A method for interpreting the shape of busbar power variation curve of lunar probe

Through sliding window moving average, baseline estimation, sparse noise denoising processing and other technologies, the lunar probe bus power curve is decomposed, single-shot and coupling events are identified, and the rapid analysis problem of complex changes in the lunar probe bus power curve is solved, and the correct The depth and rapid interpretation of the multi-source coupled power curve support data analysis and testing of deep space exploration.

CN114021310BActive Publication Date: 2024-12-03BEIJING INST OF SPACECRAFT SYST ENG
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111221492.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-10-20
Publication Date
2024-12-03
Estimated Expiration
2041-10-20

AI Technical Summary

Technical Problem

The complex changes in the power curve of the busbar of the lunar probe are difficult to quickly analyze, and the existing technology is difficult to effectively identify the multi-source coupling effect, affecting the satellite's data interpretation and testing process.

Method used

Using the sliding window moving average analysis method, combined with baseline estimation and sparse denoising treatment, the power curve is decomposed into the long-term trend baseline and short-term change curve, and single-shot and coupling events are identified through multivariate statistical analysis, and nominal curve set is constructed for secondary Fusion realizes rapid interpretation of multi-source coupled power curves.

Benefits of technology

It realizes rapid and in-depth analysis of the power change curve of the busbar of the lunar probe, provides automatic retrieval and data cleaning methods, supports data interpretation and functional testing in deep space exploration, and improves data accuracy and efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114021310B_ABST
    Figure CN114021310B_ABST
Patent Text Reader

Abstract

The present invention provides a method for interpreting the shape of a busbar power change curve of a lunar probe. The method includes: sampling and processing, obtaining a parameter set representing the power state of the entire satellite and recording it as an output power data set of each power supply array; using a sliding window moving average analysis method to calculate the power data set of a channel within a selected time, and obtaining the moving average MA of each window; judging whether there is a fluctuation change within the window interval according to the comparison between the MA value and the threshold; for a case where the window data fluctuation exceeds the threshold, it is necessary to collect the corresponding event elements in the window data; judging whether the parameter is a single event or a coupling effect within the time window; performing secondary fusion on the nominal curve set of multiple events to obtain a coupled nominal curve P of the influence of multiple events on the power parameter. merge . The short-term change curve C and P merge Conduct correlation analysis.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of signal recognition and analysis of multi-source coupling characteristics, and in particular to a method for quickly judging the curve shape of a lunar probe bus power. Background Art

[0002] The unmanned lunar sample return mission is a typical representative of deep space exploration missions. Its energy supply system has a complex working mode, involving multiple devices connected to the grid and other modes. There are many devices in the whole device, and the working mode is difficult to analyze in a model. The bus power coupling is deep, and the power curve often shows obvious drastic changes, making it difficult to identify known power models. The power curve is also a crucial interpretation data for the satellite, which can objectively reflect whether the core components are operating within the interpretation envelope under the working mode. In order to ensure that the data can be fully and quickly analyzed and interpreted in the comprehensive test phase, a data interpretation method that matches the intelligent computing capability of the detector is needed to better support the conventional interpretation and functional interpretation during the satellite comprehensive test process. Summary of the invention

[0003] The technical problem solved by the present invention is: to overcome the shortcomings of the prior art and provide a method for quickly analyzing the complex change curve of spacecraft bus power, which utilizes the effective decomposition of the fusion characteristics of multi-source coupled power curves to facilitate the subsequent use of a pattern recognition classifier to trace the source of external excitation data.

[0004] The technical solution of the present invention is: a method for interpreting the shape of a busbar power variation curve of a lunar probe, the steps of which are as follows:

[0005] Step 1: By monitoring the output signal of the satellite ground power array equipment, the characterization parameter set of the satellite whole device power state is obtained and recorded as the output power data set of each power array {W1, W2...W t}, where W is the multiplication result of the voltage and current output by the ground power supply array device in real time, and t is the number of channels of the ground power supply array device; the sum of the moving averages of different parameters in the data set is calculated as the measurement value;

[0006] Step 2: Use the sliding window moving average analysis method to calculate the power data set of a channel within the selected time to obtain the moving average data MA; if the MA value exceeds the specified threshold, it is considered that there is a fluctuation change in the window interval, and enter step 3; otherwise, it is considered that the window data fluctuation has no large fluctuation change, and exit this method;

[0007] Step 3: Collect the corresponding event elements in the window data; define the event element M n ={t0,t f ,A,S,P}, including the event start time t0, end time t f, the target single device A, the event effect correlation parameter set S, and the nominal curve P of the event's impact on the power parameter;

[0008] Step 4: Determine whether the single-shot effect or the coupling effect of the event within the time window for the parameter;

[0009] Step 5: Use baseline estimation and sparsity denoising processing to decompose the power window time series data W affected by the coupling of multiple event elements t into the long-term changing trend baseline B and the short-term changing curve C, eliminating the influence of the long-term changing trend baseline B on the instantaneous event; Extract the nominal curve P of the power parameter affected from multiple event elements M n ={t0, t f , A, S, P}, and form the nominal curve set Q = {P1, P2,..., P m}}, and perform secondary fusion on the nominal curve sets of m events according to the start time t0 and end time t in the event elements f , to obtain the coupling nominal curve P of the multiple events' impact on the power parameter merge ;

[0010] Step 6: Perform correlation analysis on the short-term changing curve C and P merge .

[0011] The specific process of the above Step 2 is as follows:

[0012] Assume that the length of the power data of a certain channel in the ground simulation is T, and it is segmented with a sliding window of length n, where n < T, to obtain T - n + 1 segmented window data. For each segmented subsequence obtained, the sliding window moves forward by n positions, and then the subsequences W t0 , W t1 , W t2 , … W tT-n ; are obtained. Calculate the moving average data MA of the subsequences.

[0013] The specific process of the above Step 4 is as follows:

[0014] Analyze the event elements M n ={t0, t f , A, S, P} collected under the fluctuating window data obtained in Step 3. If there are multiple event elements within the window length w, then directly determine that the event within this time window is a coupling effect, and the time series data W of this section of the window tThe fluctuation of is caused by the coupling of multiple event elements; if there is a single event element within the window length w, then the event in the time window is considered to be a single event; for the coupling effect caused by the close occurrence time of the event, the power supply data of the channel is decoupled and analyzed; for the data fluctuation caused by the single event, the event element M is further determined through evidence theory. n ={t0,t f ,A,S,P} whether the associated parameter set S is strongly correlated with the fluctuation.

[0015] The detailed process of step 5 is as follows:

[0016] Step 5-1: Construct a model for sparse peak data;

[0017] Step 5-2: Design high-pass and low-pass filters;

[0018] Step 5-3: Composite sparse derivative modeling;

[0019] Step 5-4: Power window timing data W t After processing in steps 5-1, 5-2, and 5-3, it can be decomposed into long-term changing trend baseline data B and short-term changing sparse peak feature data C.

[0020] The specific process of step 5-1 is as follows:

[0021] Baseline estimation and sparsity noise reduction methods are used to decompose the trend term and transient term of bus power change;

[0022] Modeling based on N-point noise-free data vector

[0023] s=x+f,s∈R N (1)

[0024] Among them, vector x represents a vector composed of many peaks in the bus power curve, which is identified as a sparse derivative signal; the bus power curve baseline vector f is a low-pass signal;

[0025] For noisy bus power curve data, the model is:

[0026] y=s+w (2)

[0027] =x+f+w,y∈R N (3)

[0028] Where w is a stationary white Gaussian noise with variance σ;

[0029] To determine the long-term trend baseline B of the bus power, the component f should be estimated correctly; in order to obtain the short-term variation characteristics C, the vector x needs to be estimated.

[0030] The specific process of step 5-2 is as follows:

[0031] Consider filters L and H as zero-phase, non-causal recursive filters; the high-pass filter H is described in the form

[0032] H=A -1 B (4)

[0033] Where A and B are banded convolution Toeplitz matrices; H is represented by a banded matrix, and the algorithm of the fast solver is used for the banded linear system as the convolution matrix. A and B represent linear time-invariant LTI systems, and H represents the cascade of LTI systems. Using the exchange property, the above expression is expressed as

[0034] H=BA -1 (5)

[0035] The specific process of step 5-3 is as follows:

[0036] Assume that there are M peaks in the transient term vector x, and the first-order derivative of vector x i = 0, ... M should be sparse; in order to obtain the estimated value x^ of the transient term vector x, the optimization expression is obtained as

[0037]

[0038] Di represents the first-order difference operator; function It is expressed as:

[0039]

[0040] Subtracting formula (7) from formula (6), we get the following formula:

[0041]

[0042] Where N i Indicates D i The length of x; the constant λ i >0 is the regularization parameter.

[0043] The specific process of step 6 is: if the result after comparison and analysis is within the set threshold, it is considered that the fluctuation curve data is consistent with the expected model; otherwise, it is considered that the power curve changes caused by multiple events deviate from the nominal curve model and a reminder should be issued.

[0044] The advantages of the present invention compared with the prior art are:

[0045] (1) The present invention provides, for the first time, an automatic retrieval marking, data cleaning, and feature decoupling processing method for easily disturbed and easily coupled telemetry data, providing a reference for the in-depth and rapid interpretation of data during the AIT comprehensive test process of deep space exploration and during on-orbit flight.

[0046] (2) The present invention considers the decoupling and in-depth analysis strategy of abnormal data in the case of concurrent multiple trigger sources, which is in line with the change of the key parameter curve caused by the continuous sending of multiple instructions in the flight procedure. Through the multivariate statistical analysis method, it can be further confirmed whether there is a hidden trigger event affecting the nominal curve of the parameter.

[0047] (3) The present invention uses the ground sub-array power supply curve to describe the satellite power. The sampling frequency of the ground sub-array power supply data is higher than that of the telemetry data block, and the ground sampling frequencies of different satellites are unified, making it easier and more unified for horizontal comparison and analog analysis between satellites, and the model transplantation is faster. Description of the Drawings

[0048] Figure 1 Power curve of the ground square array acquisition channel. Among them, (a) is the nominal curve after the square array finishes charging the battery, and (b) is the change curve after multiple events are continuously triggered.

[0049] Figure 2 Judgment process of the multi-coupled key parameter curve.

[0050] Figure 3 Decoupling processing result of the data.

[0051] Figure 4 Nominal curve coupling model when the same device switches between different modes.

[0052] Figure 5 Nominal curve coupling model of different devices with event changes. Detailed Implementation Manner

[0053] A method for judging the shape of the bus power change curve of a lunar probe according to the present invention is as follows:

[0054] Step 1: As Figure 1 shown, by monitoring the output signals of the satellite ground power supply square array equipment, the characterization parameter set of the satellite integrator power state is obtained and recorded as the output power data set of each power supply sub-array {W1, W2... W t}, where W is the product of the voltage and current output by the ground power supply square array equipment in real time, and t is the number of channels of the ground power supply square array equipment. Calculate the sum of the moving averages of different parameters in the data set as the measurement value of the parameter over time.

[0055] Step 2: Adopt the moving average analysis method of the sliding window to calculate the power data set of a certain channel within the selected time to obtain the data after moving average. Assume that the length of the power data of a certain channel in the ground simulation is T, and it is divided by a sliding window with a length of n (n < T), and T - n + 1 divided window data can be obtained. Each time a subsequence is obtained after division, the sliding window moves forward by n positions, and then the subsequences Wt0 ,W t1 ,W t2 ,…W tT-n . Calculate the average value of the subsequence as the measurement value of the subsequence, recorded as MA. Then, if the MA value of the subsequence exceeds the specified threshold, it is considered that there is a fluctuation change in the window interval, and further analysis and interpretation are required. Otherwise, it is considered that the window data fluctuation has not fluctuated significantly.

[0056] Step 3: When the window data fluctuation exceeds the threshold, it is necessary to collect the corresponding event elements in the window data, such as satellite remote control commands, ground equipment operations, and other events that change the working status of the entire satellite. Define event element M n ={t0,t f ,A,S,P}, including the event start time t0, end time t f , target single-machine equipment A, event effect associated parameter set S, nominal curve P of the event's impact on power parameters.

[0057] Step 4: Figure 2 As shown, the parameter is judged as a single event or a coupling effect in the time window. The event element M collected under the fluctuation of the window data obtained in step 3 n ={t0,t f ,A,S,P} for analysis. If there are multiple event elements within the window length w, then the event in the time window is directly considered to be a coupling effect. The time series data W of this window t The fluctuation of is caused by the coupling of multiple event elements; if there is a single event element within the window length w, then the event in the time window is considered to be a single event. For the coupling effect caused by the close occurrence of events, the power supply data of the channel should be decoupled and analyzed; for the data fluctuation caused by a single event, the event element M should be further determined through evidence theory. n ={t0,t f ,A,S,P} whether the associated parameter set S is strongly correlated with the fluctuation.

[0058] According to the theory of evidence, determine the event element M n ={t0,t f ,A,S,P}, the correlation between the fusion features of the associated parameter set S and the power curve fluctuation features. The method is as follows.

[0059] (1) Extract the characteristics of the power curve data at multiple levels of the parameter set and the fluctuation within the window, and use the characteristics of the power curve data at multiple levels as the reference data column, denoted as X t (k).

[0060] (2) Extract the time domain characteristics and frequency domain characteristics of each parameter of the time series telemetry parameter set based on the time domain characteristic statistical digital model. The time domain characteristic statistical model includes peak-to-peak value, kurtosis, root mean square, C factor, L factor, variance, deviation, S factor, I factor and other statistical models.

[0061] (3) Use grey theory to generate data. Common data generation methods include cumulative generation, cumulative subtraction generation, average generation and normalization generation. Cumulative generation means that the first original data remains unchanged, the second data is the first and second original data, and the third data is the first, second and third original data added together. It can be described as

[0062]

[0063] (4) The generated data is used to calculate the correlation coefficient between the parameter set data in the event element and the reference data column at different times. The calculation formula is:

[0064]

[0065] Among them, ζ(k) is the associated data column X at the kth comparison time t The relative difference with the reference data X0 is called X t The correlation coefficient of X0 at time k.

[0066] (5) Actual effect. Assume that in a simulated flight mission, the control subsystem operating mode is switched by sending a command event, and the resulting single event causes fluctuations in the power parameter data of a certain channel. According to (1), the time domain characteristics and frequency domain characteristics of each parameter are extracted respectively. The calculation results are obtained.

[0067]

[0068] Construct feature set A.

[0069] A=[{μ MWi ,x skwMWi},{A MWi},{ca MWi},{cd MWi}]

[0070] At the same time, the features of the power curve data at multiple levels are constructed and a feature set B is formed.

[0071] B=[{μ pwr ,x pwr},{A pwr},{ca pwr},{cd pwr}]

[0072] Where:

[0073] μ—mean;

[0074] x skw x skw — skewness;

[0075] A—frequency domain amplitude;

[0076] ca—approximation coefficient;

[0077] cd—detail coefficient;

[0078] Perform correlation analysis on feature set A and feature set B to obtain a correlation matrix, where each row vector is the correlation between the injected data and an event, and any element in the row vector is the correlation corresponding to the feature of a single telemetry parameter at a single level. The order of the elements in the row vector is consistent with the above formula.

[0079] [0.932, 0.854, 0.653, 0.705;

[0080] 0.064, 0.746, 0.065, 0.005;

[0081] 0.302, 0.606, 0.730, 0.122]

[0082] (3-2)

[0083] The correlation degree in the above formula is subjected to information fusion analysis to obtain the event M when the infrared attitude determination mode switches to the star-sensitivity attitude determination mode. n ={t0,t f ,A,S,P} and the power curve data fluctuation characteristics are 0.7709. It can be considered that the power curve fluctuation characteristics are highly correlated with the parameter set fusion characteristics in the single event caused by the switch from infrared attitude determination to star sensor attitude determination mode.

[0084] Step 5: Use baseline estimation and sparsity denoising to convert the power window time series data W affected by the coupling of multiple event elements into t Decompose it into a long-term trend baseline B and a short-term change curve C, eliminating the impact of the long-term trend baseline B on instantaneous events. n ={t0,t f , A, S, P}, the nominal curve P affected by the power parameters in the extraction, for example, if there are m events, then m nominal curves can be extracted to form a nominal curve set Q = {P1, P2, ..., P m}, for this set, according to the start time t0 and end time t f , the nominal curve set of m events is fused twice to obtain the coupled nominal curve P of the influence of multiple events on power parameters mergeThe secondary fusion effect of multiple time nominal curve sets is as follows: Figure 4 and Figure 5 . Figure 4 It is a secondary fusion model of nominal curves when switching between different modes for the same device. Figure 5 It is a quadratic fusion model of nominal curves for event changes of different devices.

[0085] like Figure 3 As shown in Figure 5, the decoupling process of the impact of multiple event factors on power time series data is as follows. Step 5-1: Model construction of sparse peak data

[0086] Baseline estimation and sparsity denoising methods are used to decompose the trend term and transient term of bus power variation.

[0087] This method is based on modeling of N-point noise-free data vectors.

[0088] s=x+f,s∈R N (1)

[0089] Among them, vector x represents a vector composed of many peaks in the bus power curve, which is identified as a sparse derivative signal. The bus power curve baseline vector f is a low-pass signal.

[0090] For noisy bus power curve data, the model is:

[0091] y=s+w (2)

[0092] =x+f+w,y∈R N (3)

[0093] where w is a stationary white Gaussian noise with variance σ.

[0094] To determine the long-term trend baseline B of the bus power, the component f should be estimated correctly; in order to obtain the short-term variation characteristics C, the vector x needs to be estimated.

[0095] Step 5-2: High-pass and low-pass filter design

[0096] In order to filter out the influence of noise and obtain higher fidelity data, it is necessary to design appropriate low-pass filter L and high-pass filter H. Filters L and H are regarded as zero-phase, non-causal recursive filters.

[0097] The high-pass filter H described is of the form

[0098] H=A -1 B (4)

[0099] Where A and B are banded convolution (Toeplitz) matrices. H is represented by a banded matrix and the fast solver algorithm is used for banded linear systems. As convolution matrices, A and B represent linear time-invariant (LTI) systems and H represents the cascade of LTI systems. Using the commutative property, the above expression can be expressed as

[0100] H=BA -1 (5)

[0101] Since A and B are finite matrices, they are not completely commutative. However, the above two expressions are limited to the beginning and end of the signal and can be ignored for long signals. The actual formula (5) is more suitable for deriving an optimization algorithm with high computational efficiency.

[0102] Step 5-3: Composite sparse derivative modeling

[0103] According to formula (1), assuming that there are M peaks in the transient term vector x, the first-order derivatives of vector x i = 0, ... M should be sparse. In sparse signal processing, the behavior of sparse signals is usually achieved by using appropriate non-quadratic regularization terms. Therefore, in order to obtain the estimated value x^ of the transient term vector x, the optimization expression is obtained as

[0104]

[0105] The above formula uses the classic quadratic data fitting term, which reflects the result after the observed data is added with Gaussian white noise (AWGN). Di represents the first-order difference operator. Function It can be expressed as:

[0106]

[0107] function It belongs to the cost function and can promote sparsity. Subtracting formula (7) from formula (6), we can get the following formula.

[0108]

[0109] Where N i Indicates D i The length of x. The constant λi>0 is the regularization parameter.

[0110] Step 5-4: Power window timing data W t After processing in steps 5-1, 5-2, and 5-3, it can be decomposed into long-term changing trend baseline data B and short-term changing sparse peak feature data C.

[0111] Step 6: Compare the short-term change curves C and P mergeConduct correlation analysis; if the results after comparison and analysis are within the set threshold, it is considered that the fluctuation curve data is consistent with the expected model; otherwise, it is considered that the power curve changes caused by multiple events deviate from the nominal curve model and a reminder should be issued.

[0112] The short-term variation curve C and the coupled nominal curve P of the influence of multiple events on power parameters are merge As input, the canonical correlation analysis method is used to carry out the correlation analysis of multivariate parameters.

[0113] According to the canonical correlation analysis method, two sets of variables x and y are assumed.

[0114] x n Indicates n known trigger sources. y is the variation curve data.

[0115] To study the relationship between two sets of variables x and y, consider their linear combination:

[0116]

[0117] In the formula, a=(a 11 a 12 ... a 1p )', b=(b 11 b 12 ... b 1q )', is an arbitrary non-zero constant vector. Under the given conditions of x, y and the covariance matrix ∑ between them, select a and b to maximize the correlation coefficient between U1 and V1.

[0118]

[0119] Although the present invention has been disclosed as above in the form of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications to the technical solution of the present invention by using the technical contents disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope of the technical solution of the present invention.

Claims

1. A method for interpreting the shape of a lunar probe bus power variation curve, characterized in that Here are the steps: Step 1: By monitoring the output signal of the satellite ground power array equipment, the characterization parameter set of the satellite whole device power state is obtained and recorded as the output power data set of each power array {W1, W2...W t }, where W is the multiplication result of the voltage and current output by the ground power supply array device in real time, and t is the number of channels of the ground power supply array device; the sum of the moving averages of different parameters in the data set is calculated as the measurement value; Step 2: Use the sliding window moving average analysis method to calculate the power data set of a channel within the selected time to obtain the moving average data MA; if the MA value exceeds the specified threshold, it is considered that there is a fluctuation change in the window interval, and enter step 3; otherwise, it is considered that the window data fluctuation has no large fluctuation change, and exit this method; Step 3: Collect the corresponding event elements in the window data; define the event element M n ={t0,t f ,A,S,P}, including the event start time t0, end time t f , target single-machine device A, event effect associated parameter set S, nominal curve P of event impact on power parameters; Step 4: Determine whether the parameter is a single effect or a coupled effect of the event within the time window; Step 5: Use baseline estimation and sparsity denoising to convert the power window time series data W affected by the coupling of multiple event elements into t Decompose into long-term changing trend baseline B and short-term changing curve C, eliminating the influence of long-term changing trend baseline B on instantaneous events; Multiple event elements M n ={t0,t f , A, S, P}, and the nominal curve P affected by the power parameters is extracted to form a nominal curve set Q = {P1, P2, ..., P m }, for this set, according to the start time t0 and end time t f , the nominal curve set of m events is fused twice to obtain the coupled nominal curve P of the influence of multiple events on power parameters merge ; Step 6: Compare the short-term change curves C and P merge Conduct correlation analysis; The detailed process of step 5 is as follows: Step 5-1: Construct a model for sparse peak data; Step 5-2: Design high-pass and low-pass filters; Step 5-3: Composite sparse derivative modeling; Step 5-4: Power window timing data W t After processing in steps 5-1, 5-2, and 5-3, it can be decomposed into long-term trend baseline data B and short-term sparse peak feature data C; The specific process of step 5-1 is as follows: Baseline estimation and sparsity noise reduction methods are used to decompose the trend term and transient term of bus power change; Modeling based on N-point noise-free data vector s=x+f,s∈R N (1) Among them, vector x represents a vector composed of many peaks in the bus power curve, which is identified as a sparse derivative signal; the bus power curve baseline vector f is a low-pass signal; For noisy bus power curve data, the model is: y=s+w (2) =x+f+w,y∈R N (3) Where w is a stationary white Gaussian noise with variance σ; Then to determine the long-term trend baseline B of the bus power, the component f should be estimated correctly; in order to obtain the short-term variation characteristics C, the vector x needs to be estimated; The specific process of step 5-2 is as follows: Consider filters L and H as zero-phase, non-causal recursive filters; the high-pass filter H is described in the form H=A -1 B (4) Where A and B are banded convolution Toeplitz matrices; H is represented by a banded matrix, and the algorithm of the fast solver is used for the banded linear system as the convolution matrix. A and B represent linear time-invariant LTI systems, and H represents the cascade of LTI systems. Using the exchange property, the above expression is expressed as H=BA -1 (5); The specific process of step 5-3 is as follows: Assume that there are M peaks in the transient term vector x, and the first-order derivative of vector x i = 0, ... M should be sparse; in order to obtain the estimated value x^ of the transient term vector x, the optimization expression is obtained as Di represents the first-order difference operator; function R i : It is expressed as: Subtracting formula (7) from formula (6), we get the following formula: Where N i Indicates D i The length of x; the constant λ i >0 is the regularization parameter.

2. The method for interpreting the bus power variation curve of a lunar probe according to claim 1, characterized in that: The specific process of step 2 is: Assume that the length of the power data of a certain channel in ground simulation is T, and it is segmented by a sliding window of length n, where n < T, to obtain T - n + 1 segmented window data. Each time a subsequence is obtained, the sliding window moves forward by n positions, and then subsequences W t0 , W t1 , W t2 , … W tT-n ; Calculate the moving average data MA of the subsequences.

3. The method for interpreting the bus power variation curve of a lunar probe according to claim 2 is characterized in that: The specific process of step 4 is as follows: The event elements M collected under the window data fluctuation condition obtained in step 3 n ={t0,t f ,A,S,P} for analysis. If there are multiple event elements within the window length w, then the event in the time window is directly considered to be a coupling effect. The time series data W of this window t The fluctuation of is caused by the coupling of multiple event elements; if there is a single event element within the window length w, then the event in the time window is considered to be a single event; for the coupling effect caused by the close occurrence time of the event, the power supply data of the channel is decoupled and analyzed; for the data fluctuation caused by the single event, the event element M is determined through evidence theory. n ={t0,t f ,A,S,P} whether the associated parameter set S is strongly correlated with the fluctuation.

4. The method for interpreting the bus power variation curve of a lunar probe according to claim 1 is characterized in that: The specific process of step 6 is: if the result after comparison and analysis is within the set threshold, it is considered that the fluctuation curve data is consistent with the expected model; otherwise, it is considered that the power curve changes caused by multiple events deviate from the nominal curve model and a reminder should be issued.