A method for updating AT1 model based on prior information

Abstract

The present application belongs to the field of time and frequency technology. The present application provides a method for correcting the AT1 model of prior information. The disclosed embodiment sets the clock difference data of a preset time length before the experimental clock difference data segment to constitute the prior information for least squares polynomial fitting, and then sets the initial value according to the fitting coefficient. During the iteration of the AT1 algorithm, the frequency calculation model is improved, and the dynamic estimation of the frequency drift is added, wherein the prior information is dynamically changed and the least squares polynomial fitting is dynamically performed, the frequency and frequency drift values ​​are updated in real time, and then the next moment value is continuously iterated and predicted until the operation is completed. During the iteration of the algorithm, the prior information is set to be dynamically changed to ensure the accuracy of the dynamic estimation of the atomic clock frequency and frequency drift, so that the trend of the change of the time scale within the calculation interval is greatly weakened and the stability is improved.

Description

Technical Field

[0001] The disclosed embodiments relate to the field of time-frequency technology, and in particular to an AT1 model correction method based on prior information. Background Art

[0002] The calculation of time scale can provide an important reference for the stable operation and real-time control of the timekeeping system. Common time scale calculation methods include ALGOS algorithm, AT1 algorithm, Kalman algorithm, wavelet change algorithm, etc. Different time scales have different characteristics. Among them, the ALGOS algorithm has good long-term stability as a whole because the estimation of its weight parameters is based on the long-term stability of the atomic clock; and the frequency and weight parameters of the AT1 time scale algorithm are iteratively calculated based on the short-term and real-time data of the atomic clock, and the algorithm as a whole has good short-term stability and real-time performance.

[0003] AT1 is a time scale algorithm proposed by the National Institute of Standards and Technology (NIST). As time goes by, some scholars have pointed out the shortcomings of this algorithm. Assuming that the frequency drift term of the timekeeping clock group in the timekeeping laboratory is generally small, then even if there are atomic clocks with large short-term frequency changes, the use of the exponential filtering type AT1 time scale algorithm to calculate the atomic time scale can also achieve good results. However, when there are atomic clocks with large frequency drift in the timekeeping clock group, the use of the AT1 algorithm to calculate the atomic time scale will show a fluctuation trend with the continuous development of iterative calculations, and the long-term stability will deteriorate. For example, Judah Levine and others pointed out that the AT1 algorithm does not estimate or update the frequency, which will lead to inaccurate frequency estimation and then cause frequency drift in the time scale calculated by the entire clock group. They specifically pointed out that the frequency drift of the hydrogen clock is on the order of 10 −21 s −1 , it is very important to include its estimation in the AT1 algorithm model. At the same time, it is also recognized that even if the aging of the atomic clock frequency is small, if the calculation process is not estimated and corrected in time, it will have a great impact on the result. In response to this problem, Polish scholars Michał Marszalec, Marzenna Lusawa and others proposed the AT2 algorithm to improve the AT1 algorithm, that is, using Kalman to calculate the frequency deviation and frequency drift, and achieved certain results. However, Kalman modeling is relatively complex, and it is necessary to construct a reasonable atomic clock noise model based on the characteristics of the atomic clock noise distribution, and then estimate the state process covariance matrix and the observation noise covariance matrix. Therefore, it is necessary to further solve the disturbance problem of the AT1 algorithm due to frequency drift, ensure the robustness of the algorithm, and achieve good long-term and short-term stability. Summary of the invention

[0004] In order to avoid the shortcomings of the prior art, the present application provides an AT1 model correction method based on prior information to solve the problem of disturbance of the AT1 algorithm due to frequency drift in the prior art.

[0005] According to an embodiment of the present disclosure, a method for modifying an AT1 model based on prior information is provided, the method comprising:

[0006] S1: Determine the calculation interval of the time scale of the AT1 algorithm according to the comparison cycle of the timekeeping system;

[0007] S2: Based on the prior information of several atomic clocks, the initial frequency, initial frequency drift value and initial weight of the AT1 algorithm are obtained;

[0008] S3: preprocess the test data set of the current time period of all atomic clocks to handle missing data and outliers;

[0009] S4: Obtain the clock error prediction correction item of the current time period according to the initial frequency and the initial frequency drift value, and obtain the time scale of the current time period according to the clock error prediction correction item of the current time period, the initial weight and the test data set of the current time period;

[0010] S5: obtaining the actual value of the clock error correction item of the current time period according to the time scale of the current time period;

[0011] S6: updating the prior information according to the test data set of the current time period, obtaining the model correction estimate item of the next time period according to the updated prior information, and obtaining the clock error prediction correction item of the next time period according to the model correction estimate item of the next time period;

[0012] S7: Obtaining the weight of the current time period according to the actual value of the clock error correction item of the current time period and the clock error prediction correction item of the current time period;

[0013] S8: Obtain the time scale of the next time period according to the clock error prediction correction term of the next time period, the test data set of the next time period and the weight of the current time period;

[0014] S9: Repeat the above steps S5 to S8 to perform iterative calculation until all time scale calculations are completed.

[0015] Furthermore, based on the prior information of several atomic clocks, the step of obtaining the initial frequency, initial frequency drift value and initial weight of the AT1 algorithm includes:

[0016] Performing quadratic fitting on the prior information of each atomic clock to obtain the initial frequency and initial frequency drift value;

[0017] Compare the data based on prior information to obtain the Allan variance of each atomic clock;

[0018] The initial weights are obtained according to the Allan variance and calculation interval of each atomic clock.

[0019] Furthermore, the step of preprocessing the test data set of the current time period of all atomic clocks to handle missing data and outliers includes:

[0020] The first moment mark of the test data set in the current time period is used as the starting point of data calculation, and the time unit of the test data set in the current time period is unified as nanoseconds;

[0021] Perform integrity check and outlier check on the test data set for the current time period;

[0022] If the test data set of the current time period is missing, the missing data is estimated using linear prediction and added to the test data set of the current time period;

[0023] If there are outliers in the test data set of the current time period, the least squares fitting is used to obtain the fitting value, and the fitting value is used to replace the outlier.

[0024] Furthermore, the expression of the time scale is:

[0025]

[0026]

[0027] in, , Indicates the previous time period. t Indicates the current time period. For the The correction term for the predicted clock error of the atomic clock in the current time period, For the The weight or initial weight of the atomic clock in the previous time period, An atomic clock indicating the current time period and atomic clocks The clock difference, For the Atomic clock The phase difference of time, is the frequency of the current time period, is the frequency drift value.

[0028] Furthermore, the step of obtaining the actual value of the clock error correction term of the current time period according to the time scale of the current time period includes:

[0029] The actual value of the clock error correction term at the current moment is obtained according to the time scale of the current time period and the test data set of the current time period.

[0030] Further, the step of updating the prior information according to the test data set of the current time period, obtaining the model correction estimate item of the next time period according to the updated prior information, and obtaining the clock error prediction correction item of the next time period according to the model correction estimate item of the next time period includes:

[0031] Update the prior information using the test data set of the current time period;

[0032] The model correction estimate for the next time period is obtained based on the updated prior information;

[0033] The frequency of the clock error prediction correction item in the next time period is obtained according to the model correction estimation item in the next time period and the frequency in the current time period;

[0034] Obtaining the frequency drift value of the clock error prediction correction term for the next time period according to the model correction estimation term for the next time period;

[0035] The clock error prediction correction item for the next time period is obtained according to the frequency of the clock error prediction correction item for the next time period and the frequency drift value of the clock error prediction correction item for the next time period.

[0036] Furthermore, when the calculation interval is When , the frequency calculation model for the next time period is:

[0037]

[0038] in, is the exponential filter coefficient, and , For the clock The time interval corresponding to the minimum value of the Allan variance within the calculation interval;

[0039] is the model correction estimate, and its expression is:

[0040]

[0041]

[0042] in, Phase The estimated value of The frequency for the next time period The estimated value of The frequency drift value for the next time period The estimated value of represents the atomic clock comparison time scale, express Comparison data for time periods, is the dynamically updated prior information.

[0043] Furthermore, the expression of the weight of the current time period is:

[0044]

[0045] in, , is the deviation root of the error estimate, , is the clock error prediction correction item for the current time period, is the actual value of the clock correction term in the current time period, and , is the exponential filter time constant.

[0046] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects:

[0047] In the embodiment of the present disclosure, through the above-mentioned AT1 model correction method based on prior information, on the one hand, the clock difference data of a preset time length before the experimental clock difference data segment is set to constitute the prior information for least squares polynomial fitting, and then the initial value is set according to the fitting coefficient. During the iteration process of the AT1 algorithm, the calculation expression of the clock speed is transformed, and the dynamic estimation of the frequency drift is added, wherein the prior information is dynamically changed and the least squares polynomial fitting is dynamically performed, the frequency and frequency drift values ​​are updated in real time, and then the next moment value is continuously iterated and predicted until the operation is completed. On the other hand, during the iteration process of the algorithm, the prior information is set to be dynamically changed to ensure the accuracy of the dynamic estimation of the atomic clock frequency and frequency drift, so that the trend of the change of the time scale within the calculation interval is greatly weakened and the stability is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0049] Figure 1 A step diagram showing a method for correcting an AT1 model based on prior information in an exemplary embodiment of the present disclosure;

[0050] Figure 2 A diagram showing a timekeeping system structure in an exemplary embodiment of the present disclosure is shown;

[0051] Figure 3A specific flow chart showing the AT1 model correction method based on prior information in an exemplary embodiment of the present disclosure;

[0052] Figure 4 A graph showing the stability comparison results of two time scales of the AT1 model correction method based on prior information and the weighted average WAT in an exemplary embodiment of the present disclosure is shown;

[0053] Figure 5 A graph showing the stability comparison results of two time scales of the AT1 model correction method based on prior information and the AT1 algorithm in an exemplary embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0054] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the disclosure will be more comprehensive and complete and to fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0055] In addition, the accompanying drawings are only schematic illustrations of the embodiments of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated descriptions will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities.

[0056] This example implementation provides an AT1 model correction method based on prior information. Figure 1 As shown in , the AT1 model correction method based on prior information may include:

[0057] S1: Determine the calculation interval of the time scale of the AT1 algorithm according to the comparison cycle of the timekeeping system;

[0058] S2: Based on the prior information of several atomic clocks, the initial frequency, initial frequency drift value and initial weight of the AT1 algorithm are obtained;

[0059] S3: preprocess the test data set of the current time period of all atomic clocks to handle missing data and outliers;

[0060] S4: Obtain the clock error prediction correction item of the current time period according to the initial frequency and the initial frequency drift value, and obtain the time scale of the current time period according to the clock error prediction correction item of the current time period, the initial weight and the test data set of the current time period;

[0061] S5: obtaining the actual value of the clock error correction item of the current time period according to the time scale of the current time period;

[0062] S6: updating the prior information according to the test data set of the current time period, obtaining the model correction estimate item of the next time period according to the updated prior information, and obtaining the clock error prediction correction item of the next time period according to the model correction estimate item of the next time period;

[0063] S7: Obtaining the weight of the current time period according to the actual value of the clock error correction item of the current time period and the clock error prediction correction item of the current time period;

[0064] S8: Obtain the time scale of the next time period according to the clock error prediction correction term of the next time period, the test data set of the next time period and the weight of the current time period;

[0065] S9: Repeat the above steps S5 to S8 to perform iterative calculation until all time scale calculations are completed.

[0066] Through the above-mentioned AT1 model correction method based on prior information, on the one hand, the forecast setting uses the clock difference data of a preset time length before the experimental clock difference data segment to form the prior information for least squares polynomial fitting, and then sets the initial value according to the fitting coefficient. During the iteration of the AT1 algorithm, the calculation expression of the clock speed is transformed, and the dynamic estimation of the frequency drift is added. The prior information is dynamically changed and the least squares polynomial fitting is performed dynamically, and the frequency and frequency drift values ​​are updated in real time, and then the next moment value is predicted continuously until the operation is completed. On the other hand, during the iteration of the algorithm, the prior information is set to be dynamically changed to ensure the accuracy of the dynamic estimation of the atomic clock frequency and frequency drift, so that the trend of changes in the time scale within the calculation interval is greatly weakened and the stability is improved.

[0067] Next, we will refer to Figures 1 to 5 Each step of the above-mentioned AT1 model modification method based on prior information in this exemplary embodiment is described in more detail.

[0068] In step S1, the time scale calculation cycle of the AT1 algorithm is set to be the same as the comparison cycle of the timekeeping system. Set to 1 hour, no specific limit. Figure 2 As shown, it is a timekeeping system structure diagram. The principle of the timekeeping system is the existing technology and will not be described in detail here.

[0069] In step S2, if Figure 3 FIG. 1 is a flow chart of the AT1 model correction method based on prior information. First, the initial frequency and the frequency drift value are set.

[0070] A test data set of a preset time length before the current moment is selected as prior information, and the prior information is used as the input of the least squares fitting model. The frequency and frequency drift values ​​of each atomic clock are obtained through quadratic fitting, and used as the initial frequency and initial frequency drift value of the AT1 algorithm iteration.

[0071] In step S3, if Figure 3 As shown, the test data set of the current time period is preprocessed.

[0072] The original clock error data is in seconds. To facilitate further calculations, the mark of the first moment of the data is used as the starting point of the data calculation. At the same time, the test data set is unified into nanoseconds. The integrity of the test data set is then checked and outliers are detected. Among them, the integrity test can be performed by subtracting the previous data from the latter data to determine whether the jump is completed, or by drawing a graph. For the missing data detected, linear prediction is used to estimate and supplement it, and the outlier detection is tested using the Wright criterion. The specific principles are as follows:

[0073] First, fit the test dataset as , then the residual between the measured value and the fitted data is: ;in, is the measured value, is the fitting value at the corresponding time scale.

[0074] Taking the residuals as the basis, the variance is then calculated: ;

[0075] in, is the residual mean, N is the number of atomic clocks, assuming that t 1 Moment, satisfaction , then For the detected outliers, the outliers were then replaced by the least squares fitted values.

[0076] In step S4 to step S6, the time scale calculation formula is:

[0077]

[0078] in, For the The calculation expression of the clock error prediction correction term of the table clock is:

[0079]

[0080] in, express time, is the frequency term (i.e., frequency), is the frequency drift term (i.e., frequency drift value).

[0081] Indicates the clock He Zhong The phase difference data is in the sampling interval The frequency calculation model is improved as follows:

[0082]

[0083] in, is the exponential filter coefficient, which is expressed as follows:

[0084]

[0085] in, It can be obtained by selecting the sampling interval corresponding to the minimum difference of the independent Allan variance of the atomic clock.

[0086] is the model correction estimate, The value of is calculated as follows:

[0087]

[0088] in, is the amount of prior information. This value changes dynamically over time to ensure the use of the latest comparison results, discard the comparison values ​​that are far away from the current time, and ensure the accuracy of the forecast estimate. Represents the atomic clock comparison time scale.

[0089] use , , Phase ,frequency , frequency drift value The estimated value of , using the least squares estimation method, the polynomial coefficients can be obtained as:

[0090]

[0091] in, express Comparison data at the moment.

[0092] In step S7, Figure 3 As shown, calculate the weights.

[0093] According to the prediction correction items of the clock errors of each atomic clock in the current time period The actual clock error correction term of each atomic clock in the current time period The difference is calculated, which involves the selection of exponential constants as follows:

[0094] The expression of weight is: ,in, ,in is the deviation root of the error estimate, and the calculation formula is ,and .

[0095] Assuming that the formula Take 20 days, The initial value is , The calculation is based on the tricorn hat method to calculate the Allan variance of a single atomic clock. The calculation process is as follows:

[0096] The Allan variance of the three atomic clocks compared in pairs over a certain period of time is expressed as , , , assuming that the Allan variance of each single atomic clock at a certain time is expressed as , , , then:

[0097]

[0098] The solution to this system of equations is:

[0099]

[0100] If the atomic clock If the number of atomic clocks is greater than 3, the Allan variance expansion method for a single atomic clock is the polygonal hat method, which is expressed as follows:

[0101]

[0102]

[0103] In step S8 and step S9, the time scale of the next time period is obtained according to the clock error prediction correction item of the next time period, the test data set of the next time period and the weight of the current time period; the above steps S5 to S8 are repeated for iterative calculation until all time scale calculations are completed.

[0104] Specifically, the clock error prediction correction term of N atomic clocks is calculated for the first time using the initial frequency, frequency drift value and clock error prediction correction term formula. The initial weight of the time scale is calculated based on the Allan variance of a single atomic clock using the comparison data of prior information. ,use set up , and then the initial weight is obtained according to the weight calculation formula.

[0105] Obtain the clock error prediction correction item of each atomic clock and the weight distribution of each atomic clock. At the same time, combine the preprocessed test data set at the current moment and use the time scale calculation formula to obtain the first time scale. After obtaining the time scale, the actual value of the clock error correction item can be obtained. Subsequently, the clock error prediction correction item and the actual clock error correction item of each atomic clock in the current time period are used to calculate the weights to prepare for the calculation of the time scale at the next moment, and this is iterated. The specific calculation process of the weight is as shown in step S7. In the calculation of the clock error prediction correction item at the next moment, the frequency of the clock error prediction correction item is calculated by the model correction estimation item combined with the frequency item at the previous moment, and the frequency drift is estimated by the model correction estimation item. At subsequent moments, iterative calculations are performed in this way.

[0106] In a specific embodiment, three hydrogen atomic clocks H1, H2, and H3 of a certain institution are selected for illustration of this application. The calculation interval of the data sample is 1 hour, and the sample length is one month. First, the data of the previous day of the experimental data sample is selected as the prior information, and the prior information is used as the input of the least squares fitting model. The frequency and frequency drift values ​​of each atomic clock are obtained by quadratic fitting as the initial value of the algorithm iteration.

[0107] The data integrity of the experimental sample data was checked, and the Wright criterion was used to detect data outliers. The fitting polynomial coefficients for outlier detection were set to 2. First, the initial clock error prediction correction term was calculated using the initial value, and then the exponential filter coefficient was calculated. It is calculated based on the data from the two months before the calculation sample. Specifically, the data of the three clocks are converted to compare with each other, and then the Allan variance is calculated. The Allan variance of a single atomic clock is obtained using the triangular hat method. Finally, the sampling interval corresponding to the minimum difference in the independent Allan variance of the atomic clocks is .

[0108] Another parameter is the initial value of the atomic clock. The determination is based on the data from one month before the sample is calculated. First, the data of the three atomic clocks are converted for pairwise comparison, and then the Allan variance is calculated. The triangular hat method is used to obtain the Allan variance of a single atomic clock, and the value is obtained by combining the time interval.

[0109] After the exponential filter coefficient is determined, the atomic clock frequency is calculated The model correction estimate is obtained by the least squares estimation of the dynamic prior information, which is calculated as follows:

[0110]

[0111]

[0112] The final results are similar to those of the weighted average WAT algorithm and the AT1 algorithm. Figure 4 and Figure 5 As shown in the figure. The long-term and short-term stability of the time scale formed by the AT1 correction algorithm based on prior information is better than that of the weighted average WAT time scale. The long-term stability of the time scale formed by the AT1 correction algorithm based on prior information is significantly better than that of the AT1 time scale. The effectiveness of the method proposed in this application is verified.

[0113] In addition, in order to further optimize the solution of this application, the length of the prior information data is scaled for different types of timekeeping atomic clocks. Specifically, the stability index of the atomic clock is calculated to determine the optimal corresponding time scale for the stability of different types of atomic clocks. This time scale is used as the length of the prior information to ensure the accuracy of the calculation.

[0114] In terms of data preprocessing, this application mainly considers the anomaly of the atomic clock phase jump. In order to further enhance the robustness of this application and ensure the accuracy of the calculation time scale, a frequency jump test is performed at the same time. The Allan variance factor of the calculated sample data is used as the threshold to traverse and test all frequency data. After the test is completed, the frequency data is converted into phase data for the next step of calculation.

[0115] It should be emphasized that when the overall data volume is limited, this application can flexibly adjust the sampling interval to ensure a smaller sampling interval while having sufficient data volume.

[0116] The experimental results show that the time scale formed by the Improved ATl Algorithm based on Prior Information (IAAPI) has better short-term and long-term stability than the weighted average time scale WAT, and better long-term stability than the AT1 algorithm time scale. This invention can provide a more stable reference for the real-time control of the master clock frequency source of the timekeeping system, and can be used as a supplementary means for the control of standard time output.

[0117] Through the above-mentioned AT1 model correction method based on prior information, this method has simple modeling and strong real-time performance, which weakens the changing trend of the AT1 time scale caused by the atomic clock frequency drift and alleviates the problem of long-term stability deterioration. Specifically, this method is carried out from the head and middle parts of the original algorithm. Since the AT1 algorithm is a continuous iterative process, inaccurate estimation of the initial value will cause the error of the overall calculation result to spread. Therefore, this step is predicted and set based on the prior information, deducting the changing trend from the beginning of the calculation and reducing the error effect. During the iterative process of the algorithm, the prior information is set to dynamically change to ensure the accuracy of the dynamic estimation of the atomic clock rate and frequency drift, so that the changing trend of the time scale within the calculation interval is greatly weakened and the stability is improved.

[0118] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification.

[0119] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the appended claims.

Claims

1. A method for modifying an AT1 model based on prior information, characterized in that: The method includes: S1: Determine the calculation interval of the time scale of the AT1 algorithm according to the comparison cycle of the timekeeping system; S2: Based on the prior information of several atomic clocks, the initial frequency, initial frequency drift value and initial weight of the AT1 algorithm are obtained; S3: preprocess the test data set of the current time period of all atomic clocks to handle missing data and outliers; S4: Obtain the clock error prediction correction item of the current time period according to the initial frequency and the initial frequency drift value, and obtain the time scale of the current time period according to the clock error prediction correction item of the current time period, the initial weight and the test data set of the current time period; S5: obtaining the actual value of the clock error correction item of the current time period according to the time scale of the current time period; S6: updating the prior information according to the test data set of the current time period, obtaining the model correction estimate item of the next time period according to the updated prior information, and obtaining the clock error prediction correction item of the next time period according to the model correction estimate item of the next time period; S7: Obtaining the weight of the current time period according to the actual value of the clock error correction item of the current time period and the clock error prediction correction item of the current time period; S8: Obtain the time scale of the next time period according to the clock error prediction correction term of the next time period, the test data set of the next time period and the weight of the current time period; S9: Repeat the above steps S5 to S8 to perform iterative calculation until all time scale calculations are completed.

2. The AT1 model correction method based on prior information according to claim 1, characterized in that: Based on the prior information of several atomic clocks, the steps of obtaining the initial frequency, initial frequency drift value and initial weight of the AT1 algorithm include: Performing quadratic fitting on the prior information of each atomic clock to obtain the initial frequency and initial frequency drift value; Compare the data based on prior information to obtain the Allan variance of each atomic clock; The initial weights are obtained according to the Allan variance and calculation interval of each atomic clock.

3. The AT1 model correction method based on prior information according to claim 2 is characterized in that: The steps of preprocessing the test data set of the current time period of all atomic clocks to handle missing data and outliers include: The first moment mark of the test data set in the current time period is used as the starting point of data calculation, and the time unit of the test data set in the current time period is unified as nanoseconds; Perform integrity check and outlier check on the test data set for the current time period; If the test data set of the current time period is missing, the missing data is estimated using linear prediction and added to the test data set of the current time period; If there are outliers in the test data set of the current time period, the least squares fitting is used to obtain the fitting value, and the fitting value is used to replace the outlier.

4. The AT1 model correction method based on prior information according to claim 3 is characterized in that: The expression of time scale is: in, , Indicates the previous time period. t Indicates the current time period. For the The correction term for the predicted clock error of the atomic clock in the current time period, For the The weight or initial weight of the atomic clock in the previous time period, An atomic clock indicating the current time period and atomic clocks The clock difference, For the Atomic clock The phase difference of time, is the frequency of the current time period, is the frequency drift value.

5. The AT1 model correction method based on prior information according to claim 4 is characterized in that: The step of obtaining the actual value of the clock error correction term of the current time period according to the time scale of the current time period includes: The actual value of the clock error correction term at the current moment is obtained according to the time scale of the current time period and the test data set of the current time period.

6. The AT1 model correction method based on prior information according to claim 5 is characterized in that: The steps of updating the prior information according to the test data set of the current time period, obtaining the model correction estimate item of the next time period according to the updated prior information, and obtaining the clock error prediction correction item of the next time period according to the model correction estimate item of the next time period include: Update the prior information using the test data set of the current time period; The model correction estimate for the next time period is obtained based on the updated prior information; Obtain the frequency of the clock error prediction correction term for the next time period according to the model correction estimate term for the next time period and the frequency of the current time period; Obtaining the frequency drift value of the clock error prediction correction term for the next time period according to the model correction estimation term for the next time period; The clock error prediction correction item for the next time period is obtained according to the frequency of the clock error prediction correction item for the next time period and the frequency drift value of the clock error prediction correction item for the next time period.

7. The AT1 model correction method based on prior information according to claim 6 is characterized in that: The calculation interval is When , the frequency calculation model for the next time period is: in, is the exponential filter coefficient, and , For atomic clock The time interval corresponding to the minimum value of the Allan variance within the calculation interval; is the model correction estimate, and its expression is: in, Phase The estimated value of The frequency for the next time period The estimated value of The frequency drift value for the next time period The estimated value of represents the atomic clock comparison time scale, express Comparison data for time periods, is the dynamically updated prior information.

8. The AT1 model correction method based on prior information according to claim 7 is characterized in that: The expression of the weight of the current time period is: in, , is the deviation root of the error estimate, , is the clock error prediction correction item for the current time period, is the actual value of the clock correction term in the current time period, and , is the exponential filter time constant.

Images