizing the model parameters by the least squares method, outputting a first clock drift value and a first coefficient, and using a machine learning model to predict a second clock drift value and a second coefficient; then, using a Bayesian fusion model to combine the Beidou receiver clock model and the machine learning model, outputting a clock drift prediction value, and calculating a time offset prediction value; calculating a carrier-to-noise ratio loss value, an autocorrelation function distortion value, and a phase change value, generating a signal quality index, and using a weight allocation mechanism to calculate an error compensation value; finally, correcting the clock deviation according to the error compensation value. Through the combination of multiple models, the accuracy of clock drift prediction is improved, and the clock deviation is effectively corrected, thereby improving the accuracy of Beidou satellite precise timing in mountainous areas.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An enhanced Beidou precise timing method based on a receiver clock model, comprising:

[0008] Step S10: Acquire multi-source data, where the multi-source data includes environmental data, GIS data, meteorological data, signal data, and clock history data; preprocess the multi-source data to obtain processed data;

[0009] Step S20: Establish a Beidou receiver clock model, where the Beidou receiver clock model is used to calculate the clock drift caused by the processed data; optimize the parameters of the Beidou receiver clock model according to the processed data by the least squares method, output a first clock drift value and a first coefficient; use a machine learning model for the processed data to output a second clock drift value and a second coefficient;

[0010] Step S30: According to the first clock drift value, the first coefficient, the second clock drift value, and the second coefficient, use a Bayesian fusion model to output a clock drift prediction value; calculate a time offset prediction value according to the clock drift prediction value;

[0011] Step S40: Calculate a carrier-to-noise ratio loss value, an autocorrelation function distortion value, and a phase change value for the processed data to generate a signal quality index; calculate an error compensation value according to the signal quality index and the time offset prediction value by using a weight allocation mechanism;

[0012] Step S50: Correct the clock deviation according to the error compensation value.

[0013] Further, the multi-source data includes:

[0014] The environmental data includes altitude and receiver data;

[0015] The GIS data includes topographic features, geological structures, and vegetation cover;

[0016] The meteorological data includes rainfall, wind speed, temperature, humidity, and air pressure;

[0017] The signal data includes signal power, noise power, direct-path signals, and multipath signals;

[0018] The clock history data includes the historical drift information of the clock.

[0019] Furthermore, the establishment process of the Beidou receiver clock model includes:

[0020] Establish a temperature influence model for the processed data and output the temperature frequency drift; the temperature influence model is a linear model of the crystal oscillator frequency varying with temperature;

[0021] Establish a vibration influence model for the processed data and output the vibration frequency drift; the vibration influence model is a model of the crystal oscillator frequency varying with the vibration intensity;

[0022] Establish a power supply fluctuation influence model for the processed data and output the power supply frequency drift; the power supply fluctuation influence model is a model of the crystal oscillator frequency varying with the power supply fluctuation;

[0023] Sum up the temperature frequency drift, the vibration frequency drift, the power supply frequency drift, and the random error to obtain the Beidou receiver clock model.

[0024] Furthermore, the implementation process of the least squares method includes:

[0025] Obtain the processed data and construct a design matrix according to the temperature sequence, vibration intensity sequence, and power supply fluctuation amplitude sequence, expressed as:

[0026]

[0027] where X is the design matrix, [T1, T2…T n is the temperature sequence, T0 is the reference temperature, [V1, V2…V n is the vibration intensity sequence, [ΔPS1, ΔPS2…ΔPS n is the power supply fluctuation amplitude sequence, and 1, 2, and n are serial numbers;

[0028] Based on the design matrix, minimize the sum of squared errors according to the Beidou receiver clock model, expressed as:

[0029] S=(y - X×k) t ·(y - X×k);

[0030] Among them, S is the least squares objective value, y is the actual clock drift vector, X is the design matrix, k is the parameter vector, and t is the transpose symbol;

[0031] Derive S with respect to k and set the derivative to zero, and output the parameter vector k, which is expressed as:

[0032] k = (X T × X) -1 × X T × y;

[0033] Among them, X T × X is the covariance matrix of the design matrix, and X T × y is the inner product of the design matrix and the actual clock drift vector.

[0034] Furthermore, the implementation process of the Bayesian fusion model includes:

[0035] Define the prior distribution, which is a normal distribution generated according to the first clock drift value and the first coefficient;

[0036] Define the likelihood function, which is the normal distribution generated according to the second clock drift value and the second coefficient;

[0037] According to Bayes' theorem, combine the prior distribution with the likelihood function to generate the posterior distribution;

[0038] Use the mean value of the posterior distribution as the final clock drift prediction value.

[0039] Furthermore, the calculation process of the signal quality index includes:

[0040] Divide the signal power by the noise power to obtain the received signal carrier-to-noise ratio; obtain the ideal carrier-to-noise ratio to obtain the carrier-to-noise ratio loss value;

[0041] Calculate the autocorrelation function of the received signal and the autocorrelation function of the ideal signal to obtain the autocorrelation function distortion value;

[0042] Use the Hilbert transform to extract the instantaneous phase of the received signal to obtain the phase change value;

[0043] Add the carrier-to-noise ratio loss value, the autocorrelation function distortion value, and the phase change value to obtain the signal quality index.

[0044] Furthermore, the weight allocation mechanism includes:

[0045] Map the signal quality index to a signal error, and perform weighted summation with the time offset prediction value to obtain the error compensation value;

[0046] If the change amplitude of the signal quality index is greater than a preset threshold, the gradient descent algorithm is used to adjust the signal quality weight and the time offset prediction value weight; otherwise, no adjustment is made.

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0048] 1. The present invention adopts a Beidou receiver clock model, models the linear effects of environmental and equipment factors such as temperature, vibration, and power supply fluctuations on the crystal oscillator frequency drift in the complex mountain environment, and combines the least squares method to optimize parameters, accurately capturing the linear clock drift in the mountains. At the same time, a machine learning model is introduced to non-linearly predict the complex clock drift relationship using multi-source data, comprehensively modeling the complex non-linear factors in the mountain environment, effectively predicting the non-linear clock drift caused by environmental disturbances, and thus enhancing the stability and accuracy of Beidou precise timing.

[0049] 2. Based on Bayesian theory, the present invention integrates the output of the Beidou receiver clock model and the output of the machine learning model, comprehensively utilizes the advantages of model-driven and data-driven, and reduces the bias that may be introduced by single-model prediction through probability inference in the complex signal environment in the mountains. Then, the prior distribution and likelihood function are defined, the information from the two sources is integrated into the posterior distribution, and the mean value of the posterior distribution is used as the final clock drift prediction value. Especially in the case of high uncertainty or data loss, the accuracy of clock drift prediction is improved, and thus the stability and accuracy of Beidou precise timing are enhanced.

[0050] 3. By defining the signal quality index, the present invention comprehensively considers factors such as the carrier-to-noise ratio loss value, the autocorrelation function distortion value, and the phase change value, comprehensively reflecting the signal transmission quality in the mountains. Then, a weight allocation mechanism is introduced, the signal quality index and the time offset prediction value are weighted and summed to calculate the error compensation value, and the weight is dynamically adjusted according to the change amplitude of the signal quality, effectively coping with the signal quality fluctuations in mountain communication, reasonably allocating the influence weights of signal quality and time offset, reducing the dependence on low-quality signals, and thus improving the timing accuracy and adaptive ability in the complex signal environment. Brief Description of the Drawings

[0051] Figure 1 It is a schematic flowchart of a method for enhancing Beidou precise timing based on a Beidou receiver clock model provided by the present invention;

[0052] Figure 2 It is a schematic flowchart of constructing a Beidou receiver clock model provided by the present invention. Detailed Embodiment

[0053] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0054] Please refer to Figures 1 to 2 , the present invention provides an enhanced Beidou precise timing method based on a receiver clock model, and the technical solution is as follows:

[0055] The first embodiment is as follows:

[0056] An enhanced Beidou precise timing method based on a receiver clock model includes:

[0057] Step S10: Obtain multi-source data, where the multi-source data includes environmental data, GIS data, meteorological data, signal data, and clock history data; preprocess the multi-source data to obtain processed data.

[0058] Further, the multi-source data includes:

[0059] The environmental data includes altitude and receiver data;

[0060] The GIS data includes terrain features, geological structures, and vegetation coverage;

[0061] The meteorological data includes rainfall, wind speed, temperature, humidity, and air pressure;

[0062] The signal data includes signal power, noise power, direct path signals, and multipath signals;

[0063] The clock history data includes historical drift information of the clock.

[0064] Specifically, the mountain altitude in the environmental data affects the signal propagation path and attenuation, and the state parameters such as the operating temperature and voltage of the receiver affect the clock stability. The GIS data affects the signal propagation path and multipath effect, and understanding the terrain helps to optimize signal processing. The meteorological data affects the performance of the receiver and the clock stability, and the signal data directly affects the signal quality and the accuracy of clock synchronization, and needs to be monitored in real time to optimize error compensation. The clock history data is used to identify the long-term trend and periodic changes of clock drift, which is convenient for subsequent analysis and modeling. The acquisition of these multi-source data makes the entire clock synchronization and error compensation process have higher accuracy, is applicable to complex and changeable mountain communication environments, and thus meets the requirements for Beidou high-precision clock synchronization and stable communication.

[0065] Step S20: Establish a Beidou receiver clock model, which is used to calculate the clock drift caused by the processed data; optimize the parameters of the Beidou receiver clock model using the least squares method according to the processed data, and output the first clock drift value and the first coefficient; use a machine learning model for the processed data to output the second clock drift value and the second coefficient.

[0066] Further, the establishment process of the Beidou receiver clock model includes:

[0067] Establish a temperature influence model for the processed data, and output the temperature frequency drift. The temperature influence model is a linear model of the crystal oscillator frequency varying with temperature, expressed as:

[0068] Δf T =α T ×(T - T0);

[0069] where, Δf T is the temperature frequency drift, α T is the temperature coefficient, T is the current temperature, and T0 is the reference temperature.

[0070] Among them, if the temperature range in the mountainous area is 5 to 25 degrees Celsius, the reference temperature can be selected as the median of this range, which is 15 degrees Celsius, or the reference temperature recommended by the crystal oscillator manufacturer can also be used.

[0071] Establish a vibration influence model for the processed data, and output the vibration frequency drift. The vibration influence model is a model of the crystal oscillator frequency varying with the vibration intensity, expressed as:

[0072] Δf V =α V ×V;

[0073] where, Δf V is the vibration frequency drift, α V is the vibration coefficient, and V is the vibration intensity.

[0074] Establish a power supply fluctuation influence model for the processed data, and output the power supply frequency drift. The power supply fluctuation influence model is a model of the crystal oscillator frequency varying with the power supply fluctuation, expressed as:

[0075] Δf PS =α PS ×ΔPS;

[0076] where, Δf PS is the power supply frequency drift, α PS is the power supply fluctuation coefficient, and ΔPS is the power supply fluctuation amplitude.

[0077] Sum up the temperature frequency drift, the vibration amplitude frequency drift, the power supply frequency drift and the random error to obtain the Beidou receiver clock model, which is expressed as:

[0078] Δf = Δf T + Δf V + Δf PS + ∈;

[0079] Among them, Δf is the clock drift, Δf T is the temperature frequency drift, Δf V is the vibration amplitude frequency drift, Δf PS is the power supply frequency drift, and ∈ is the random error caused by other unconsidered factors, which can be set to a small range of fluctuations between -10 -8 and 10 -8 seconds.

[0080] Specifically, in areas with large temperature differences such as mountainous areas, the crystal oscillator frequency of the receiver is particularly significantly affected by temperature changes, and temperature drift becomes one of the key factors of clock drift. By establishing a temperature influence model, the influence of temperature on frequency can be quantified, so as to accurately calculate the clock drift caused by environmental temperature changes. Then, the vibration intensity refers to the acceleration of vibration in the environment where the crystal oscillator is located. Vibration will affect the physical performance of the crystal oscillator. By introducing the influence model of vibration intensity, the frequency drift caused by vibration can be effectively estimated, so as to improve the clock accuracy. In addition, power supply fluctuations will cause unstable power supply of the crystal oscillator, which in turn affects the frequency output. By establishing a power supply fluctuation influence model, the frequency drift characteristics of the crystal oscillator can be captured more comprehensively. In addition, there are also some uncontrollable or unmodeled random errors (such as environmental noise and system noise, etc.). By introducing a random error term, the Beidou receiver clock model can be ensured to have better adaptability to the real-time situation in mountainous areas.

[0081] By quantitatively modeling temperature, vibration, power supply fluctuations and random errors, and constructing a comprehensive Beidou receiver clock model, not only the accuracy of calculation is improved theoretically, but also a reliable data basis is laid for the accurate prediction and effective compensation of clock drift.

[0082] Furthermore, the implementation process of the least squares method includes:

[0083] Obtain sufficient processing data, including temperature, reference temperature, power supply fluctuations, vibration intensity and actual clock drift values, organize them into a format suitable for regression analysis, and construct a design matrix, which is expressed as:

[0084]

[0085] Among them, X is the design matrix, [T1, T2…T n is the temperature sequence, T0 is the reference temperature, [V1, V2…Vn is the vibration intensity sequence, [ΔPS1, ΔPS2…ΔPS n is the power supply fluctuation amplitude sequence, 1, 2, and n are serial numbers;

[0086] Minimize the sum of squared errors according to the Beidou receiver clock model, expressed as:

[0087] S = (y - X × k) t ·(y - X × k);

[0088] where S is the least squares objective value, y is the actual clock drift vector, X is the design matrix, k is the parameter vector and k = [α T , α V …α PS , α T is the temperature coefficient, α V is the vibration coefficient, α PS is the power supply fluctuation coefficient, t is the transpose symbol;

[0089] Take the derivative of S with respect to k and set the derivative to zero, output the parameter vector k, expressed as:

[0090] k = (X T × X) -1 × X T × y;

[0091] where X T × X is the covariance matrix of the design matrix, X T × y is the inner product of the design matrix and the actual clock drift vector.

[0092] The least squares method constructs a design matrix that comprehensively reflects environmental factors by obtaining multi-dimensional data such as temperature, reference temperature, power supply fluctuation, and vibration intensity, and can accurately capture the influence of these factors on clock drift, providing a strong basis for precise time service in mountainous areas. In addition, the least squares method can adjust the model parameters under the changing mountain environment, enabling the Beidou receiver clock model to adapt to environmental changes, thus ensuring the accuracy of the receiver clock.

[0093] Next, subtract the clock drift output by the optimized Beidou receiver clock model from the actual clock drift to obtain the residual. Then, calculate the degrees of freedom according to the number of parameters and the total number of data points, and calculate the residual standard deviation according to the degrees of freedom and the residual to obtain the first coefficient, which is the overall uncertainty of the Beidou receiver clock model, representing the typical deviation between the model prediction value and the observed value, expressed as:

[0094]

[0095] where σ1 is the first coefficient, e iis the i-th residual, n is the total number of data points, m is the number of parameters, and n - m is the degree of freedom.

[0096] Further, the machine learning model can choose to use a random forest model to predict the clock drift in the mountain area. Calculate the mean and standard deviation of the prediction results of all decision trees in the random forest. The mean is used as the second clock drift value of the model, and the standard deviation is used as the uncertainty quantification index of the prediction result, reflecting the fluctuation range of the predicted value. Then, using the normal distribution hypothesis, calculate the 95% prediction interval, so as to effectively quantify the uncertainty of the prediction result and provide more comprehensive information. Finally, take the average of all standard deviations to obtain the second coefficient.

[0097] Step S30: According to the first clock drift value, the first coefficient, the second clock drift value, and the second coefficient, use the Bayesian fusion model to output the clock drift prediction value; calculate the time offset prediction value according to the clock drift prediction value.

[0098] Further, the implementation process of the Bayesian fusion model includes:

[0099] Define the prior distribution, and the prior distribution is a normal distribution generated according to the first clock drift value and the first coefficient, expressed as:

[0100]

[0101] where θ is the target variable of Bayesian inference, that is, the final clock drift prediction value, P(θ) is the prior distribution, y1 is the first clock drift value, σ1 is the first coefficient, is the normal distribution of y1 and

[0102] Define the likelihood function, and the likelihood function is the normal distribution generated according to the second clock drift value and the second coefficient, expressed as:

[0103]

[0104] where, is the clock offset value obtained from experimental measurement, and can also be equal to σ2, is the likelihood function, y2 is the second clock drift value, σ2 is the second coefficient, is the normal distribution of y2 and

[0105] According to Bayes' theorem, combine the prior distribution with the likelihood function to generate the posterior distribution;

[0106] Use the mean of the posterior distribution as the final clock drift prediction value.

[0107] Specifically, in a scenario with complex and non-linearly changing environment such as mountainous areas, the machine learning model can accurately capture the special features of mountainous areas, while the Beidou receiver clock model can calculate the overall trend. With the help of Bayesian fusion, the certainty of the first clock drift value is fused with the flexibility of the second clock drift value. When new influencing factors appear in the mountainous area environment, causing changes in clock drift, it can quickly adapt and adjust the prediction strategy, automatically optimizing the model weights. Thus, in the changing environment of mountainous areas, not only the accuracy of clock drift prediction is improved, but also the robustness of the model is enhanced.

[0108] Step S40: Calculate the carrier-to-noise ratio loss value, autocorrelation function distortion value, and phase change value for the processed data to generate a signal quality index; according to the signal quality index and the time offset prediction value, use a weight allocation mechanism to calculate an error compensation value.

[0109] Furthermore, the calculation process of the signal quality index includes:

[0110] Divide the signal power by the noise power to obtain the received signal carrier-to-noise ratio; obtain the ideal carrier-to-noise ratio, and subtract the ideal carrier-to-noise ratio from the received signal carrier-to-noise ratio to obtain the carrier-to-noise ratio loss value;

[0111] Calculate the autocorrelation function of the received signal and the autocorrelation function of the ideal signal to obtain the autocorrelation function distortion value, expressed as:

[0112]

[0113] ΔR(τ) = R y (τ) - R s (τ);

[0114] where y(t) is the signal captured by the receiver, s(t) is the direct path signal, a i is the amplitude factor of the multipath signal, τ i is the delay of the multipath signal, i is the index of the multipath signal, N is the total number of multipath signals, t is the time variable, R y (τ) is the autocorrelation function of the received signal, R s (τ) is the autocorrelation function of the ideal signal, with a peak at τ = 0, ∞ is the infinity symbol, dt is the time differential, and ΔR(τ) is the autocorrelation function distortion value.

[0115] Use the Hilbert transform to extract the instantaneous phase of the received signal to obtain the phase change value;

[0116] Add the carrier-to-noise ratio loss value, the autocorrelation function distortion value, and the phase change value to obtain the signal quality index.

[0117] Specifically, the carrier-to-noise ratio loss value is used to quantitatively evaluate the impact of environmental noise and interference on signal quality, thereby helping to optimize the noise resistance performance of the receiver. The autocorrelation function distortion value can quantitatively describe the impact of multipath interference on the time characteristics of the signal. The phase change value is used to detect the phase stability of the signal, providing a basis for subsequent phase compensation and correction. By synthesizing the carrier-to-noise ratio loss value, the autocorrelation function distortion value, and the phase change value, the overall quality of the signal is quantitatively evaluated, providing an overall signal quality evaluation criterion, facilitating the dynamic optimization of subsequent parameters, and thus ensuring the accuracy of Beidou time service in mountainous areas.

[0118] Furthermore, the weight allocation mechanism includes:

[0119] Mapping the signal quality index to a signal error, and performing weighted summation with the time offset prediction value to obtain the error compensation value;

[0120] If the change amplitude of the signal quality index is greater than a preset threshold, use the gradient descent algorithm to adjust the signal quality weight and the time offset prediction value weight; otherwise, do not make adjustments.

[0121] Specifically, the signal quality index can reflect the current signal condition, covering carrier-to-noise ratio loss, autocorrelation function distortion, and phase change, etc. These factors are directly related to the accuracy of clock synchronization. The time offset prediction value is obtained based on the Beidou receiver clock model and historical data, revealing the long-term trend and prediction of clock drift. Through weighted summation, the real-time signal quality can be fused with the time offset predicted by the model to comprehensively evaluate the required error compensation. For example, in an environment with large fluctuations in signal quality, the weight of the signal quality index can be increased; conversely, in a scenario with stable signal quality, more reliance can be placed on the time offset prediction value. Flexibly adjust the proportion of each index to adapt to the diverse performance requirements in mountainous areas, thereby improving the accuracy of Beidou time service in mountainous areas.

[0122] Step S50: Correct the clock deviation according to the error compensation value.

[0123] In summary, the present invention calculates the clock drift value and the corresponding coefficient by means of the Beidou receiver clock model and the machine learning model respectively, giving full play to the advantages of different models, thereby improving the accuracy and stability of clock drift prediction. In mountainous areas where the signal propagation path is complex and the uncertainty is high, by Bayesian fusing the prediction results of different models and comprehensively considering the uncertainty and reliability of each model, the prediction accuracy is further improved. In addition, by combining real-time signal quality and time offset prediction and dynamically adjusting the weights, it is ensured that in the mountainous area environment with large signal quality fluctuations, clock synchronization can respond and adjust quickly to adapt to different mountain communication requirements and changes, thereby improving the accuracy of Beidou time service.

[0124] Embodiment 2 is as follows:

[0125] In mountainous area A, the terrain is complex, with valleys and ridges distributed irregularly, and signals are easily blocked. At the same time, the climate changes greatly and has strong seasonality, which may cause fluctuations in the receiver. To ensure the high real-time scenario requirements such as Beidou time service in emergency rescue and time synchronization of mountain communication base stations in mountainous area A, based on Embodiment 1, an enhanced Beidou precise time service method based on the receiver clock model is given, including:

[0126] Figure 1 It is a schematic flow diagram of an enhanced Beidou precise time service method based on the receiver clock model provided by the present invention.

[0127] Refer to Figure 1 Step S10 in [reference]: Obtain multi-source data, where the multi-source data includes environmental data, GIS data, meteorological data, signal data, and clock history data; preprocess the multi-source data to obtain processed data;

[0128] Refer to Figure 1 Step S20 in [reference]: Establish a Beidou receiver clock model, where the Beidou receiver clock model is used to calculate the clock drift caused by the processed data; optimize the parameters of the Beidou receiver clock model using the least squares method according to the processed data, and output a first clock drift value and a first coefficient; use a machine learning model for the processed data to output a second clock drift value and a second coefficient;

[0129] Refer to Figure 1 Step S30 in [reference]: According to the first clock drift value, the first coefficient, the second clock drift value, and the second coefficient, use a Bayesian fusion model to output a clock drift prediction value; calculate a time offset prediction value according to the clock drift prediction value;

[0130] Refer to Figure 1 Step S40 in [reference]: Calculate the carrier-to-noise ratio loss value, the autocorrelation function distortion value, and the phase change value for the processed data to generate a signal quality index; calculate an error compensation value using a weight allocation mechanism according to the signal quality index and the time offset prediction value;

[0131] Refer to Figure 1 Step S50 in [reference]: Correct the clock deviation according to the error compensation value.

[0132] Furthermore, the multi-source data includes:

[0133] The environmental data includes altitude and receiver data;

[0134] The GIS data includes terrain features, geological structures, and vegetation coverage;

[0135] The meteorological data includes rainfall, wind speed, temperature, humidity, and air pressure;

[0136] The signal data includes signal power, noise power, direct path signal, and multipath signal;

[0137] The clock history data includes the historical drift information of the clock.

[0138] Specifically, in this embodiment, the altitude and GIS data of each receiver within the A mountainous area are obtained using an existing GIS database and updated monthly. The real-time working status data is obtained through the temperature sensor and power status monitoring inside the receiver and updated every 10 seconds. The meteorological data is monitored in real-time using a local weather station near the communication base station and updated every 10 minutes. The signal power and noise power are measured in real-time through a power meter built into the receiver and updated every 10 seconds. The clock history data is obtained based on the clock synchronization log and updated every 12 hours. After cleaning and unifying these data according to the minimum time step, the processed data is obtained.

[0139] Figure 2 This is a schematic flow diagram for constructing the Beidou receiver clock model provided by the present invention.

[0140] Further, as Figure 2 shown, the establishment process of the Beidou receiver clock model includes:

[0141] Establish a temperature influence model for the processed data and output the temperature frequency drift. The temperature influence model is a linear model of the crystal oscillator frequency varying with temperature;

[0142] Establish a vibration influence model for the processed data and output the vibration frequency drift. The vibration influence model is a model of the crystal oscillator frequency varying with the vibration intensity;

[0143] Establish a power supply fluctuation influence model for the processed data and output the power supply frequency drift. The power supply fluctuation influence model is a model of the crystal oscillator frequency varying with the power supply fluctuation;

[0144] Sum up the temperature frequency drift, the vibration frequency drift, the power supply frequency drift, and the random error to obtain the Beidou receiver clock model.

[0145] Further, the implementation process of the least squares method includes:

[0146] Obtain the processed data and construct a design matrix, expressed as:

[0147]

[0148] where X is the design matrix, [T1,T2…T nis the temperature sequence, T0 is the reference temperature, [V1, V2…V n is the vibration intensity sequence, [ΔPS1, ΔPS2…ΔPS n is the power supply fluctuation amplitude sequence, 1, 2, and n are serial numbers;

[0149] Minimize the sum of squared errors according to the Beidou receiver clock model, expressed as:

[0150] S = (y - X × k) t ·(y - X × k);

[0151] where S is the least squares objective value, y is the actual clock drift vector, X is the design matrix, k is the parameter vector k = [α T , α V …α PS , t is the transpose symbol;

[0152] Take the derivative of S with respect to k and set the derivative to zero to output the parameter vector k, expressed as:

[0153] k = (X T × X) -1 × X T × y;

[0154] where X T × X is the covariance matrix of the design matrix, X T × y is the inner product of the design matrix and the actual clock drift vector.

[0155] Table 1 Example of input data

[0156] Serial number Temperature Reference temperature Power supply fluctuation Vibration intensity Actual clock drift value 1 30 25 5 2 1.2 2 35 25 10 4 2.4 3 25 25 5 1 1.1 4 40 25 15 3 3.2

[0157] Table 2 Example of output data

[0158] Parameter name Temperature coefficient Power supply fluctuation coefficient Vibration intensity coefficient Estimated value 0.1 0.1 0.25

[0159] Table 3 Example of output results of the machine learning model

[0160] Serial number Predicted mean Predicted standard deviation Lower bound of 95% prediction interval Upper bound of 95% prediction interval 1 0.12 0.02 0.08 0.16 2 0.15 0.03 0.09 0.21 3 0.10 0.01 0.08 0.12

[0161] Specifically, obtain sufficient clock history data, as shown in Table 1, including temperature, reference temperature, power supply fluctuation, vibration intensity, and actual clock drift values, organize them into a format suitable for regression analysis, and obtain the parameters of the optimized Beidou receiver clock model according to the least squares method, as shown in Table 2. For example, the temperature coefficient is 0.1, which means that whenever the temperature increases by 1°C compared to the reference temperature, the clock drift value increases by 0.1 unit, thereby improving the calculation accuracy of the Beidou receiver clock model. If the accuracy of the model does not meet the requirements, then use the least squares method again to output the updated parameters.

[0162] Further, the random forest model is used for the processed data to output a second clock drift value and a second coefficient. The output results are shown in Table 3. The predicted values of each sample have a 95% probability of falling within the prediction interval, indicating that the reliability of each prediction result is relatively high. Then, the mean value of all predicted standard deviations is taken, and the second coefficient is 0.02.

[0163] Further, the implementation process of the Bayesian fusion model includes:

[0164] Define a prior distribution, which is a normal distribution generated based on the first clock drift value and the first coefficient;

[0165] Define a likelihood function, which is the normal distribution generated based on the second clock drift value and the second coefficient;

[0166] According to Bayes' theorem, combine the prior distribution with the likelihood function to generate a posterior distribution;

[0167] Use the mean value of the posterior distribution as the final clock drift prediction value.

[0168] Specifically, in order to prove the effect of the Bayesian fusion model, the random forest model and the Beidou receiver clock model before fusion of the present invention, as well as the SVM model in machine learning, are respectively used for comparison. The model prediction results are shown in Table 4. It can be seen that the Bayesian fusion model is superior to the single Beidou receiver clock model, random forest model, and SVM model in terms of both the mean absolute error and the root mean square error, improving the prediction accuracy.

[0169] Table 4 Model Prediction Results

[0170] Model type Mean absolute error Root mean square error BeiDou receiver clock model 0.18 0.20 Random forest model 0.15 0.17 Support vector machine model 0.16 0.18 Bayesian fusion model 0.13 0.14

[0171] Then, calculate the time offset prediction value according to the clock drift prediction value, expressed as:

[0172]

[0173] where DEV(Δt) is the time offset prediction value, Δf is the clock drift prediction value, t0 is the starting time, t is the current time, and dt is the differential of time.

[0174] Further, the calculation process of the signal quality index includes:

[0175] Divide the signal power by the noise power to obtain the carrier-to-noise ratio of the received signal; obtain the ideal carrier-to-noise ratio to obtain the carrier-to-noise ratio loss value;

[0176] Calculate the autocorrelation function of the received signal and the autocorrelation function of the ideal signal to obtain the autocorrelation function distortion value;

[0177] Use the Hilbert transform to extract the instantaneous phase of the received signal to obtain the phase change value;

[0178] Add the carrier-to-noise ratio loss value, the autocorrelation function distortion value, and the phase change value to obtain the signal quality index.

[0179] Further, the weight allocation mechanism includes:

[0180] Map the signal quality index to a signal error, and perform weighted summation with the time offset prediction value to obtain the error compensation value;

[0181] If the change amplitude of the signal quality index is greater than a preset threshold, use the gradient descent algorithm to adjust the signal quality weight and the time offset prediction value weight; otherwise, do not adjust.

[0182] Specifically, use a machine learning model, such as an LSTM model, to map the signal quality index to a possible time error range, that is, predict the current signal error. When running for the first time, set initial weights for the signal quality index and the time offset prediction value, and satisfy the sum of 1. In this embodiment, the initial signal quality weight W q,old and the time offset prediction value weight W t,old are set to 0.5 and 0.5 respectively. Then, calculate the change amplitude ΔQ of the signal quality index, and compare it with the preset threshold of 0.05 to determine whether the weights need to be adjusted. If the change amplitude is greater than 0.05, perform weight adjustment; otherwise, keep the weights unchanged.

[0183] Define the loss function, expressed as:

[0184] L=(O current -E previous ) 2 ;

[0185] Where L is the loss value, O current is the currently actually measured time offset value, and E previous is the error compensation value calculated at the previous moment.

[0186] Then, use the gradient descent algorithm to calculate the updated signal quality weight W q,new and the time offset prediction value weight W t,new . As shown in Table 5, an example of the weight allocation mechanism process is given. By combining the signal quality index with the clock offset prediction value and calculating the error compensation value in a weighted summation manner, it is possible to comprehensively consider the real-time signal conditions and the long-term clock offset trend, ensuring that when the signal quality changes significantly, it can respond quickly, optimize the error compensation effect, and improve the accuracy of Beidou time service in the mountain communication environment.

[0187] Table 5 Example of the Process of the Weight Allocation Mechanism

[0188] Time step ΔQ L <![CDATA[W q,old > <![CDATA[W t,old > <![CDATA[W q,new > <![CDATA[W t,new > 1 - - 0.5 0.5 0.5 0.5 2 0.02 0.0004 0.5 0.5 0.5 0.5 3 0.04 0.0001 0.5 0.5 0.5 0.5 4 0.14 0.0289 0.5 0.5 0.5005 0.4995 5 0.07 0.01 0.5005 0.4995 0.50106 0.49894

[0189] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An enhanced Beidou precise time service method based on the receiver clock model, characterized in that, Including: Obtain multi-source data, where the multi-source data includes environmental data, GIS data, meteorological data, signal data, and clock history data; Preprocess the multi-source data to obtain processed data; Establish a Beidou receiver clock model, where the Beidou receiver clock model is used to calculate the clock drift caused by the processed data; optimize the parameters of the Beidou receiver clock model using the least squares method according to the processed data, and output a first clock drift value and a first coefficient; use a machine learning model for the processed data to output a second clock drift value and a second coefficient; According to the first clock drift value, the first coefficient, the second clock drift value, and the second coefficient, use a Bayesian fusion model to output a clock drift prediction value; Calculate a time offset prediction value according to the clock drift prediction value; Calculate a carrier-to-noise ratio loss value, an autocorrelation function distortion value, and a phase change value for the processed data to generate a signal quality index; according to the signal quality index and the time offset prediction value, use a weight allocation mechanism to calculate an error compensation value; Correct the clock deviation according to the error compensation value.

2. The enhanced Beidou precise time service method based on the receiver clock model according to claim 1, wherein, The multi-source data includes: The environmental data includes altitude and receiver data; The GIS data includes terrain features, geological structures, and vegetation coverage; The meteorological data includes rainfall, wind speed, temperature, humidity, and air pressure; The signal data includes signal power, noise power, direct path signals, and multipath signals; The clock history data includes the historical drift information of the clock.

3. An enhanced Beidou precise timing method based on a receiver clock model according to claim 1, characterized in that, The establishment process of the receiver clock model includes: Establish a temperature influence model for the processed data and output a temperature frequency drift; the temperature influence model is a linear model of the crystal oscillator frequency varying with temperature; Establish a vibration influence model for the processed data and output a vibration frequency drift; the vibration influence model is a model of the crystal oscillator frequency varying with the vibration intensity; Establish a power supply fluctuation influence model for the processed data and output a power supply frequency drift; the power supply fluctuation influence model is a model of the crystal oscillator frequency varying with the power supply fluctuation; Sum the temperature frequency drift, the vibration frequency drift, the power supply frequency drift, and the random error to obtain the Beidou receiver clock model.

4. An enhanced Beidou precise timing method based on a receiver clock model according to claim 1, characterized in that, The implementation process of the least squares method includes: Obtain the processed data and construct a design matrix according to the temperature sequence, the vibration intensity sequence, and the vibration intensity sequence; Based on the design matrix, minimize the sum of squared errors for the Beidou receiver clock model to obtain a least squares method objective value; Take the derivative of the least squares method objective value with respect to the parameter vector and set the derivative to zero to output the parameter vector.

5. An enhanced Beidou precise timing method based on the receiver clock model according to claim 1, characterized in that The implementation process of the Bayesian fusion model includes: Define a prior distribution, where the prior distribution is a normal distribution generated according to the first clock drift value and the first coefficient; Define a likelihood function, where the likelihood function is the normal distribution generated according to the second clock drift value and the second coefficient; According to Bayes' theorem, combine the prior distribution with the likelihood function to generate a posterior distribution; Use the mean of the posterior distribution as the final clock drift prediction value.

6. The enhanced Beidou precise timing method based on the receiver clock model according to claim 1, wherein The calculation process of the signal quality index includes: Divide the signal power by the noise power to obtain the received signal carrier-to-noise ratio; obtain the ideal carrier-to-noise ratio to get the carrier-to-noise ratio loss value; Calculate the autocorrelation function of the received signal and the autocorrelation function of the ideal signal to obtain the autocorrelation function distortion value; Use the Hilbert transform to extract the instantaneous phase of the received signal to obtain the phase change value; Add the carrier-to-noise ratio loss value, the autocorrelation function distortion value, and the phase change value to obtain the signal quality index.

7. An enhanced Beidou precise time service method based on a receiver clock model according to claim 1, characterized in that, The weight allocation mechanism includes: Map the signal quality index to a signal error and perform a weighted sum with the time offset prediction value to obtain the error compensation value; If the change amplitude of the signal quality index is greater than a preset threshold, use the gradient descent algorithm to adjust the signal quality weight and the time offset prediction value weight; otherwise, do not make adjustments.

Images