Two-way prediction intelligent handover method and system for deep space narrowband doppler measurement long-term interruption data

Abstract

The application provides a bidirectional prediction intelligent connection method and system for long-term interruption data of deep space narrowband Doppler measurement, which comprises the following steps: S1, channel decomposition and aggregation of bidirectional prediction data before and after interruption based on time-frequency domain multi-angle characteristics; S2, one-way data modeling and prediction based on multi-channel decomposition and aggregation; S3, variable weight fusion of bidirectional prediction data based on confidence interval and frequency domain decomposition; and S4, polynomial fitting and prediction curve smoothing of integrated data. The application performs adaptive decomposition and channel aggregation on the data segments before and after long-term interruption of deep space exploration based on the characteristics of the data segments, and respectively performs modeling and prediction on the channel data from the positive direction and the reverse direction, and performs data fusion and trend smoothing under the guidance of the confidence interval, so that the interpolation accuracy is significantly higher than that of the traditional one-way extrapolation fitting method, and the application has good generality and stability, especially in the long-term deep space data interruption state.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of deep space exploration and intelligent signal processing, and particularly relates to a bidirectional prediction intelligent connection method and system for long-term interruption data of deep space narrowband Doppler measurement. BACKGROUND

[0002] Continuous tracking of the phase and Doppler frequency changes of the downlink narrowband signal of a deep space probe is one of the important tasks of a ground observation station. Changes in the phase, frequency and other parameters of the downlink signal help to calculate the position, velocity and other motion parameters of the deep space probe, understand the subtle changes in the deep space environment around the probe, and help to solve the navigation and orbit determination of the deep space probe and the research of planetary geodesy and related space science.

[0003] However, due to the complexity of the deep space observation environment and the diversity requirements of the task itself, such as signal blocking in a short time, strong radio noise interference or target observation alternation from the task requirements, the ground station equipment inevitably appears to be interrupted for a period of time during the tracking of the downlink signal of the probe. Frequent interruptions significantly reduce the availability of collected data, and at the same time pose a threat to deep space missions and increase the uncertainty in the execution of observation tasks. These difficult-to-eliminate interruptions pose technical challenges to current deep space exploration tasks. The current technology such as fitting extrapolation has problems such as poor stability and poor generalization in predicting and connecting interrupted data, and the connection accuracy needs to be further improved. SUMMARY

[0004] The present application aims to provide a bidirectional prediction intelligent connection method and system for long-term interruption data of deep space narrowband Doppler measurement, which intelligently connects long-term interrupted observation data with high precision due to continuous strong noise interference, task observation requirements and other reasons, improves the robustness of the task and the availability of collected data.

[0005] To solve the above technical problems, the present application provides a bidirectional prediction intelligent connection method and system for long-term interruption data of deep space narrowband Doppler measurement, which can simultaneously perform high-precision modeling from both positive and negative directions according to the data characteristics before and after the interruption when a long-term data observation interruption is detected, and perform high-precision connection of the missing data through bidirectional intelligent prediction, data fusion and fitting smoothing. The specific technical solutions are as follows:

[0006] A bidirectional prediction intelligent connection method for long-term interruption data of deep space narrowband Doppler measurement, comprising the following steps:

[0007] S1, bidirectional prediction of data before and after interruption based on channel decomposition and aggregation of time-frequency domain multi-angle characteristics;

[0008] S2, one-way data modeling and prediction based on step S1 multi-channel decomposition;

[0009] S3, bidirectional prediction data variable weight fusion based on confidence interval and frequency domain decomposition;

[0010] S4, polynomial fitting and prediction curve smoothing of integrated data.

[0011] Preferably, in step S1, the channel decomposition based on time-frequency domain multi-angle features is specifically:

[0012] The overall decomposition process is divided into two levels.

[0013] First, select the time backtracking window size, and select the data segment before and after the interruption as the network input data for prediction and continuation; for the data in the selected backtracking window, design the STL decomposition mode to separate the trend item and the periodic item.

[0014] Second, for the periodic item after STL decomposition, convert it to the frequency domain through discrete Fourier transform, and divide it into high, medium and low frequency bands in the frequency spectrum, and then set up independent data channels.

[0015] Third, for each independent data channel designed, the high, medium and low frequency data are converted to the time domain through inverse Fourier transform for channel aggregation, completing the channel separation and analysis of the periodic item part of the input data.

[0016] Preferably, in step S2, the one-way data modeling and prediction based on multi-channel decomposition is specifically:

[0017] S21, simple network modeling of the trend item part: the trend item part of the input data after STL decomposition is modeled and predicted by GRBF network. The network output uses the iterative input prediction method, that is, The data pred(t-1) obtained by prediction at time t-1 will be used as the input for the next time prediction.

[0018] S22, multi-channel aggregation and network analysis modeling of the periodic item part: the periodic item part of the data obtained by STL decomposition is converted to the frequency domain after discrete Fourier transform, and the entire normalized frequency spectrum is divided into high frequency [0.7-1], medium frequency [0.4-0.7] and low frequency [0-0.4] three intervals, that is, the input periodic item data is divided into three independent channels of high, medium and low frequency according to the spectrum characteristics. Then, after converting the high, medium and low frequency data information back to the time domain through inverse discrete Fourier transform (IDFT), an LSTM model is established for each independent channel data for separate prediction.

[0019] S23, and the total data output of the periodic term part is determined by the weighted fusion of the outputs of the individual channels.

[0020] S24, when detecting that the set time data observation interruption occurs, simultaneously performing forward and reverse one-way prediction; wherein, the total output result d F (t) of the forward and reverse one-way prediction is determined by the addition of the periodic term prediction result and the trend term prediction result of the respective prediction direction. I (t) is determined by the addition of the periodic term prediction result and the trend term prediction result of the respective prediction direction.

[0021] Preferably, in step S3, the variable weight fusion of the bidirectional prediction data based on the confidence interval is specifically:

[0022] First, based on the calculation of the forward and reverse data prediction, the confidence interval of the forward data prediction and the reverse data prediction is calculated respectively.

[0023] Secondly, the data segment with the intersection of the bidirectional prediction confidence interval is found to carry out data fusion. The data fusion weight is according to the area ratio of each other in the intersection interval, that is, d'(t) = α × d F (t) + (1-α) × d I (t), Wherein, S F and S I are the areas occupied by the forward prediction and the reverse prediction in the intersection of the confidence interval.

[0024] Preferably, in step S4, the polynomial fitting and the smoothing of the prediction curve are specifically:

[0025] First, integrate the prediction data in multiple segments. Specifically, the prediction data used for connection is composed of the following three parts, respectively: 1) the data before the confidence interval intersection adopts the forward prediction result d F (t), 2) the data after the confidence interval intersection adopts the reverse prediction result d I (t), and 3) the part of the intersection adopts the fused data result d'(t). The above three data segments are integrated as the preliminary result of data connection.

[0026] Secondly, the polynomial fitting is applied to the above three data segments, and the smoothing processing is carried out to obtain the final connection data segment.

[0027] The application also discloses a bidirectional prediction intelligent connection system for long-term interruption data of deep space narrowband Doppler measurement, which is used for executing the above method, and comprises the following modules:

[0028] Channel decomposition module: channel decomposition and aggregation of bidirectional prediction data before and after interruption based on time-frequency domain multi-angle characteristics;

[0029] Unidirectional data modeling and prediction module; unidirectional data modeling and prediction for multi-channel decomposition;

[0030] Data fusion module: bidirectional prediction data variable weight fusion based on confidence interval and frequency domain decomposition;

[0031] Continuation data acquisition module: polynomial fitting and prediction curve smoothing of integrated data.

[0032] The application provides a bidirectional prediction intelligent continuation method and system for long-term interrupted data of deep space narrowband Doppler measurement, which realizes high-precision unidirectional prediction by adaptively decomposing data in a backtracking window and setting up independent channels, and designs a data fusion scheme based on a confidence interval for forward and reverse unidirectional prediction data, thereby successfully suppressing error accumulation in the multi-step prediction continuation process. The application has the advantages of high continuation precision and good generalization in the continuation of long-term interrupted data. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 A flow chart of a bidirectional prediction intelligent continuation method for long-term interrupted data of deep space narrowband Doppler measurement is provided for the preferred embodiment of the application.

[0034] Figure 2 A multi-level adaptive decomposition and channel aggregation diagram of input data is provided for the preferred embodiment of the application.

[0035] Figure 3 An input data independent channel modeling and prediction diagram is provided for the preferred embodiment of the application.

[0036] Figure 4a A single-channel long short-term memory artificial neural network (LSTM) modeling based on frequency domain decomposition and periodic term data prediction and fusion diagram is provided for the preferred embodiment of the application.

[0037] Figure 5a A gradient radial basis function neural network (GRBF) modeling and iterative prediction diagram of trend term data is provided for the preferred embodiment of the application.

[0038] Figure 6 A bidirectional data fusion and data integration diagram based on a confidence interval is provided for the preferred embodiment of the application.

[0039] Figure 7 A cumulative error suppression effect diagram of the bidirectional prediction intelligent continuation method for long-term interrupted data prediction is provided for the preferred embodiment of the application. Figure 8A block diagram of a bidirectional prediction intelligent splicing system for long-term interruption data of deep space narrowband Doppler measurement for a preferred embodiment of the present application. DETAILED DESCRIPTION

[0040] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0041] As shown in the figure, the embodiment of the present application is a bidirectional prediction intelligent splicing method for long-term interruption data of deep space narrowband Doppler measurement, and the specific implementation steps are as follows: Figure 1

[0042] S1, channel decomposition of bidirectional prediction data before and after interruption based on time-frequency domain multi-angle features;

[0043] S2, one-way data modeling and prediction based on the decomposition mechanism;

[0044] S3, variable weight fusion of bidirectional prediction data based on confidence interval and frequency domain decomposition;

[0045] S4, polynomial fitting and prediction curve smoothing of integrated data.

[0046] In step S1, in the data preprocessing and modeling process, after detecting the occurrence of long-time data interruption, the backtracking window width is first determined. And on this basis, a section of data is selected on both sides of the data to be spliced as network input.

[0047] Secondly, the multi-level adaptive feature decomposition is performed on the bidirectional input data to divide the input data channels. The first level is STL decomposition, which decomposes the input data into trend items and periodic items. The second level decomposition converts the relatively complex features in the periodic items to the frequency domain, and according to the different performances of the periodic frequencies, the adaptive high-frequency, medium-frequency and low-frequency channels are divided.

[0048] Finally, the high, medium and low frequency data are converted to the frequency domain again through the inverse discrete Fourier transform, and the channel separation and aggregation of the periodic item part of the input data are completed.

[0049] In step S2, after the adaptive decomposition and channel aggregation are completed, the one-way prediction is performed on each independent channel by designing a prediction network. Among them, the weights of the high, medium and low frequency channels are determined according to the amplitude ratio of each in the frequency spectrum when the data is fused.

[0050] Specifically:​

[0051] S21, for the decomposed trend term, a simple model based on GRBF is designed to model and iteratively predict it;

[0052] S22, for the input data periodic term part, LSTM network is designed to model and iteratively predict it according to high, medium and low frequency channels;

[0053] S23, the overall output of the input data periodic term part is obtained by weighting the prediction results of the high, medium and low frequency channels, and the weight of each channel is determined by the amplitude ratio of the high, medium and low frequency components in the frequency spectrum;

[0054] S24, when the set time data observation interruption is detected, forward and reverse one-way prediction is performed simultaneously; wherein, the output results d F (t) and d I (t) are determined by adding the periodic term prediction results and the trend term prediction results of each prediction direction.

[0055] Step S3, after completing the one-way prediction of data, the confidence interval of each prediction direction is calculated. For the intersection part of the interval, the weighted fusion is performed according to the area ratio of the confidence interval of each direction in the intersection. Specifically:

[0056] S31, after the one-way prediction is completed, the confidence interval of the forward data prediction and the reverse data prediction is calculated respectively;

[0057] S32, the intersection part of the forward and reverse data prediction confidence intervals is taken for data fusion, and the fused data is used to calibrate the prediction data, and the fusion rule is d'(t) = α × d F (t) + (1-α) × d I (t); wherein, the fusion factor S F and S I are the areas occupied by the forward prediction and the reverse prediction in the confidence interval intersection.

[0058] Step S4, the forward prediction data, the bidirectional interval fusion data and the reverse prediction data are integrated to form a rough connection data segment. The rough connection data segment is polynomial fitted to smooth its numerical trend, and it is used as the final connection data segment. Specifically:

[0059] S41, prediction data integration; the prediction data used for connection is composed of the following three parts, which are: 1) the data before the confidence interval intersection is composed of the forward prediction d F (t); 2) the confidence interval intersection part is composed of the fusion data d'(t); 3) the data after the confidence interval intersection is composed of the reverse prediction d I(t) constitute;

[0060] S42, apply polynomial fitting to the three pieces of data arranged in step S41 to smooth the curve, and output the result as the final connection data.

[0061] For step S1, Figure 2 Multi-level adaptive decomposition and channel aggregation of input data are described.

[0062] Among them, the first level decomposition is STL decomposition as shown in formula (1), (2), which is to divide the input data into trend term X Trend and periodic term X Period . Among them, Avgpool represents the average pooling operation on the input data, and Padding represents the padding operation on the input data sequence X in the pooling process. x is a specific value of the input data sequence, N is the length of the average pooling data, x k is a specific data in the pooling sequence.

[0063]

[0064] X period = X-X trend (2)

[0065] For the periodic part of the data, because of its complex characteristics, a secondary frequency domain channel decomposition mechanism is designed, and the specific steps are as follows:

[0066] First, the periodic part is converted to the frequency domain by the discrete Fourier transform of the N-point sequence as shown in formula (3). Among them, X[k] is the corresponding frequency domain sequence value, N DFT is the Fourier transform point number, x w [n] is the corresponding Fourier transform coefficient, e is the natural logarithm base, j is the imaginary unit, n is the time domain sequence number, and k is the frequency domain sequence number.

[0067]

[0068] Second, normalize the entire transformed spectrum and divide the frequency domain interval, where low frequency [0~0.4], medium frequency [0.4~0.7], and high frequency [0.7~1]. According to this division interval, the data in the frequency domain range is aggregated, and it is converted back to the time domain through inverse Fourier transform, thereby dividing the data channel. The specific mathematical expression is shown in formula (4), where D HF ,D MF ,D LF respectively represent high, medium and low frequency data channels.

[0069] D HF = IDFT{Spec[0.7:1]}, DMF = IDFT {Spec [0.4:0.7]}, D LF = X Peridical - D HF - D MF (4)

[0070] For step S2, Figure 3 Modeling and prediction for independent channels of input data. After adaptive decomposition and channel aggregation of input data, modeling is performed on each independent channel data.

[0071] Among them, for the periodic term data, its channel adaptive decomposition, aggregation and independent prediction process is as shown in Figure 4a The LSTM network as shown in Figure 4b is designed for each high, medium and low frequency independent channel after the periodic term is decomposed again to perform iterative multi-step prediction. At the output end, the weighted fusion is performed according to the amplitude ratio of each component in the frequency spectrum.

[0072] Among them, for the trend term data, the GRBF network as shown in Figure 5a is designed for iterative prediction. In the GRBF network, in order to better perceive the gradient change of the input data and deduce the data change trend at the future time point, a difference layer is added before the overall GRBF network for differential processing of the input data. The data prediction rule is established as shown in equation (5), and the iterative prediction method is used in the data prediction process, that is, the data pred(t-1) obtained by prediction at t-1 time is used as the input of the next time prediction data.

[0073]

[0074] Among them, is the corresponding relationship between the output trend term result and the input sequence, and n is the network backtracking window width. The overall output of one-way prediction is obtained by adding the periodic term prediction result and the trend term prediction result.

[0075] Figure 6 is the bidirectional data fusion and data integration based on confidence interval. For step S4, after the forward and reverse one-way data prediction is completed, the confidence interval of each is calculated. Find the part where the confidence intervals intersect, and integrate the data into three parts, namely: 1) the data segment before the confidence interval intersects, 2) the data segment where the confidence intervals intersect, and 3) the data segment after the confidence interval intersects.

[0076] For the data segment before the confidence interval intersects, the forward prediction data is used because the forward prediction is more reliable. For the data segment where the confidence intervals intersect, the weighted fusion is performed according to the proportion of the area of each confidence interval in the intersection.

[0077] The fusion rule is shown as formula (6),

[0078] d'(t)=α×d F (t)+(1-α)×d I (t) (6)

[0079] Wherein, the fusion factor alpha is calculated as formula (7),

[0080]

[0081] Wherein, S F And S I The area occupied by the forward prediction and the backward prediction in the intersection of the confidence interval.

[0082] Figure 7 For typical satellite narrowband Doppler frequency measurement after interruption, the cumulative error of the bidirectional prediction intelligent connection method and the conventional one-way fitting extrapolation under different connection step conditions is compared. From the results, under the conditions of 60, 90, 150, 300 steps, compared with the conventional one-way fitting extrapolation method, the proposed bidirectional prediction intelligent connection method improves the cumulative error by 4.73, 5.46, 8.85, and 9.29 times respectively. Therefore, the bidirectional prediction intelligent connection method for long-term interruption of deep space narrowband Doppler measurement data has the characteristics of high connection accuracy, stable performance and strong generalization ability when dealing with long-term data interruption connection tasks.

[0083] As Figure 8 shown, the embodiment discloses a bidirectional prediction intelligent connection system for long-term interruption of deep space narrowband Doppler measurement data, for executing the above method, comprising the following modules:

[0084] Channel decomposition module: channel decomposition and aggregation of bidirectional prediction data before and after interruption based on time-frequency domain multi-angle features;

[0085] One-way data modeling and prediction module; one-way data modeling and prediction after multi-channel decomposition;

[0086] Data fusion module: variable weight fusion of bidirectional prediction data based on confidence interval and frequency domain decomposition;

[0087] Connection data acquisition module: polynomial fitting and prediction curve smoothing of integrated data.

[0088] Other contents of the embodiment can refer to the above method embodiment.

[0089] The application can intelligently splice the deep space downlink narrowband Doppler track change data collected by the station equipment with high precision, which can significantly improve the reliability of the observation task. At the same time, it can help the ground station to recapture the information such as the position of the aircraft after the data interruption, improve the availability of the collected signal data, and has very important theoretical and practical significance for the smooth progress of the deep space exploration task. It is an effective technical compensation means under the extreme tracking conditions such as deep space extremely low signal-to-noise ratio.

[0090] In summary, the application adaptively decomposes and aggregates the data segments before and after the long-term interruption of deep space exploration based on their own characteristics, models and predicts each channel data from the positive and negative directions respectively, and carries out data fusion and trend smoothing under the guidance of the confidence interval. The interpolation accuracy is significantly higher than that of the traditional one-way extrapolation fitting method, especially in the long-term deep space data interruption state, the application has good generality and stability.

[0091] The above only describes the preferred embodiments of the application, and the protection scope of the application is not limited to the above embodiments. Any equivalent modifications or changes made by those skilled in the art according to the disclosed content of the application shall be included in the protection scope recited in the claims.

Claims

1. A two-way predictive intelligent handover method for deep space narrowband Doppler measurement long-term outage data, characterized in that, Comprising the following steps: S1, channel decomposition and aggregation of bidirectional prediction data before and after interruption based on time-frequency domain multi-angle features; S2, one-way data modeling and prediction based on multi-channel decomposition of step S1; this step is as follows: S21, for the decomposed trend term, a simple model based on GRBF is designed to model and iteratively predict it; S22, for the input data periodic term part, according to high, medium and low frequency three channels, LSTM network is designed respectively to model and iteratively predict; S23, the overall output of the input data periodic term part is obtained by weighting the prediction results of the high, medium and low frequency three channels, and the weight size of each is determined by the amplitude proportion of the high, medium and low frequency components in the frequency spectrum; S24, when detecting that the set time data observation interruption occurs, simultaneously performing forward and reverse one-way prediction; wherein, the forward and reverse one-way prediction output results d F (t) and d I (t) are determined by adding the periodic term prediction results and the trend term prediction results of respective prediction directions; S3, variable weight fusion of bidirectional prediction data based on confidence interval and frequency domain decomposition; this step is as follows: S31, after one-way prediction, the confidence interval of forward data prediction and reverse data prediction is calculated respectively; S32, taking the intersection part of the positive and reverse data prediction confidence intervals to perform data fusion, and using the fused data to calibrate the prediction data, and the fusion rule is ; wherein the fusion factor ; S F and S I is the area occupied by the positive prediction and the reverse prediction in the confidence interval intersection; S4, polynomial fitting and prediction curve smoothing of integrated data.

2. The two-way predictive intelligent handover method for deep space narrowband Doppler measurement long-term outage data according to claim 1, characterized in that, Step S1 is as follows: S11, select the time backtracking window size, and select the data segments before and after the interruption as network input data for prediction and continuation; S12, STL decomposition is performed on the input data to decompose it into periodic term and trend term; S13, for the periodic component, it is transformed to the frequency domain; based on the transformed frequency spectrum, it is divided into three channels of high frequency, medium frequency and low frequency; S14, after the high, medium and low frequency data are converted to the time domain through inverse discrete Fourier transform, the channel decomposition and aggregation of the input data periodic term part are completed.

3. The two-way predictive intelligent handover method for deep space narrowband Doppler measurement long-term outage data according to claim 2, characterized in that, In step S12, the STL decomposition is as follows: wherein X Trend is a trend term, X Period is a periodic term, Avgpool denotes an average pooling operation on the input data, Padding denotes a padding operation on the input data sequence X in the pooling process; x is a specific value of the input data sequence, N is the length of the average pooled data, x k is a specific data in the pooled sequence.

4. The two-way predictive intelligent handover method for deep space narrowband Doppler measurement long-term outage data according to claim 3, characterized in that, In step S13, the periodic term part is converted to the frequency domain by discrete Fourier transform, wherein X[k] is the number of frequency domain sequence values, N DFT is the number of Fourier transform points, x w [n] is the Fourier transform coefficient, e is the natural logarithm base, j is the imaginary unit, n is the time domain sequence index, k is the frequency domain sequence index.

5. The two-way predictive intelligent handover method for deep space narrowband Doppler measurement long-term outage data according to claim 1, characterized in that, Step S4 is as follows: S41, prediction data integration; the prediction data for the next time interval consists of three parts, respectively: 1) the data before the intersection of the confidence intervals is composed of the positive prediction d F (t) ; 2) the data at the intersection of the confidence intervals is composed of the fusion data d'(t) ; 3) the data after the intersection of the confidence intervals is composed of the negative prediction d I (t). S42, apply polynomial fitting to the three data segments to smooth the curve, and output the final continuation data.

6. A bidirectional predictive intelligent handover system for deep space narrowband Doppler measurement long-term outage data for performing the method according to any one of claims 1 to 5, characterized in that Comprising the following modules: Channel decomposition module: channel decomposition and aggregation of bidirectional prediction data before and after interruption based on time-frequency domain multi-angle features; One-way data modeling and prediction module; one-way data modeling and prediction after multi-channel decomposition; Data fusion module: variable weight fusion of bidirectional prediction data based on confidence interval and frequency domain decomposition; Continuation data acquisition module: polynomial fitting and prediction curve smoothing of integrated data.

Images