A satellite-borne clock driving method, device and system based on model prediction

Abstract

The present application relates to a satellite clock control method, device and system based on model prediction. The method includes: performing state estimation based on a pre-constructed high-precision clock model and historical clock error data to obtain the clock state at each moment in the prediction interval; constructing a quadratic cost function based on the clock state at each moment in the prediction interval and a weight matrix, minimizing the quadratic cost function using a rolling optimization method, outputting the control amount at the current control moment, optimizing the control amount using a differential control amount and a differential coefficient, and obtaining the optimal control amount; obtaining a clock error sequence to be adjusted based on the clock error data corresponding to future moments in the prediction interval, performing phase adjustment on the clock error sequence to be adjusted based on the optimal control amount at the current control moment and the time interval between two adjacent clock errors, obtaining an adjusted clock error sequence, and iteratively updating the clock error sequence according to the timing sequence to realize satellite clock control. The use of this method can enhance the performance of the satellite-borne secondary frequency standard.

Description

Technical Field

[0001] The present application relates to the field of satellite-borne clock control technology, and in particular to a satellite-borne clock control method, device and system based on model prediction. Background Art

[0002] With the establishment and improvement of the comprehensive PNT (positioning, navigation and timing) system, the navigation augmentation system is also developing rapidly as an important part of the future comprehensive PNT system. With the maturity of technology and the reduction of costs in various links of the industrial chain, technological breakthroughs have a very significant effect on reducing the cost of satellite construction. At present, the cost reduction mainly involves three aspects: satellite launch technology, satellite recovery technology, and satellite payload. From the perspective of satellite payload, if the secondary frequency standard can be used as the onboard clock, the payload cost will be significantly reduced. However, if the secondary frequency standard is carried, on the one hand, the importance of autonomous timekeeping begins to emerge, and on the other hand, industrial applications have high requirements for time synchronization accuracy, which means that the performance of the clock carried by itself is better and can meet higher accuracy requirements, while relying solely on the secondary frequency standard cannot meet these requirements. For example, high-stability crystal oscillators and chip atomic clocks have good short-term stability, but poor long-term stability. Therefore, how to use satellite clock driving technology to achieve low-cost clocks that meet higher accuracy requirements is a technical problem that needs to be solved urgently.

[0003] Existing clock control algorithms, such as PID (Proportional-Integral-Derivative Controller) and ping-pong method, are commonly used for rubidium clock crystal control and are very mature. The parameter adjustment of GPS disciplined clock based on PID control algorithm is very complicated because the three PID parameters are interrelated and it is impossible to determine whether they are the optimal control results after adjustment. The advantage of using ping-pong method to control rubidium clock using GPS is that compared with PID algorithm, the control amount calculation process is relatively simple and the design complexity is low. The disadvantage is that it is still a proportional control algorithm in essence. On the one hand, it will take a long time for the system to reach steady state. On the other hand, there will still be large fluctuations after stable control, which cannot meet higher performance requirements. Summary of the invention

[0004] Based on this, it is necessary to provide a satellite clock control method, device and system based on model prediction to address the above technical problems.

[0005] A satellite-borne clock driving method based on model prediction, the method comprising:

[0006] The state is estimated based on the pre-built high-precision clock model and historical clock error data to obtain the clock state at each moment in the prediction interval; the lower limit of the prediction interval is the current driving moment, and the upper limit is the rolling length;

[0007] A quadratic cost function is constructed according to the clock state and the weight matrix at each moment in the prediction interval, a rolling optimization method is used to minimize the quadratic cost function, a control amount at the current driving moment is output, and the control amount is optimized using a differential control amount and a differential coefficient to obtain an optimal control amount;

[0008] The clock error sequence to be adjusted is obtained according to the clock error data corresponding to the future time in the prediction interval. The phase of the clock error sequence to be adjusted is adjusted according to the optimal control amount of the current steering moment and the time interval between two adjacent clock errors to obtain the adjusted clock error sequence. The clock error sequence is updated iteratively according to the time sequence to realize the satellite clock steering.

[0009] In one embodiment, it also includes: constructing a trend term model according to the clock error data of the previous moment; obtaining historical clock error data, removing the trend term from the historical clock error data according to the trend term model, and obtaining the historical clock error data after detrending; performing spectrum analysis on the historical clock error data after detrending, extracting the frequency by using the fast Fourier transform method, estimating the amplitude and phase of the frequency by using the least squares fitting, and constructing a periodic term model; removing the periodic term from the historical clock error data after detrending according to the periodic term model, and obtaining the corresponding historical residual data; obtaining a pre-trained residual prediction model, inputting the historical residual data and the historical steering adjustment amount into the pre-trained residual prediction model, obtaining the residual data at the future moment, and constructing an uncertainty term model according to the predicted residual data;

[0010] A high-precision clock model is constructed according to the trend item model, the period item model and the uncertainty item model.

[0011] In one of the embodiments, the high-precision clock model is:

[0012] ;

[0013] in, represents the current clock error, Represents the existing clock error at the previous moment, represents the frequency deviation, represents the frequency drift, represents the periodic term, Represents the uncertain term, Represents the superposition of five power-law spectrum noises that vary with time.

[0014] In one of the embodiments, it also includes: traversing the prediction interval, using the Kalman filter algorithm to process the clock difference data of the previous moment of the current traversal moment, obtaining the clock state of the current traversal moment based on the high-precision clock model, until the traversal of each moment in the prediction interval is completed, stopping the traversal, and outputting the clock state of each moment in the prediction interval.

[0015] In one embodiment, it further includes: constructing a quadratic cost function according to the clock state at each moment in the prediction interval and the weight matrix:

[0016] ;

[0017] ;

[0018] ;

[0019] in, is a quadratic cost function, for The prediction interval at the moment The clock state at a moment, is the terminal state weight matrix, for The state error weight matrix at time , for The control weight matrix at the moment, for The prediction interval at the moment The amount of control at a moment, , N is the length of the prediction interval, and the prediction interval is , Represents a transpose operation.

[0020] In one embodiment, the method further includes: obtaining a differential control amount according to the clock state change rate, weighting the differential control amount using a differential coefficient and then fusing it with the control amount, optimizing the differential coefficient using a dichotomy method, and obtaining an optimal control amount according to the optimization result:

[0021] ;

[0022] in, is the optimal control quantity, To control the amount, is the differential coefficient, is the differential control quantity.

[0023] In one embodiment, the method further includes: performing phase adjustment on the clock difference sequence to be adjusted according to the optimal control amount at the current driving moment and the time interval between two adjacent clock differences, so that the adjusted clock difference sequence is:

[0024] ;

[0025] in, is the adjusted clock error sequence, is the clock error sequence to be adjusted, for The optimal control quantity at the moment, is the time interval between two adjacent clock differences, is the length of the prediction interval.

[0026] A satellite-borne clock driving device based on model prediction, the device comprising:

[0027] The state estimation module is used to perform state estimation based on the pre-built high-precision clock model and historical clock error data to obtain the clock state at each moment in the prediction interval; the lower limit of the prediction interval is the current driving moment, and the upper limit is the rolling length;

[0028] A control quantity optimization module is used to construct a quadratic cost function according to the clock state and the weight matrix at each moment in the prediction interval, minimize the quadratic cost function using a rolling optimization method, output the control quantity at the current driving moment, and optimize the control quantity using a differential control quantity and a differential coefficient to obtain an optimal control quantity;

[0029] The clock control module is used to obtain the clock error sequence to be adjusted according to the clock error data corresponding to the future time in the prediction interval, and to adjust the phase of the clock error sequence to be adjusted according to the optimal control amount at the current control time and the time interval between two adjacent clock errors to obtain the adjusted clock error sequence, and to iteratively update the clock error sequence according to the time sequence to realize the on-board clock control.

[0030] A satellite-borne clock steering system based on model prediction, the system comprising:

[0031] Host computer, onboard clock, digital-to-analog converter and voltage-controlled crystal oscillator;

[0032] The upper level performs state estimation according to the historical clock error data and the high-precision clock model to obtain the clock state at each moment in the prediction interval, constructs a quadratic cost function according to the clock state at each moment in the prediction interval and a weight matrix, minimizes the quadratic cost function using a rolling optimization method, outputs the control amount at the current driving moment, optimizes the control amount using a differential control amount and a differential coefficient, and obtains an optimal control amount; the lower limit of the prediction interval is the current driving moment, and the upper limit is the rolling length;

[0033] The digital-to-analog converter is used to convert the optimal control quantity output by the host computer into an analog voltage signal, and output it to the voltage-controlled crystal oscillator;

[0034] The voltage-controlled crystal oscillator is used to generate a corresponding optimal frequency adjustment value according to the analog voltage signal;

[0035] The satellite-borne clock is used to accumulate the optimal frequency adjustment amount over time, convert it into a clock difference adjustment amount, and perform clock control according to the clock difference adjustment amount.

[0036] The above-mentioned satellite clock control method, device and system based on model prediction calculates the clock error data of the prediction interval through a high-precision clock model, performs state estimation based on the clock error data, and obtains the clock state at each moment in the prediction interval; constructs a quadratic cost function based on the clock state at each moment in the prediction interval and the weight matrix, uses a rolling optimization method to minimize the quadratic cost function, outputs the control amount at the current control moment, optimizes the control amount using the differential control amount and the differential coefficient, and obtains the optimal control amount; obtains the clock error sequence to be adjusted based on the clock error data corresponding to the future moment in the prediction interval, and performs phase adjustment on the clock error sequence to be adjusted based on the optimal control amount at the current control moment and the time interval between two adjacent clock errors to obtain the adjusted clock error sequence, and updates the clock error sequence according to the time sequence iteration to realize satellite clock control. The embodiment of the present invention has simple calculation and low cost, can reduce clock error fluctuation, improve frequency accuracy and frequency stability, enhance the performance of satellite-borne secondary frequency standards in complex environments, and can meet the real-time, high-precision and low-cost requirements of low-orbit satellite time synchronization. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flowchart of a satellite-borne clock driving method based on model prediction in one embodiment;

[0038] Figure 2 A schematic diagram of a clock control process in an embodiment;

[0039] Figure 3 A flowchart of a clock control algorithm based on MPC (Model Predictive Controller) in one embodiment;

[0040] Figure 4 This is a diagram of clock error simulation results in an embodiment, where: Figure 4 (a) is a schematic diagram showing the comparison of theoretical and simulated Allan deviation curves when the rubidium clock is applied to the method of the present invention. Figure 4 (b) is a schematic diagram showing the comparison of theoretical and simulated Allan deviation curves when the crystal oscillator is applied to the method of the present invention;

[0041] Figure 5 FIG. 1 is a schematic diagram of clock driving frequency stability results under different algorithms in one embodiment, wherein: Figure 5 (a) is a schematic diagram of the crystal oscillator driving frequency stability results under different algorithms. Figure 5 (b) is a schematic diagram of the frequency stability results of the rubidium clock under different algorithms;

[0042] Figure 6 is a structural block diagram of a satellite-borne clock driving device based on model prediction in one embodiment;

[0043] Figure 7 The present invention is a structural block diagram of a satellite-borne clock steering system based on model prediction in one embodiment. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0045] In one embodiment, Figure 1 As shown, a satellite-borne clock steering method based on model prediction is provided, comprising the following steps:

[0046] Step 102, perform state estimation based on the pre-built high-precision clock model and historical clock error data to obtain the clock state at each moment in the prediction interval.

[0047] Among them, the lower limit of the prediction interval is the current driving moment, and the upper limit is the rolling length.

[0048] There are different types of satellite clocks, which are generally divided into two categories: primary frequency standards, including hydrogen clocks, cesium clocks and high-performance rubidium clocks, and secondary frequency standards, including slightly inferior rubidium clocks, chip atomic clocks and high-performance crystal oscillators. Both types of clocks conform to the polynomial model of atomic clocks, and the difference lies in the frequency drift and five power-law spectrum noise coefficients in the model. Different clock signals can be simulated according to their respective characteristics.

[0049] For the satellite clock, based on its original physical characteristics, the reference clock signal is used to solve the clock error, and the real-time and correctness of the control method are improved based on the short-term prediction of the clock error. Then, the optimal adjustment amount is calculated through the model prediction algorithm to achieve accurate synchronization between the satellite clock and the reference clock, thereby improving the performance of the satellite clock. It should be noted that although MPC is a powerful control method in many systems, when applied to clock systems in complex environments, due to the complexity and uncertainty of clock errors, the accuracy of the model is difficult to guarantee, and the prediction results will be biased, resulting in unsatisfactory control effects. In order to solve the challenges brought by unclear models and uncertainties, the method of the present invention designs a high-precision clock model. Using the method of the present invention, it is possible to ensure that the performance of the satellite clock meets the requirements while carrying the secondary frequency standard in a complex environment.

[0050] Step 104, construct a quadratic cost function according to the clock state and weight matrix at each moment in the prediction interval, use the rolling optimization method to minimize the quadratic cost function, output the control amount at the current driving moment, optimize the control amount using the differential control amount and differential coefficient, and obtain the optimal control amount.

[0051] The control amount during clock driving is optimized through the cost function, which has a clear optimization strategy and parameter adjustment rules. The rolling optimization strategy is used to better balance the relationship between short-term stability and long-term stability in clock performance during clock driving. The differential term is introduced to comprehensively calculate the control adjustment amount, which reduces the fluctuation of the clock phase. At the same time, due to the introduction of differentials, the taming system can converge faster.

[0052] Step 106, obtain the clock error sequence to be adjusted according to the clock error data corresponding to the future time in the prediction interval, perform phase adjustment on the clock error sequence to be adjusted according to the optimal control amount at the current steering moment and the time interval between two adjacent clock errors, obtain the adjusted clock error sequence, and iteratively update the clock error sequence according to the time sequence to realize the satellite clock steering.

[0053] like Figure 2 The clock driving process diagram shown in the figure includes the following steps: clock error generation; clock state estimation: using the established high-precision clock model and historical clock error data to estimate the clock error data state; designing the cost function of the algorithm; using rolling optimization to calculate the optimal control at the current moment; introducing the differential term and the differential coefficient, and optimizing and updating the optimal control quantity obtained in the previous step; calculating the optimal control quantity and converting it into phase change; repeating the above steps in a time sequence. The method of the present invention can fully improve the performance of the satellite-borne secondary frequency standard clock in a complex environment, while maintaining the low-cost advantage of the satellite-borne secondary frequency standard clock. In practical applications, the rubidium clock and crystal oscillator can be driven to keep them synchronized with the GNSS system time.

[0054] In the above-mentioned satellite clock steering method based on model prediction, the clock error data of the prediction interval is calculated by a high-precision clock model, and the state is estimated according to the clock error data to obtain the clock state at each moment in the prediction interval; a quadratic cost function is constructed according to the clock state at each moment in the prediction interval and the weight matrix, and the quadratic cost function is minimized by a rolling optimization method, and the control amount at the current steering moment is output, and the control amount is optimized by using the differential control amount and the differential coefficient to obtain the optimal control amount; the clock error sequence to be adjusted is obtained according to the clock error data corresponding to the future moment in the prediction interval, and the phase of the clock error sequence to be adjusted is adjusted according to the optimal control amount at the current steering moment and the time interval between two adjacent clock errors to obtain the adjusted clock error sequence, and the clock error sequence is updated according to the time sequence iteration to realize satellite clock steering. The embodiment of the present invention has simple calculation and low cost, can reduce clock error fluctuations, improve frequency accuracy and frequency stability, and enhance the performance of satellite-borne secondary frequency standards in complex environments. Realize the real-time, high-precision and low-cost requirements of low-orbit satellite time synchronization.

[0055] In one embodiment, the steps of constructing a high-precision clock model include: constructing a trend term model based on the clock error data of the previous moment; obtaining historical clock error data, removing the trend term from the historical clock error data according to the trend term model, and obtaining the de-trended historical clock error data; performing spectral analysis on the de-trended historical clock error data, extracting the frequency using the fast Fourier transform method, estimating the amplitude and phase of the frequency using least squares fitting, and constructing a periodic term model; removing the periodic term from the de-trended historical clock error data according to the periodic term model, and obtaining the corresponding historical residual data; obtaining a pre-trained residual prediction model, inputting the historical residual data and the historical driving adjustment amount into the pre-trained residual prediction model, and obtaining the residual data of the future moment, and constructing an uncertainty term model according to the predicted residual data; and constructing a high-precision clock model according to the trend term model, the periodic term model, and the uncertainty term model.

[0056] In this embodiment, the model prediction algorithm is very sensitive to the model when it is controlling. Affected by the complex environment, its clock error model is complex and uncertain. Directly applying the model control algorithm to clock control will encounter problems such as unclear modeling and wrong control. The present invention re-models the clock error based on the clock error data and separates the clock error into three parts: trend term, period term, and uncertainty term caused by the complex environment in which the satellite is located, including noise, interference, temperature, ionosphere, etc.

[0057] Specifically, a clock trend term model is constructed based on the initial clock error, frequency deviation, frequency drift, power-law spectrum noise function, and clock error data of the previous moment. Based on historical data, the trend term is removed, and the clock error data is subjected to FFT spectrum analysis. Then, based on the least squares estimation, its amplitude and phase are estimated and aligned with the residual data. On this basis, the periodic term is further removed to obtain residual data containing only uncertain terms after removing the trend term and the periodic term. The residual data that cannot be described in the state space is predicted using a pre-trained residual prediction model. The residual prediction model can use a model architecture such as LSTM or Trasformer, whose input is historical residual data and historical steering adjustments, and the output is the clock error residual caused by the uncertain terms for a period of time in the future.

[0058] The high-precision clock model is:

[0059] ;

[0060] in, represents the current clock error, Represents the existing clock error at the previous moment, represents the frequency deviation, represents the frequency drift, represents the periodic term. The state of the periodic term is different in different orbits. Represents the uncertain term, It represents the superposition of five types of power-law spectrum noise that change with time, including phase white noise, phase flicker noise, frequency white noise, frequency flicker noise, and random walk noise. The size of the noise impact can be determined according to different clock types. , and According to these parameters, the clock is modeled and the actual clock error data is generated. The method of the present invention uses the generated rubidium clock crystal oscillator and two types of clock errors as input and uses an algorithm to control and adjust.

[0061] In one embodiment, state estimation is performed based on clock difference data to obtain the clock state at each moment in the prediction interval, including: traversing the prediction interval, using a Kalman filter algorithm to process the clock difference data of the previous moment of the current traversal moment to obtain the clock state at the current traversal moment, until the traversal of each moment in the prediction interval is completed, stopping the traversal, and outputting the clock state at each moment in the prediction interval.

[0062] In this embodiment, the onboard clock control has a high requirement for the real-time performance of the algorithm. In order to meet the real-time performance requirement of the clock control, the method of the present invention uses the Kalman algorithm to predict the short-term clock error based on the clock error obtained after comparing the historical clock signals. The Kalman algorithm can control the different types of clocks generated above according to the clock theory. Perform real-time estimation processing as follows:

[0063] ;

[0064] in, , , is the state transfer matrix, , is the clock error noise. Using the above formula, we can estimate the clock state at each moment.

[0065] In one embodiment, constructing a quadratic cost function according to the clock state at each moment in the prediction interval and the weight matrix includes: constructing a quadratic cost function according to the clock state at each moment in the prediction interval and the weight matrix as follows:

[0066] ;

[0067] ;

[0068] ;

[0069] in, is a quadratic cost function, for The prediction interval at the moment The clock state at a moment, is the terminal state weight matrix, for The state error weight matrix at time , for The control weight matrix at the moment, for The prediction interval at the moment The amount of control at a moment, , is the length of the prediction interval, and the prediction interval is , Represents a transpose operation.

[0070] In this embodiment, when the control quantity is solved by using the model prediction algorithm, the optimal control quantity is found by minimizing the quadratic loss function. Then, the optimal control quantity at the current moment is calculated by using rolling optimization. Start driving at any time, the initial state is , within the prediction interval, based on the cost function, the optimal control sequence is calculated , then discard except The control amount at other times other than Moment, repeat Time operation will be performed after the last time After the effect As The initial state at time t, the prediction interval is from Start by calculating a new set of , also discard except The remaining control quantities are repeated as time goes by to calculate the optimal control quantity at each moment.

[0071] Satellite clock steering also has requirements for the robustness of the method. The model prediction algorithm adopts a rolling optimization method based on the prediction interval. At the same time, constraints can be added directly before calculating the adjustment amount, and then the steering adjustment amount is calculated. Due to the existence of rolling optimization, the algorithm will calculate the adjustment amount based on this larger input value for the next period of time when abnormal jumps and other problems occur. This is equivalent to averaging the impact of the error to the next period of time for adjustment. However, when it reaches the next moment, the adjustment amount will be recalculated instead of using the previous moment, which has a robustness effect and thus ensures the correctness of the steering.

[0072] like Figure 3 The flowchart of the MPC-based clock control algorithm is shown in Figure 3In the calculation, the control matrix is ​​converted into a standard quadratic solution mode based on the system matrix and the weight matrix. The optimal clock adjustment amount is calculated by minimizing the quadratic loss function. Each adjustment amount is calculated based on the solved clock error signal, and the optimal adjustment amount that meets the comprehensive consideration of the expected control target value, the control amount, and the clock error fluctuation during the process is calculated in real time.

[0073] In one embodiment, the control amount is optimized by using the differential control amount and the differential coefficient to obtain the optimal control amount, which includes: obtaining the differential control amount according to the clock state change rate, weighting the differential control amount by using the differential coefficient and then fusing it with the control amount, and optimizing the differential coefficient by using the dichotomy method, and obtaining the optimal control amount according to the optimization result as follows:

[0074] ;

[0075] in, is the optimal control quantity, To control the amount, is the differential coefficient, is the differential control quantity.

[0076] In this embodiment, a differential term is introduced, and the optimal control quantity obtained in the previous step is optimized and updated. Since the clock itself has a large change problem during the clock control process, differential control is added on the basis of the original MPC control theory, and a weight coefficient is set, which is weighted with the control quantity calculated by the classical model prediction algorithm, and finally a better control value is obtained. Specifically, after determining the Q, R, S matrix and other parameters in MPC control, the parameter is adjusted from 0, and the parameter is increased to a value that is worse than when the differential term is not added, as the maximum value, and based on the dichotomy, the differential term coefficient is adjusted. Conduct optimization.

[0077] In one embodiment, the phase of the clock difference sequence to be adjusted is adjusted according to the new optimal control amount at the current driving moment and the time interval between two adjacent clock differences, and the adjusted clock difference sequence is obtained, which includes: the phase of the clock difference sequence to be adjusted is adjusted according to the optimal control amount at the current driving moment and the time interval between two adjacent clock differences, and the adjusted clock difference sequence is obtained as follows:

[0078] ;

[0079] in, is the adjusted clock error sequence, is the clock error sequence to be adjusted, for The optimal control quantity at the moment, is the time interval between two adjacent clock difference data sampling moments, is the length of the prediction interval.

[0080] In this embodiment, the optimal control amount obtained in the above steps is the frequency control amount. When verifying the simulation experiment, it needs to be converted into a phase control amount to adjust the current clock error and subsequent clock errors. Specifically, the corresponding frequency adjustment amount is accumulated and summed according to time, and the corresponding phase adjustment amount calculated is added to the corresponding clock error data. In this way, the frequency adjustment amount is applied to the simulated phase data.

[0081] In a specific embodiment, the key issues are modeled and analyzed theoretically, and then the analysis and verification are carried out by building a simulation environment. Finally, the theoretical analysis and simulation results are verified by measured data, and the model and algorithm are adjusted according to the measured data analysis and verification results. The overall technical route is verified by theoretical analysis and simulation verification, combined with actual system verification. By comparing the differences between medium and high orbits and low orbits, such as the small relativistic effect of high-orbit satellites, and the differences between low orbits and the ground, such as the different ionospheric effects caused by the locations, these two angles are used to compare and analyze the time effects caused by the differences between the two categories, and then based on these low-orbit measured data, a time series analysis is performed on them, and the characteristics of the data analysis and the characteristics of the theoretical influencing factors are compared to determine the influencing factors of time synchronization in the low-orbit environment one by one. For the clock driving algorithm, especially the clock driving algorithm for the secondary frequency standard, through a relatively comprehensive understanding of the secondary frequency standard driving algorithm and the low-orbit clock driving method at home and abroad, combined with the content of optimal control theory, the applicability of the model-based prediction algorithm to the clock model in the clock field is analyzed, and the applicability of the algorithm considering various low-orbit influencing factors in the low-orbit case is verified through simulation experiments. Figure 4 The clock error simulation result diagram is shown in Figure 1, where Figure 4 (a) is a schematic diagram showing the comparison of theoretical and simulated Allan deviation curves when the rubidium clock is applied to the method of the present invention. Figure 4 (b) is a schematic diagram comparing the theoretical and simulated Allan deviation curves when the crystal oscillator is applied to the method of the present invention, wherein the theoretical Allan method represents the theoretical Allan variance calculation curve, the overlapping Allan method represents the result curve of the overlapping Allan variance calculation of the simulated clock error sequence, and the modified Allan method represents the result curve of the modified Allan variance calculation of the simulated clock error sequence. Figure 5 The schematic diagram of the clock driving frequency stability results under different algorithms is shown in the figure, where: Figure 5 (a) is a schematic diagram of the crystal oscillator driving frequency stability results under different algorithms. Figure 5(b) is a schematic diagram of the rubidium clock driving frequency stability results under different algorithms, I, J, K, L and M respectively represent the crystal oscillator driving frequency stability results when not driven and the crystal oscillator driving frequency stability results using PID, PP (Proportional-Plus-Controller), LQR (Linear Quadratic Regulator) and the algorithm of the present invention, N, O, P, Q and R respectively represent the rubidium clock driving frequency stability results when not driven and the rubidium clock driving frequency stability results using PID, PP, LQR and the algorithm of the present invention. The above simulation results prove that the simulated clock driving is in line with the actual control process and fits the driving scenario, so that the driving method can be accurately verified.

[0082] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0083] In one embodiment, Figure 6 As shown, a satellite-borne clock driving device based on model prediction is provided, comprising:

[0084] The state estimation module 602 is used to perform state estimation based on the pre-built high-precision clock model and historical clock error data to obtain the clock state at each moment in the prediction interval; the lower limit of the prediction interval is the current driving moment, and the upper limit is the rolling length;

[0085] The control amount optimization module 604 is used to construct a quadratic cost function according to the clock state and the weight matrix at each moment in the prediction interval, minimize the quadratic cost function using a rolling optimization method, output the control amount at the current driving moment, and optimize the control amount using a differential control amount and a differential coefficient to obtain an optimal control amount;

[0086] The clock control module 606 is used to obtain the clock error sequence to be adjusted according to the clock error data corresponding to the future time in the prediction interval, and to perform phase adjustment on the clock error sequence to be adjusted according to the optimal control amount at the current control time and the time interval between two adjacent clock errors to obtain the adjusted clock error sequence, and to iteratively update the clock error sequence according to the time sequence to realize the onboard clock control.

[0087] For the specific limitations of the satellite-borne clock control device based on model prediction, please refer to the limitations of the satellite-borne clock control method based on model prediction above, which will not be repeated here. Each module in the above-mentioned satellite-borne clock control device based on model prediction can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0088] In one embodiment, Figure 7 As shown, a satellite-borne clock driving system based on model prediction is provided, including a host computer, a satellite-borne clock, a digital-to-analog converter and a voltage-controlled crystal oscillator;

[0089] The host computer is used to perform state estimation based on the pre-built high-precision clock model and historical clock error data, obtain the clock state at each moment in the prediction interval, construct a quadratic cost function based on the clock state at each moment in the prediction interval and the weight matrix, use the rolling optimization method to minimize the quadratic cost function, output the control amount at the current driving moment, optimize the control amount using the differential control amount and differential coefficient, and obtain the optimal control amount; the lower limit of the prediction interval is the current driving moment, and the upper limit is the rolling length;

[0090] The digital-to-analog converter is used to convert the optimal control quantity output by the host computer into an analog voltage signal and output it to the voltage-controlled crystal oscillator;

[0091] The voltage-controlled crystal oscillator is used to generate the corresponding optimal frequency adjustment amount according to the analog voltage signal;

[0092] The onboard clock is used to accumulate the optimal frequency adjustment amount over time, convert it into a clock error adjustment amount, and perform clock control based on the clock error adjustment amount.

[0093] The onboard clock control system calculates the optimal adjustment amount through the onboard clock control algorithm, uses DAC to calculate the corresponding voltage control amount, and then uses the voltage-controlled crystal oscillator to generate the corresponding optimal frequency adjustment amount. During simulation implementation, both DAC and voltage-controlled crystal oscillator are converted into system transfer functions, and then through time accumulation, the optimal frequency adjustment amount is converted into the optimal clock difference adjustment amount to realize the construction of the control system, thereby making the algorithm implementation consistent with reality.

[0094] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0095] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A satellite-borne clock driving method based on model prediction, characterized in that: The method comprises: The state is estimated based on the pre-built high-precision clock model and historical clock error data to obtain the clock state at each moment in the prediction interval; the lower limit of the prediction interval is the current driving moment, and the upper limit is the rolling length; A quadratic cost function is constructed according to the clock state and the weight matrix at each moment in the prediction interval, the quadratic cost function is minimized by numerical solution, a preliminary control amount at the current driving moment is output, the control amount is optimized by using the differential control amount and the differential coefficient to obtain the optimal control amount, and the optimal control amount at each moment is gradually calculated based on rolling optimization; The clock error sequence to be adjusted is obtained according to the clock error data corresponding to the future time in the prediction interval. The phase of the clock error sequence to be adjusted is adjusted according to the optimal control amount of the current steering moment and the time interval between two adjacent clock errors to obtain the adjusted clock error sequence. The clock error sequence is updated iteratively according to the time sequence to realize the satellite clock steering.

2. The method according to claim 1, characterized in that The steps to build a high-precision clock model include: Construct a trend term model based on the clock difference data of the previous moment; Acquire historical clock error data, and remove trend items from the historical clock error data according to the trend item model to obtain detrended historical clock error data; The frequency spectrum of the detrended historical clock error data is analyzed, the frequency is extracted using the fast Fourier transform method, the amplitude and phase of the frequency are estimated using the least squares fitting method, and the periodic term model is constructed. Removing the periodic term from the detrended historical clock error data according to the periodic term model to obtain corresponding historical residual data; Obtaining a pre-trained residual prediction model, inputting historical residual data and historical driving adjustment amounts into the pre-trained residual prediction model to obtain residual data at a future time, and constructing an uncertainty model based on the predicted residual data; A high-precision clock model is constructed according to the trend item model, the period item model and the uncertainty item model.

3. The method according to claim 1, characterized in that The high-precision clock model is: in, represents the current clock error, Represents the existing clock error at the previous moment, represents the frequency deviation, represents the frequency drift, represents the periodic term, Represents the uncertainty, Represents the superposition of five power-law spectrum noises that vary with time.

4. The method according to claim 1, characterized in that Based on the pre-built high-precision clock model and historical clock error data, the state estimation is performed to obtain the clock state at each moment in the prediction interval, including: Traverse the prediction interval, use the Kalman filter algorithm to process the clock difference data of the previous moment of the current traversal, and obtain the clock state of the current traversal moment based on the high-precision clock model. When the traversal of each moment in the prediction interval is completed, stop the traversal and output the clock state of each moment in the prediction interval.

5. The method according to claim 1, characterized in that Constructing a quadratic cost function according to the clock state at each moment in the prediction interval and the weight matrix includes: The quadratic cost function is constructed according to the clock state at each moment in the prediction interval and the weight matrix: ； ； ； in, is a quadratic cost function, for The prediction interval at the moment The clock state at a moment, is the terminal state weight matrix, for The state error weight matrix at time , for The control weight matrix at the moment, for The prediction interval at the moment The amount of control at a moment, , is the length of the prediction interval, and the prediction interval is , Represents a transpose operation.

6. The method according to claim 1, characterized in that The method of optimizing the control quantity by using the differential control quantity and the differential coefficient to obtain the optimal control quantity includes: The differential control quantity is obtained according to the clock state change rate, and the differential control quantity is weighted by the differential coefficient and then integrated with the control quantity. The differential coefficient is optimized by the dichotomy method, and the optimal control quantity is obtained according to the optimization result: in, is the optimal control quantity, To control the amount, is the differential coefficient, is the differential control quantity.

7. The method according to claim 1, characterized in that According to the new optimal control quantity at the current driving time and the time interval between two adjacent clock errors, the phase of the clock error sequence to be adjusted is adjusted, and the adjusted clock error sequence includes: According to the optimal control quantity of the current driving time and the time interval between two adjacent clock errors, the phase of the clock error sequence to be adjusted is adjusted, and the adjusted clock error sequence is obtained as follows: ； in, is the adjusted clock error sequence, is the clock error sequence to be adjusted, for The optimal control quantity at the moment, is the time interval between two adjacent clock differences, is the length of the prediction interval.

8. A satellite-borne clock driving device based on model prediction, characterized in that: The device comprises: The state estimation module is used to perform state estimation based on the pre-built high-precision clock model and historical clock error data to obtain the clock state at each moment in the prediction interval; the lower limit of the prediction interval is the current driving moment, and the upper limit is the rolling length; A control quantity optimization module is used to construct a quadratic cost function according to the clock state at each moment in the prediction interval and the weight matrix, minimize the quadratic cost function using a numerical solution method, output the control quantity at the current driving moment, and optimize the control quantity using a differential control quantity and a differential coefficient to obtain an optimal control quantity; The clock control module is used to obtain the clock error sequence to be adjusted according to the clock error data corresponding to the future time in the prediction interval, and to adjust the phase of the clock error sequence to be adjusted according to the optimal control amount at the current control time and the time interval between two adjacent clock errors to obtain the adjusted clock error sequence, and to iteratively update the clock error sequence according to the time sequence to realize the on-board clock control.

9. A system applied to the on-board clock driving method based on model prediction according to any one of claims 1 to 7, characterized in that: The system includes a host computer, a satellite-borne clock, a digital-to-analog converter and a voltage-controlled crystal oscillator; The host computer performs state estimation according to the clock difference data and the high-precision clock model to obtain the clock state at each moment in the prediction interval, constructs a quadratic cost function according to the clock state at each moment in the prediction interval and a weight matrix, minimizes the quadratic cost function using a numerical solution method, outputs the control amount at the current driving moment, optimizes the control amount using a differential control amount and a differential coefficient, and obtains an optimal control amount; the lower limit of the prediction interval is the current driving moment, and the upper limit is the rolling length; The digital-to-analog converter is used to convert the optimal control quantity output by the host computer into an analog voltage signal, and output it to the voltage-controlled crystal oscillator; The voltage-controlled crystal oscillator is used to generate a corresponding optimal frequency adjustment value according to the analog voltage signal; The satellite-borne clock is used to accumulate the optimal frequency adjustment amount over time, convert it into a clock difference adjustment amount, and perform clock control according to the clock difference adjustment amount.

Images