A multi-source fusion closed-loop safety monitoring method suitable for high-frequency injection-production working conditions of a salt cavern gas storage

Abstract

This invention relates to the field of safety detection technology, specifically to a multi-source fusion closed-loop safety monitoring method for salt cavern gas storage facilities applicable to high-frequency injection and production conditions. In this invention, the Mankendall trend test outputs trend direction labels and significant trend indicators on the cavity pressure change value sequence, the annular pressure change value sequence, and the event increment value sequence. The trend direction labels are combined with the consistent out-of-bounds item count to form a stability direction discrimination sequence. The Viterbi algorithm outputs a time step hidden state label sequence based on the stability direction discrimination sequence and completes the location of continuous consistent sections. The location of continuous consistent sections separates short-term fluctuations from persistent anomalies and reduces the frequency of triggering actions under high-frequency injection and production conditions. The hidden state label sequence and the dominant direction of the section are mapped to the differential pressure reduction amount, the upper limit contraction amount, and the pressure stabilization extension amount to form an injection and production control action set. The action parameters are bound to the number of execution cycles and solidified into the execution trajectory of the operating state.

Description

Technical Field

[0001] This invention relates to the field of safety detection technology, and in particular to a multi-source fusion closed-loop safety monitoring method for salt cavern gas storage facilities applicable to high-frequency injection and production conditions. Background Technology

[0002] The field of safety detection technology aims to form a quantifiable and reproducible risk control link by measuring, integrating, calculating, determining thresholds and trends of multiple safety status quantities and triggering response feedback, thereby reducing the probability of accidents and limiting the consequences of accidents to preset boundaries.

[0003] The purpose of a multi-source fusion closed-loop safety monitoring method for salt cavern gas storage suitable for high-frequency injection and production conditions is to form a monitoring-judgment-disposal-verification closed loop by synchronously fusing and coupling multiple safety status quantities, determining and outputting executable actions. This method identifies signs of wellbore integrity degradation, cavity stability deterioration, and gas leakage before risks occur, and controls cavity pressure fluctuations, wellbore annular pressure drift, microseismic event increments, and surface subsidence rates within preset threshold ranges. This ensures that the gas storage maintains a controllable operating state and reduces the probability of instability, leakage, and shutdown events.

[0004] Existing technologies primarily rely on multiple safety status quantity measurements, fusion calculations, threshold determination, and trend-based triggering for response feedback. The fusion process often uses a combination of single-time-step threshold exceeding limits and simple upward / downward trend criteria, resulting in a separation between time-step boundaries and response boundaries. The lack of a unified index for these boundaries leads to discrepancies in criteria when different statistical windows are used for different status quantities within the same injection / production cycle. Fusion results drift with window shifts, causing unstable triggering. Response feedback is often output as a single trigger corresponding to a single action, lacking continuous segment screening and action duration binding. Under high-frequency injection / production conditions, short-term pressure curve reversals, short-term annular pressure drifts, and instantaneous microseismic count aggregations easily trigger multiple actions and multiple responses. The repeated cancellations and increased frequency of action switching led to sawtooth-shaped jumps in the sequence of operating parameters. The wellbore annulus and the pressure zone of the cavity were repeatedly disturbed. The verification methods mostly focused on whether a single indicator had fallen back, making it difficult to establish stable criteria for directional consistency and multi-indicator resonance. The cycle connection often adopted the method of directly using the end value of the previous cycle. The starting parameters lacked constraints on the number of switching times and reset rules. After the cross-cycle parameter drift accumulated, the boundary convergence path became uncontrollable. After several small reductions, the injection-production pressure difference entered the low pressure difference range and was superimposed with the accumulation of pressure stabilization time. The injection-production efficiency decreased and induced more frequent parameter adjustments, forming a phenomenon of frequent handling and operational boundary drift superimposed on each other. It was difficult to keep the risk control link in a consistent and reproducible state. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a multi-source fusion closed-loop safety monitoring method for salt cavern gas storage suitable for high-frequency injection and production conditions.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a multi-source fusion closed-loop safety monitoring method for salt cavern gas storage suitable for high-frequency injection and production conditions, comprising the following steps: S1: Read the high-frequency injection and production condition constraint set and determine the injection and production time step length. Around the salt cavern cavity, wellbore annulus section, injection and production valve group and pressure stabilization control section, limit the upper limit of operation, injection and production pressure difference, pressure stabilization time and annulus drift interval and combine them to generate an operation boundary combination table. S2: Based on the aforementioned operational boundary combination table, the Mankendall trend test is used to compare the pressure change direction, annular change direction, and event increment direction step by step around the pressure-bearing zone of the salt cavern cavity, the annulus section of the wellbore, and the microseismic influence zone. The number of consistent out-of-bounds items is counted and the consistency of the direction is determined to generate a stability direction discrimination sequence. S3: Based on the stability direction discrimination sequence, the Viterbi algorithm is used to determine the number of consecutive consistent time steps around the cavity stable section, the wellbore intact section, and the injection-production disturbance section, and select the section that meets the threshold. For the consistent direction section, the combination of pressure difference reduction, upper limit contraction, and pressure stabilization extension is selected to generate the injection-production control action set. S4: Based on the injection and extraction control action set, around the injection and extraction valve group, cavity operating area, and pressure stabilization control area, replace the cavity operating upper limit value and the injection and extraction pressure difference value and accumulate the pressure stabilization time, keep the value unchanged in continuous time steps, and generate the operating state execution trajectory. S5: Based on the execution trajectory of the operating status, around the salt cavern operation stage, the wellbore safety stage, and the injection-production cycle connection segment, count the number of status switching times and compare them with the threshold, select to continue or reset the action parameters and write them into the starting point of the next cycle to obtain the injection-production cycle connection parameter set.

[0007] As a further embodiment of the present invention, the operating boundary combination table includes the upper limit of cavity operating pressure, single-cycle injection-production pressure difference, stabilization period duration, annular pressure drift tolerance range, and cavity pressure fluctuation amplitude tolerance range. The stability direction discrimination sequence includes a consistent out-of-bounds status flag, a direction consistency flag, a change direction flag, and the number of out-of-bounds items. The injection-production control action set includes the lowering amount of the upper limit of cavity operating pressure, the lowering amount of the single-cycle injection-production pressure difference, the extension amount of the stabilization period, and the number of execution cycles. The operating status execution trajectory includes the upper limit of cavity operating pressure value, the single-cycle injection-production pressure difference value, the stabilization period duration value, and the operating status number. The injection-production cycle inheriting parameter set includes the starting value of the upper limit of cavity operating pressure, the starting value of the single-cycle injection-production pressure difference, the starting value of the stabilization period duration, and a continuation flag.

[0008] As a further aspect of the present invention, the specific steps for generating the running boundary combination table are as follows: Based on reading the high-frequency injection and extraction condition constraint set and determining the injection and extraction time step length, an injection and extraction time step number sequence is generated, the start and end times of the injection and extraction time steps are checked and the time granularity is unified, missing time steps are identified and continuity identifiers are added, the range of available time steps is locked, and the injection and extraction time step reference sequence is obtained. Based on the injection-production time step benchmark sequence, the upper limit value of the salt cavern operation is matched, the single-cycle injection-production pressure difference value is matched, the duration of the pressure stabilization period is matched, and the wellbore annular pressure drift range value is matched. These are then merged into a parameter set according to the time step and written into the index key to generate an operation boundary combination table.

[0009] As a further aspect of the present invention, the specific steps for generating the stability direction discrimination sequence are as follows: Based on the aforementioned operational boundary combination table, the Mankendall trend test is used to read the cavity pressure change value and annulus pressure change value in the order of injection and production time steps around the pressure-bearing zone of the salt cavern cavity, the annulus section of the wellbore, and the microseismic influence zone. The event increment value corresponding to the same time step is also read. The change values ​​of adjacent time steps are marked with positive and negative directions and the change direction label is recorded to obtain the direction label sequence. Based on the directional labeling sequence, the cavity pressure change amplitude is compared with the pressure fluctuation tolerance range in the operating boundary combination table at each time step, the annular pressure change amplitude is compared with the annular drift tolerance range, the event increment is compared with the event threshold, the number of boundary items that occur at each time step is recorded, and the boundary count sequence is obtained. Based on the out-of-bounds counting sequence, the consistency of the cavity pressure direction, annular pressure direction and event increment direction in the same time step is compared. The number of out-of-bounds items is judged against the trigger number threshold. A direction consistency flag is output and a consistent out-of-bounds flag is output to generate a stability direction discrimination sequence.

[0010] As a further aspect of the present invention, the Mankendall trend test selects the cavity pressure change value sequence, the annular pressure change value sequence, and the event increment value sequence. The number of injection / import time steps is set to be no less than a preset length. For any sequence, pairwise comparisons are performed according to the time step index, ensuring that the previous index is less than the next index. The difference is calculated as the difference between the next sequence value and the previous sequence value. A difference greater than zero is recorded as positive, a difference equal to zero as zero, and a difference less than zero as negative. All pairwise comparison signs are summed to obtain a statistic. The variance is calculated, and the summation of duplicate value counts is corrected. A standardized statistic is calculated based on the statistic and variance. If the standardized statistic is greater than zero, an upward trend label is output; if the standardized statistic is less than zero, a downward trend label is output; if the standardized statistic is equal to zero, a no-trend label is output. A significance level is set, and a critical value is calculated. If the absolute value of the standardized statistic is greater than the critical value, a significant trend indicator is output; if the absolute value of the standardized statistic is less than or equal to the critical value, a significant trend indicator is output.

[0011] As a further aspect of the present invention, the specific steps for generating the injection-import control action set are as follows: Based on the stability direction discrimination sequence, the Viterbi algorithm is used to scan the consistent out-of-bounds flags along the time axis, record the number of time steps that are continuously maintained in the consistent out-of-bounds state, mark the start and end points of the continuous segments, filter the segments with the number of consecutive segments reaching the threshold, and obtain the set of continuous consistent segments. Based on the continuous and consistent segment set, the number of times the direction consistency mark appears in each segment is counted. The direction is selected as the dominant direction of the segment based on the number of occurrences. The dominant direction is then mapped to the injection-production pressure difference adjustment level, the cavity operation upper limit adjustment level, and the pressure stabilization period adjustment level to form a segment parameter combination and obtain an action parameter combination set. Based on the set of action parameters, the number of execution cycles is assigned to each segment and the effective time step range is bound. The action parameters of adjacent segments are checked for conflicts and replacement or merging is performed to form segment action entries arranged in chronological order, and an injection and extraction control action set is generated.

[0012] As a further aspect of the present invention, the Viterbi algorithm sets a set of hidden states and an observation sequence, sets an initial state probability table, a state transition probability matrix, and an observation probability table. At time step one, it calculates the path probability for each hidden state and writes it into the path probability table and the predecessor index table. From time step two to the end of time step one, it calculates a set of candidate path probabilities for each hidden state. The set of candidate path probabilities is obtained by multiplying the path probability, state transition probability, and observation probability of the previous time step. It sorts the candidate path probability set in descending order, selects the first path probability in the sorted list and writes it into the path probability table, and writes the corresponding predecessor index into the predecessor index table. At the end of time step one, it sorts the path probability table in descending order, selects the first hidden state in the sorted list as the final hidden state, and backtracks along the predecessor index table to time step one to obtain the hidden state sequence and outputs the hidden state label sequence of the time step.

[0013] As a further aspect of the present invention, the specific steps for generating the execution trajectory of the running state are as follows: Based on the injection and extraction control action set, align the action entries with the injection and extraction time step numbers, mark the effective start point and the ineffective end point of the action, check the number of time steps covered by the action and remove overlapping and conflicting segments, splice the continuous coverage range in time order, and obtain the action parameter execution segment. Based on the action parameters, the upper limit value of the cavity operation is replaced with the corresponding value after downward adjustment in the segment, the single-cycle injection-production pressure difference value is replaced with the corresponding value after downward adjustment in the segment, the duration of the pressure stabilization period is replaced with the corresponding value after accumulation in the segment, the fixed parameter values ​​within the segment are marked with the operation status number, and the operation status execution trajectory is generated.

[0014] As a further aspect of the present invention, the specific steps for generating the injection-production cycle acceptance parameter set are as follows: Based on the execution trajectory of the running status, the running status number is read step by step and compared with the previous time step number. The same identifier is recorded and the change identifier is recorded. The number of changes within the cycle is accumulated and the time step of the change is located. The cycle switching statistics are summarized to obtain the state switching statistics. Based on the state switching statistics, the number of switching times is compared with the threshold and a continuation flag or a reset flag is output. The continuation flag corresponds to retaining the current cavity operating limit, injection-production pressure difference and stabilization period duration values. The reset flag corresponds to selecting the next cycle's starting cavity operating limit, injection-production pressure difference and stabilization period duration values. The selected values ​​are written into the next cycle's starting point record to obtain the injection-production cycle succession parameter set.

[0015] As a further aspect of the present invention, the operating status number is specifically obtained by performing a ternary concatenation of the upper limit value of the cavity operating pressure, the single-cycle injection-production pressure difference value, and the stabilization period duration value within the same injection-production time step. The operating status number includes an upper limit level code, a pressure difference level code, and a stabilization level code. The upper limit level code is determined by dividing the upper limit value of the cavity operating pressure and the upper limit value of the cavity operating pressure into multiple intervals and then assigning the corresponding interval number. The pressure difference level code is determined by dividing the single-cycle injection-production pressure difference value and the single-cycle injection-production pressure difference upper limit into multiple intervals and then assigning the corresponding interval number. The stabilization level code is determined by dividing the stabilization period duration value and the lower limit of the stabilization period duration into multiple intervals and then assigning the corresponding interval number.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, the Mankendall trend test outputs trend direction labels and significant trend indicators on the cavity pressure change value sequence, the annular pressure change value sequence and the event increment value sequence. The trend direction labels are combined with the consistent outbound term count to form a stability direction discrimination sequence. The stability direction discrimination sequence suppresses the influence of single-point anomalies on the judgment results at the time step scale and strengthens the constraint of multi-source same-direction changes. In this invention, the Viterbi algorithm outputs a time-step hidden state label sequence based on the stability direction discrimination sequence and completes the location of continuous and consistent segments. The location of continuous and consistent segments separates short-term fluctuations from persistent anomalies and reduces the frequency of triggering treatments under high-frequency injection and production conditions. The hidden state label sequence and the dominant direction of the segment are mapped to the pressure difference reduction amount, the upper limit contraction amount, and the pressure stabilization extension amount, and then combined to form an injection and production control action set. The action parameters are bound to the number of execution cycles and solidified into the execution trajectory of the running state. In this invention, the execution trajectory of the running state maintains constant values ​​in continuous time steps and forms a traceable parameter sequence. The number of state switching times is statistically analyzed and compared with the threshold to output a continuation flag, which is then reset and written into the injection and collection cycle parameter set. The parameter values ​​at the starting point of the cross-cycle have a unified caliber and suppress the cumulative parameter drift. Compared with the link that only relies on threshold overrun and simple trend judgment, the three types of results—trend direction label, significant trend flag, and hidden state label sequence—jointly constrain the risk identification boundary and the continuous boundary of the treatment. The triggering of the treatment action changes from single-point triggering to segment triggering, and the treatment output changes from single parameter adjustment to triple parameter linkage and binding with the number of execution cycles. The risk control link is quantifiable and reproducible. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the main steps of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] Example 1 Please see Figure 1 This invention provides a technical solution: a multi-source fusion closed-loop safety monitoring method for salt cavern gas storage suitable for high-frequency injection and production conditions, comprising the following steps: S1: Read the high-frequency injection and production condition constraint set and determine the injection and production time step length. Around the salt cavern cavity, wellbore annulus section, injection and production valve group and pressure stabilization control section, limit the upper limit of operation, injection and production pressure difference, pressure stabilization time and annulus drift interval and combine them to generate an operation boundary combination table. S2: Based on the operational boundary combination table, the Mankendall trend test is used to compare the pressure change direction, annular change direction, and event increment direction step by step around the pressure-bearing zone of the salt cavern cavity, the annulus section of the wellbore, and the microseismic influence zone. The number of consistent out-of-bounds items is counted and the consistency of the direction is determined to generate a stability direction discrimination sequence. S3: Based on the stability direction discrimination sequence, the Viterbi algorithm is used to determine the number of consecutive consistent time steps around the stable cavity section, the intact wellbore section, and the injection-production disturbance section, and select the section that meets the threshold. For the consistent direction section, the combination of differential pressure reduction, upper limit contraction, and pressure stabilization extension is selected to generate a set of injection-production control actions. S4: Based on the injection and production control action set, around the injection and production valve group, cavity operating area, and pressure stabilization control area, replace the upper limit value of cavity operation and the injection and production pressure difference value and accumulate the pressure stabilization time, keep the value unchanged in continuous time steps, and generate the operating status execution trajectory. S5: Based on the execution trajectory of the operating status, around the salt cavern operation stage, the wellbore safety stage, and the injection-production cycle connection segment, count the number of status switching times and compare them with the threshold, select to continue or reset the action parameters and write them into the starting point of the next cycle to obtain the injection-production cycle connection parameter set.

[0020] The operational boundary combination table includes the upper limit of cavity operating pressure, single-cycle injection-production pressure difference, stabilization period duration, annular pressure drift tolerance range, and cavity pressure fluctuation range tolerance range. The stability direction discrimination sequence includes consistent boundary status flags, direction consistency flags, change direction flags, and the number of boundary items. The injection-production control action set includes the amount of reduction of the upper limit of cavity operating pressure, the amount of reduction of the single-cycle injection-production pressure difference, the amount of extension of the stabilization period, and the number of execution cycles. The operational status execution trajectory includes the upper limit of cavity operating pressure value, the single-cycle injection-production pressure difference value, the stabilization period duration value, and the operational status number. The injection-production cycle inheriting parameter set includes the starting value of the upper limit of cavity operating pressure, the starting value of the single-cycle injection-production pressure difference, the starting value of the stabilization period duration, and the continuation flag.

[0021] The specific steps for generating the run boundary combination table are as follows: Based on reading the high-frequency injection and extraction condition constraint set and determining the injection and extraction time step length, an injection and extraction time step number sequence is generated, the start and end times of the injection and extraction time steps are checked and the time granularity is unified, missing time steps are identified and continuity identifiers are added, the range of available time steps is locked, and the injection and extraction time step reference sequence is obtained. Based on the injection-production time step benchmark sequence, the upper limit value of the salt cavern operation is matched, the single-cycle injection-production pressure difference value is matched, the duration of the pressure stabilization period is matched, and the wellbore annular pressure drift range value is matched. The parameters are then merged into a parameter set according to the time step and written into the index key to generate an operation boundary combination table. Based on reading the high-frequency injection and extraction condition constraint set and determining the injection and extraction time step length, date and time parsing rules are used to perform format validation on the injection and extraction cycle start time string and injection and extraction cycle end time string. The format field uses 4-digit year, 2-digit month, 2-digit day connected with the letter T, and 2-digit hour, 2-digit minute, and 2-digit second. The time zone field uses +08:00 as the identifier. The start time and end time are converted to second-level integers using second-level time conversion rules. The injection and extraction time step length is 60 seconds and used as the step size value. An equal step size increment generation rule is used to generate a sequence from the start second-level integer to the end second-level integer, which is incremented by the step size value. The sequence elements are converted into date and time strings and written into the time step number field. The difference between adjacent sequence elements is checked and equal to 60 and written into the continuity flag with a value of 1. The difference is checked and equal to 60 and written into the continuity flag with a value of 0. The gap start second-level integer and gap end second-level integer are written into the gap. The gap segment is filled and inserted into the sequence according to the equal step size increment generation rule. The first and last elements of the sequence are written into the available flag with a value of 1 and written into the boundary flag field to obtain the injection and extraction time step reference sequence. Based on the injection-production time step benchmark sequence, a key-value matching rule is adopted. Using the time step number field as the index key, the index key is matched with the time key field in the operating upper limit configuration table to extract the maximum cavity operating pressure upper limit value field. The index key is also matched with the time key field in the injection-production pressure difference configuration table to extract the single-cycle injection-production pressure difference value field. Furthermore, the index key is matched with the time key field in the pressure stabilization configuration table to extract the pressure stabilization period duration value field. Finally, the index key is matched with the time key field in the annular drift configuration table to extract the annular pressure drift lower limit value. The field and the annular pressure drift upper limit value field are set up using the default value rules. For unmatched index keys, the maximum cavity operating pressure upper limit is written with a default value of 20 MPa, the single-cycle injection-production pressure difference is written with a default value of 5 MPa, the stabilization period duration is written with a default value of 900 seconds, the annular pressure drift lower limit is written with a default value of -0.2 MPa, and the annular pressure drift upper limit is written with a default value of 0.2 MPa. The aforementioned value fields are merged by the time step number field and written into the parameter set field, the index key field, and the version number field with a value of 1, generating the operating boundary combination table.

[0022] The specific steps for generating the stability direction discrimination sequence are as follows: Based on the operational boundary combination table, the Mankendall trend test was used to read the cavity pressure change value and annulus pressure change value in the order of injection and production time steps around the pressure-bearing zone of the salt cavern cavity, the annulus section of the wellbore, and the microseismic influence zone. The event increment value corresponding to the same time step was also read. The change values ​​of adjacent time steps were marked with positive and negative directions and the change direction labels were recorded to obtain the direction label sequence. Based on the directional labeling sequence, the cavity pressure change amplitude is compared with the pressure fluctuation tolerance range in the operating boundary combination table at each time step, the annular pressure change amplitude is compared with the annular drift tolerance range, the event increment is compared with the event threshold, and the number of boundary-exceeding items at each time step is recorded to obtain the boundary count sequence. Based on the out-of-bounds counting sequence, the consistency of the cavity pressure direction, annular pressure direction and event increment direction in the same time step is compared. The number of out-of-bounds items is judged against the trigger number threshold. A direction consistency flag is output and a consistent out-of-bounds flag is output to generate a stability direction discrimination sequence. Based on the operational boundary combination table, the Mankendall trend test was used. Focusing on the pressure-bearing zone of the salt cavern, the annulus section of the wellbore, and the microseismic influence zone, the pressure change sequence of the cavern, the annulus pressure change sequence, and the event increment sequence were read sequentially according to the injection-production time steps. A significance level of 0.05 and a test window length of 30 injection-production time steps were set. For any sequence within the window, pairwise time-series comparisons were performed, ensuring that the index of the preceding time step was less than the index of the following time step. The difference between the preceding and following sequence values ​​was calculated, and the sign of the difference was determined. A difference greater than 0 was recorded as positive and assigned the value 1. A difference of 0 is recorded as zero and assigned the value 0; a difference less than 0 is recorded as negative and assigned the value -1. The sum of all assigned values ​​is used to obtain the statistical value. Repeated values ​​are counted and written to the repeat count table. The variance of the statistical value is corrected and written to the variance value. The statistical value and variance value are standardized to obtain the standardized statistical value. If the standardized statistical value is greater than 0, it is written to the upward direction label. If the standardized statistical value is less than 0, it is written to the downward direction label. If the standardized statistical value is equal to 0, it is written to the no-direction label. The change direction label is recorded to obtain the direction label sequence. Based on the directional labeling sequence, using the interval comparison judgment rule, the cavity pressure change amplitude is read step by step, and the lower limit and upper limit of the pressure fluctuation allowable interval are read. If the cavity pressure change amplitude is less than the lower limit, an out-of-bounds flag 1 is written and the out-of-bounds type is written as "lower out-of-bounds". If the cavity pressure change amplitude is greater than the upper limit, an out-of-bounds flag 1 is written and the out-of-bounds type is written as "upper out-of-bounds". If the cavity pressure change amplitude falls within the interval, an out-of-bounds flag 0 is written. The annular pressure change amplitude is read step by step, and the lower limit and upper limit of the annular drift allowable interval are read and written as "out-of-bounds" according to the same comparison rule. The event increment value and event threshold value are read step by step. If the event increment value is greater than the event threshold value, an out-of-bounds flag 1 is written and the out-of-bounds type is written as "exceeding threshold". If the event increment value is less than or equal to the event threshold value, an out-of-bounds flag 0 is written. The cavity out-of-bounds flag, annular out-of-bounds flag, and event out-of-bounds flag are summed step by step to obtain the number of out-of-bounds items. The number of out-of-bounds items that occur at each time step is recorded to obtain the out-of-bounds count sequence. Based on the out-of-bounds counting sequence, a consistency comparison rule is adopted to perform an item-by-item equality judgment on the cavity pressure direction label, annular pressure direction label and event increment direction label within the same time step. If the three direction labels are equal, a direction consistency flag of 1 is written; if any direction label is not equal, a direction consistency flag of 0 is written. A threshold comparison rule is adopted to compare the number of out-of-bounds items with the trigger number threshold value. The trigger number threshold value is set to 2. If the number of out-of-bounds items is greater than or equal to 2, a consistency out-of-bounds flag of 1 is written; if the number of out-of-bounds items is less than 2, a consistency out-of-bounds flag of 0 is written. The direction consistency flag and the consistency out-of-bounds flag are written into the discrimination field and the time step number field according to the time step, generating a stable direction discrimination sequence.

[0023] The Mankendall trend test selects sequences of cavity pressure change values, annular pressure change values, and event increment values. The number of injection / impulse time steps is set to be no less than a preset length. For any sequence, pairwise comparisons are performed according to the time step index, ensuring the preceding index is less than the following index. The difference is calculated as the difference between the following and preceding sequence values. A difference greater than zero is marked as positive, a difference equal to zero as zero, and a difference less than zero as negative. All pairwise comparison signs are summed to obtain the statistic. The variance is calculated, and the summation of duplicate counts is corrected. A standardized statistic is calculated based on the statistic and variance. If the standardized statistic is greater than zero, an upward trend label is output; if less than zero, a downward trend label is output; if equal to zero, no trend label is output. A significance level is set, and a critical value is calculated. If the absolute value of the standardized statistic is greater than the critical value, a significant trend indicator is output; if the absolute value is less than or equal to the critical value, a significant trend indicator is output. Mankendall trend test, according to the formula: in: This indicates the first step in the multi-source integrated closed-loop safety monitoring method for salt cavern gas storage under high-frequency injection and production conditions. and The direction of change between each injection and sampling time step is marked with the value. This represents the index of discrete time steps divided according to the injection and extraction operation sequence. This represents the weighting coefficient of the pressure change in the pressure-bearing zone of the salt cavern in the multi-source fusion calculation. This represents the weighting coefficient of the pressure change in the wellbore annulus section in the multi-source fusion calculation. This represents the weighting coefficient of the microseismic event increment in the multi-source fusion calculation. Indicates the first The change in pressure within the pressure-bearing zone of the salt cavern cavity during each injection-production time step. Indicates the first Changes in annular pressure in the wellbore annulus section within each injection-production time step. Indicates the first The event increment value corresponding to the microseismic influence zone within each injection-production time step. Indicates the first Pressure change in the pressure-bearing zone of the salt cavern cavity within each injection-production time step Indicates the first Changes in annular pressure in the wellbore annulus section within each injection-production time step. Indicates the first The event increment value corresponding to the microseismic influence zone within each injection and production time step; Execution process: Construct a discrete injection-production time step index according to the injection-production operation sequence. Based on the operational boundary combination table, the pressure change value of the salt cavern cavity corresponding to the pressure-bearing zone is read in each injection-production time step. Changes in annular pressure corresponding to the annular section of the wellbore and the event increment value corresponding to the microseismic influence zone The three changes are respectively compared with the pre-set weighting coefficients. , , Multiply and linearly superimpose to obtain the multi-source fusion change at the current injection-import time step, then apply the changes to the adjacent... With the The difference between the multi-source fusion changes corresponding to each injection and extraction time step is calculated, and the difference result is input into the sign determination function to determine the direction of change. When the difference is greater than zero, a positive direction label value is output; when the difference is less than zero, a negative direction label value is output; and when the difference is equal to zero, a zero direction label value is output. Thus, a direction label sequence for closed-loop safety monitoring is formed according to the injection and extraction time step sequence.

[0024] The specific steps for generating the injection-progression control action set are as follows: Based on the stability direction discrimination sequence, the Viterbi algorithm is used to scan the consistent out-of-bounds flags along the time axis, record the number of time steps that are continuously maintained in the consistent out-of-bounds state, mark the start and end points of the continuous segments, filter the segments with the number of consecutive segments reaching the threshold, and obtain the set of continuous consistent segments. Based on the continuous and consistent segment set, the number of times the consistent direction marker appears in each segment is counted. The direction is selected as the dominant direction of the segment based on the number of occurrences. The dominant direction is then mapped to the injection-production pressure difference adjustment level, the cavity operation upper limit adjustment level, and the pressure stabilization period adjustment level to form a segment parameter combination and obtain an action parameter combination set. Based on the action parameter combination set, the number of execution cycles is assigned to each segment and the effective time step range is bound. The action parameters of adjacent segments are checked for conflicts and replacement or merging is performed to form segment action items arranged in time order and generate injection and extraction control action set. Based on the stability direction discrimination sequence, the Viterbi algorithm is used to scan for consistent out-of-bounds flags along the time axis. The hidden state set is set to 9 items, obtained by combining 3 items from the wellbore integrity level and 3 items from the cavity stability level. The wellbore integrity level includes normal, degradation, and failure warnings; the cavity stability level includes stable, degradation, and instability warnings. An initial state probability table is set to 9 items, all uniformly set to 0.111. The state transition probability matrix is ​​set to 9 x 9 items, allowing only same-level maintenance and adjacent-level migration, with a maintenance migration probability of 0.60. The migration probability of adjacent levels is set to 0.40, with 0.20 evenly distributed in both adjacent directions. Cross-level transitions are prohibited and are set to 0. The observation symbol set is defined with 6 items, discretely obtained by setting the direction consistency flag to 0 or 1, the consistency boundary crossing flag to 0 or 1, and the number of boundary crossing items to 0 to 3. An observation probability table is set to establish an 18-item mapping between hidden states and observation symbols, and the observation probability values ​​are written into the table. The observation probability values ​​are set to three levels: 0.70, 0.20, and 0.10. When the consistency boundary crossing flag is 1, 0.70 is allocated to high-risk hidden states and then to non-high-risk hidden states. Allocate 0.10 to the remaining hidden states and 0.20 to the remaining hidden states. When the consistent out-of-bounds flag is 0, allocate 0.70 to the non-high-risk hidden states, 0.10 to the high-risk hidden states, and 0.20 to the remaining hidden states. In time step 1, calculate the path probability for each of the 9 hidden states and write it into the path probability table and the predecessor index table. From time step 2 to the end of time step 1, generate a candidate path probability set for each hidden state. Multiply the 9 path probabilities from the previous time step by the state transition probability and then by the observation probability to obtain 9 candidate values. Sort the candidate values ​​in descending order and select... The first candidate value is sorted and written into the path probability table and the corresponding predecessor index. At the end of the time step, the 9 termination path probabilities are sorted in descending order and the first hidden state is selected as the termination hidden state. The hidden state label sequence is obtained by backtracking along the predecessor index table to time step 1. The set of hidden states that meet the high-risk hidden state in the hidden state label sequence is marked as a consistent out-of-bounds state and continuous counting is performed. The continuous counting threshold is 5 and the start time step number and end time step number of the continuous segment are marked. The segments with a continuous count of not less than 5 are filtered to obtain the set of continuous consistent segments. Based on a continuous and consistent segment set, a counting comparison rule is adopted. Within each segment, the cumulative number of time steps for the directional consistency flag is set to 1 and recorded as the consistency count value. Within each segment, the cumulative number of time steps for the directional change label is set as the rising time step count value, and the cumulative number of time steps for the directional change label is set as the falling time step count value. The rising and falling count values ​​are compared, and the direction label with the larger number is output as the dominant direction of the segment. When the rising count value equals the falling count value, no direction label is output as the dominant direction of the segment. The dominant direction of the segment is mapped to the injection-production pressure difference adjustment level, the cavity operation upper limit adjustment level, and the pressure stabilization period adjustment level. The injection-production pressure difference adjustment level is set to 3 levels, with level 1 decreasing by 0.5 MPa, level 2 decreasing by 1.0 MPa, and level 3 decreasing by 1.0 MPa. The pressure is lowered by 1.5 MPa. The upper limit of cavity operation is set to level 3, with level 1 lowered by 0.5 MPa, level 2 lowered by 1.0 MPa, and level 3 lowered by 1.5 MPa. The stabilization period is set to level 3, with level 1 extended by 300 seconds, level 2 extended by 600 seconds, and level 3 extended by 900 seconds. The upward direction label is mapped to injection-production pressure difference adjustment level 2, cavity operation upper limit adjustment level 2, and stabilization period adjustment level 1. The downward direction label is mapped to injection-production pressure difference adjustment level 1, cavity operation upper limit adjustment level 1, and stabilization period adjustment level 2. The no-direction label is mapped to injection-production pressure difference adjustment level 1, cavity operation upper limit adjustment level 1, and stabilization period adjustment level 1. This forms a combination of segment parameters, resulting in a set of action parameter combinations. Based on the action parameter combination set, a segment constraint scheduling rule is adopted to allocate an execution cycle number to each segment and bind an effective time step range. The execution cycle number is set to 3 and written into the segment execution cycle number field. The effective time step range is written into the segment start time step number field and the segment end time step number field. The action parameter conflict verification between adjacent segments is carried out by comparing fields and determining the conflict relationship. The conflict relationship is set as follows: the adjustment level of the stabilization period is greater than 1 and the injection-production pressure difference adjustment level is less than 2, which is determined to be conflict relationship A. The conflict relationship is set as follows: the upper limit adjustment level of the cavity operation is greater than 1 and the injection-production pressure difference adjustment level is equal to 1, which is determined to be conflict relationship B. When conflict relationship A is triggered, the injection-production pressure difference adjustment level is replaced with 2 while the adjustment level during the stabilization period remains unchanged. When conflict relationship B is triggered, the upper limit adjustment level of the cavity operation is replaced with 1 while the injection-production pressure difference adjustment level remains unchanged. The segment merging condition is set as follows: adjacent segments have the same dominant direction, the same injection-production pressure difference adjustment level, the same upper limit adjustment level of the cavity operation, and the same adjustment level during the stabilization period. When the merging condition is met, the end time step number of the previous segment is replaced with the end time step number of the next segment, and the entry of the next segment is deleted, forming segment action entries arranged in chronological order, and generating an injection-production control action set.

[0025] The Viterbi algorithm sets a set of hidden states and an observation sequence, and sets an initial state probability table, a state transition probability matrix, and an observation probability table. At time step one, it calculates the path probability for each hidden state and writes it into the path probability table and the predecessor index table. From time step two to the end of time step one, it calculates a set of candidate path probabilities for each hidden state. The set of candidate path probabilities is obtained by multiplying the path probability, state transition probability, and observation probability of the previous time step. The set of candidate path probabilities is sorted in descending order, and the path probability of the first-ranked path is selected and written into the path probability table, and the corresponding predecessor index is written into the predecessor index table. At the end of time step one, the path probability table is sorted in descending order, and the hidden state of the first-ranked path is selected as the final hidden state. The algorithm backtracks along the predecessor index table to time step one to obtain the hidden state sequence and outputs the hidden state label sequence of the time step. Viterbi algorithm, according to the formula: in: This represents the optimal consistent out-of-bounds state sequence obtained by decoding using an improved Viterbi algorithm in a multi-source fusion closed-loop safety monitoring method for salt cavern gas storage applicable to high-frequency injection and production conditions. Represents the set of candidate hidden state sequences. This represents the maximization operator used to determine the sequence of hidden states that maximizes the objective probability function. This represents the total number of monitoring time steps obtained by dividing the process according to the sequence of high-frequency injection and extraction operations. This represents the time step index that increments along the monitoring time axis. This indicates that the hidden state at the initial monitoring time step is... The prior probability value, This indicates that the hidden state at the initial time step is The observation probability value obtained by combining the consistent out-of-bounds observation indicator and multi-source correction parameters under the given conditions. Indicates the first The hidden state at each time step is The observation probability value obtained by combining the consistent out-of-bounds observation indicator and multi-source correction parameters under the given conditions. Indicates the first The hidden state at each time step is determined by To the Hidden state at each time step During the transition, the operating boundary and the state transition probability value after closed-loop control correction are introduced. This represents the consistent out-of-bounds observation flag value generated by the stability direction discrimination sequence at the initial time step. Indicates the first Consistent out-of-bounds observation flag values ​​generated by the stability direction discrimination sequence at each time step. This represents the density correction factor for consistent out-of-bounds observations within the time window at the initial time step. Indicates the first Density correction coefficients for consistent out-of-bounds observations within the time window at each time step. This represents the penalty coefficient for orientation reversal in the initial time-step stability orientation discrimination sequence on consistent out-of-bounds observations. Indicates the first The penalty coefficient for consistent out-of-bounds observations caused by orientation reversal in the time-step stability orientation discrimination sequence. This represents the suppression coefficient of the initial time-step microseismic event disturbance on the reliability of consistent out-of-bounds observations. Indicates the first The suppression coefficient of microseismic event perturbation at each time step on the reliability of consistent out-of-bounds observations. Indicates the first The boundary compression correction coefficient of the state transition probability due to the change in safety margin in the boundary combination table for each time step. Indicates the first The correction coefficient for the closed-loop control action at each time step to correct the state from uniform out-of-bounds to non-uniform out-of-bounds regression. Execution process: The monitoring process is discretized according to the high-frequency injection and extraction operation sequence. Each time step is used to construct a consistent out-of-bounds observation sequence. The initial hidden state prior probability is set based on historical statistical results. Then at each time step Integral stability direction discrimination sequence calculation consistent out-of-bounds observation flag And within a preset sliding window, the proportion of consistent out-of-bounds occurrences is statistically analyzed to form a consistent out-of-bounds density correction coefficient. The number of positive and negative flips of the directional label sequence within the same window is counted to form a trend consistency penalty coefficient. It also reads the incremental changes of events in the microseismic-affected area and generates microseismic disturbance suppression coefficients according to the sudden suppression rules. ,Will Common input observation probability model to obtain hidden state Observation probability under certain conditions Simultaneously, the safety margin status for the corresponding time step is read from the running boundary combination table, and the running boundary compression correction coefficient is constructed. The intensity of the control action is read from the closed-loop control record, and a closed-loop correction coefficient is constructed. ,Will and Introducing a hidden state transition model to obtain the result of... Towards Corrected transition probability The hidden state sequence is recursively calculated along the time axis for all candidate hidden state sequences, and the product of the initial probability, transition probability, and observation probability is accumulated. The maximization criterion is then used to determine the hidden state sequence that maximizes the overall probability function. Finally, scan along the time axis. The time steps marked as consistent out-of-bounds states are counted and their start and end positions are marked. Segments whose continuous length reaches the threshold are selected to form a set of continuous consistent segments, which are then used as input for closed-loop safety monitoring criteria.

[0026] The specific steps for generating the runtime execution trajectory are as follows: Based on the injection and release control action set, align the action items with the injection and release time step numbers, mark the effective start point and the ineffective end point of the action, check the number of time steps covered by the action and remove overlapping and conflicting sections, splice the continuous coverage range in time order, and obtain the action parameter execution section. Based on the action parameters, the upper limit value of the cavity operation is replaced with the value after the corresponding reduction in the segment, the single-cycle injection-production pressure difference value is replaced with the value after the corresponding reduction in the segment, the duration of the pressure stabilization period is replaced with the value after the corresponding accumulation in the segment, the fixed parameter values ​​within the segment are marked with the operation status number, and the operation status execution trajectory is generated. Based on the injection and collection control action set, an interval sorting and conflict resolution rule is adopted. For each action item, the starting time step number, ending time step number, and action parameter fields are read. The items are sorted in ascending order by the starting time step number, and within the same starting scenario, in ascending order by the ending time step number. The effective starting point is determined by the starting time step number, and the ineffective ending point is determined by the ending time step number, both written to the segment field. The coverage time step count is calculated by subtracting the starting time step number from the ending time step number and adding 1, then written to the coverage count field. A conflict criterion is set to consider overlapping segments and ensure that the starting time step number of the subsequent item is less than or equal to the previous one. The entry's end time step number, the conflict resolution priority field takes the upper limit reduction of the cavity operation, the injection-production pressure difference reduction, and the pressure stabilization period extension, weighted summed and weighted at 0.4, 0.4, and 0.2. Conflict resolution actions are retained and other overlapping entries are deleted in descending order of priority field. The splicing rule is to take the end time step number of the previous entry plus 1, which equals the start time step number of the next entry, to determine continuous coverage and merge the segments. When merging, the end time step number is taken from the end time step number of the next entry, and the action parameter field takes the parameter of the entry with the larger priority field value, and the action parameter is used to obtain the execution segment. Based on the action parameter execution segment, a triplet coding rule is adopted. For each segment, the maximum cavity operating pressure upper limit benchmark value is read, and the maximum cavity operating pressure upper limit reduction amount is read and subtracted to obtain the cavity operating upper limit value. The single-cycle injection-production pressure difference benchmark value is read, and the single-cycle injection-production pressure difference reduction amount is read and subtracted to obtain the single-cycle injection-production pressure difference value. The pressure stabilization period duration benchmark value is read, and the pressure stabilization period extension amount is read and added to obtain the pressure stabilization period duration value. The cavity operating upper limit value is written at each time step within the segment. The single-cycle injection-production pressure difference value is written into the stabilization period duration value. The operation status number is obtained by concatenating the upper limit level code, the pressure difference level code, and the stabilization level code. The upper limit level code segment thresholds are 18 MPa, 20 MPa, 22 MPa, and 24 MPa. The pressure difference level code segment thresholds are 3 MPa, 4 MPa, 5 MPa, and 6 MPa. The stabilization level code segment thresholds are 600 seconds, 900 seconds, 1200 seconds, and 1500 seconds. The level code is taken from the threshold interval number 1 to 5 and written into the operation status number field as three fields to generate the operation status execution trajectory.

[0027] The specific steps for generating the injection-production cycle acceptance parameter set are as follows: Based on the execution trajectory of the running status, the running status number is read step by step and compared with the previous time step number. The same identifier is recorded and the change identifier is recorded. The number of changes within the cycle is accumulated and the time step of the change is located. The cycle switching statistics are summarized to obtain the status switching statistics. Based on the state switching statistics, the number of switching times is compared with the threshold and a continuation flag or a reset flag is output. The continuation flag corresponds to retaining the current cavity operating limit, injection-production pressure difference and stabilization period duration. The reset flag corresponds to selecting the next cycle's starting cavity operating limit, injection-production pressure difference and stabilization period duration. The selected values ​​are written into the next cycle's starting point record to obtain the injection-production cycle succession parameter set. Based on the execution trajectory of the running status, the run-length statistics rule is adopted. The running status number is read step by step and compared with the previous time step number. The comparison rule is to determine the same if the strings are completely equal and to determine the change if the strings are not equal. The same identifier is written to the same identifier field with a value of 1 and changed identifier field with a value of 0. The change identifier is written to the same identifier field with a value of 0 and changed identifier field with a value of 1. The cycle boundary is the injection and sampling cycle number field, and the change of injection and sampling cycle number is used as the cycle switching point. The number of changes within the cycle is summed from the change identifier field and written to the switching number field. The time step where the change occurs is located, and the list of time step numbers corresponding to the change identifier field with a value of 1 is written to the switching time step list field. The running status number list is the running status number corresponding to all time step numbers within the cycle and written to the status sequence field. The summary field is the injection and sampling cycle number field, switching number field, switching time step list field and status sequence field and written to the statistical entry table to obtain the status switching statistics. Based on state switching statistics, a threshold branch selection rule is adopted. The switching count field is read, and the switching count threshold value is read and set to 3. The comparison rule is that if the switching count is less than or equal to 3, the continuation flag field is written with a value of 1 and the reset flag field is written with a value of 0. If the switching count is greater than 3, the continuation flag field is written with a value of 0 and the reset flag field is written with a value of 1. In the scenario where the continuation flag field is 1, the end-of-cycle time step number is read, the end-of-cycle cavity operating upper limit value is read, the end-of-cycle single-cycle injection-production pressure difference value is read, and the end-of-cycle stabilization period duration value is read. In the scenario where the reset flag field is 1, the operating boundary combination table is read, and the maximum cavity operating pressure upper limit value corresponding to the start time step number of the next injection-production cycle is read, the single-cycle injection-production pressure difference value is read, and the stabilization period duration value is read. The record is written using the start time step number field of the next injection-production cycle, the cavity operating upper limit value field, the single-cycle injection-production pressure difference value field, the stabilization period duration value field, the continuation flag field, and the reset flag field, to obtain the injection-production cycle continuity parameter set.

[0028] The operational status number is specifically obtained by concatenating the upper limit of the cavity operating pressure, the single-cycle injection-production pressure difference, and the duration of the pressure stabilization period within the same injection-production time step into a ternary digit. The operational status number includes an upper limit level code, a pressure difference level code, and a pressure stabilization level code. The upper limit level code is determined by dividing the upper limit of the cavity operating pressure into multiple intervals and then assigning the corresponding interval number. The pressure difference level code is determined by dividing the single-cycle injection-production pressure difference into multiple intervals and then assigning the corresponding interval number. The pressure stabilization level code is determined by dividing the duration of the pressure stabilization period into multiple intervals and then assigning the corresponding interval number.

[0029] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A multi-source fusion closed-loop safety monitoring method for salt cavern gas storage facilities suitable for high-frequency injection and production conditions, characterized in that, Includes the following steps: S1: Read the high-frequency injection and production condition constraint set and determine the injection and production time step length. Around the salt cavern cavity, wellbore annulus section, injection and production valve group and pressure stabilization control section, limit the upper limit of operation, injection and production pressure difference, pressure stabilization time and annulus drift interval and combine them to generate an operation boundary combination table. S2: Based on the aforementioned operational boundary combination table, the Mankendall trend test is used to compare the pressure change direction, annular change direction, and event increment direction step by step around the pressure-bearing zone of the salt cavern cavity, the annulus section of the wellbore, and the microseismic influence zone. The number of consistent out-of-bounds items is counted and the consistency of the direction is determined to generate a stability direction discrimination sequence. S3: Based on the stability direction discrimination sequence, the Viterbi algorithm is used to determine the number of consecutive consistent time steps around the cavity stable section, the wellbore intact section, and the injection-production disturbance section, and select the section that meets the threshold. For the consistent direction section, the combination of pressure difference reduction, upper limit contraction, and pressure stabilization extension is selected to generate the injection-production control action set. S4: Based on the injection and extraction control action set, around the injection and extraction valve group, cavity operating area, and pressure stabilization control area, replace the cavity operating upper limit value and the injection and extraction pressure difference value and accumulate the pressure stabilization time, keep the value unchanged in continuous time steps, and generate the operating state execution trajectory. S5: Based on the execution trajectory of the operating status, around the salt cavern operation stage, the wellbore safety stage, and the injection-production cycle connection segment, count the number of status switching times and compare them with the threshold, select to continue or reset the action parameters and write them into the starting point of the next cycle to obtain the injection-production cycle connection parameter set.

2. The multi-source fusion closed-loop safety monitoring method for salt cavern gas storage facilities applicable to high-frequency injection and production conditions as described in claim 1, characterized in that, The operational boundary combination table includes the upper limit of cavity operating pressure, single-cycle injection-production pressure difference, stabilization period duration, annular pressure drift tolerance range, and cavity pressure fluctuation amplitude tolerance range. The stability direction discrimination sequence includes a consistent out-of-bounds status flag, a direction consistency flag, a change direction flag, and the number of out-of-bounds items. The injection-production control action set includes the reduction amount of the upper limit of cavity operating pressure, the reduction amount of the single-cycle injection-production pressure difference, the extension amount of the stabilization period, and the number of execution cycles. The operational status execution trajectory includes the upper limit of cavity operating pressure value, the single-cycle injection-production pressure difference value, the stabilization period duration value, and the operational status number. The injection-production cycle inheriting parameter set includes the initial value of the upper limit of cavity operating pressure, the initial value of the single-cycle injection-production pressure difference, the initial value of the stabilization period duration, and a continuation flag.

3. The multi-source fusion closed-loop safety monitoring method for salt cavern gas storage facilities applicable to high-frequency injection and production conditions as described in claim 1, characterized in that, The specific steps for generating the running boundary combination table are as follows: Based on reading the high-frequency injection and extraction condition constraint set and determining the injection and extraction time step length, an injection and extraction time step number sequence is generated, the start and end times of the injection and extraction time steps are checked and the time granularity is unified, missing time steps are identified and continuity identifiers are added, the range of available time steps is locked, and the injection and extraction time step reference sequence is obtained. Based on the injection-production time step benchmark sequence, the upper limit value of the salt cavern operation is matched, the single-cycle injection-production pressure difference value is matched, the duration of the pressure stabilization period is matched, and the wellbore annular pressure drift range value is matched. These are then merged into a parameter set according to the time step and written into the index key to generate an operation boundary combination table.

4. The multi-source fusion closed-loop safety monitoring method for salt cavern gas storage facilities applicable to high-frequency injection and production conditions as described in claim 1, characterized in that, The specific steps for generating the stability direction discrimination sequence are as follows: Based on the aforementioned operational boundary combination table, the Mankendall trend test is used to read the cavity pressure change value and annulus pressure change value in the order of injection and production time steps around the pressure-bearing zone of the salt cavern cavity, the annulus section of the wellbore, and the microseismic influence zone. The event increment value corresponding to the same time step is also read. The change values ​​of adjacent time steps are marked with positive and negative directions and the change direction label is recorded to obtain the direction label sequence. Based on the directional labeling sequence, the cavity pressure change amplitude is compared with the pressure fluctuation tolerance range in the operating boundary combination table at each time step, the annular pressure change amplitude is compared with the annular drift tolerance range, the event increment is compared with the event threshold, the number of boundary items that occur at each time step is recorded, and the boundary count sequence is obtained. Based on the out-of-bounds counting sequence, the consistency of the cavity pressure direction, annular pressure direction and event increment direction in the same time step is compared. The number of out-of-bounds items is judged against the trigger number threshold. A direction consistency flag is output and a consistent out-of-bounds flag is output to generate a stability direction discrimination sequence.

5. The multi-source fusion closed-loop safety monitoring method for salt cavern gas storage facilities applicable to high-frequency injection and production conditions as described in claim 4, characterized in that, The Mankendall trend test selects the sequence of cavity pressure change values, the sequence of annular pressure change values, and the sequence of event increment values. The number of injection and collection time steps is set to be no less than a preset length. For any sequence, pairwise comparisons are performed according to the time step index, ensuring that the previous index is less than the next index. The difference is calculated as the difference between the next sequence value and the previous sequence value. A difference greater than zero is marked as positive, a difference equal to zero as zero, and a difference less than zero as negative. All pairwise comparison signs are summed to obtain the statistic. The variance is calculated, and the summation of duplicate value counts is corrected. Based on the statistic and variance, a standardized statistic is calculated. If the standardized statistic is greater than zero, an upward trend label is output; if the standardized statistic is less than zero, a downward trend label is output; if the standardized statistic is equal to zero, a no-trend label is output. A significance level is set, and a critical value is calculated. If the absolute value of the standardized statistic is greater than the critical value, a significant trend indicator is output; if the absolute value of the standardized statistic is less than or equal to the critical value, a significant trend indicator is output.

6. The multi-source fusion closed-loop safety monitoring method for salt cavern gas storage facilities applicable to high-frequency injection and production conditions as described in claim 1, characterized in that, The specific steps for generating the injection and extraction control action set are as follows: Based on the stability direction discrimination sequence, the Viterbi algorithm is used to scan the consistent out-of-bounds flags along the time axis, record the number of time steps that are continuously maintained in the consistent out-of-bounds state, mark the start and end points of the continuous segments, filter the segments with the number of consecutive segments reaching the threshold, and obtain the set of continuous consistent segments. Based on the continuous and consistent segment set, the number of times the direction consistency mark appears in each segment is counted. The direction is selected as the dominant direction of the segment based on the number of occurrences. The dominant direction is then mapped to the injection-production pressure difference adjustment level, the cavity operation upper limit adjustment level, and the pressure stabilization period adjustment level to form a segment parameter combination and obtain an action parameter combination set. Based on the set of action parameters, the number of execution cycles is assigned to each segment and the effective time step range is bound. The action parameters of adjacent segments are checked for conflicts and replacement or merging is performed to form segment action entries arranged in chronological order, and an injection and extraction control action set is generated.

7. The multi-source fusion closed-loop safety monitoring method for salt cavern gas storage facilities applicable to high-frequency injection and production conditions as described in claim 6, characterized in that, The Viterbi algorithm sets a set of hidden states and an observation sequence, and sets an initial state probability table, a state transition probability matrix, and an observation probability table. At time step one, it calculates the path probability for each hidden state and writes it into the path probability table and the predecessor index table. From time step two to the end of time step one, it calculates a set of candidate path probabilities for each hidden state. The set of candidate path probabilities is obtained by multiplying the path probability, state transition probability, and observation probability of the previous time step. The set of candidate path probabilities is sorted in descending order, and the path probability of the first-ranked path is selected and written into the path probability table, and the corresponding predecessor index is written into the predecessor index table. At the end of time step one, the path probability table is sorted in descending order, and the hidden state of the first-ranked path is selected as the terminating hidden state. The algorithm backtracks along the predecessor index table to time step one to obtain the hidden state sequence and outputs the hidden state label sequence of the time step.

8. The multi-source fusion closed-loop safety monitoring method for salt cavern gas storage facilities applicable to high-frequency injection and production conditions as described in claim 1, characterized in that, The specific steps for generating the execution trajectory of the running state are as follows: Based on the injection and extraction control action set, align the action entries with the injection and extraction time step numbers, mark the effective start point and the ineffective end point of the action, check the number of time steps covered by the action and remove overlapping and conflicting segments, splice the continuous coverage range in time order, and obtain the action parameter execution segment. Based on the action parameters, the upper limit value of the cavity operation is replaced with the corresponding value after downward adjustment in the segment, the single-cycle injection-production pressure difference value is replaced with the corresponding value after downward adjustment in the segment, the duration of the pressure stabilization period is replaced with the corresponding value after accumulation in the segment, the fixed parameter values ​​within the segment are marked with the operation status number, and the operation status execution trajectory is generated.

9. The multi-source fusion closed-loop safety monitoring method for salt cavern gas storage facilities applicable to high-frequency injection and production conditions as described in claim 1, characterized in that, The specific steps for generating the injection-production cycle acceptance parameter set are as follows: Based on the execution trajectory of the running status, the running status number is read step by step and compared with the previous time step number. The same identifier is recorded and the change identifier is recorded. The number of changes within the cycle is accumulated and the time step of the change is located. The cycle switching statistics are summarized to obtain the state switching statistics. Based on the state switching statistics, the number of switching times is compared with the threshold and a continuation flag or a reset flag is output. The continuation flag corresponds to retaining the current cavity operating limit, injection-production pressure difference and stabilization period duration values. The reset flag corresponds to selecting the next cycle's starting cavity operating limit, injection-production pressure difference and stabilization period duration values. The selected values ​​are written into the next cycle's starting point record to obtain the injection-production cycle succession parameter set.

10. The multi-source fusion closed-loop safety monitoring method for salt cavern gas storage facilities applicable to high-frequency injection and production conditions as described in claim 9, characterized in that, The operational status number is specifically obtained by performing a ternary concatenation operation on the upper limit of the cavity operating pressure, the single-cycle injection-production pressure difference, and the duration of the pressure stabilization period within the same injection-production time step. The operational status number includes an upper limit level code, a pressure difference level code, and a pressure stabilization level code. The upper limit level code is determined by dividing the upper limit of the cavity operating pressure into multiple intervals and then assigning the corresponding interval number. The pressure difference level code is determined by dividing the single-cycle injection-production pressure difference into multiple intervals and then assigning the corresponding interval number. The pressure stabilization level code is determined by dividing the duration of the pressure stabilization period into multiple intervals and then assigning the corresponding interval number.

Images