Method and system for completing missing meteorological monitoring data

Abstract

The application discloses a meteorological monitoring data missing completion method and system, belongs to the technical field of meteorological data processing, obtains a multi-element monitoring sequence, a local cache segment and a device running state log, reconstructs data through a unified time base, and identifies an implicit missing interval; variation slopes before and after the missing interval, peak-valley arrival sequences and element collaborative offsets are extracted, missing state fingerprints are constructed, and the missing state fingerprints are matched with state fingerprints in historical complete data to screen a candidate trajectory set; in combination with measured boundary values at two ends of the missing interval, forward deduction and backward backstepping coupled calculation is performed, cross inversion checking is performed through coupling constraints, trend constraints and range constraints, and a target completion sequence is determined; finally, the completion result is written into a corresponding missing interval, and continuous meteorological monitoring data are output; through data collaboration and device linkage capability of the Hongmeng intelligent meteorological station, the application can realize high-precision data completion in a complex environment, and improve continuity and reliability of meteorological data.

Description

Technical Field

[0001] This invention relates to the field of meteorological data processing technology, specifically to a method and system for completing missing meteorological monitoring data. Background Technology

[0002] When intelligent weather stations are networked in the field, in addition to data interruptions caused by routine network outages, they may also experience hidden gaps due to momentary congestion of the sensor bus, asynchronous edge cache flushing, or low-power switching. These gaps can result in what appears to be a continuous timestamp but actually missing data for some meteorological elements. This phenomenon is particularly pronounced in short-term scenarios such as sudden changes in wind speed and rapid fluctuations in air pressure. Existing technologies often employ linear interpolation, single-element prediction, or breakpoint recovery, which struggle to identify such pseudo-continuities and are even less capable of reliably completing data when the coupling relationships between multiple elements are disturbed. This leads to deviations between the completed data and the actual meteorological evolution, affecting subsequent early warning analysis. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for completing missing meteorological monitoring data, so as to solve the deficiencies in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for completing missing meteorological monitoring data, comprising: S100: Acquire multi-element monitoring sequences, local cache fragments, and equipment operation status logs from the weather station, and reconstruct each data based on a unified time base to identify implicit missing intervals with continuous timestamps but abnormal element sampling relationships, and obtain the set of intervals to be completed Q. S200: For each interval Q to be completed, extract the slope of changes of multiple elements before and after the interval, the order of arrival of peaks and valleys, the collaborative offset of elements, and the corresponding device status jump information to construct the missing state fingerprint K. S300, the missing state fingerprint K is matched with the state fingerprint in the historical complete data to filter out the multi-factor joint evolution trajectory under similar states, and the candidate trajectory set R is obtained. S400, based on the candidate trajectory set R and combined with the measured boundary values ​​at both ends of the interval Q to be completed, performs forward inference and backward push coupled calculation on the missing interval to generate a completion result matrix B with simultaneous boundary convergence; S500, using the coupling constraints, trend constraints and range constraints between the completion result matrix B and the non-missing elements, cross-inversion verification is performed to determine the target completion sequence D; S600: Write the target completion sequence D into the corresponding interval Q to be completed, generate a confidence flag corresponding to the completion source and constraint verification results, and output the continuous meteorological monitoring data sequence.

[0005] Preferably, the step of identifying implicit missing intervals with continuous timestamps but abnormal feature sampling relationships includes: S110 performs time alignment processing on multi-element meteorological monitoring sequences and synchronously rearranges each meteorological element according to a unified sampling interval to form a sequence of coordinated changes in elements. S120, Based on the element co-change sequence, extract the difference in change rate and phase lag characteristics between meteorological elements to generate an element coupling feature set. S130, compare the set of element coupling features with the preset normal coupling feature range, determine the abnormal time segment where the coupling deviation exceeds the threshold, and obtain the candidate abnormal interval; S140, combining the device operation status log and data cache write timing, perform consistency verification on the candidate abnormal intervals, eliminate normal fluctuation intervals caused by environmental changes, and finally determine the hidden missing intervals and form the interval set Q to be completed.

[0006] Preferably, the step of constructing the missing state fingerprint K includes: S210, for each interval to be completed, extract multi-element monitoring data within a preset time length before and after the interval to be completed, and calculate the slope of change of each meteorological element within the time range and the time sequence of the corresponding peak and valley points to form a basic change feature sequence. S220, Based on the aforementioned basic change feature sequence, according to the degree of synchronization of changes of each meteorological element on the time axis, calculate the time offset between each meteorological element and obtain the element collaborative offset feature sequence. S230, combine the basic change feature sequence with the element collaborative offset feature sequence, and align it with the status change time in the device operation status log within the corresponding time range to extract device status jump information; S240, the basic change feature sequence, the element collaborative offset feature sequence, and the device state jump information are encoded and combined in a unified time order to generate the construction missing state fingerprint K corresponding to the interval to be filled.

[0007] Preferably, the step of screening out the multi-factor joint evolutionary trajectories under similar states includes: S310, based on historical continuous and complete multi-element meteorological data, extracts the basic change feature sequence, element collaborative offset feature sequence and equipment status jump information for the corresponding time period in the same way as constructing missing state fingerprints, and combines them in a unified time order to form a historical state fingerprint set. S320, the missing state fingerprint is compared with the historical state fingerprint set segment by segment, and the difference in slope of change, the consistency of peak and valley arrival order and the difference in time offset are calculated respectively to obtain the fingerprint difference sequence. S330, based on the fingerprint difference sequence, the difference in slope of change, the consistency of peak and valley arrival order, and the difference in time offset are weighted and accumulated according to preset weights to obtain the matching degree of the corresponding historical state fingerprint; S340, sort the historical state fingerprints according to the matching degree, and filter the multi-element joint evolution trajectory corresponding to the historical state fingerprints with a matching degree higher than the matching degree threshold to form a candidate trajectory set R.

[0008] Preferably, the step of generating the completed result matrix B that converges simultaneously at the boundary includes: S410, extract the trajectory segments in each candidate trajectory that correspond to the length of the interval to be completed, and read the measured boundary values ​​at the start and end of the interval to be completed, and construct the corresponding boundary constraint sequence; S420, using the measured boundary value at the starting end as the initial value, perform forward inference point by point according to the time evolution order of each candidate trajectory segment to obtain the forward completion sequence; S430, using the measured boundary value of the end as the initial value, perform backward push point by point according to the reverse time order of each candidate trajectory segment to obtain the backward completion sequence; S440, the deviations of the forward completion sequence and the backward completion sequence at the corresponding time positions are coupled and adjusted, and completion results with boundary deviations at both ends being less than the boundary deviation threshold are selected and combined to form a completion result matrix B with simultaneous boundary convergence.

[0009] Preferably, the step of performing cross-inversion verification using the coupling constraints, trend constraints, and range constraints between the completed result matrix B and the non-missing elements includes: S510: Extract the candidate completion sequence corresponding to each completion result from the completion result matrix, and simultaneously read the measured sequence of the non-missing elements in the interval to be completed, and construct the verification data group at the same time position; S520, based on the corresponding change relationship between various meteorological elements in the historical complete data, the measured sequence of the non-missing elements is used to reverse calculate each candidate complete sequence to obtain the corresponding inversion verification sequence, and the coupling deviation value between the candidate complete sequence and the inversion verification sequence is calculated. S530, each candidate completion sequence is compared with the measured data at both ends of the interval to be completed, and the continuity of the change direction, the continuity of the slope transition and the deviation of the numerical range are calculated to obtain the trend verification result and the range verification result. S540, determine the comprehensive verification value based on the coupling deviation value, trend verification result and range verification result, and select the candidate completion sequence with the best comprehensive verification value as the target completion sequence D.

[0010] Preferably, the step of outputting the continuous meteorological monitoring data sequence includes: S610: Map the completed values ​​of each time position in the target completed sequence to the missing time positions in the interval to be completed, and replace the original missing identifiers according to the meteorological element category and time order to form the completed data to be written. S620, for the data to be written and completed, record the corresponding candidate trajectory source, boundary convergence result and cross-inversion verification result, and generate completion source information according to time and position association; S630, based on the completion source information, calculate the boundary consistency, feature coupling consistency and numerical range conformity of the completion values ​​at each time location, and generate corresponding credibility tags; S640, The data to be written and the confidence mark are synchronously written into the original multi-element monitoring sequence, the distinguishing mark between the measured data and the completed data is retained, and the continuous meteorological monitoring data sequence is output.

[0011] This invention also provides a system for completing missing meteorological monitoring data, comprising: Latent missing data identification module: acquires multi-element monitoring sequences, local cache fragments and equipment operation status logs from weather stations, and reconstructs each data based on a unified time base to identify latent missing intervals with continuous timestamps but abnormal element sampling relationships, and obtains the set of intervals to be completed Q; Fingerprint construction module: For each interval Q to be completed, extract the slope of changes in multiple elements before and after the interval, the order of arrival of peaks and valleys, the collaborative offset of elements, and the corresponding device state change information to construct the missing state fingerprint K; Trajectory filtering module: Matches the missing state fingerprint K with the state fingerprint in the historical complete data, filters out the multi-factor joint evolution trajectory under similar states, and obtains the candidate trajectory set R; Boundary convergence calculation module: Based on the candidate trajectory set R and combined with the measured boundary values ​​at both ends of the interval Q to be completed, the module performs forward deduction and backward deduction coupled calculation on the missing interval to generate a completion result matrix B with simultaneous boundary convergence. Multi-constraint cross-inversion verification module: Uses the coupling constraints, trend constraints and range constraints between the completion result matrix B and the non-missing elements to perform cross-inversion verification and determine the target completion sequence D; Credibility tag generation module: Writes the target completion sequence D into the corresponding interval Q to be completed, generates credibility tags corresponding to the completion source and constraint verification results, and outputs the continuous meteorological monitoring data sequence.

[0012] The technical effects and advantages provided by the present invention in the above technical solution are as follows: 1. The meteorological monitoring data missing completion method provided by this invention constructs a missing state fingerprint based on the collaborative change characteristics of multiple elements and performs fine matching with the state fingerprint in historical complete data. This enables effective identification and accurate completion in scenarios with implicit missing data where time stamps are continuous but data relationships are abnormal. Compared with traditional methods that rely solely on single-element interpolation or simple time series prediction, this invention introduces multi-dimensional features such as change slope, peak and valley arrival order, and element collaborative offset to characterize the inherent coupling relationship between meteorological elements. Thus, even under complex environmental disturbances and short-term abnormal fluctuations, it can still recover data sequences that conform to the actual meteorological evolution law, significantly improving the physical rationality and consistency of the completion results.

[0013] 2. This invention employs a bidirectional constraint calculation method combining forward extrapolation and backward extrapolation, along with boundary convergence judgment and multi-constraint cross-inversion verification, to perform multi-layered screening and optimization of the completed data. This not only ensures smooth connection of the completed data at time boundaries but also effectively avoids deviations from actual meteorological change ranges through a comprehensive verification mechanism of coupling constraints, trend constraints, and range constraints. Furthermore, by introducing a credibility marker and associating it with the source information of the completed data, the traceability and evaluability of the completed data are achieved, significantly improving the continuity, reliability, and engineering application value of the final output meteorological monitoring data sequence. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0015] Figure 1 This is a flowchart of the method of the present invention.

[0016] Figure 2 This is a flowchart of the system modules of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1, please refer to Figure 1As shown in this embodiment, the method for completing missing meteorological monitoring data includes: S100 acquires the multi-element monitoring sequence, local cache fragments, and equipment operation status logs of the weather station, and reconstructs each data based on a unified time base to identify implicit missing intervals with continuous timestamps but abnormal element sampling relationships, thus obtaining the set of intervals to be completed Q.

[0019] In this embodiment, the step of identifying implicit missing intervals with continuous timestamps but abnormal feature sampling relationships includes: S110 performs time alignment processing on multi-element meteorological monitoring sequences and synchronously rearranges each meteorological element according to a unified sampling interval to form a sequence of coordinated changes in elements. S120, Based on the element co-change sequence, extract the difference in change rate and phase lag characteristics between meteorological elements to generate an element coupling feature set. S130, compare the set of element coupling features with the preset normal coupling feature range, determine the abnormal time segment where the coupling deviation exceeds the threshold, and obtain the candidate abnormal interval; S140, combining the device operation status log and data cache write timing, perform consistency verification on the candidate abnormal intervals, eliminate normal fluctuation intervals caused by environmental changes, and finally determine the hidden missing intervals and form the interval set Q to be completed.

[0020] Specifically, in step S110, the multi-element monitoring sequence of the smart weather station within a preset time range, local cache fragments, and equipment operation status logs are first acquired. The multi-element monitoring sequence includes data of temperature, humidity, air pressure, wind speed, and wind direction recorded in chronological order. The local cache fragments are data records temporarily stored during network interruptions. The equipment operation status logs include sensor start-up and shutdown times, power switching times, and communication status change times.

[0021] The multi-element monitoring sequence is sorted based on timestamps, and the meteorological element data is mapped to a unified time scale based on a preset sampling interval. Missing time points are filled in with placeholders, so that each meteorological element corresponds point by point on the same time series, and then the elements are combined in chronological order to form a sequence of coordinated changes.

[0022] In step S120, the element coordinated change sequence is segmented according to a sliding time window. Within each time window, the numerical change between adjacent sampling points of each meteorological element is calculated, and the change rate is obtained by dividing the change by the corresponding time interval. At the same time, taking one meteorological element as a benchmark, the change curves of other meteorological elements are compared with time offsets to determine the time difference when they reach the same change trend, thereby obtaining the phase lag characteristic. Finally, the difference in change rate and phase lag characteristics between each meteorological element are combined according to the time window order to form an element coupling characteristic set.

[0023] In step S130, the range of values ​​for the difference in change rate and phase lag between meteorological elements under normal conditions is obtained in advance based on historical complete meteorological data, and stored as a preset normal coupling feature range; the element coupling feature set obtained in step S120 is compared with the preset normal coupling feature range window by window; when the difference in change rate or phase lag between any pair of meteorological elements exceeds the corresponding range, the time window is marked as abnormal; multiple consecutive abnormal time windows are merged to obtain candidate abnormal intervals.

[0024] In step S140, the device operation status log and data cache write sequence within the time range corresponding to the candidate abnormal interval are obtained. First, it is determined whether there are records of sensor start / stop, power supply mode switching, or communication status change within the time range. If there are, the candidate abnormal interval is retained. If not, the data write order and timestamp of the corresponding time period in the local cache segment are compared to see if they are consistent. When there is a write delay or data rewriting phenomenon, the candidate abnormal interval is also retained. For candidate abnormal intervals with no device status change records and continuous and normal data writing, they are determined to be normal environmental fluctuations and are removed. Finally, the remaining intervals are determined as implicit missing intervals and form a set of intervals to be completed Q.

[0025] S200: For each interval Q to be filled, extract the slope of changes in multiple elements before and after the interval, the order of arrival of peaks and valleys, the collaborative offset of elements, and the corresponding device state change information to construct the missing state fingerprint K.

[0026] In this embodiment, the step of constructing the missing state fingerprint K includes: S210, for each interval to be completed, extract multi-element monitoring data within a preset time length before and after the interval to be completed, and calculate the slope of change of each meteorological element within the time range and the time sequence of the corresponding peak and valley points to form a basic change feature sequence. S220, Based on the aforementioned basic change feature sequence, according to the degree of synchronization of changes of each meteorological element on the time axis, calculate the time offset between each meteorological element and obtain the element collaborative offset feature sequence. S230, combine the basic change feature sequence with the element collaborative offset feature sequence, and align it with the status change time in the device operation status log within the corresponding time range to extract device status jump information; S240, the basic change feature sequence, the element collaborative offset feature sequence, and the device state jump information are encoded and combined in a unified time order to generate the construction missing state fingerprint K corresponding to the interval to be filled.

[0027] Specifically, in step S210, for each interval to be completed, the start and end times of the interval are first determined. Then, multi-element monitoring data within a preset time length are extracted from before the start time and after the end time, respectively, to form a front data segment and a back data segment of the interval to be completed. The preset time length is preferably the duration corresponding to multiple consecutive sampling intervals to ensure that both the front and back data segments contain the complete change process. Subsequently, the sampled values ​​of temperature, humidity, air pressure, wind speed, and wind direction in the front and back data segments are arranged in chronological order. For each meteorological element, the numerical difference between two adjacent sampling points is calculated, and the numerical difference is divided by the corresponding sampling time interval to obtain the slope of change of the meteorological element between adjacent sampling points in the corresponding data segment. The slope of change is used to characterize the rate of change of the meteorological element per unit time. After obtaining the slope of change, the time points corresponding to the local maximum and local minimum values ​​of each meteorological element are searched within the preceding and following data segments. The arrival order of peak and trough points is determined according to time sequence. When a meteorological element has both peak and trough points in both the preceding and following data segments, it is recorded whether it reaches the peak or trough point first, as well as the time distances of the peak and trough points relative to the boundary of the interval to be completed. Finally, the slope of change, peak arrival time, trough arrival time, and peak-trough arrival order corresponding to each meteorological element are arranged in chronological order to form a basic change feature sequence. This basic change feature sequence refers to the data set composed of the slope of change and peak-trough arrival order extracted from each meteorological element on both sides of the same interval to be completed, arranged in a unified time sequence.

[0028] In step S220, based on the basic change feature sequence obtained in step S210, the element co-location features between various meteorological elements are further extracted. Specifically, a preset benchmark meteorological element is first selected, preferably temperature or air pressure with strong change continuity as the benchmark; then, the remaining meteorological elements are compared with the benchmark meteorological element respectively. During the comparison, the change slope sequence in the preceding and following data segments is used as the object, and the change slope sequence of the meteorological element to be compared is moved forward or backward at each sampling interval on the time axis; each time it is moved, the sum of the absolute values ​​of the numerical differences between the moved sequence and the benchmark meteorological element change slope sequence at the corresponding positions is calculated; the movement time corresponding to the minimum sum of the absolute values ​​of the numerical differences is determined as the time offset of the meteorological element relative to the benchmark meteorological element. The time offset is used to represent the sequential arrival relationship of the two meteorological elements in the same change process. Subsequently, based on the time offset of each meteorological element, the degree of synchronization between the changes of each meteorological element is determined; the smaller the time offset, the higher the degree of synchronization between the changes of the corresponding two meteorological elements is considered; the larger the time offset, the lower the degree of synchronization between the changes of the corresponding two meteorological elements is considered. Finally, the time offset, degree of synchronization of change, and corresponding preceding and following data segment identifiers of each meteorological element relative to the baseline meteorological element are arranged in chronological order to form an element collaborative offset feature sequence. This element collaborative offset feature sequence refers to a data set reflecting the sequential and synchronous relationships of changes among various meteorological elements.

[0029] In step S230, the basic change feature sequence formed in step S210 and the element collaborative offset feature sequence formed in step S220 are merged accordingly, and the equipment status change information is extracted by combining the equipment operation status log within the time range corresponding to the interval to be completed. Specifically, firstly, based on the start and end times of the interval to be completed, the equipment operation status log within a preset time length before and after the time range is extracted; then, each record in the equipment operation status log is parsed to read the sensor start / stop time, power supply status switching time, communication status change time, and cache write status change time. Further, each status change time in the equipment operation status log is aligned item by item with the peak point arrival time, valley point arrival time in the basic change feature sequence, and the time offset in the element collaborative offset feature sequence; when the equipment status change time falls between the peak point arrival time and the valley point arrival time, or falls within the time range where the element collaborative offset feature changes significantly, the correlation between the equipment status change and the corresponding meteorological change is recorded. Subsequently, the device status change information, including sensor shutdown and restart, power supply switching from normal to low power, communication link interruption and recovery, and cache switching from continuous writing to delayed writing, was compiled in chronological order. This device status change information is used to characterize whether the operating status of devices near the area to be filled has changed in accordance with abnormal changes in meteorological elements.

[0030] In step S240, based on the results obtained in step S230, a missing state fingerprint for the corresponding interval to be filled is generated. Specifically, the basic change feature sequence, the element collaborative offset feature sequence, and the equipment status jump information are first mapped onto a unified time axis, so that the three types of data correspond item by item under the same time reference. For each time position on the unified time axis, the slope of the meteorological element change, the peak or valley arrival identifier, the peak and valley arrival order, the time offset relative to the reference meteorological element, the degree of change synchronization, and the equipment status jump type are written sequentially. For time positions where no corresponding value exists, a preset null value identifier is filled in to ensure that the data structure of each time position is consistent. Subsequently, the contents on the unified time axis are arranged continuously in the order of "previous data segment features - interval boundary features to be filled - subsequent data segment features - equipment status jump information" to form a complete set of feature description results. Finally, this feature description result is used as the missing state fingerprint for the corresponding interval to be filled. The missing state fingerprint is used to characterize the combined features of the changes in multiple elements before and after the interval to be filled and the changes in the equipment operating status, and serves as the basis for subsequent matching of historical similar evolution trajectories and execution of missing state filling.

[0031] S300, the missing state fingerprint K is matched with the state fingerprint in the historical complete data to filter out the multi-factor joint evolution trajectory under similar states, and the candidate trajectory set R is obtained.

[0032] In this embodiment, the process of matching the missing state fingerprint with the state fingerprint in the historical complete data in step S300, filtering out the multi-factor joint evolution trajectory under similar states, and obtaining the candidate trajectory set specifically includes the following steps.

[0033] In step S310, firstly, historical continuous and uninterrupted multi-element meteorological data is selected as historical analysis data. This historical continuous and uninterrupted multi-element meteorological data refers to data records that do not contain missing identifiers or abnormal write-up identifiers within a continuous sampling time range, and whose equipment operating status is complete and traceable. Subsequently, based on the time length corresponding to the interval to be completed, historical time periods are segmented along the time axis from the historical continuous and uninterrupted multi-element meteorological data. Each historical time period includes a preceding time period, a middle time period, and a following time period. The lengths of the preceding and following time periods are consistent with the preset time lengths used when constructing the missing state fingerprint, and the length of the middle time period is consistent with the length of the interval to be completed. For each historical time period, following the same processing order as when constructing missing state fingerprints, firstly, the slope of change, peak arrival time, valley arrival time, and peak-valley arrival order of each meteorological element in the preceding and following time periods are extracted to form a basic change feature sequence. Then, using a preset benchmark meteorological element as a reference, the time offset and degree of synchronization of other meteorological elements relative to the benchmark meteorological element are extracted to form an element collaborative offset feature sequence. Next, the equipment operation status records corresponding to that historical time period are read, and sensor start-up and shutdown times, power supply status switching times, communication status change times, and cache write status change times are extracted to form equipment status jump information. Finally, the basic change feature sequence, element collaborative offset feature sequence, and equipment status jump information corresponding to the same historical time period are mapped onto a unified time axis and combined according to a unified chronological order to form a historical state fingerprint. This process is repeated for all historical time periods to obtain a set of historical state fingerprints. The historical state fingerprint set refers to a data set composed of historical state fingerprints formed from multiple historical time periods arranged in chronological order; each historical state fingerprint corresponds to a complete multi-element joint evolution trajectory, wherein the multi-element joint evolution trajectory refers to the continuous value sequence of temperature, humidity, air pressure, wind speed and wind direction on a unified time axis within the historical time period.

[0034] In step S320, the missing state fingerprint is compared and aligned segment by segment with each historical state fingerprint in the historical state fingerprint set. Specifically, the missing state fingerprint and the currently compared historical state fingerprint are first aligned according to a uniform time length, so that the front feature parts, boundary feature parts, and back feature parts of the two are aligned one by one; if there is a difference in the number of time positions, they are realigned to the same number of time positions according to the sampling interval. After alignment, the change slope, peak-valley arrival order, and time offset of the two are read at each corresponding time position. For the difference in change slope, the absolute value of the difference between the change slopes at the corresponding time positions is taken; for the consistency of peak-valley arrival order, it is compared whether the missing state fingerprint and the historical state fingerprint both show that the peak point is reached before the valley point, or the valley point is reached before the peak point at the corresponding time position. If the order is consistent, it is recorded as consistent; if the order is inconsistent, it is recorded as inconsistent; for the difference in time offset, the absolute value of the difference between the time offsets at the corresponding time positions is taken. Following a unified chronological order, the differences in slope, peak-to-valley arrival order, and time offset at each corresponding time position are arranged sequentially to form a fingerprint difference sequence corresponding to the current historical state fingerprint. This fingerprint difference sequence refers to an ordered data set used to characterize the degree of difference between the missing state fingerprint and a certain historical state fingerprint at each corresponding time position, and is subsequently used to calculate the matching degree.

[0035] In step S330, the matching degree of the corresponding historical state fingerprint is calculated based on the fingerprint difference sequence. Specifically, the slope difference and time offset difference in the fingerprint difference sequence are first normalized to eliminate the influence of different units on the calculation results. The normalization process is as follows: first, the maximum and minimum values ​​of the slope difference in the current fingerprint difference sequence are calculated, and the slope difference at each time position is converted into a difference ratio value between zero and one. Then, the time offset difference is processed in the same way to obtain the corresponding time offset difference ratio value. For the consistency of the peak and valley arrival order, consistency is recorded as the first order value, and inconsistency is recorded as the second order value, where the first order value is greater than the second order value. Subsequently, each indicator is weighted and accumulated according to preset weights. The difference in slope corresponds to the first weight, the consistency of peak and trough arrival order corresponds to the second weight, and the difference in time offset corresponds to the third weight. The first, second, and third weights are pre-defined non-negative numbers, and their sum is one. In specific calculations, the difference in slope and time offset are accumulated using the result of "the first baseline value minus the corresponding difference ratio value" to ensure that the smaller the difference, the greater its contribution to the matching degree. The consistency of peak and trough arrival order is directly accumulated using the corresponding order value. After completing the weighted calculation for each corresponding time position, the weighted results for all time positions are accumulated or averaged to obtain the matching degree between the current historical state fingerprint and the missing state fingerprint. The matching degree is used to characterize the similarity between the two; the larger the matching degree value, the closer the current historical state fingerprint is to the missing state fingerprint. The matching degree is used in subsequent steps as a basis for screening historical state fingerprints and extracting multi-factor joint evolution trajectories.

[0036] In step S340, based on the matching degree obtained in step S330, all historical state fingerprints in the historical state fingerprint set are sorted. Specifically, first, a correspondence is established between each historical state fingerprint and its corresponding matching degree, and then they are arranged in descending order of matching degree to obtain a matching sorting result. Subsequently, the matching degree in the matching sorting result is compared with a preset matching degree threshold one by one, retaining historical state fingerprints with matching degrees higher than the matching degree threshold and removing historical state fingerprints with matching degrees lower than or equal to the matching degree threshold. For each retained historical state fingerprint, complete multi-element meteorological data in its corresponding historical time period is read, and the data of the intermediate time period with the same length as the interval to be filled is extracted as the filling reference trajectory corresponding to the historical state fingerprint; at the same time, the continuous data before and after the intermediate time period data is retained for boundary connection verification in subsequent steps. Finally, all the retained filling reference trajectories are summarized in descending order of matching degree to form a candidate trajectory set. The candidate trajectory set refers to a set of multi-factor joint evolution trajectories corresponding to multiple historical time periods that meet the requirements of similarity to the missing state fingerprint. Each trajectory in the candidate trajectory set is used to generate candidate completion results for the interval to be completed.

[0037] S400, based on the candidate trajectory set R and combined with the measured boundary values ​​at both ends of the interval Q to be completed, performs forward inference and backward push coupled calculation on the missing interval to generate a completion result matrix B with simultaneous boundary convergence.

[0038] In this embodiment, step S400, which involves performing forward extrapolation and backward extrapolation coupled calculations on the missing interval based on the candidate trajectory set and combined with the measured boundary values ​​at both ends of the interval to be filled, to generate a completion result matrix with simultaneously converged boundaries, specifically includes the following steps.

[0039] In step S410, each candidate trajectory in the candidate trajectory set is first read, and the length of the interval to be completed is determined. The interval length is the number of missing time points within the interval to be completed, or the number of sampling intervals between the start and end times of the interval to be completed. Then, using the interval length as a standard, a continuous trajectory segment matching the length of the interval to be completed is extracted from each candidate trajectory. This continuous trajectory segment contains the values ​​of temperature, humidity, air pressure, wind speed, and wind direction at continuous time positions. If a candidate trajectory already corresponds to the length of the interval to be completed, it is directly used as the trajectory segment. Next, the measured boundary values ​​at the start and end of the interval to be completed are read. The measured boundary value at the start refers to the multi-element measured value at the most recent valid sampling time point immediately preceding the start time of the interval to be completed, and the measured boundary value at the end refers to the multi-element measured value at the most recent valid sampling time point immediately following the end time of the interval to be completed. Furthermore, the measured boundary values ​​at the starting end are arranged in order of meteorological element type and time location to form a starting boundary constraint sequence; the measured boundary values ​​at the ending end are arranged in the same order to form a ending boundary constraint sequence; then the starting boundary constraint sequence and the ending boundary constraint sequence are combined to form a boundary constraint sequence. The boundary constraint sequence is used to define the starting and ending points of each meteorological element's value at both ends of the interval to be filled during subsequent forward and backward extrapolation. The construction order of the boundary constraint sequence is as follows: first, read the measured boundary values ​​at both ends; then, arrange them according to meteorological element type; and finally, combine them in the order of starting end first, ending end last.

[0040] In step S420, the measured boundary value at the beginning of the initial boundary constraint sequence is used as the initial value, and forward extrapolation is performed on each trajectory segment. Specifically, the multi-element value corresponding to the measured boundary value at the beginning is first set as the first time position value in the forward extrapolation; then, the change between adjacent time positions in the current trajectory segment is read, where the change refers to the difference between the value of the same meteorological element at the later time position and the value at the previous time position. For the first missing time position in the trajectory segment, the first change of the corresponding meteorological element in the current trajectory segment is added to the measured boundary value at the beginning to obtain the forward extrapolation value for that missing time position; for the second and subsequent missing time positions, the forward extrapolation value already obtained for the previous missing time position is added to the change corresponding to the current time position, and the forward extrapolation value for the subsequent missing time positions is obtained point by point. In the above order, calculations are performed on temperature, humidity, air pressure, wind speed, and wind direction for each time position until the extrapolation of values ​​for all missing time positions in the entire interval to be filled is completed, forming the forward completion sequence corresponding to the current trajectory segment. The forward completion sequence refers to a sequence of missing values ​​for multiple elements obtained by recursively extrapolating point by point according to the original time evolution order of the trajectory segment, starting from the measured boundary value at the initial end. The calculation order of the forward extrapolation is as follows: first determine the initial value, then read the change corresponding to the current time position, then perform the addition operation, and finally write the extrapolation result at the current time position. After the forward completion sequence is formed, it is used in step S440 for corresponding position comparison and coupling adjustment with the backward completion sequence.

[0041] In step S430, the measured boundary value at the end of the termination boundary constraint sequence is used as the initial value, and backward regression is performed on each trajectory segment. Specifically, the multi-element value corresponding to the measured boundary value at the end is first set as the last time position value for backward regression; then, the current trajectory segment is read in reverse order according to time, and the reverse change between adjacent time positions is extracted. The reverse change refers to the difference between the value of the same meteorological element at the current time position and the value at the next time position, or it is obtained by taking the opposite of the positive change of the trajectory segment. For the last missing time position in the interval to be filled, the measured boundary value at the end is subtracted from the change corresponding to the end at that time position to obtain the backward regression value of the last missing time position; for the second to last and the preceding missing time positions, the backward regression value already obtained for the next missing time position is subtracted from the change corresponding to the current time position, and the backward regression values ​​for the remaining missing time positions are obtained point by point. Following the above sequence, the backward calculations for each meteorological element are performed sequentially from the end point to the beginning point until a backward completion sequence covering the entire interval to be completed is obtained. The backward completion sequence refers to a sequence of missing values ​​for multiple elements obtained by backtracking point by point according to the reverse time sequence of the trajectory segments, starting with the measured boundary value at the end point. The calculation order for the backward backtracking is as follows: first, determine the initial value at the end point; then, read the change in reverse; subsequently, perform subtraction operations; and finally, write the backtracking result at the current time position. After being formed, the backward completion sequence serves as the input data for performing bidirectional deviation verification and boundary convergence judgment in step S440.

[0042] In step S440, the deviations between the forward and backward completion sequences at corresponding time positions are coupled and adjusted, and a completion result matrix with simultaneous boundary convergence is generated. Specifically, the forward and backward completion sequences corresponding to the same trajectory segment are first aligned point by point according to a unified time position; for each missing time position within the interval to be completed, the forward extrapolation value in the forward completion sequence and the backward retrospective value in the backward completion sequence are read respectively, and the deviation value between them is calculated. The deviation value is the absolute value of the difference between the forward extrapolation value and the backward retrospective value of the same meteorological element at the same time position. Subsequently, based on the time distance from the current time position to the starting point and the time distance to the ending point, the forward weight and backward weight are determined; the closer to the starting point, the greater the forward weight; the closer to the ending point, the greater the backward weight. Then, using the forward weight and backward weight, the forward extrapolation value and the backward retrospective value at the same time position are weighted and combined to obtain the coupling adjustment value for that time position. Following the above method, calculations are performed point-by-point on all missing time locations and all meteorological elements within the interval to be completed, forming the coupled completion result corresponding to the current trajectory segment. Then, the initial boundary deviation between the starting point and the measured boundary value at the starting point, and the ending boundary deviation between the ending point and the measured boundary value at the ending point are calculated. The initial boundary deviation refers to the difference between the time location value closest to the starting point and the measured boundary value at the starting point in the coupled completion result; the ending boundary deviation refers to the difference between the time location value closest to the ending point and the measured boundary value at the ending point in the coupled completion result. When both the initial boundary deviation and the ending boundary deviation are less than the boundary deviation threshold, the current coupled completion result is retained; when either boundary deviation is greater than or equal to the boundary deviation threshold, the current coupled completion result is discarded. Finally, all coupled completion results that meet the conditions are arranged in the order of "trajectory segment number—time location—meteorological element value" to form a completion result matrix with simultaneously converged boundaries. The completion result matrix refers to a data set formed by summarizing multiple coupled completion results that satisfy double-ended boundary constraints in a unified format. Each row corresponds to a set of multi-element completion results obtained by coupling forward extrapolation and backward extrapolation of a candidate trajectory, and each column corresponds to the value of a certain meteorological element at a certain time position within the interval to be completed. The completion result matrix is ​​used in subsequent processing to perform consistency verification, trend constraint screening, and determination of the target completion sequence.

[0043] S500, using the coupling constraints, trend constraints, and range constraints between the completed result matrix B and the non-missing elements, perform cross-inversion verification to determine the target completed sequence D.

[0044] In this embodiment, step S500 utilizes the coupling constraints, trend constraints, and range constraints between the completed result matrix B and the non-missing elements to perform cross-inversion verification and determine the implementation process of the target completed sequence D. Specifically, it includes the following steps.

[0045] In step S510, all completion results in the completion result matrix B are first read. The completion result matrix B is the data set output in step S400, constructed as follows: a set of complete completion results formed by coupling forward extrapolation and backward extrapolation for each candidate trajectory is sequentially written into the same data structure according to a unified time position and a unified meteorological element order. Each set of complete completion results corresponds to a candidate completion sequence. The candidate completion sequence refers to the set of completed values ​​for temperature, humidity, air pressure, wind speed, and wind direction arranged in chronological order for the same interval to be completed. Subsequently, the measured sequences of non-missing elements within the interval to be completed are extracted from the original multi-element monitoring data. Non-missing elements refer to data elements within the interval to be completed that have not experienced data interruptions, have not shown missing identifiers, and have passed the aforementioned quality screening; the measured sequence of non-missing elements refers to the data sequence formed by arranging the measured values ​​of the element at each time position within the interval to be completed in chronological order. Subsequently, the candidate completed sequences and the measured sequences of non-missing elements are matched point-by-point according to a unified time position, so that each missing time position corresponds to both a candidate completed value and a measured value of a non-missing element. Finally, the candidate completed values, measured values ​​of non-missing elements, time position numbers, and adjacent boundary time information at the same time position are combined to form a verification data group. The verification data group is used for subsequent coupling constraint verification, trend constraint verification, and range constraint verification. The construction order of the verification data group is as follows: first, candidate completed sequences are extracted from the completion result matrix B; then, the measured sequences of non-missing elements are read; then, they are aligned point-by-point according to time position; and finally, they are combined to form a verification data group at the same time position.

[0046] In step S520, based on the corresponding changes between meteorological elements in the historical complete data, the measured sequences of the elements without missing data are used to reverse-engineer each candidate completion sequence to obtain the corresponding inversion verification sequence, and the coupling deviation value between the candidate completion sequence and the inversion verification sequence is calculated. Specifically, firstly, historically continuous, non-missing, non-abnormally written, and equipment-stable multi-element meteorological data are selected as historical complete data; then, data segments with the same or similar meteorological states to the current interval to be completed are extracted from the historical complete data. The criteria for determining the meteorological state include the temperature level, humidity level, air pressure change direction, wind speed change intensity, and day / night information at both ends of the interval to be completed. Then, for each pair of "non-missing element - element to be completed", the synchronous and lagging changes of the two at the same and adjacent time positions in the historical complete data are statistically analyzed. The corresponding change relationship is constructed as follows: First, the continuous values ​​of the non-missing elements within the same historical time period are read, and then the continuous values ​​of the corresponding elements to be completed are read. Next, the directions of increase, decrease, or stability of the two values ​​are compared at each sampling interval. The main direction and range of change of the element to be completed within the same or subsequent sampling interval are recorded when the non-missing element changes in a certain direction. This process is repeated for multiple historical time periods, and the corresponding change direction with the highest statistical frequency and the most frequently occurring range of change are determined as the corresponding change relationship. After constructing the corresponding change relationship, reverse calculation is performed using the measured sequences of the non-missing elements within the current interval to be completed. The specific procedure is as follows: The measured sequence of the missing elements is read point by point in chronological order. First, it is determined whether the current time position is increasing, decreasing, or remaining stable relative to the previous time position. Then, based on the corresponding change relationship, the direction of change that the element to be completed should have at that time position is determined. After determining the direction of change, the amplitude value closest to the boundary changes at both ends of the interval to be completed is selected from the range of change amplitudes and used as the inversion change amount for that time position. Subsequently, this inversion change amount is superimposed or subtracted from the reference value of the element to be completed at the previous time position to obtain the inversion value for the current time position. By proceeding point by point according to the above steps, a corresponding inversion verification sequence can be formed. The inversion verification sequence refers to the sequence of values ​​for the element to be completed, calculated backward from the measured sequence of the missing elements and the corresponding change relationship in the historical complete data. Next, the candidate completed sequence and the inverted verification sequence are compared point-by-point at the same time position. The numerical difference between the two sequences is calculated for each time position, and the absolute value of this difference is taken to obtain the coupling deviation for that time position. Then, the coupling deviations at all time positions are accumulated or averaged in chronological order to obtain the coupling deviation value for the corresponding candidate completed sequence. The smaller the coupling deviation value, the closer the coupling relationship between the candidate completed sequence and the non-missing elements is to the historical true relationship. The coupling deviation value is used in step S540 to participate in the calculation of the comprehensive verification value.

[0047] In step S530, each candidate completion sequence is compared with the measured data at both ends of the interval to be completed, and the continuity of the direction of change, the continuity of the slope transition, and the deviation of the numerical range are calculated to obtain the trend verification result and the range verification result. Specifically, firstly, the measured values ​​of multiple elements at the most recent valid measured time position before the start of the interval to be completed, and the measured values ​​of multiple elements at the most recent valid measured time position after the end of the interval to be completed are read; then, the first completion value closest to the start and the last completion value closest to the end of the candidate completion sequence are read. For the continuity of the direction of change, the direction of change from the measured value at the start to the first completion value and the direction of change from the last completion value to the measured value at the end are compared respectively; if the directions at both ends are consistent with the adjacent measured trends outside the interval to be completed, the candidate completion sequence is considered to be continuous in the direction of change; if the direction is reversed at either end, it is recorded as discontinuous in direction. For slope transition continuity, first calculate the slope change of the measured value within the previous sampling interval at the starting end, then calculate the connection slope between the measured value at the starting end and the first completed value; subsequently, compare the difference between the two, and take the absolute value of the difference to obtain the slope transition deviation at the starting end; using the same method, calculate the slope transition deviation at the ending end between adjacent positions. Combine the slope transition deviations at the starting end and the ending end according to time position weights to form the slope transition continuity evaluation result. The time position weights are determined as follows: the closer the position is to the boundary, the greater the weight; the farther the position is from the boundary, the smaller the weight, thus highlighting the smoothness of the boundary connection. For the numerical range deviation, firstly, historical data segments consistent with the current season, current time period, and current weather change type are extracted from the complete historical data. Then, the common lower and upper limits of the corresponding meteorological elements under the same conditions are statistically analyzed, and these common lower and upper limits are determined as the range constraint interval. Next, it is checked whether the value of the candidate completion sequence at each time position in the interval to be completed falls within the range constraint interval. When the value at a certain time position is lower than the common lower limit, the difference between the common lower limit and the value is calculated. When the value at a certain time position is higher than the common upper limit, the difference between the value and the common upper limit is calculated. When the value falls within the range constraint interval, the range deviation at that time position is recorded as 0. The range deviations at all time positions are accumulated or averaged in chronological order to obtain the numerical range deviation. Finally, the continuity of the direction of change and the continuity of the slope transition are combined into a trend verification result, and the numerical range deviation is combined into a range verification result. The trend verification result is used to reflect whether the candidate complete sequence naturally connects with the boundary measured value in terms of time evolution, and the range verification result is used to reflect whether the candidate complete sequence exceeds the historical real meteorological change range.

[0048] In step S540, a comprehensive verification value is determined based on the coupling deviation value, trend verification result, and range verification result, and the candidate completion sequence with the optimal comprehensive verification value is selected as the target completion sequence D. Specifically, the coupling deviation value, slope transition deviation, and numerical range deviation are first subjected to unified scaling. The unified scaling process is as follows: first, the maximum and minimum values ​​of all candidate completion sequences on the corresponding indicators are statistically analyzed; then, the original indicator values ​​of each candidate completion sequence are converted into relative difference values ​​between 0 and 1, so that different indicators can be compared under the same evaluation standard. For the continuity of the direction of change, a discrete assignment method is used, that is, a higher continuity value is assigned when both ends are continuous, an intermediate continuity value is assigned when only one end is continuous, and a lower continuity value is assigned when neither end is continuous. Subsequently, the evaluation weights corresponding to the coupling constraint, trend constraint, and range constraint are determined. The evaluation weights are determined as follows: based on the meteorological scenario to which the interval to be completed belongs, when the interval is in a scenario of strong convection, sudden wind speed changes, or rapid pressure jumps, the evaluation weights of coupling constraints and trend constraints are increased; when the interval is in a stable weather scenario, the evaluation weights of range constraints are increased. After the weights are set, a comprehensive verification value is calculated for each candidate completion sequence. The specific calculation order is as follows: first, the relative difference value corresponding to the coupling deviation value is converted into a coupling consistency score, i.e., the smaller the difference, the higher the score; then, a trend consistency score is formed based on the evaluation results of the continuity value of the direction of change and the continuity of the slope transition; then, the relative difference value corresponding to the numerical range deviation is converted into a range consistency score, i.e., the smaller the deviation, the higher the score; finally, the coupling consistency score, trend consistency score, and range consistency score are weighted and summarized according to the preset evaluation weights to obtain the comprehensive verification value of the candidate completion sequence. The larger the comprehensive verification value, the higher the degree to which the candidate completion sequence simultaneously satisfies coupling constraints, trend constraints, and range constraints. After calculating the comprehensive verification value of all candidate completion sequences, they are sorted from highest to lowest comprehensive verification value, and the candidate completion sequence with the highest comprehensive verification value is selected as the target completion sequence D. If two or more candidate completion sequences have the same comprehensive verification value or the difference is less than the preset micro-difference range, their coupling deviation values ​​are further compared, and the one with the smaller coupling deviation value is selected first. If the coupling deviation values ​​are still the same, the slope transition deviation at the boundary connection position is further compared, and the one with the smaller slope transition deviation is selected first. The final determined target completion sequence D is the final completion result used to write back the interval to be completed.

[0049] S600: Write the target completion sequence D into the corresponding interval Q to be completed, generate a confidence flag corresponding to the completion source and constraint verification results, and output the continuous meteorological monitoring data sequence.

[0050] In this embodiment, the process of writing the target completion sequence D into the corresponding interval Q to be completed in step S600, generating a confidence marker corresponding to the completion source and constraint verification result, and outputting the continuous meteorological monitoring data sequence specifically includes the following steps.

[0051] In step S610, the target completion sequence D and the corresponding missing interval Q are first read. The target completion sequence D is the final completion result determined after cross-inversion verification in step S500, and the missing interval Q is the latent missing interval identified in step S100. Then, the start time, end time, and timestamps of each missing time position in the original multi-element monitoring sequence of the missing interval Q are read and arranged in ascending order of timestamps to form a missing time position list. The missing time position list refers to an ordered set of all missing sampling positions within the missing interval Q arranged in chronological order. Afterwards, the completed values ​​for each time position in the target completion sequence D are read in the same chronological order as the missing time position list, and each value is matched sequentially according to a fixed order of temperature, humidity, air pressure, wind speed, and wind direction. If the number of time positions in the target completed sequence D matches the number of time positions in the missing time position list, a one-to-one mapping relationship is directly established. If the number of time positions is inconsistent, the target completed sequence D is first rearranged chronologically based on a uniform sampling interval to ensure it completely corresponds to the time positions in the missing time position list before establishing the mapping relationship. After mapping, the missing identifiers at each missing time position in the original multi-element monitoring sequence are replaced with the completed values ​​at the corresponding time positions in the target completed sequence D, forming the completed data to be written. The completed data to be written refers to the data set that has been one-to-one matched with each missing time position in the interval Q to be completed and can be directly written back to the original multi-element monitoring sequence. To ensure traceability of the writing process, while replacing the missing identifiers, the original timestamp, sampling order number, and element category identifier of each time position are retained, so that the completed data to be written can still maintain the same sorting rules as the original record after being written back.

[0052] In step S620, for the data to be completed, the corresponding completion source, boundary convergence result, and cross-inversion verification result are recorded, and completion source information is generated by associating them according to time position. Specifically, the formation process record of the target completion sequence D in steps S400 and S500 is first read, including the sorting position of the candidate trajectory corresponding to the target completion sequence D in the candidate trajectory set R, the historical time period identifier of the candidate trajectory, the source position in the completion result matrix B of simultaneous boundary convergence, and the comprehensive verification value. The completion source refers to which candidate trajectory, which set of simultaneous boundary convergence completion results, and which comprehensive verification and screening result the target completion sequence D originates from. Subsequently, the boundary convergence result corresponding to the target completion sequence D is read. The boundary convergence result includes at least the initial boundary deviation and the final boundary deviation. The initial boundary deviation refers to the difference between the completed value closest to the beginning of the interval Q to be completed in the target completed sequence D and the measured boundary value at the beginning. The final boundary deviation refers to the difference between the completed value closest to the end of the interval Q to be completed in the target completed sequence D and the measured boundary value at the end. The aforementioned differences are obtained by reading two values ​​of the same meteorological element at the boundary position, subtracting the two values, and using the absolute value as the boundary deviation of that meteorological element. Finally, the boundary deviations of all meteorological elements are accumulated or averaged to obtain the boundary convergence result corresponding to the current time position. Afterwards, the cross-inversion verification result is read. The cross-inversion verification result includes coupling deviation value, trend verification result, and range verification result. The coupling deviation value reflects the degree of difference between the target completed sequence D and the inversion verification sequence; the trend verification result reflects the smoothness of the connection between the target completed sequence D and the measured data at both ends of the interval Q to be completed; and the range verification result reflects whether the target completed sequence D falls within a reasonable value range obtained from historical statistics. After reading is complete, the source of completion, boundary convergence results, and cross-inversion verification results are associated with meteorological element categories according to time location. For example, for the first missing time location in the interval Q to be completed, its corresponding completed value, source candidate trajectory number, start or end distance, boundary deviation, coupling deviation, trend verification result, and range verification result are all written into the same record; the remaining missing time locations are processed in the same way, ultimately forming the source information of completion. The source information of completion refers to the set of additional records that correspond one-to-one with the data to be completed, used to explain where each completed value comes from, what constraints it is subject to, and how it performs during verification.

[0053] In step S630, based on the completion source information, the boundary consistency degree, element coupling consistency degree, and numerical range conformity degree of the completion values ​​at each time location are calculated respectively, and corresponding confidence labels are generated. Specifically, the boundary consistency degree is calculated first. The boundary consistency degree is used to characterize the naturalness of the connection between the completion value at a certain time location and the measured boundary values ​​at both ends of the interval Q to be completed. During the calculation, the time distance from the time location to the starting end and the time distance to the ending end are read first, and then the starting boundary deviation and ending boundary deviation of the meteorological element to which the time location belongs are weighted. The closer to the starting end, the greater the weight of the starting boundary deviation; the closer to the ending end, the greater the weight of the ending boundary deviation. The weight is constructed as follows: the distance from the time location to a certain boundary is compared with the total length of the interval Q to be completed. The smaller the distance, the higher the corresponding weight; then the starting end weight and the ending end weight are normalized to be added together to 1. After the weight setting is completed, the starting boundary deviation and the ending boundary deviation are combined according to the above weights, and the combined deviation result is converted into the boundary consistency degree. The conversion method is as follows: the smaller the deviation, the higher the boundary consistency; the larger the deviation, the lower the boundary consistency. Next, the element coupling consistency is calculated. This element coupling consistency characterizes whether the current time-position completed value and the measured value of the missing element satisfy the corresponding change relationship in historical complete data. During calculation, the coupling deviation value of the current time-position completed value is first read, then the change direction of the measured value of the missing element at the same time position and the corresponding change direction obtained from historical complete data statistics are read. If the change direction of the completed value is consistent with the corresponding change direction obtained from historical statistics, a higher coupling consistency score is assigned; if the directions are inconsistent, a lower coupling consistency score is assigned. Then, the magnitude of the coupling deviation value is used for correction; the smaller the coupling deviation value, the higher the corrected element coupling consistency; the larger the coupling deviation value, the lower the corrected element coupling consistency. Next, the numerical range conformity is calculated. This numerical range conformity characterizes whether the current time-position completed value is within a reasonable numerical range obtained from historical statistics. The implementation method is as follows: First, based on the current season, current time period, and current weather change type, extract measured records under the same conditions from historical complete data; then, statistically analyze the common lower and upper limits of the corresponding meteorological elements under the same conditions, and determine these common lower and upper limits as the reasonable value range of the current meteorological element; next, determine whether the current time and location completion value falls within this reasonable value range. If it does, the degree of conformity is recorded as high; if it exceeds the reasonable value range, deductions are made based on the extent of the exceedance, with a larger exceedance indicating a lower degree of conformity. After completing the calculation of the aforementioned three degree values, the degree of boundary consistency, the degree of element coupling consistency, and the degree of conformity of the value range are summarized to generate a confidence label.The credibility marker is generated as follows: First, set weights for boundary consistency, element coupling consistency, and numerical range conformity, with the sum of these three weights being 1. Then, sum the three weights according to their respective weights to obtain the credibility score. To facilitate write-back and subsequent retrieval, the credibility score can be further divided into three levels. A high credibility marker is generated when the credibility score is not less than 0.85; a medium credibility marker is generated when the credibility score is greater than or equal to 0.60 and less than 0.85; and a low credibility marker is generated when the credibility score is less than 0.60. The aforementioned 0.85 and 0.60 are credibility marker thresholds. The credibility marker thresholds are determined as follows: Several known missing intervals are artificially constructed from historical complete data, and then the completion process is performed according to the method of this invention. The deviation distribution between the completion result and the actual measured value is compared. The lower bound of the sample score with smaller deviation is determined as the high credibility marker threshold, and the lower bound of the sample score with acceptable deviation is determined as the medium credibility marker threshold. This ensures that the credibility marker corresponds to the actual completion accuracy.

[0054] In step S640, the data to be completed and the confidence marker are synchronously written into the original multi-element monitoring sequence, retaining the distinguishing identifier between the measured data and the completed data, and outputting a continuous meteorological monitoring data sequence. Specifically, records are read one by one according to the time order of the original multi-element monitoring sequence. When a missing time position within the interval Q to be completed is read, the missing identifier at that position is replaced with the corresponding completed value in the data to be completed. Subsequently, the corresponding confidence marker and completion source information are written into the additional field at that time position. The additional field includes at least the data source identifier, confidence marker, candidate trajectory source number, boundary convergence result summary, and cross-inversion verification result summary. The data source identifier distinguishes between measured data and supplementary data, preferably set to two values: "measured" and "supplementary". The credibility flag indicates the credibility of the current supplementary value. The candidate trajectory source number traces which candidate trajectory the supplementary value originates from. The boundary convergence result summary records the connection status of the supplementary value under double-ended boundary constraints. The cross-inversion verification result summary records the comprehensive performance of the supplementary value under coupling constraints, trend constraints, and range constraints. After replacing and writing all missing time positions within the interval Q to be supplemented, a continuity check is performed on the entire multi-factor monitoring sequence. The continuity check includes: whether the timestamps are continuously increasing, whether missing identifiers still exist at each time position, whether new abnormal jumps occur before and after the supplementary value, and whether the boundary connection between the measured data and the supplementary data remains smooth. If the inspection results meet the continuity requirements, the processed complete sequence is identified as a continuous meteorological monitoring data sequence. If missing identifiers that have not been replaced are found, or new abnormal jumps occur after completion, the corresponding time position is repositioned, and the target completed sequence D and confidence marker are reread. A second check is then performed to complete the correction. The final output continuous meteorological monitoring data sequence refers to a complete data sequence on a unified time axis where temperature, humidity, air pressure, wind speed, and wind direction all have continuous values, and each time position is accompanied by a data source identifier and a confidence marker. This continuous meteorological monitoring data sequence can be directly used for subsequent data transmission, storage, statistical analysis, and early warning judgment processing.

[0055] Example 2, please refer to Figure 2 As shown, the meteorological monitoring data completion system described in this embodiment includes: Latent missing data identification module: acquires multi-element monitoring sequences, local cache fragments and equipment operation status logs from weather stations, and reconstructs each data based on a unified time base to identify latent missing intervals with continuous timestamps but abnormal element sampling relationships, and obtains the set of intervals to be completed Q; Fingerprint construction module: For each interval Q to be completed, extract the slope of changes in multiple elements before and after the interval, the order of arrival of peaks and valleys, the collaborative offset of elements, and the corresponding device state change information to construct the missing state fingerprint K; Trajectory filtering module: Matches the missing state fingerprint K with the state fingerprint in the historical complete data, filters out the multi-factor joint evolution trajectory under similar states, and obtains the candidate trajectory set R; Boundary convergence calculation module: Based on the candidate trajectory set R and combined with the measured boundary values ​​at both ends of the interval Q to be completed, the module performs forward deduction and backward deduction coupled calculation on the missing interval to generate a completion result matrix B with simultaneous boundary convergence. Multi-constraint cross-inversion verification module: Uses the coupling constraints, trend constraints and range constraints between the completion result matrix B and the non-missing elements to perform cross-inversion verification and determine the target completion sequence D; Credibility tag generation module: Writes the target completion sequence D into the corresponding interval Q to be completed, generates credibility tags corresponding to the completion source and constraint verification results, and outputs the continuous meteorological monitoring data sequence.

[0056] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for completing missing meteorological monitoring data, characterized by: include: S100: Acquire multi-element monitoring sequences, local cache fragments, and equipment operation status logs from the weather station, and reconstruct each data based on a unified time base to identify implicit missing intervals with continuous timestamps but abnormal element sampling relationships, and obtain the set of intervals to be completed Q. The meteorological monitoring data includes temperature, humidity, air pressure, wind speed, and wind direction data. S200: For each interval Q to be completed, extract the slope of changes of multiple elements before and after the interval, the order of arrival of peaks and valleys, the collaborative offset of elements, and the corresponding device status jump information to construct the missing state fingerprint K. S300, the missing state fingerprint K is matched with the state fingerprint in the historical complete data to filter out the multi-factor joint evolution trajectory under similar states, and the candidate trajectory set R is obtained. S400, based on the candidate trajectory set R and combined with the measured boundary values ​​at both ends of the interval Q to be completed, performs forward inference and backward push coupled calculation on the missing interval to generate a completion result matrix B with simultaneous boundary convergence; S500, using the coupling constraints, trend constraints, and range constraints between the completed result matrix B and the non-missing elements, a cross-inversion verification is performed to determine the target completed sequence D, including: S510: Extract the candidate completion sequence corresponding to each completion result from the completion result matrix, and simultaneously read the measured sequence of the non-missing elements in the interval to be completed, and construct the verification data group at the same time position; S520, based on the corresponding change relationship between various meteorological elements in the historical complete data, the measured sequence of the non-missing elements is used to reverse calculate each candidate complete sequence to obtain the corresponding inversion verification sequence, and the coupling deviation value between the candidate complete sequence and the inversion verification sequence is calculated. S530, each candidate completion sequence is compared with the measured data at both ends of the interval to be completed, and the continuity of the change direction, the continuity of the slope transition and the deviation of the numerical range are calculated to obtain the trend verification result and the range verification result. For the continuity of the direction of change, the direction of change from the initial measured value to the first completed value and the direction of change from the last completed value to the final measured value are compared respectively. If the directions at both ends are consistent with the adjacent measured trends outside the interval to be completed, the candidate completed sequence is considered to be continuous in the direction of change. If the direction is reversed at either end, it is recorded as discontinuous in direction. For the continuity of the direction of change, a discrete assignment method is used to obtain the continuity value of the direction of change. Specifically, for slope transition continuity, the slope of the measured value change within the previous sampling interval before the starting end is calculated, and then the connection slope between the measured value at the starting end and the first completed value is calculated; the difference between the two is compared, and the absolute value of the difference is taken to obtain the slope transition deviation at the starting end; the slope transition deviation between adjacent positions at the ending end is calculated; the slope transition deviation at the starting end and the slope transition deviation at the ending end are combined according to the time position weight to form the slope transition continuity evaluation result; S540, determine the comprehensive verification value based on the coupling deviation value, trend verification result and range verification result, and select the candidate completion sequence with the best comprehensive verification value as the target completion sequence D; Specifically, the relative difference value corresponding to the coupling deviation value is converted into a coupling consistency score, and then a trend consistency score is formed based on the evaluation results of the continuity value of the direction of change and the continuity of the slope transition; the relative difference value corresponding to the numerical range deviation is converted into a range consistency score, and the coupling consistency score, trend consistency score and range consistency score are weighted and summarized according to the preset evaluation weights to obtain the comprehensive verification value of the candidate completion sequence; S600: Write the target completion sequence D into the corresponding interval Q to be completed, generate a confidence flag corresponding to the completion source and constraint verification results, and output the continuous meteorological monitoring data sequence.

2. The method for completing missing meteorological monitoring data according to claim 1, characterized in that: The step of identifying implicitly missing intervals with continuous timestamps but abnormal feature sampling relationships includes: S110 performs time alignment processing on multi-element meteorological monitoring sequences and synchronously rearranges each meteorological element according to a unified sampling interval to form a sequence of coordinated changes in elements. S120, Based on the element co-change sequence, extract the difference in change rate and phase lag characteristics between meteorological elements to generate an element coupling feature set. S130, compare the set of element coupling features with the preset normal coupling feature range, determine the abnormal time segment where the coupling deviation exceeds the threshold, and obtain the candidate abnormal interval; S140, combining the device operation status log and data cache write timing, perform consistency verification on the candidate abnormal intervals, eliminate normal fluctuation intervals caused by environmental changes, and finally determine the hidden missing intervals and form the interval set Q to be completed.

3. The method for completing missing meteorological monitoring data according to claim 1, characterized in that: The steps for constructing the missing state fingerprint K include: S210, for each interval to be completed, extract multi-element monitoring data within a preset time length before and after the interval to be completed, and calculate the slope of change of each meteorological element within the time range and the time sequence of the corresponding peak and valley points to form a basic change feature sequence. S220, Based on the aforementioned basic change feature sequence, according to the degree of synchronization of changes of each meteorological element on the time axis, calculate the time offset between each meteorological element and obtain the element collaborative offset feature sequence. S230, combine the basic change feature sequence with the element collaborative offset feature sequence, and align it with the status change time in the device operation status log within the corresponding time range to extract device status jump information; S240, the basic change feature sequence, the element collaborative offset feature sequence, and the device state jump information are encoded and combined in a unified time order to generate the construction missing state fingerprint K corresponding to the interval to be filled.

4. The method for completing missing meteorological monitoring data according to claim 1, characterized in that: The step of selecting multi-factor joint evolutionary trajectories under similar states includes: S310, based on historical continuous and complete multi-element meteorological data, extracts the basic change feature sequence, element collaborative offset feature sequence and equipment status jump information for the corresponding time period in the same way as constructing missing state fingerprints, and combines them in a unified time order to form a historical state fingerprint set. S320, the missing state fingerprint is compared with the historical state fingerprint set segment by segment, and the difference in slope of change, the consistency of peak and valley arrival order and the difference in time offset are calculated respectively to obtain the fingerprint difference sequence. S330, based on the fingerprint difference sequence, the difference in slope of change, the consistency of peak and valley arrival order, and the difference in time offset are weighted and accumulated according to preset weights to obtain the matching degree of the corresponding historical state fingerprint; S340, sort the historical state fingerprints according to the matching degree, and filter the multi-element joint evolution trajectory corresponding to the historical state fingerprints with a matching degree higher than the matching degree threshold to form a candidate trajectory set R.

5. The method for completing missing meteorological monitoring data according to claim 1, characterized in that: The steps for generating the completed result matrix B that converges simultaneously at the boundary include: S410, extract the trajectory segments in each candidate trajectory that correspond to the length of the interval to be completed, and read the measured boundary values ​​at the start and end of the interval to be completed, and construct the corresponding boundary constraint sequence; S420, using the measured boundary value at the starting end as the initial value, perform forward inference point by point according to the time evolution order of each candidate trajectory segment to obtain the forward completion sequence; S430, using the measured boundary value of the end as the initial value, perform backward push point by point according to the reverse time order of each candidate trajectory segment to obtain the backward completion sequence; S440, the deviations of the forward completion sequence and the backward completion sequence at the corresponding time positions are coupled and adjusted, and completion results with boundary deviations at both ends being less than the boundary deviation threshold are selected and combined to form a completion result matrix B with simultaneous boundary convergence.

6. The method for completing missing meteorological monitoring data according to claim 1, characterized in that: The step of outputting the continuous meteorological monitoring data sequence includes: S610: Map the completed values ​​of each time position in the target completed sequence to the missing time positions in the interval to be completed, and replace the original missing identifiers according to the meteorological element category and time order to form the completed data to be written. S620, for the data to be written and completed, record the corresponding candidate trajectory source, boundary convergence result and cross-inversion verification result, and generate completion source information according to time and position association; S630, based on the completion source information, calculate the boundary consistency, feature coupling consistency and numerical range conformity of the completion values ​​at each time location, and generate corresponding credibility tags; S640, The data to be written and the confidence mark are synchronously written into the original multi-element monitoring sequence, the distinguishing mark between the measured data and the completed data is retained, and the continuous meteorological monitoring data sequence is output.

7. A meteorological monitoring data completion system, used to implement the meteorological monitoring data completion method according to any one of claims 1-6, characterized in that: include: Latent missing data identification module: acquires multi-element monitoring sequences, local cache fragments and equipment operation status logs from weather stations, and reconstructs each data based on a unified time base to identify latent missing intervals with continuous timestamps but abnormal element sampling relationships, and obtains the set of intervals to be completed Q; Fingerprint construction module: For each interval Q to be completed, extract the slope of changes in multiple elements before and after the interval, the order of arrival of peaks and valleys, the collaborative offset of elements, and the corresponding device state change information to construct the missing state fingerprint K; Trajectory filtering module: Matches the missing state fingerprint K with the state fingerprint in the historical complete data, filters out the multi-factor joint evolution trajectory under similar states, and obtains the candidate trajectory set R; Boundary convergence calculation module: Based on the candidate trajectory set R and combined with the measured boundary values ​​at both ends of the interval Q to be completed, the module performs forward deduction and backward deduction coupled calculation on the missing interval to generate a completion result matrix B with simultaneous boundary convergence. Multi-constraint cross-inversion verification module: Uses the coupling constraints, trend constraints and range constraints between the completion result matrix B and the non-missing elements to perform cross-inversion verification and determine the target completion sequence D; Credibility tag generation module: Writes the target completion sequence D into the corresponding interval Q to be completed, generates credibility tags corresponding to the completion source and constraint verification results, and outputs the continuous meteorological monitoring data sequence.

Images