Coal bed gas well production dynamic analysis method and system based on AI time series prediction

Abstract

The present application provides a kind of coalbed gas well production dynamic analysis method and system based on AI timing prediction, first, the production dynamic data sequence of coalbed gas well, corresponding well site geological condition data sequence and surrounding environment data sequence are collected, and the ternary timing dynamic association network containing node dynamic timing segment and connection real-time association is constructed.Then call the pre-trained AI timing coupling deviation evolution mining model to process the ternary timing dynamic association network, obtain the ternary timing coupling deviation result.According to this ternary timing coupling deviation result, the deviation conduction closed loop path is determined by combining forward conduction positioning and reverse tracing verification, and the deviation conduction analysis report is generated.Finally, based on the deviation conduction analysis report, the dynamic control strategy is generated in combination with the production dynamic prediction threshold and output to the production control terminal, while the network node association relationship is updated by feedback control data, to realize the accurate analysis and control of coalbed gas well production dynamics.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to a coalbed methane well production dynamic analysis method and system based on AI time series prediction. BACKGROUND

[0002] In the production process of a coalbed methane well, accurate analysis of production dynamics and formulation of reasonable control strategies are crucial for improving production efficiency and ensuring production safety. Traditional coalbed methane well production dynamic analysis methods mainly analyze based on a single data type. For example, only production dynamic data such as gas production rate and bottom hole pressure changes are used to evaluate production conditions, but the above method ignores the influence of geological conditions and environmental factors on production. In terms of geological conditions, changes in coal seam permeability and porosity directly affect the migration and production of coalbed methane, but the traditional method fails to effectively correlate them with production dynamic data. The surrounding environmental factors, such as surface temperature and humidity, also indirectly affect the production of coalbed methane wells, which are also not fully considered in traditional analysis. In addition, existing technologies are relatively superficial in analyzing the relationship between different data sequences, lacking in-depth mining of dynamic correlation and coupling deviation evolution between data, making it difficult to accurately grasp the key factors and transmission paths affecting production dynamics, resulting in a lack of scientific basis for formulating production control strategies, and failing to achieve efficient and accurate management of coalbed methane well production. SUMMARY

[0003] In view of the above-mentioned problems, in combination with the first aspect of the present application, the present application embodiment provides a coalbed methane well production dynamic analysis method based on AI time series prediction, which comprises:

[0004] Collecting production dynamic data sequences of a coalbed methane well, corresponding geological condition data sequences of the well site, and surrounding environmental data sequences, the production dynamic data sequences containing gas production rate change information and bottom hole pressure change information, the geological condition data sequences containing coal seam permeability change information and coal seam porosity change information, and the environmental data sequences containing surface temperature change information and surface humidity change information;

[0005] Based on the production dynamic data sequences, the geological condition data sequences and the environmental data sequences, a ternary time series dynamic correlation network of a coalbed methane well is constructed, each node in the ternary time series dynamic correlation network corresponding to a dynamic time series segment of a single data sequence, the connection between nodes corresponding to the real-time correlation relationship between dynamic time series segments of different data sequences, and the connection weight evolving and updating over time;

[0006] A pre-trained AI time series coupling deviation evolution mining model is called to process the ternary time series dynamic correlation network, mine the coupling deviation evolution information between different data sequence dynamic time series segments, and obtain a ternary time series coupling deviation result containing deviation evolution trend.

[0007] According to the ternary time sequence coupling deviation result, combined with forward conduction positioning and reverse trace verification, a deviation conduction closed loop path affecting the production dynamics is determined, and a deviation conduction analysis report containing the deviation conduction direction, the correlation strength and the trace verification result is generated;

[0008] Based on the deviation conduction analysis report, combined with a production dynamic prediction threshold, a dynamic regulation strategy for the coal bed gas well production parameters is generated, and the dynamic regulation strategy is output to a production control terminal, and the regulated data is fed back to the ternary time sequence dynamic correlation network to update the node correlation relationship.

[0009] In another aspect, the embodiment of the present application also provides a coal bed gas well production dynamic analysis system based on AI time sequence prediction, which comprises a processor and a machine readable storage medium, the machine readable storage medium is connected with the processor, the machine readable storage medium is used for storing programs, instructions or codes, and the processor is used for executing the programs, instructions or codes in the machine readable storage medium to realize the above-mentioned method.

[0010] Based on the above aspects, by comprehensively collecting the production dynamic data sequence of the coal bed gas well, the geological condition data sequence corresponding to the well site and the surrounding environment data sequence, a ternary time sequence dynamic correlation network of the coal bed gas well is constructed, the organic integration and dynamic correlation analysis of the multi-source heterogeneous data are realized, the pre-trained AI time sequence coupling deviation evolution mining model is called to process the network, the coupling deviation evolution information between different data sequence dynamic time sequence segments can be deeply mined, based on the obtained deviation evolution result, combined with forward conduction positioning and reverse trace verification, the deviation conduction closed loop path affecting the production dynamics can be accurately determined, and a comprehensive and accurate deviation conduction analysis report is generated. According to the deviation conduction analysis report, a dynamic regulation strategy is generated combined with a production dynamic prediction threshold, and the regulation data is fed back to update the network, thereby effectively improving the accuracy of the coal bed gas well production dynamic analysis. BRIEF DESCRIPTION OF DRAWINGS

[0011] Figure 1 is the execution flow schematic diagram of the coal bed gas well production dynamic analysis method based on AI time sequence prediction provided by the embodiment of the present application.

[0012] Figure 2 is the schematic diagram of the exemplary hardware and software components of the coal bed gas well production dynamic analysis system based on AI time sequence prediction provided by the embodiment of the present application. DETAILED DESCRIPTION

[0013] The present application will be specifically described below in conjunction with the drawings of the specification, Figure 1is a flowchart of a coalbed methane well production dynamic analysis method based on AI time series prediction provided by an embodiment of the present application. The coalbed methane well production dynamic analysis method based on AI time series prediction will be described in detail below.

[0014] Step S110: Collecting a production dynamic data sequence of a coalbed methane well, a corresponding well site geological condition data sequence, and a surrounding environment data sequence, the production dynamic data sequence containing gas production rate change information and well bottom pressure change information, the geological condition data sequence containing coal seam permeability change information and coal seam porosity change information, and the environment data sequence containing surface temperature change information and surface humidity change information.

[0015] In this embodiment, a production well in a certain coalbed methane field is taken as an application object, and data collection is completed by various sensors and monitoring devices deployed in the production well. The production dynamic data sequence is collected by a gas production rate sensor and a well bottom pressure sensor installed at the wellhead. The gas production rate sensor records the gas volume change in unit time in real time, and the well bottom pressure sensor transmits the well bottom pressure change data through the downhole cable. Both of them generate gas production rate change information and well bottom pressure change information containing multiple time data points according to a preset collection interval, and together constitute the production dynamic data sequence.

[0016] The geological condition data sequence is collected relying on a downhole geological monitoring device. The coal seam permeability change information is obtained by a permeability tester, which calculates the permeability change by injecting a test fluid into the coal seam and monitoring the pressure response. The coal seam porosity change information is collected by a porosity logging instrument, which obtains porosity data using acoustic or neutron logging principles. Both types of data are recorded at a fixed period to form the geological condition data sequence.

[0017] The surrounding environment data sequence is collected by an environmental monitoring station installed around the well site. The surface temperature change information is recorded in real time by a temperature sensor, and the surface humidity change information is obtained by a humidity sensor. Both types of environmental parameters generate time series data at the same collection interval to form the environment data sequence. During the collection process, all data are attached with time stamps and sent to the data processing center through an encrypted transmission protocol to prevent data leakage and ensure data security.

[0018] Step S120: Based on the production dynamic data sequence, the geological condition data sequence, and the environment data sequence, a ternary time series dynamic correlation network of the coalbed methane well is constructed. Each node in the ternary time series dynamic correlation network corresponds to a dynamic time series segment of a single data sequence, the connection between nodes corresponds to the real-time correlation relationship between dynamic time series segments of different types of data sequences, and the connection weight evolves and updates over time.

[0019] In this embodiment, the construction of ternary time sequence dynamic correlation network needs to go through data preprocessing, time sequence segment division, feature extraction, correlation strength calculation, network building and weight updating, which is realized through the following sub-steps.

[0020] Step S121: Time sequence continuity detection is performed on the production dynamic data sequence, the geological condition data sequence and the environmental data sequence, and continuous data segments are retained as effective data sequences.

[0021] The three types of collected data sequences are detected by using a time sequence continuity detection algorithm, the timestamps in each data sequence are traversed, and it is judged whether the interval between two adjacent timestamps conforms to the preset collection interval. If the interval of multiple continuous timestamps in a data segment is within the allowable error range and there is no data loss, it is determined that the data segment is continuous data segment; if there is an abnormal interval between timestamps or data loss, the data segment is truncated and only the continuous data segment is retained. For example, the collection interval of multiple continuous time points in a data segment in the production dynamic data sequence conforms to the requirement and there is no missing value, so the data segment is regarded as effective production dynamic data sequence; there is a missing data segment in the geological condition data sequence due to equipment failure, the missing segment is removed and the remaining continuous data segments are retained as effective geological condition data sequences.

[0022] Step S122: According to the change fluctuation frequency of the effective data sequence, a dynamic time interval division rule is determined, the effective production dynamic data sequence is divided into multiple production dynamic time sequence segments according to the dynamic time interval division rule, the effective geological condition data sequence is divided into multiple geological condition dynamic time sequence segments, and the effective environmental data sequence is divided into multiple environmental dynamic time sequence segments, to obtain a production dynamic time sequence segment set, a geological condition dynamic time sequence segment set and an environmental dynamic time sequence segment set, and the higher the change fluctuation frequency, the smaller the time interval.

[0023] The change fluctuation frequency of each effective data sequence is calculated, and the change fluctuation frequency is determined by counting the number of data value changes in a unit time. According to the fluctuation frequency, a dynamic time interval division rule is set: for data sequences with high fluctuation frequency, such as production dynamic data sequences with frequent gas production rate changes, smaller time intervals are used for division; for data sequences with low fluctuation frequency, such as geological condition data sequences (coal seam permeability and porosity change slowly), larger time intervals are used for division.

[0024] According to the division rule, the effective production dynamic data sequence is divided into a plurality of production dynamic time sequence segments with smaller time intervals, each segment containing gas production rate change information and bottom hole pressure change information in the time period; the effective geological condition data sequence is divided into a plurality of geological condition dynamic time sequence segments with larger time intervals, each geological condition dynamic time sequence segment containing coal seam permeability change information and coal seam porosity change information in the corresponding time period; the effective environmental data sequence is divided into a plurality of environmental dynamic time sequence segments according to the time interval determined by its fluctuation frequency, each environmental dynamic time sequence segment containing surface temperature change information and surface humidity change information. After the division is completed, all production dynamic time sequence segments are classified into a production dynamic time sequence segment set, and similarly, a geological condition dynamic time sequence segment set and an environmental dynamic time sequence segment set are formed.

[0025] Step S123: Extract the gas production rate change trend feature and the bottom hole pressure change trend feature of each production dynamic time sequence segment, extract the coal seam permeability change trend feature and the coal seam porosity change trend feature of each geological condition dynamic time sequence segment, and extract the surface temperature change trend feature and the surface humidity change trend feature of each environmental dynamic time sequence segment.

[0026] For each production dynamic time sequence segment, the gas production rate change trend feature and the bottom hole pressure change trend feature are extracted by a trend analysis algorithm. Specifically, the gas production rate data points in the segment are fitted to determine the overall upward, downward or stable trend, and the inflection point position and change amplitude of the trend are recorded to form the gas production rate change trend feature; the same method is used to process the bottom hole pressure data to obtain the bottom hole pressure change trend feature.

[0027] For each geological condition dynamic time sequence segment, the coal seam permeability data sequence is analyzed to determine its long-term change trend and short-term fluctuation feature, and the coal seam permeability change trend feature is formed; the coal seam porosity data is analyzed in the same way to extract the coal seam porosity change trend feature.

[0028] For each environmental dynamic time sequence segment, the surface temperature data sequence is processed to extract its change trend over time, including the diurnal fluctuation rule and the long-term change trend, to form the surface temperature change trend feature; the change mode of the surface humidity data is analyzed to obtain the surface humidity change trend feature.

[0029] Step S124: calculating the real-time correlation strength between the gas production rate change trend feature and the coal seam permeability change trend feature, calculating the real-time correlation strength between the gas production rate change trend feature and the surface temperature change trend feature, calculating the real-time correlation strength between the bottom hole pressure change trend feature and the coal seam porosity change trend feature, calculating the real-time correlation strength between the bottom hole pressure change trend feature and the surface humidity change trend feature, calculating the real-time correlation strength between the coal seam permeability change trend feature and the surface temperature change trend feature, and calculating the real-time correlation strength between the coal seam porosity change trend feature and the surface humidity change trend feature.

[0030] The real-time correlation strength calculation between each type of trend feature follows similar analysis logic. Taking the real-time correlation strength calculation between the gas production rate change trend feature and the coal seam permeability change trend feature as an example, the calculation is implemented through the following sub-steps.

[0031] Step S1241: obtaining a first dynamic time sequence vector corresponding to the gas production rate change trend feature, each element of the first dynamic time sequence vector corresponding to the gas production rate value at different time in the production dynamic time sequence segment.

[0032] The specific value of the gas production rate at each time in the production dynamic time sequence segment is extracted, and the values are arranged in chronological order to form the first dynamic time sequence vector. The length of the vector is equal to the number of time in the segment, and each element corresponds to the gas production rate value at a time.

[0033] Step S1242: obtaining a second dynamic time sequence vector corresponding to the coal seam permeability change trend feature, each element of the second dynamic time sequence vector corresponding to the coal seam permeability value at different time in the geological condition dynamic time sequence segment.

[0034] The specific value of the coal seam permeability at each time in the geological condition dynamic time sequence segment is extracted, and the values are arranged in chronological order to form the second dynamic time sequence vector. The length of the vector is equal to the number of time in the segment, and each element corresponds to the coal seam permeability value at a time.

[0035] Step S1243: inputting the first dynamic time sequence vector and the second dynamic time sequence vector into a time sequence correlation analysis module, and performing elastic alignment processing on the time dimension of the first dynamic time sequence vector and the second dynamic time sequence vector based on the dynamic time warping algorithm to obtain the aligned first dynamic time sequence vector and the aligned second dynamic time sequence vector.

[0036] After receiving the first dynamic time series vector and the second dynamic time series vector, the time correlation analysis module starts the dynamic time warping algorithm, which performs elastic stretching or compression on the time axes of the two vectors by finding the optimal time mapping relationship, so that the time points of the corresponding trends in the vectors are matched with each other. For example, if the time point of a trend in the first dynamic time series vector is earlier than the time point of the corresponding trend in the second dynamic time series vector, the algorithm adjusts the time axis of the second dynamic time series vector so that the time points of the corresponding trends are aligned, and finally outputs the aligned first dynamic time series vector and the aligned second dynamic time series vector with consistent lengths and matched trend time points.

[0037] Step S1244: Extract the rising inflection point time and the falling inflection point time of the aligned first dynamic time series vector, and extract the rising inflection point time and the falling inflection point time of the aligned second dynamic time series vector.

[0038] Perform inflection point detection on the aligned first dynamic time series vector. When the element value of the vector changes from rising to falling, the time point is the falling inflection point time. When the element value changes from falling to rising, the time point is the rising inflection point time. Record all rising inflection point times and falling inflection point times. Use the same inflection point detection method to process the aligned second dynamic time series vector, and extract and record the rising inflection point times and the falling inflection point times.

[0039] Step S1245: Count the number of coincidences of the rising inflection point times of the first dynamic time series vector and the rising inflection point times of the second dynamic time series vector, count the number of coincidences of the falling inflection point times of the first dynamic time series vector and the falling inflection point times of the second dynamic time series vector, and add the two coincidence numbers to obtain the total number of coincident inflection points.

[0040] Compare the rising inflection point times of the first dynamic time series vector with the rising inflection point times of the second dynamic time series vector, count the number of time stamps that are completely consistent, and take the number of coincidences of the rising inflection points as the number of coincidences of the rising inflection points. Similarly, count the number of coincidences of the falling inflection point times of the two vectors, and add the number of coincidences of the rising inflection points and the number of coincidences of the falling inflection points to obtain the total number of coincident inflection points.

[0041] Step S1246: Count the total number of inflection points of the first dynamic time series vector, which is the sum of the number of rising inflection points and the number of falling inflection points, count the total number of inflection points of the second dynamic time series vector, which is the sum of the number of rising inflection points and the number of falling inflection points, and add the two total inflection point numbers to obtain the total number of inflection points.

[0042] The number of all rising inflection points and falling inflection points in the first dynamic time sequence vector is counted respectively, and the sum of the two is the total number of inflection points of the first dynamic time sequence vector; the total number of inflection points of the second dynamic time sequence vector is calculated in the same way, and the sum of the two total numbers of inflection points is obtained.

[0043] Step S1247: The proportion of the total number of coincident inflection points to the total number of inflection points is calculated as the inflection point coordination ratio.

[0044] The ratio of the total number of coincident inflection points to the total number of inflection points is obtained, which is the inflection point coordination ratio. The inflection point coordination ratio reflects the coordination degree of the two dynamic time sequence vectors at the trend turning point.

[0045] Step S1248: The average change amplitude of the aligned first dynamic time sequence vector is calculated, which is the average of the absolute value of the difference between each time point and the previous time point. The average change amplitude of the aligned second dynamic time sequence vector is calculated, which is the average of the absolute value of the difference between each time point and the previous time point. The ratio of the two average change amplitudes is taken as the amplitude coordination coefficient.

[0046] For the aligned first dynamic time sequence vector, the difference between the value of each time point and the value of the previous time point is calculated, the absolute value of the difference is taken, and the average of all absolute values is obtained as the average change amplitude of the first dynamic time sequence vector. The average change amplitude of the second dynamic time sequence vector is calculated in the same way. The average change amplitude of the first dynamic time sequence vector is divided by the average change amplitude of the second dynamic time sequence vector to obtain the amplitude coordination coefficient, which reflects the matching degree of the two vectors in the change amplitude.

[0047] Step S1249: The inflection point coordination ratio and the amplitude coordination coefficient are normalized respectively to convert them into dimensionless score values. The inflection point coordination ratio weight and the amplitude coordination coefficient weight are set, and the sum of the two is 1. The normalized inflection point coordination ratio is multiplied by the inflection point coordination ratio weight, and the normalized amplitude coordination coefficient is multiplied by the amplitude coordination coefficient weight to obtain the real-time correlation strength between the gas production rate change trend feature and the coal seam permeability change trend feature.

[0048] The min-max normalization method is used to convert the inflection point coordination ratio and the amplitude coordination coefficient into the interval of 0 to 1 to obtain dimensionless score values. According to the analysis results of historical data, the inflection point coordination ratio weight and the amplitude coordination coefficient weight are set, and the sum of the two is 1. The normalized inflection point coordination ratio is multiplied by its weight, and the normalized amplitude coordination coefficient is multiplied by its weight. The sum of the two products is the real-time correlation strength between the gas production rate change trend feature and the coal seam permeability change trend feature.

[0049] The real-time correlation strength between each of the remaining types of trend features, such as the gas production rate change trend feature and the surface temperature change trend feature, the bottom hole pressure change trend feature and the coal seam porosity change trend feature, is calculated according to the method of steps S1241 to S1249.

[0050] Step S125: Introducing a time-series correlation strength correction coefficient determined based on the fluctuation range of the correlation strength of the historical same-period same-type data sequence, multiplying each real-time correlation strength by the corresponding time-series correlation strength correction coefficient to obtain the corrected correlation strength.

[0051] The same-period same-type trend feature correlation strength is calculated by collecting the same-period production performance data sequence, geological condition data sequence and environmental data sequence of the coalbed methane well in past years, and the fluctuation range of the correlation strength is counted, and the time-series correlation strength correction coefficient is determined according to the fluctuation range. For example, if the correlation strength of the historical same-period gas production rate and coal seam permeability fluctuates less, the correction coefficient is close to 1; if the fluctuation is larger, the corresponding correction coefficient is set according to the fluctuation amplitude.

[0052] Each real-time correlation strength is matched with the corresponding time-series correlation strength correction coefficient, and the real-time correlation strength is multiplied by the correction coefficient to obtain the corrected correlation strength, so as to eliminate the influence of seasonal, annual and other periodic factors on the correlation strength.

[0053] Step S126: Building an initial ternary time-series dynamic correlation network with each production performance dynamic time-series segment, each geological condition dynamic time-series segment and each environmental dynamic time-series segment as a network node, and the corrected correlation strength as the initial weight of the connection between nodes.

[0054] Each segment in the production performance dynamic time-series segment set, the geological condition dynamic time-series segment set and the environmental dynamic time-series segment set is mapped to a node in the network, and each node is assigned a unique node identifier. For nodes of different categories, such as production performance nodes and geological condition nodes, production performance nodes and environmental nodes, and geological condition nodes and environmental nodes, connections are established between nodes according to the corresponding corrected correlation strength, and the corrected correlation strength is taken as the initial weight of the connection.

[0055] For example, the connection weight between a certain production performance dynamic time-series segment node and a certain geological condition dynamic time-series segment node is the corrected correlation strength between the trend features of the two; the connection weight between a certain production performance dynamic time-series segment node and a certain environmental dynamic time-series segment node is the corresponding corrected correlation strength. The initial ternary time-series dynamic correlation network is built in the above manner and stored in the graph database.

[0056] Step S127: Set a weight update period, recalculate the real-time correlation strength between nodes according to the weight update period, and correct and update the connection weight of the initial ternary time sequence dynamic correlation network, to obtain the final coalbed methane well ternary time sequence dynamic correlation network.

[0057] The weight update period is set according to the frequency of data changes, and the weight update period can be adjusted according to actual production needs. In each weight update period, the latest production dynamic data sequence, geological condition data sequence and environmental data sequence are re-collected, and the real-time correlation strength between nodes of various types is recalculated and corrected according to the method of steps S121 to S125, to obtain the updated correlation strength.

[0058] The initial weight of the connection between the corresponding nodes in the initial ternary time sequence dynamic correlation network is replaced by the updated correlation strength, to realize dynamic updating of the network weight. After multiple weight updates, the final coalbed methane well ternary time sequence dynamic correlation network is formed, which can reflect the changes in the correlation between different types of data time sequence fragments in real time.

[0059] Step S130: Call the pre-trained AI time sequence coupling deviation evolution mining model to process the ternary time sequence dynamic correlation network, mine the coupling deviation evolution information between different data sequence dynamic time sequence fragments, and obtain a ternary time sequence coupling deviation result containing a deviation evolution trend.

[0060] In this embodiment, the pre-trained AI model is used to perform deep analysis on the ternary time sequence dynamic correlation network and mine coupling deviation information. The following sub-steps are used to achieve this.

[0061] Step S131: Input the ternary time sequence dynamic correlation network into the network analysis layer of the AI time sequence coupling deviation evolution mining model, analyze the dynamic time sequence fragment data corresponding to each node, the dynamic time sequence fragment data corresponding to each node containing specific data values at each time in the fragment, analyze the corrected correlation strength data and weight update records corresponding to the connection between nodes, and obtain network analysis data.

[0062] The network analysis layer of the AI time sequence coupling deviation evolution mining model receives the ternary time sequence dynamic correlation network and analyzes the network structure and data. First, the identification of each node and the corresponding dynamic time sequence fragment data are extracted, and the dynamic time sequence fragment data contains specific numerical values at each time in the fragment, such as the gas production rate and bottom hole pressure value at each time in the production dynamic node fragment, the permeability and porosity value at each time in the geological condition node fragment, etc.

[0063] Then the attribute information of the connection between nodes is parsed, including the corrected correlation strength data and the record of the weight update in the past, which contains the update time and the weight value of each update. The node data, connection data and weight update record parsed are integrated into network parsing data and transmitted to the next layer of the model.

[0064] Step S132: input the network parsing data into the time series coupling analysis layer of the AI time series coupling bias evolution mining model, and analyze the coupling matching degree of the production dynamic time series segment and the geological condition dynamic time series segment from three dimensions of trend consistency, phase difference synergy and amplitude synergy; wherein the trend consistency is determined by calculating the coincidence length proportion of the change trend of the two types of dynamic time series segments, the phase difference synergy is determined by calculating the time difference average of the inflection point time of the two types of dynamic time series segments, and the amplitude synergy is determined by calculating the fluctuation range of the change amplitude ratio of the two types of dynamic time series segments.

[0065] After the time series coupling analysis layer receives the network parsing data, the coupling matching degree is calculated from three dimensions for the combination of the production dynamic time series segment and the geological condition dynamic time series segment.

[0066] In the trend consistency dimension, the change trend curves of the two types of segments are compared, the length of the consistent curve trend is counted, and the proportion obtained by dividing the total length of the segment by the length is the trend consistency index.

[0067] In the phase difference synergy dimension, the inflection point time of the two types of segments is extracted, the time difference of the corresponding inflection point time is calculated, and the phase difference synergy index is obtained by averaging all the time differences. The smaller the average time difference is, the higher the phase difference synergy is.

[0068] In the amplitude synergy dimension, the change amplitudes of the two types of segments are calculated respectively to obtain the amplitude ratio sequence, and the fluctuation range of the sequence is counted. The smaller the fluctuation range is, the higher the amplitude synergy is.

[0069] Through the calculation of the above three dimensions, the coupling matching degree basic data of the production dynamic time series segment and the geological condition dynamic time series segment are obtained.

[0070] Step S133: normalize the coupling matching degree of each dimension, set the weight of the normalized coupling matching degree of each dimension, calculate the weighted sum of the three types of dimension coupling matching degree score values according to the set weight, and take it as the comprehensive coupling matching degree of the two types of dynamic time series segments.

[0071] The indexes of the three dimensions of trend consistency, phase difference synergy and amplitude synergy are normalized to convert the score values between 0 and 1. According to the influence degree of each dimension on the coupling matching degree, a weight is set for each normalized score value, and the sum of the three weights is 1. The normalized trend consistency score value is multiplied by its corresponding weight, the normalized phase difference synergy score value is multiplied by its corresponding weight, and the normalized amplitude synergy score value is multiplied by its corresponding weight. The sum of the three products is the comprehensive coupling matching degree of the production dynamic time sequence segment and the geological condition dynamic time sequence segment. Using the same method, the comprehensive coupling matching degrees of the production dynamic time sequence segment and the environment dynamic time sequence segment, and the geological condition dynamic time sequence segment and the environment dynamic time sequence segment are calculated respectively, so as to realize the quantitative evaluation of the coupling matching of all different types of dynamic time sequence segment combinations.

[0072] Step S134: identifying a dynamic time sequence segment combination with a comprehensive coupling matching degree lower than a preset matching threshold, and marking the dynamic time sequence segment combination as a deviation candidate combination.

[0073] According to the historical production data and the coupling matching under normal production conditions, a preset matching threshold is set. All different types of dynamic time sequence segment combinations and their corresponding comprehensive coupling matching degrees are traversed, and combinations with a comprehensive coupling matching degree less than the preset matching threshold are screened out. For example, the comprehensive coupling matching degree of a certain production dynamic time sequence segment and a certain geological condition dynamic time sequence segment is lower than the preset threshold, indicating that the coupling relationship deviates from the normal level. The combination is marked as a deviation candidate combination, and the node identifier, category and comprehensive coupling matching degree value of the two dynamic time sequence segments in the combination are recorded.

[0074] Step S135: for each deviation candidate combination, extracting the comprehensive coupling matching degree change data of the combination in multiple continuous weight update periods, and constructing a deviation evolution curve, wherein the horizontal coordinate of the deviation evolution curve is the weight update period, and the vertical coordinate of the deviation evolution curve is the comprehensive coupling matching degree.

[0075] From the weight update record of the ternary time sequence dynamic association network, the comprehensive coupling matching degree data of each deviation candidate combination in the past multiple continuous weight update periods is extracted. Taking the weight update period as the horizontal coordinate and the comprehensive coupling matching degree corresponding to each period as the vertical coordinate, the data points are plotted in the coordinate system, and then the curve fitting algorithm is used to connect the above data points into a smooth curve to form the deviation evolution curve of the deviation candidate combination. Through the curve, the change trajectory of the comprehensive coupling matching degree with time can be directly observed.

[0076] Step S136: Analyze the slope change trend of the deviation evolution curve. If the slope is negative and the absolute value gradually increases, it is determined that the deviation is expanding; if the slope is negative and the absolute value gradually decreases, it is determined that the deviation is slowing down; if the slope is positive, it is determined that the deviation is repairing.

[0077] The slope of the deviation evolution curve is analyzed, and the slope of the curve in each weight update period interval is calculated. If the slopes of multiple consecutive intervals are negative, and the absolute value of the slope gradually increases, it means that the comprehensive coupling matching degree is declining at an accelerated rate, and it is determined that the deviation of the deviation candidate combination is expanding; if the slopes of multiple consecutive intervals are negative, but the absolute value of the slope gradually decreases, it means that the comprehensive coupling matching degree is declining at a slower rate, and it is determined that the deviation is slowing down; if the slopes of multiple consecutive intervals are positive, it means that the comprehensive coupling matching degree is rising, and it is determined that the deviation is repairing.

[0078] Step S137: Calculate the deviation degree of the dynamic timing segment in the deviation candidate combination to obtain the deviation value corresponding to each deviation candidate combination.

[0079] The calculation of the deviation degree needs to refer to the standard coupling relationship for multi-dimensional comparison, which is realized by the following sub-steps.

[0080] Step S1371: Collect the production dynamic data sequence, geological condition data sequence, and environmental data sequence of the coalbed methane well in the historical normal production state, extract the coupling data of the same type of dynamic timing segment, and construct a standard coupling relationship library. The standard coupling relationship library stores the standard trend consistency range, standard phase difference cooperativity range, and standard amplitude cooperativity range of different types of dynamic timing segment combinations.

[0081] Collect the production dynamic data sequence, geological condition data sequence, and environmental data sequence of the coalbed methane well in the historical normal production state, and divide the above data sequences into dynamic timing segments according to the division rules of step S122. Extract the coupling data of the same type of dynamic timing segment combination (such as production-geology, production-environment, and geology-environment), including the historical normal values of trend consistency, phase difference cooperativity, and amplitude cooperativity.

[0082] Statistically analyze the above historical normal values to determine the normal fluctuation range of different types of dynamic timing segment combinations in three dimensions, i.e., the standard trend consistency range, the standard phase difference cooperativity range, and the standard amplitude cooperativity range. Store the above ranges and corresponding dynamic timing segment combination types in the standard coupling relationship library.

[0083] Step S1372: According to the type of the deviation candidate combination, the corresponding standard trend consistency range, the standard phase difference consistency range, and the standard amplitude consistency range in the standard coupling relationship library are retrieved as the standard coupling relationship of the deviation candidate combination. The type of the deviation candidate combination includes production-geology deviation, production-environment deviation, and geology-environment deviation.

[0084] The type of each deviation candidate combination is determined. For example, the combination type composed of the production dynamic dynamic time sequence segment and the geology condition dynamic time sequence segment is production-geology deviation, the combination type composed of the production dynamic dynamic time sequence segment and the environment dynamic time sequence segment is production-environment deviation, and the combination type composed of the geology condition dynamic time sequence segment and the environment dynamic time sequence segment is geology-environment deviation.

[0085] According to the determined type, a matching query is performed in the standard coupling relationship library, and the corresponding standard trend consistency range, the standard phase difference consistency range, and the standard amplitude consistency range of the type are retrieved as the standard coupling relationship for evaluating the deviation degree of the deviation candidate combination.

[0086] Step S1373: The actual change curve of the first type of dynamic time sequence segment in the deviation candidate combination is extracted, and the actual change curve of the second type of dynamic time sequence segment in the deviation candidate combination is extracted. The horizontal coordinate of the actual change curve is the time within the segment, and the vertical coordinate of the actual change curve is the data value. The horizontal coordinate of the actual change curve is the time within the segment, and the vertical coordinate of the actual change curve is the data value.

[0087] Taking two dynamic time sequence segments in the deviation candidate combination as objects, the specific data value of each time within each segment is extracted, the time within the segment is taken as the horizontal coordinate, and the data value is taken as the vertical coordinate. The actual change curves of the two dynamic time sequence segments, i.e., the actual change curve of the first type of dynamic time sequence segment and the actual change curve of the second type of dynamic time sequence segment, are drawn. For example, in the production-geology type deviation candidate combination, the first type is the production dynamic dynamic time sequence segment, and its actual change curve takes time as the horizontal axis and gas production rate and bottom hole pressure as the vertical axis. The second type is the geology condition dynamic time sequence segment, and its actual change curve takes time as the horizontal axis and coal seam permeability and porosity as the vertical axis.

[0088] Step S1374: According to the standard trend consistency range in the standard coupling relationship, the standard trend curve range corresponding to the first type of dynamic time sequence segment is generated, and the standard trend curve range corresponding to the second type of dynamic time sequence segment is generated. The standard trend curve range includes an upper limit trend curve and a lower limit trend curve. The standard trend curve range includes an upper limit trend curve and a lower limit trend curve.

[0089] According to the standard trend consistency range in the standard coupling relationship, a standard trend curve range corresponding to the first type of dynamic time sequence segment is generated based on the basic data (such as initial data values, average change rates) of the first type of dynamic time sequence segment, the standard trend curve range is composed of an upper limit trend curve and a lower limit trend curve, and the area between the two curves is the data fluctuation interval of the segment under the normal trend. Using the same method, the standard trend curve range corresponding to the second type of dynamic time sequence segment is generated based on the standard trend consistency range, and also contains the upper limit and lower limit trend curves.

[0090] Step S1375: Calculate the difference area between the actual change curve of the first type of dynamic time sequence segment and the upper limit of the standard trend curve range, and calculate the difference area between the actual change curve of the first type of dynamic time sequence segment and the lower limit of the standard trend curve range. The maximum of the two difference areas is taken as the trend deviation area of the first type of segment.

[0091] The actual change curve of the first type of dynamic time sequence segment is compared with the upper limit curve of the standard trend curve range, and the area enclosed between the two curves is calculated, which is the difference area between the actual change curve and the upper limit curve. The area enclosed between the actual change curve and the lower limit curve is calculated as the difference area between the actual change curve and the lower limit curve. The larger of the two difference areas is selected as the trend deviation area of the first type of segment. The larger the area is, the more serious the trend deviation of the first type of dynamic time sequence segment is.

[0092] Step S1376: Calculate the difference area between the actual change curve of the second type of dynamic time sequence segment and the upper limit of the standard trend curve range, and calculate the difference area between the actual change curve of the second type of dynamic time sequence segment and the lower limit of the standard trend curve range. The maximum of the two difference areas is taken as the trend deviation area of the second type of segment.

[0093] Using the same method as step S1375, the difference areas between the actual change curve of the second type of dynamic time sequence segment and the upper limit and lower limit curves of the standard trend curve range are calculated, and the maximum of the two is taken as the trend deviation area of the second type of segment, so as to quantify the trend deviation degree of the second type of dynamic time sequence segment.

[0094] Step S1377: According to the standard phase difference consistency range in the standard coupling relationship, the absolute value of the difference between the actual phase difference of the two types of dynamic time sequence segments in the deviation candidate combination and the upper limit of the standard phase difference consistency range is calculated, and the absolute value of the difference between the actual phase difference and the lower limit of the standard phase difference consistency range is calculated. The maximum of the two difference absolute values is taken as the phase difference deviation value.

[0095] The actual phase difference of the two types of dynamic timing segments in the deviation candidate combination (i.e. the difference value corresponding to the inflection point time) is extracted, and the actual phase difference is compared with the upper limit value and the lower limit value of the standard phase difference coordination range respectively, the absolute value of the difference between the actual phase difference and the upper limit value is calculated, and the absolute value of the difference between the actual phase difference and the lower limit value is calculated. Compare the two absolute values, and select the larger one as the phase difference deviation value. The larger the phase difference deviation value is, the more significant the phase difference deviation from the normal range is.

[0096] Step S1378: According to the standard amplitude coordination range in the standard coupling relationship, the absolute value of the difference between the actual amplitude ratio of the two types of dynamic timing segments in the deviation candidate combination and the upper limit of the standard amplitude coordination range is calculated, and the absolute value of the difference between the actual amplitude ratio and the lower limit of the standard amplitude coordination range is calculated. Take the maximum of the two absolute values as the amplitude coordination deviation value.

[0097] The actual amplitude ratio of the two types of dynamic timing segments in the deviation candidate combination (i.e. the ratio of the average change amplitude of the two types of segments) is calculated, and the actual amplitude ratio is calculated with the upper limit value and the lower limit value of the standard amplitude coordination range, respectively. The absolute value of the difference between the actual amplitude ratio and the upper limit value is calculated, and the absolute value of the difference between the actual amplitude ratio and the lower limit value is calculated. Compare the two absolute values, and select the larger one as the amplitude coordination deviation value, which reflects the deviation degree of amplitude coordination.

[0098] Step S1379: Set the trend deviation weight, the phase difference deviation weight, and the amplitude coordination deviation weight, and the sum of the trend deviation weight, the phase difference deviation weight, and the amplitude coordination deviation weight is 1; The average value of the trend deviation area of the first type of segment, the trend deviation area of the second type of segment, the phase difference deviation value and the amplitude coordination deviation value are normalized, the normalized trend deviation area score value is multiplied by the trend deviation weight, and the normalized phase difference deviation score value is multiplied by the phase difference deviation weight. Add the normalized amplitude coordination deviation score value multiplied by the amplitude coordination deviation weight to obtain the deviation value corresponding to the deviation candidate combination.

[0099] According to the influence weight of the three deviation indexes on the overall deviation degree, set the trend deviation weight, the phase difference deviation weight, and the amplitude coordination deviation weight, and the sum of the three is 1. Calculate the average value of the trend deviation area of the first type of segment and the trend deviation area of the second type of segment as the comprehensive trend deviation area.

[0100] The integrated trend deviation area, phase difference deviation value and amplitude cooperation deviation value are normalized respectively to convert into score values between 0 and 1. The score value of the normalized integrated trend deviation area is multiplied by the trend deviation weight, the score value of the normalized phase difference deviation is multiplied by the phase difference deviation weight, and the score value of the normalized amplitude cooperation deviation is multiplied by the amplitude cooperation deviation weight. The sum of the three products is the deviation value corresponding to the deviation candidate combination. The greater the deviation value, the more serious the overall deviation.

[0101] Step S138: The deviation candidate combination, the deviation value and the deviation evolution trend are integrated to generate a ternary time sequence coupled deviation result including a deviation position, a deviation type, a deviation degree and a deviation evolution trend. The deviation position corresponds to a dynamic time sequence segment number. The deviation type includes production-geology deviation, production-environment deviation and geology-environment deviation. The deviation degree is the deviation value.

[0102] The information of each deviation candidate combination is integrated. The deviation position is represented by the node numbers of the two dynamic time sequence segments in the combination. The deviation type is determined as production-geology deviation, production-environment deviation or geology-environment deviation according to the combination category. The deviation degree is the deviation value calculated in step S1379. The deviation evolution trend is the expansion, slowing down or repair trend determined in step S136.

[0103] The above information is arranged in a unified format to form a ternary time sequence coupled deviation result. The ternary time sequence coupled deviation result presents the specific position, type, severity and development trend of each deviation.

[0104] Step S140: According to the ternary time sequence coupled deviation result, a deviation conduction closed loop path affecting production dynamics is determined by combining forward conduction positioning and reverse tracing verification to generate a deviation conduction analysis report including a deviation conduction direction, a correlation strength and a tracing verification result.

[0105] Based on the ternary time sequence coupled deviation result, the deviation conduction path is sorted out by combining forward positioning and reverse verification. This is achieved by the following sub-steps.

[0106] Step S141: All deviation candidate combinations and corresponding deviation values and deviation evolution trends are extracted from the ternary time sequence coupled deviation result. Deviation candidate combinations with an expansion trend are screened out. The screened deviation candidate combinations are sorted in descending order of deviation value.

[0107] All the bias candidate combinations in the ternary time sequence coupling bias result are traversed, and the bias values and bias evolution trends of each combination are extracted. The combinations with expanding trend are screened out, and the bias of these combinations is intensifying, so the transmission path thereof needs to be analyzed in priority. The screened combinations are sorted in descending order of bias values, and the combination with larger bias value is arranged in front, and the transmission path analysis is carried out in priority.

[0108] Step S142: Selecting a preset number of bias candidate combinations arranged in front as key bias combinations, extracting the timestamp information of two types of dynamic time sequence segments in each key bias combination, and determining the time sequence of the two types of dynamic time sequence segments.

[0109] The preset number is set according to the actual analysis requirement, and the combinations of the number arranged in front are selected as key bias combinations. The timestamp information of two dynamic time sequence segments in each key bias combination, i.e. the starting time and ending time of the segments, is extracted, the time sequence of the two segments is determined by comparing the starting time of the two segments, the dynamic time sequence segment with earlier starting time is the first appearing dynamic time sequence segment, and the dynamic time sequence segment with later starting time is the later appearing dynamic time sequence segment.

[0110] Step S143: According to the time sequence, the first appearing dynamic time sequence segment is marked as a starting dynamic time sequence segment, and the later appearing dynamic time sequence segment is marked as an affected dynamic time sequence segment, and the positive transmission direction of the bias from the starting dynamic time sequence segment to the affected dynamic time sequence segment is preliminarily determined.

[0111] Based on the time sequence determined in step S142, the first appearing dynamic time sequence segment is defined as a starting dynamic time sequence segment, and the starting dynamic time sequence segment is considered as the starting source of the bias; the later appearing dynamic time sequence segment is defined as an affected dynamic time sequence segment, and the bias of the affected dynamic time sequence segment is considered to be transmitted from the starting dynamic time sequence segment. Accordingly, the positive transmission direction of the bias is preliminarily determined as from the starting dynamic time sequence segment to the affected dynamic time sequence segment. For example, in a certain key bias combination, the geological condition dynamic time sequence segment appears first, and the production dynamic time sequence segment appears later, the starting dynamic time sequence segment is the geological condition dynamic time sequence segment, the affected dynamic time sequence segment is the production dynamic time sequence segment, and the positive transmission direction is from the geological condition dynamic time sequence segment to the production dynamic time sequence segment.

[0112] Step S144: Extracting the link association strength between the node corresponding to the starting dynamic time sequence segment and the node corresponding to the affected dynamic time sequence segment from the ternary time sequence dynamic correlation network, and simultaneously extracting the link association strength between the node corresponding to the starting dynamic time sequence segment and other related nodes, the other related nodes being the nodes associated with the affected dynamic time sequence segment, and extracting the link association strength between the node corresponding to the affected dynamic time sequence segment and other related nodes.

[0113] In the ternary time sequence dynamic association network, according to the node identification of the starting dynamic time sequence segment and the affected dynamic time sequence segment, the association strength of the connection between the two is found, which reflects the closeness of the direct association between the two. At the same time, all other nodes associated with the affected dynamic time sequence segment node are found, which are other related nodes. The association strength between the starting dynamic time sequence segment node and each other related node, and the association strength between the affected dynamic time sequence segment node and each other related node are extracted, which are prepared for constructing the conduction strength matrix.

[0114] Step S145: Constructing a forward conduction strength matrix, the rows of the forward conduction strength matrix represent the starting nodes and the related nodes, the columns of the forward conduction strength matrix represent the affected nodes and the related nodes, and the elements of the forward conduction strength matrix are the association strengths between the corresponding nodes.

[0115] Constructing the forward conduction strength matrix needs to determine the nodes corresponding to the rows and columns of the matrix, and fill in the corresponding association strengths, which is realized by the following sub-steps.

[0116] Step S1451: Determine the starting node number corresponding to the starting dynamic time sequence segment, and determine the affected node number corresponding to the affected dynamic time sequence segment.

[0117] Determine the node identification corresponding to the starting dynamic time sequence segment in the ternary time sequence dynamic association network as the starting node number, and determine the node identification corresponding to the affected dynamic time sequence segment as the affected node number, to ensure that the number is completely consistent with the node identification in the network.

[0118] Step S1452: From the node association list of the ternary time sequence dynamic association network, query all nodes associated with the starting node as starting related nodes, and query all nodes associated with the affected node as affected related nodes.

[0119] Access the node association list of the ternary time sequence dynamic association network, which records all the association node information of each node. According to the starting node number, query all nodes associated with the node in the list as the starting related nodes; according to the affected node number, query all nodes associated with the node in the list as the affected related nodes.

[0120] Step S1453: Integrate the starting node and the starting related node to form a matrix row node set, and arrange the nodes in the matrix row node set in the order from small to large according to the node number; integrate the affected node and the affected related node to form a matrix column node set, and arrange the nodes in the matrix column node set in the order from small to large according to the node number.

[0121] The starting node and all starting related nodes are combined to form a matrix row node set, and a sorting algorithm is used to arrange the nodes in the set in ascending order of node number to determine the node order corresponding to each row of the matrix. The affected nodes and all affected related nodes are combined to form a matrix column node set, and are also arranged in ascending order of node number to determine the node order corresponding to each column of the matrix.

[0122] Step S1454: A blank forward conduction intensity matrix is constructed, the number of rows of the blank forward conduction intensity matrix is equal to the number of nodes in the matrix row node set, and the number of columns of the blank forward conduction intensity matrix is equal to the number of nodes in the matrix column node set.

[0123] The number of rows of the blank matrix is determined according to the number of nodes in the matrix row node set, the number of columns of the blank matrix is determined according to the number of nodes in the matrix column node set, and a blank forward conduction intensity matrix with matching number of rows and columns is created, and all elements in the matrix are initially set to 0.

[0124] Step S1455: For each node in the matrix row node set, the node is recorded as a row node, and for each node in the matrix column node set, the node is recorded as a column node, and the modified correlation strength between the row node and the column node in the ternary time sequence dynamic correlation network is queried; if there is a connection correlation between the row node and the column node, the modified correlation strength is filled into the element position of the row corresponding to the row node and the column corresponding to the column node in the forward conduction intensity matrix; if there is no connection correlation between the row node and the column node, the element value is set to 0 and filled into the element position of the row corresponding to the row node and the column corresponding to the column node in the forward conduction intensity matrix, to obtain a complete forward conduction intensity matrix.

[0125] Each row node in the matrix row node set and each column node in the matrix column node set are sequentially traversed, and for each pair of row node and column node, it is queried in the ternary time sequence dynamic correlation network whether there is a connection correlation between the two. If there is an association, the modified correlation strength of the connection is extracted, and the strength value is filled into the element position intersected by the row corresponding to the row node and the column corresponding to the column node in the forward conduction intensity matrix; if there is no association, the value of the element position is kept as 0. After the traversal is completed, a complete forward conduction intensity matrix is obtained. The forward conduction intensity matrix comprehensively presents the correlation strength distribution between the starting node, the starting related node and the affected node, the affected related node. For example, a certain row node in the matrix corresponds to the starting node, a certain column node corresponds to the affected node, and the value of the element intersected by the two is the modified correlation strength between the starting node and the affected node; if the starting related node and the affected related node are associated, the corresponding element is filled with the corresponding correlation strength, otherwise it is 0.

[0126] Step S146: Calculate the cumulative correlation strength of the forward conduction path, and screen the path with cumulative correlation strength higher than the forward conduction threshold as the candidate forward conduction path. The cumulative correlation strength is the average of the correlation strength between all nodes on the path.

[0127] Based on the forward conduction strength matrix, all possible paths from the starting node to the affected node through the starting correlation node or the affected correlation node are enumerated. For each path, the correlation strength values between adjacent nodes on the path are extracted, and the average of the above values is obtained by dividing the sum of the values by the number of connections between nodes on the path, thereby obtaining the cumulative correlation strength of the path, i.e., the average of all correlation strengths on the path.

[0128] The forward conduction threshold is set based on the statistical results of the cumulative correlation strength of the historical normal conduction path. The cumulative correlation strength of each path is compared with the forward conduction threshold, and the paths with cumulative correlation strength higher than the threshold are screened out. These paths are considered to have strong conduction possibility and are regarded as candidate forward conduction paths.

[0129] Step S147: For each candidate forward conduction path, a reverse tracing verification is started: taking the affected dynamic timing segment as the starting point and the starting dynamic timing segment as the ending point, the reverse correlation path from the affected node to the starting node in the ternary timing dynamic correlation network is queried, and the cumulative correlation strength of the reverse correlation path is calculated. The cumulative correlation strength is the average of the correlation strength between all nodes on the path.

[0130] For each candidate forward conduction path, the reverse tracing verification process takes the affected node corresponding to the affected dynamic timing segment as the starting point and the starting node corresponding to the starting dynamic timing segment as the ending point, and finds all possible reverse correlation paths in the ternary timing dynamic correlation network, i.e., paths with the opposite node connection direction to the forward path.

[0131] For each reverse correlation path, the same calculation method as the forward path is used to extract the correlation strength values between adjacent nodes on the path, and the average of these values is calculated as the cumulative correlation strength of the reverse correlation path.

[0132] Step S148: If the cumulative correlation strength of the reverse correlation path is higher than the reverse tracing threshold, and the reverse correlation path forms a node-interconnected closed loop structure with the candidate forward conduction path, the candidate forward conduction path is determined as an effective deviation conduction closed loop path; if the cumulative correlation strength of the reverse correlation path is lower than the reverse tracing threshold, or the closed loop structure cannot be formed, the candidate forward conduction path is eliminated.

[0133] A reverse tracing threshold is set, which is set with reference to the forward conduction threshold and the characteristics of reverse conduction. The cumulative correlation strength of the reverse correlation path is compared with the reverse tracing threshold, and it is checked whether the reverse correlation path and the corresponding candidate forward conduction path form a node interworking closed loop structure, i.e. the starting point of the forward path is the end point of the reverse path, the end point of the forward path is the starting point of the reverse path, and the nodes passed by the two paths exist overlap or complement, forming a complete closed loop.

[0134] If the cumulative correlation strength of the reverse correlation path is higher than the reverse tracing threshold, and a closed loop structure is formed with the candidate forward conduction path, it is determined that the candidate forward conduction path is an effective deviation conduction closed loop path; if the cumulative correlation strength of the reverse correlation path is lower than the reverse tracing threshold, or the two paths cannot form a closed loop, the candidate forward conduction path is eliminated.

[0135] Step S149: The forward cumulative correlation strength, reverse cumulative correlation strength, number of nodes involved and dynamic timing segment type of each effective deviation conduction closed loop path are counted, the deviation conduction direction, correlation strength and tracing verification result of the effective deviation conduction closed loop path are integrated, and a deviation conduction analysis report is generated, the deviation conduction direction is the forward conduction direction, the correlation strength is the forward cumulative correlation strength, and the tracing verification result includes the reverse cumulative correlation strength and the closed loop verification result.

[0136] The information of each effective deviation conduction closed loop path is counted, including the forward cumulative correlation strength (i.e. the cumulative correlation strength of the candidate forward conduction path), the reverse cumulative correlation strength (i.e. the cumulative correlation strength of the corresponding reverse correlation path), the number of all nodes involved in the path, and the dynamic timing segment type (production dynamics, geological conditions or environment) corresponding to these nodes.

[0137] The deviation conduction direction (i.e. the forward conduction direction), the correlation strength (i.e. the forward cumulative correlation strength), and the tracing verification result (including the reverse cumulative correlation strength and the closed loop verification result, which is "pass" or "fail") of the effective deviation conduction closed loop path are integrated, and are arranged into a deviation conduction analysis report according to a preset report format, which clearly presents the key information of each effective deviation conduction path.

[0138] Step S150: Based on the deviation conduction analysis report, a dynamic control strategy for the production parameters of the coalbed methane well is generated in combination with the production dynamics prediction threshold, and the dynamic control strategy is output to a production control terminal, and the regulated data is fed back to the ternary timing dynamic correlation network to update the node correlation.

[0139] The generation and implementation of the dynamic control strategy and the network updating process are realized through the following sub-steps.

[0140] Step S151: Analyzing the effective deviation conduction closed loop paths in the deviation conduction analysis report, determining the starting deviation type corresponding to each effective deviation conduction closed loop path, the starting deviation type including geological condition deviation and environmental factor deviation, extracting the forward cumulative correlation strength and the reverse cumulative correlation strength of each effective deviation conduction closed loop path.

[0141] Analyzing the deviation conduction analysis report, extracting all effective deviation conduction closed loop paths therein, and determining the starting deviation type according to the starting dynamic time sequence segment type of each path: if the starting dynamic time sequence segment is a geological condition dynamic time sequence segment, the starting deviation type is a geological condition deviation; if the starting dynamic time sequence segment is an environmental dynamic time sequence segment, the starting deviation type is an environmental factor deviation. At the same time, the forward cumulative correlation strength and the reverse cumulative correlation strength of each path are extracted as the basis for subsequent determination of the control priority.

[0142] Step S152: Calling a pre-trained production dynamic prediction model, inputting the current production dynamic data sequence, the current geological condition data sequence and the current environmental data sequence, predicting the gas production rate prediction value and the bottom hole pressure prediction value in a future preset time period as the production dynamic prediction result.

[0143] The construction, training and prediction process of the production dynamic prediction model is as follows.

[0144] Step S1521: Collecting the production dynamic data sequence, the geological condition data sequence, the environmental data sequence and the actual gas production rate data and the actual bottom hole pressure data in a future preset time period corresponding thereto of a historical coalbed methane well, and constructing a production dynamic prediction model training data set.

[0145] Collecting the long-term production dynamic data sequence, the geological condition data sequence and the environmental data sequence of the coalbed methane well in the past, and collecting the actual gas production rate data and the actual bottom hole pressure data in a future preset time period corresponding to these historical data. Pairing each set of historical input data (production dynamic, geological condition, environmental data sequence) with the corresponding future actual output data (gas production rate, bottom hole pressure data) to construct a training data set of the production dynamic prediction model.

[0146] Step S1522: Dividing the production dynamic prediction model training data set into a training set, a validation set and a test set, the training set being used for model parameter learning, the validation set being used for model hyperparameter adjustment, and the test set being used for model performance evaluation.

[0147] The training data set is divided into a training set, a validation set and a test set in a random division manner. The training set is used to learn the potential rules in the data and determine the basic parameters of the model. The validation set is used to adjust the hyperparameters of the model, such as the number of network layers and the number of hidden layer nodes, during the training process. The test set is used to evaluate the generalization performance of the model after the training is completed, to ensure that the model can maintain good prediction effect on unseen data.

[0148] Step S1523: constructing a network structure of the production dynamic prediction model, the network structure of the production dynamic prediction model comprising a time sequence feature encoding layer, a multi-source feature fusion layer and a prediction output layer; the time sequence feature encoding layer adopts a bidirectional long short-term memory network structure and is used for respectively performing time sequence feature extraction on input production dynamic data sequences, geological condition data sequences and environment data sequences, and outputting production dynamic time sequence encoding features, geological condition time sequence encoding features and environment time sequence encoding features; the multi-source feature fusion layer adopts an attention mechanism fusion structure, and is used for performing weighted fusion on the production dynamic time sequence encoding features, the geological condition time sequence encoding features and the environment time sequence encoding features, the weights being dynamically adjusted based on the influence degree of each feature on a production dynamic prediction result, and outputting multi-source fusion features; and the prediction output layer adopts a fully connected network structure, and is used for processing the multi-source fusion features and outputting a gas production rate prediction value and a bottom hole pressure prediction value at each time point in a future preset time period.

[0149] The time sequence feature encoding layer is composed of a bidirectional long short-term memory network, which contains memory units in two directions, forward and backward, and can capture past and future information of an input sequence at the same time. The production dynamic data sequences, the geological condition data sequences and the environment data sequences are respectively input into three parallel branches of the layer, and the time sequence features of the corresponding sequences are independently extracted in each branch to output the production dynamic time sequence encoding features, the geological condition time sequence encoding features and the environment time sequence encoding features, which are all in the form of high-dimensional vectors.

[0150] The multi-source feature fusion layer adopts an attention mechanism, and calculates the correlation between each time sequence encoding feature and a prediction target to assign different weights to different features. The higher the correlation of a feature, the greater the weight, and the greater the contribution of the feature to the fusion result. The three time sequence encoding features are weighted and spliced according to their respective weights to obtain multi-source fusion features.

[0151] The prediction output layer is composed of multiple fully connected layers, which process the multi-source fusion features through linear transformation and activation functions, map the high-dimensional features to a low-dimensional prediction space, and finally output the gas production rate prediction value and the bottom hole pressure prediction value at each time point in a future preset time period.

[0152] Step S1524: training the production dynamic prediction model using the training set, calculating the prediction error of the production dynamic prediction model using the validation set after each preset training round, the prediction error being the mean square error of the predicted and actual gas production rate and the mean square error of the predicted and actual bottom hole pressure; if the prediction error is higher than a preset error threshold, adjusting the number of hidden layer nodes of the bidirectional long short-term memory network, the weight calculation method of the attention mechanism, and the number of layers of the fully connected network, and retraining the production dynamic prediction model; if the prediction error is lower than the preset error threshold, performing final performance verification using the test set, and if the prediction error of the test set is also lower than the preset error threshold, obtaining the pre-trained production dynamic prediction model.

[0153] The input data in the training set is input into the production dynamic prediction model, and the prediction output of the model is calculated through forward propagation. The mean square error is used as the loss function, the mean square error of the predicted and actual gas production rate and the mean square error of the predicted and actual bottom hole pressure are calculated, and the total loss value is obtained by adding the two errors. Through the back propagation algorithm, the parameters of each layer of the model are adjusted according to the total loss value, including the weights and biases of the bidirectional long short-term memory network, the parameters of the attention mechanism, and the weights and biases of the fully connected network.

[0154] After each preset training round, the model is tested using the validation set, and the prediction error on the validation set is calculated. If the prediction error of the validation set is higher than the preset error threshold, it indicates that the model may have overfitting or unreasonable parameter settings, and the number of hidden layer nodes of the bidirectional long short-term memory network, the weight calculation method of the attention mechanism (such as changing the function of correlation calculation), or the number of layers of the fully connected network need to be adjusted, and then retraining is performed.

[0155] If the prediction error of the validation set is lower than the preset error threshold, the model is finally verified for performance using the test set. If the prediction error of the test set is also lower than the preset error threshold, it indicates that the model has good performance, and the model parameters at this time are saved to obtain the pre-trained production dynamic prediction model.

[0156] Step S1525: inputting the current production dynamic data sequence, the current geological condition data sequence, and the current environmental data sequence into the pre-trained production dynamic prediction model, extracting the current production dynamic time series coding features, the current geological condition time series coding features, and the current environmental time series coding features through the time series feature coding layer; performing weighted fusion on the current production dynamic time series coding features, the current geological condition time series coding features, and the current environmental time series coding features through the multi-source feature fusion layer to obtain the current multi-source fusion features; and processing the current multi-source fusion features through the prediction output layer to obtain the predicted values of the gas production rate and the bottom hole pressure at each time in the future preset time period as the production dynamic prediction result.

[0157] The collected current production dynamic data sequence, current geological condition data sequence, and current environment data sequence are input into the pre-trained production dynamic prediction model. First, the three parallel branches of the time sequence feature encoding layer respectively extract features from the three types of current data sequences, outputting current production dynamic time sequence encoding features, current geological condition time sequence encoding features, and current environment time sequence encoding features.

[0158] Subsequently, the multi-source feature fusion layer calculates the attention weights of the three types of current time sequence encoding features, performs weighted fusion, and generates current multi-source fusion features. Finally, the prediction output layer processes the current multi-source fusion features and outputs the gas production rate prediction values and the bottom hole pressure prediction values at each time point in the future preset time period, which together constitute the production dynamic prediction results.

[0159] Step S153: setting a production dynamic prediction threshold, which includes a gas production rate prediction threshold range and a bottom hole pressure prediction threshold range, the gas production rate prediction threshold range including an upper threshold and a lower threshold, and the bottom hole pressure prediction threshold range including an upper threshold and a lower threshold.

[0160] According to the normal production standard of the coalbed methane well, the safe operation parameters of the equipment, and the production target requirements, the production dynamic prediction threshold is set. The gas production rate prediction threshold range is composed of an upper threshold and a lower threshold, the upper threshold being the maximum allowed value of the gas production rate under normal production conditions, and the lower threshold being the minimum gas production rate value to ensure production efficiency; the bottom hole pressure prediction threshold range also includes an upper threshold and a lower threshold, the upper threshold being the maximum allowed bottom hole pressure to prevent equipment damage, and the lower threshold being the minimum bottom hole pressure to ensure gas production efficiency. The above threshold ranges need to be determined in combination with field production experience and equipment technical parameters.

[0161] Step S154: comparing the gas production rate prediction values in the production dynamic prediction results with the gas production rate prediction threshold range, and marking the gas production rate as abnormal if the gas production rate prediction values are lower than the lower threshold or higher than the upper threshold; comparing the bottom hole pressure prediction values in the production dynamic prediction results with the bottom hole pressure prediction threshold range, and marking the bottom hole pressure as abnormal if the bottom hole pressure prediction values are lower than the lower threshold or higher than the upper threshold.

[0162] Each time point's gas production rate prediction value in the production dynamic prediction results is compared with the gas production rate prediction threshold range one by one. If the gas production rate prediction value at a certain time point is lower than the lower threshold or higher than the upper threshold, the gas production rate at that time point is marked as abnormal; if it is within the normal range, it is marked as normal.

[0163] The same method is used to compare the predicted value of the bottom hole pressure at each time with the predicted threshold range of the bottom hole pressure, and if it is out of range, it is marked as bottom hole pressure anomaly, otherwise it is marked as normal. The number of all abnormal moments and the corresponding predicted values are counted to determine the abnormal situation.

[0164] Step S155: For the effective deviation conduction closed loop path with the starting deviation type being the geological condition deviation, if there is a gas production rate anomaly, the preset geological-gas production regulation rule base is queried to extract the production parameter adjustment direction corresponding to the geological condition deviation and the gas production rate anomaly type; if there is a bottom hole pressure anomaly, the production parameter adjustment direction corresponding to the geological condition deviation and the bottom hole pressure anomaly type is extracted; for the effective deviation conduction closed loop path with the starting deviation type being the environmental factor deviation, if there is a gas production rate anomaly, the preset regulation rule base is queried to extract the production parameter adjustment direction corresponding to the environmental factor deviation and the gas production rate anomaly type; if there is a bottom hole pressure anomaly, the production parameter adjustment direction corresponding to the environmental factor deviation and the bottom hole pressure anomaly type is extracted.

[0165] The preset geological-gas production regulation rule base and the environmental-production regulation rule base store the corresponding relationship between different starting deviation types, different anomaly types and production parameter adjustment directions.

[0166] For the effective deviation conduction closed loop path with the starting deviation type being the geological condition deviation, if there is a gas production rate anomaly, the production parameter adjustment direction corresponding to the geological condition deviation (such as low coal seam permeability) and the gas production rate anomaly type (such as below the lower limit) is found in the geological-gas production regulation rule base (such as adjusting the extraction pressure); if there is a bottom hole pressure anomaly, the corresponding bottom hole pressure anomaly adjustment direction (such as adjusting the water injection amount) is found.

[0167] For the effective deviation conduction closed loop path with the starting deviation type being the environmental factor deviation, if there is a gas production rate anomaly or a bottom hole pressure anomaly, the corresponding production parameter adjustment direction is found in the environmental-production regulation rule base, such as adjusting the cooling system parameters when the surface temperature is too high to cause a gas production rate anomaly.

[0168] Step S156: Determine the adjustment priority coefficient according to the forward cumulative correlation strength of the effective deviation conduction closed loop path. The higher the forward cumulative correlation strength, the greater the adjustment priority coefficient.

[0169] The positive cumulative correlation strength of the effective deviation conduction closed loop path is taken as the core basis of the adjustment priority coefficient. The positive cumulative correlation strength reflects the closeness and influence degree of deviation conduction. The higher the strength, the greater the influence of the deviation on the production dynamics, which needs to be processed in priority. Therefore, the higher the positive cumulative correlation strength, the greater the adjustment priority coefficient corresponding to it; otherwise, the smaller. The adjustment priority coefficient needs to be normalized to convert it into a value between 0 and 1.

[0170] Step S157: Determine the abnormal influence coefficient in combination with the production dynamic abnormality degree. The production dynamic abnormality degree is the absolute value of the difference between the gas production rate prediction value and the threshold value, and the absolute value of the difference between the bottom hole pressure prediction value and the threshold value. The higher the abnormality degree, the greater the abnormal influence coefficient.

[0171] Calculate the production dynamic abnormality degree. For gas production rate abnormality, the abnormality degree is the absolute value of the difference between the gas production rate prediction value and the corresponding threshold value (lower limit or upper limit); for bottom hole pressure abnormality, the absolute value of the difference between the bottom hole pressure prediction value and the corresponding threshold value. The greater the absolute value, the more serious the abnormality degree, and the greater the influence on production.

[0172] Determine the abnormal influence coefficient according to the size of the abnormality degree. The higher the abnormality degree, the greater the abnormal influence coefficient. The abnormal influence coefficient also needs to be normalized to convert it into a value between 0 and 1.

[0173] Step S158: Normalize the adjustment priority coefficient and the abnormal influence coefficient respectively, add the normalized adjustment priority coefficient and the normalized abnormal influence coefficient to obtain the comprehensive priority score of each production parameter adjustment direction.

[0174] If the adjustment priority coefficient and the abnormal influence coefficient have not been normalized, first convert them into values between 0 and 1 using the min-max normalization method. Then, add the normalized adjustment priority coefficient and the normalized abnormal influence coefficient to obtain the total sum, which is the comprehensive priority score of each production parameter adjustment direction. The higher the comprehensive priority score, the more the adjustment direction needs to be implemented in priority.

[0175] Step S159: Sort the production parameter adjustment directions in the order from high to low according to the comprehensive priority score, and mark the effective deviation conduction closed loop path number, correlation strength data and predicted abnormal type corresponding to each production parameter adjustment direction.

[0176] The adjustment direction of all production parameters is sorted according to the comprehensive priority score from high to low, and the adjustment direction with the highest score is ranked first and executed preferentially. In the sorting process, the corresponding effective deviation conduction closed loop path number is marked for each production parameter adjustment direction to trace its source; the positive and negative cumulative correlation strength and other correlation strength data of the path are marked; and the corresponding predicted abnormal type, such as gas production rate lower than the lower limit, bottom hole pressure higher than the upper limit, etc. is marked.

[0177] Step S1510: The sorted production parameter adjustment direction, marked information and adjustment implementation period are integrated, the adjustment implementation period is determined based on the deviation evolution trend, the faster the deviation expands, the shorter the implementation period, and a dynamic control strategy containing adjustment sequence, adjustment object, adjustment direction, implementation period and correlation basis is generated, and the adjustment object is a specific production parameter.

[0178] According to the deviation evolution trend of the effective deviation conduction closed loop path in the deviation conduction analysis report, the adjustment implementation period of each production parameter adjustment direction is determined. If the deviation evolution trend is an expansion trend and the expansion speed is faster, the implementation period is set to be shorter to quickly suppress the deviation expansion; if the deviation expansion speed is slower or shows a slowing trend, the implementation period can be appropriately extended.

[0179] The sorted production parameter adjustment direction, corresponding marked information (path number, correlation strength data, predicted abnormal type) and adjustment implementation period are integrated, and a dynamic control strategy is generated according to a preset format. The dynamic control strategy clearly defines the adjustment sequence (sorted according to the comprehensive priority score), the adjustment object (such as specific production parameters such as drainage pressure, water injection volume, cooling system parameters, etc.), the adjustment direction (such as increasing, decreasing, maintaining, etc.), the implementation period (such as immediate execution, execution within 24 hours, phased execution within 72 hours, etc.) and the correlation basis (corresponding effective deviation conduction closed loop path information and production dynamic prediction abnormal information). For example, the comprehensive priority score of a certain production parameter adjustment direction is the highest, the adjustment object is drainage pressure, the adjustment direction is to increase, the implementation period is immediate execution, and the correlation basis is the positive cumulative correlation strength of a certain effective deviation conduction closed loop path and the abnormal information that the predicted gas production rate is lower than the lower limit, which all need to be clearly reflected in the dynamic control strategy.

[0180] For example, step S1511: converting the dynamic control strategy into an instruction format recognizable by a production control terminal, the instruction format including adjustment parameter identification, adjustment direction code, and adjustment implementation time node.

[0181] According to the communication protocol and instruction specification of the production control terminal, the contents in the dynamic regulation strategy are converted into instruction formats recognizable by the terminal. Among them, the adjustment parameter identifier is the unique code of the specific production parameter (such as the extraction pressure, water injection volume); the adjustment direction code is the preset digital or character code corresponding to the adjustment direction of increasing, decreasing, maintaining, etc.; the adjustment implementation time node is the specific execution time, such as the corresponding specific time stamp for immediate execution or the start and end time stamps of each stage for phased execution. The instruction format needs to meet the parsing requirements of the production control terminal to ensure that the instruction can be accurately recognized.

[0182] Step S1512: Send the converted dynamic regulation strategy instruction to the production control terminal. After receiving the instruction, the production control terminal executes the production parameter adjustment operation according to the adjustment implementation time node in the instruction.

[0183] The converted dynamic regulation strategy instruction is sent to the production control terminal through industrial Ethernet or a dedicated communication line. After receiving the instruction, the production control terminal first verifies the integrity and legality of the instruction, and after verification, it calls the internal parameter adjustment module and executes the adjustment operation on the corresponding production parameter according to the adjustment implementation time node in the instruction. For example, for the immediate execution of the extraction pressure increase instruction, the terminal immediately sends a control signal to the extraction pressure adjustment device, and the adjustment device changes the extraction pressure value according to the signal; for the phased execution of the water injection volume adjustment instruction, the terminal sends adjustment signals in sequence according to the preset stage time to gradually complete the water injection volume adjustment.

[0184] Step S1513: After the production parameter adjustment operation is executed, collect the regulated production dynamic data sequence, the regulated geological condition data sequence, and the regulated environmental data sequence of the coalbed methane well according to the preset data collection frequency.

[0185] A preset data collection frequency is set, which is consistent with the data collection frequency before regulation to ensure the comparability of the data. After the production parameter adjustment operation is executed, the data collection device is started to continuously collect the regulated production dynamic data (gas production rate, wellbore pressure), geological condition data (coal seam permeability, porosity), and environmental data (surface temperature, humidity) according to the collection frequency, forming the regulated production dynamic data sequence, the regulated geological condition data sequence, and the regulated environmental data sequence, all of which are attached with corresponding time stamps.

[0186] Step S1514: Perform validity detection on the regulated production dynamic data sequence, the regulated geological condition data sequence, and the regulated environmental data sequence, and retain the effective regulated data sequence.

[0187] The same time sequence continuity detection method as step S121 is used to detect the effectiveness of the three types of data sequences after regulation. The time stamps in the data sequence are traversed to check whether the interval between adjacent time stamps meets the preset acquisition frequency and whether the data value is within a reasonable range (such as the gas production rate being not less than 0 and the pressure being not exceeding the maximum bearing value of the equipment). For a data segment that is continuous and has reasonable data values, it is determined as an effective regulated data sequence; for a data segment that has time stamp abnormalities, data missing or abnormal data values, it is rejected and only the effective part is retained.

[0188] Step S1515: The effective regulated data sequence is divided into new production dynamic time sequence segments, new geological condition dynamic time sequence segments and new environment dynamic time sequence segments according to the dynamic time interval division rule.

[0189] The dynamic time interval division rule determined in step S122 is called to divide the effective regulated production dynamic data sequence, the effective regulated geological condition data sequence and the effective regulated environment data sequence into time sequence segments. The division process is consistent with that before regulation: the time interval is determined according to the fluctuation frequency of the data, and small intervals are used for high fluctuation frequencies and large intervals are used for low fluctuation frequencies. After division, new production dynamic time sequence segments, new geological condition dynamic time sequence segments and new environment dynamic time sequence segments are formed, each of which contains complete data in the corresponding time period after regulation.

[0190] Step S1516: The features of the new production dynamic time sequence segments, the features of the new geological condition dynamic time sequence segments and the features of the new environment dynamic time sequence segments are extracted, the features of the new production dynamic time sequence segments include the gas production rate change trend feature and the bottom hole pressure change trend feature, the correlation strength between the new production dynamic time sequence segments and the original production dynamic time sequence segments, the correlation strength between the new geological condition dynamic time sequence segments and the original geological condition dynamic time sequence segments, and the correlation strength between the new environment dynamic time sequence segments and the original environment dynamic time sequence segments are recalculated and corrected.

[0191] The feature extraction method of step S123 is used to extract the gas production rate change trend feature and the bottom hole pressure change trend feature of the new production dynamic time sequence segments, the coal seam permeability change trend feature and the coal seam porosity change trend feature of the new geological condition dynamic time sequence segments, and the surface temperature change trend feature and the surface humidity change trend feature of the new environment dynamic time sequence segments.

[0192] According to the correlation strength calculation method of step S124 and the correction method of step S125, the correlation strength between the new production dynamic dynamic time sequence segment and the original production dynamic dynamic time sequence segment (the segment divided before regulation), the correlation strength between the new geological condition dynamic time sequence segment and the original geological condition dynamic time sequence segment, and the correlation strength between the new environmental dynamic time sequence segment and the original environmental dynamic time sequence segment are calculated respectively, and the time sequence correlation strength correction coefficient is introduced for correction to obtain the corrected new correlation strength.

[0193] Step S1517: According to the new production dynamic dynamic time sequence segment, the new geological condition dynamic time sequence segment, the new environmental dynamic time sequence segment, and the corrected correlation strength, the node set and the node connection weight of the ternary time sequence dynamic correlation network are updated, and the node set of the ternary time sequence dynamic correlation network adds the nodes corresponding to the new production dynamic dynamic time sequence segment, the new geological condition dynamic time sequence segment, and the new environmental dynamic time sequence segment.

[0194] In the node set of the ternary time sequence dynamic correlation network, the nodes corresponding to the new production dynamic dynamic time sequence segment, the new geological condition dynamic time sequence segment, and the new environmental dynamic time sequence segment are added, and the new nodes are assigned with unique identifiers. According to the corrected new correlation strength, a connection is established between the new nodes and the original nodes, and the new correlation strength is taken as the connection weight; at the same time, for the correlation relationship between the new nodes (such as the new production dynamic node and the new geological condition node), the correlation strength is calculated and corrected to establish the connection and the weight. In addition, according to the correlation of the new nodes, the weight of the connection between the original nodes is updated (if the original correlation strength is changed due to the regulation), and the update of the entire network node set and the connection weight is completed.

[0195] Step S1518: At the same time, the regulated production dynamic data sequence, the regulated geological condition data sequence, the regulated environmental data sequence, and the ternary time sequence dynamic correlation network update record are stored in the historical database.

[0196] The three types of data sequences collected after the regulation and the update record of the ternary time sequence dynamic correlation network (including the new node information, the new connection information, the weight before and after the update, the update time, etc.) are transmitted to the historical database for storage. The historical database adopts a distributed storage architecture to ensure the safety and scalability of data storage.

[0197] Step S1519: A network update report is generated, and the network update report and the execution feedback information of the dynamic regulation strategy are output to the monitoring terminal together, the network update report contains the number of new nodes, the weight update proportion, and the updated correlation strength statistical information, and the execution feedback information of the dynamic regulation strategy includes whether the adjustment is completed and the data collection state after the adjustment.

[0198] The update information of the integrated ternary time sequence dynamic correlation network is generated to generate a network update report. The report includes the number of newly added production dynamics, geological conditions, and environmental nodes, the proportion of weight updates (the proportion of the number of updated links to the total number of links), and the statistical information of the correlation strength of each type after the update (such as the average value, the maximum value, the minimum value, the distribution, etc.).

[0199] At the same time, the dynamic regulation and control strategy execution information fed back by the production control terminal is collected, including whether each adjustment operation is completed according to the plan, whether an exception occurs in the adjustment process, whether the data acquisition after the regulation and control is normal, etc., to form execution feedback information. The network update report and the execution feedback information are combined, and are sent to the monitoring terminal through a data transmission protocol. After being received by the monitoring terminal, the network update report and the execution feedback information are displayed to enable the operation and maintenance personnel to master the regulation and control execution situation and the network update state in real time.

[0200] Based on the same inventive concept, please refer to Figure 2 , a structure schematic block diagram of a coalbed methane well production dynamic analysis system 100 based on AI time sequence prediction provided by an embodiment of the present application is shown. The coalbed methane well production dynamic analysis system 100 based on AI time sequence prediction can include a communication unit 110, a machine readable storage medium 120, and a processor 130.

[0201] In the embodiment, the machine readable storage medium 120 and the processor 130 are both located in the coalbed methane well production dynamic analysis system 100 based on AI time sequence prediction and are separately arranged. However, it should be understood that the machine readable storage medium 120 can also be independent of the coalbed methane well production dynamic analysis system 100 based on AI time sequence prediction, and can be accessed by the processor 130 through a bus interface. Alternatively, the machine readable storage medium 120 can also be integrated into the processor 130, and can communicate and interact with an external system through the communication unit 110.

[0202] The processor 130 is the control center of the AI time series prediction based coalbed methane well production dynamic analysis system 100, connects each part of the AI time series prediction based coalbed methane well production dynamic analysis system 100 by using various interfaces and lines, executes various functions of the AI time series prediction based coalbed methane well production dynamic analysis system 100 and processes data by running or executing the software program and / or module stored in the machine readable storage medium 120 and calling the data stored in the machine readable storage medium 120, thereby monitoring the AI time series prediction based coalbed methane well production dynamic analysis system 100 as a whole. Optionally, the processor 130 can include one or more processing cores; for example, the processor 130 can integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface and application program, etc., and the modem processor mainly processes wireless communication. It can be understood that the above-mentioned modem processor can also not be integrated into the processor. The machine readable storage medium 120 is used to store machine executable instructions for executing the scheme of the present application, and the processor 130 is used to execute the machine executable instructions stored in the machine readable storage medium 120 to realize the method of the above-mentioned method embodiment.

[0203] It should be noted that, in order to simplify the expression of the present disclosure and help to understand one or more embodiments of the present application, in the foregoing description of the embodiments of the present application, various features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A coalbed methane well production dynamic analysis method based on AI timing prediction, characterized in that, The method comprises: Collecting production dynamic data sequences of coalbed methane wells, geological condition data sequences corresponding to well locations, and surrounding environment data sequences, wherein the production dynamic data sequences contain gas production rate change information and well bottom pressure change information, the geological condition data sequences contain coal seam permeability change information and coal seam porosity change information, and the environment data sequences contain surface temperature change information and surface humidity change information; Based on the production dynamic data sequences, the geological condition data sequences, and the environment data sequences, a ternary time sequence dynamic correlation network of coalbed methane wells is constructed, wherein each node in the ternary time sequence dynamic correlation network corresponds to a dynamic time sequence segment of a single type of data sequence, the connection between nodes corresponds to the real-time correlation relationship between dynamic time sequence segments of different types of data sequences, and the connection weight evolves and updates over time; A pre-trained AI time sequence coupling bias evolution mining model is called to process the ternary time sequence dynamic correlation network, mine the coupling bias evolution information between dynamic time sequence segments of different data sequences, and obtain a ternary time sequence coupling bias result containing a bias evolution trend; According to the ternary time sequence coupling bias result, a bias conduction closed loop path affecting production dynamics is determined by combining forward conduction positioning and reverse tracing verification, and a bias conduction analysis report containing bias conduction direction, correlation strength, and reverse tracing verification result is generated; Based on the bias conduction analysis report, a dynamic control strategy for the production parameters of the coalbed methane well is generated by combining a production dynamic prediction threshold, and the dynamic control strategy is output to a production control terminal, and the controlled data is fed back to the ternary time sequence dynamic correlation network to update the node correlation relationship. 2.The AI timing prediction-based coalbed gas well production dynamic analysis method according to claim 1, characterized in that, The construction of the ternary time sequence dynamic correlation network of the coalbed methane well based on the production dynamic data sequences, the geological condition data sequences, and the environment data sequences comprises: Time sequence continuity detection is performed on the production dynamic data sequences, the geological condition data sequences, and the environment data sequences, and continuous data segments are retained as effective data sequences; According to the change fluctuation frequency of the effective data sequences, a dynamic time interval division rule is determined, the effective production dynamic data sequences are divided into a plurality of production dynamic time sequence segments, the effective geological condition data sequences are divided into a plurality of geological condition dynamic time sequence segments, and the effective environment data sequences are divided into a plurality of environment dynamic time sequence segments, to obtain a production dynamic time sequence segment set, a geological condition dynamic time sequence segment set, and an environment dynamic time sequence segment set, and the higher the change fluctuation frequency, the smaller the time interval; The gas production rate change trend feature and the well bottom pressure change trend feature of each production dynamic time sequence segment are extracted, the coal seam permeability change trend feature and the coal seam porosity change trend feature of each geological condition dynamic time sequence segment are extracted, and the surface temperature change trend feature and the surface humidity change trend feature of each environment dynamic time sequence segment are extracted; calculating the real-time correlation strength between the gas production rate change trend feature and the coal seam permeability change trend feature, calculating the real-time correlation strength between the gas production rate change trend feature and the surface temperature change trend feature, calculating the real-time correlation strength between the bottom hole pressure change trend feature and the coal seam porosity change trend feature, calculating the real-time correlation strength between the bottom hole pressure change trend feature and the surface humidity change trend feature, calculating the real-time correlation strength between the coal seam permeability change trend feature and the surface temperature change trend feature, and calculating the real-time correlation strength between the coal seam porosity change trend feature and the surface humidity change trend feature; a time sequence correlation strength correction coefficient is introduced, the time sequence correlation strength correction coefficient is determined based on the correlation strength fluctuation range of the historical same period same type data sequence, each real-time correlation strength is multiplied by the corresponding time sequence correlation strength correction coefficient to obtain a corrected correlation strength; each production dynamic time sequence segment, each geological condition dynamic time sequence segment, and each environmental dynamic time sequence segment are taken as network nodes, and the corrected correlation strength is taken as the initial weight of the connection line between the nodes to build an initial ternary time sequence dynamic correlation network; a weight updating period is set, the real-time correlation strength between the nodes is recalculated and corrected according to the weight updating period, the connection line weight of the initial ternary time sequence dynamic correlation network is updated, and a final ternary time sequence dynamic correlation network of the coal seam gas well is obtained. 3.The AI timing prediction-based coalbed gas well production performance analysis method according to claim 2, characterized in that, The calculation of the real-time correlation strength between the gas production rate change trend feature and the coal seam permeability change trend feature includes: a first dynamic time sequence vector corresponding to the gas production rate change trend feature is obtained, each element of the first dynamic time sequence vector corresponds to the gas production rate value at different time in the production dynamic time sequence segment; a second dynamic time sequence vector corresponding to the coal seam permeability change trend feature is obtained, each element of the second dynamic time sequence vector corresponds to the coal seam permeability value at different time in the geological condition dynamic time sequence segment; the first dynamic time sequence vector and the second dynamic time sequence vector are input into a time sequence correlation analysis module, and the first dynamic time sequence vector and the second dynamic time sequence vector are subjected to elastic alignment processing in the time dimension based on a dynamic time warping algorithm to obtain an aligned first dynamic time sequence vector and an aligned second dynamic time sequence vector; rising inflection point time and falling inflection point time of the aligned first dynamic time sequence vector are extracted, and rising inflection point time and falling inflection point time of the aligned second dynamic time sequence vector are extracted; the number of coincidences of the rising inflection point time of the first dynamic time sequence vector and the rising inflection point time of the second dynamic time sequence vector is counted, and the number of coincidences of the falling inflection point time of the first dynamic time sequence vector and the falling inflection point time of the second dynamic time sequence vector is counted, and the two numbers of coincidences are added to obtain the total number of coincident inflection points; counting a total number of inflection points of the first dynamic time series vector, the total number of inflection points of the first dynamic time series vector being a sum of the number of rising inflection points and the number of falling inflection points, counting a total number of inflection points of the second dynamic time series vector, the total number of inflection points of the second dynamic time series vector being a sum of the number of rising inflection points and the number of falling inflection points, and adding the two total numbers of inflection points to obtain a total number of inflection points; calculating a proportion of the total number of coinciding inflection points to the total number of inflection points as an inflection point coordination proportion; calculating a change amplitude mean value of the aligned first dynamic time series vector, the change amplitude mean value being an average of absolute values of differences between each time point and a previous time point, calculating a change amplitude mean value of the aligned second dynamic time series vector, the change amplitude mean value being an average of absolute values of differences between each time point and a previous time point, and taking a ratio of the two change amplitude mean values as an amplitude coordination coefficient; respectively normalizing the inflection point coordination proportion and the amplitude coordination coefficient to convert them into dimensionless score values, setting an inflection point coordination proportion weight and an amplitude coordination coefficient weight, the sum of the inflection point coordination proportion weight and the amplitude coordination coefficient weight being 1, multiplying the normalized inflection point coordination proportion by the inflection point coordination proportion weight, adding the normalized amplitude coordination coefficient multiplied by the amplitude coordination coefficient weight, and obtaining a real-time correlation strength between the gas production rate change trend feature and the coal seam permeability change trend feature. 4.The AI timing prediction-based coalbed gas well production performance analysis method according to claim 1, characterized in that, The pre-trained AI time series coupling deviation evolution mining model is called to process the ternary time series dynamic correlation network, mine coupling deviation evolution information between different data sequence dynamic time series segments, and obtain a ternary time series coupling deviation result containing a deviation evolution trend, including: The ternary time series dynamic correlation network is input into a network analysis layer of the AI time series coupling deviation evolution mining model, dynamic time series segment data corresponding to each node, dynamic time series segment data corresponding to each node containing specific data values of each time point in the segment, modified correlation strength data corresponding to the connection between nodes and weight update records are analyzed, and network analysis data is obtained. The network analysis data is input into a time series coupling analysis layer of the AI time series coupling deviation evolution mining model, and the coupling matching degrees of production dynamic time series segments and geological condition dynamic time series segments are analyzed from three dimensions of trend consistency, phase difference coordination and amplitude coordination. The trend consistency is determined by calculating the coincidence length proportion of the change trends of the two types of dynamic time series segments, the phase difference coordination is determined by calculating the time difference mean value of the inflection point time of the two types of dynamic time series segments, and the amplitude coordination is determined by calculating the fluctuation range of the change amplitude ratio of the two types of dynamic time series segments. The coupling matching degrees of the three dimensions are respectively normalized, a weight is set for the normalized coupling matching degree of each dimension, a weighted sum of the three types of dimension coupling matching degree score values is calculated according to the set weight, and the weighted sum is taken as the comprehensive coupling matching degree of the two types of dynamic time series segments. Dynamic time series segment combinations with a comprehensive coupling matching degree lower than a preset matching threshold are identified, and the dynamic time series segment combinations are marked as deviation candidate combinations. For each deviation candidate combination, extract its comprehensive coupling matching degree change data in multiple consecutive weight update periods, construct a deviation evolution curve, the horizontal coordinate of the deviation evolution curve is the weight update period, and the vertical coordinate of the deviation evolution curve is the comprehensive coupling matching degree; Analyze the slope change trend of the deviation evolution curve, if the slope is negative and the absolute value gradually increases, it is determined that the deviation is expanding; if the slope is negative and the absolute value gradually decreases, it is determined that the deviation is slowing down; if the slope is positive, it is determined that the deviation is repairing; For the dynamic timing segment in the deviation candidate combination, the deviation degree is calculated, and the deviation value corresponding to each deviation candidate combination is obtained; Integrate the deviation candidate combination, the deviation value and the deviation evolution trend to generate a ternary timing coupling deviation result containing the deviation position, the deviation type, the deviation degree and the deviation evolution trend, the deviation position corresponds to the dynamic timing segment number, the deviation type includes production-geology deviation, production-environment deviation and geology-environment deviation, and the deviation degree is the deviation value. 5.The AI timing prediction-based coalbed gas well production performance analysis method according to claim 4, characterized in that, The deviation degree calculation of the dynamic timing segment in the deviation candidate combination to obtain the deviation value corresponding to each deviation candidate combination comprises: Collect the production dynamic data sequence, the geological condition data sequence and the environment data sequence of the coal bed methane well in the historical normal production state, extract the coupling data of the same type dynamic timing segment, and construct a standard coupling relationship library, the standard coupling relationship library stores the standard trend consistency range, the standard phase difference cooperativity range and the standard amplitude cooperativity range of different types of dynamic timing segment combinations; According to the type of the deviation candidate combination, the corresponding standard trend consistency range, the standard phase difference cooperativity range and the standard amplitude cooperativity range are called from the standard coupling relationship library as the standard coupling relationship of the deviation candidate combination, and the type of the deviation candidate combination includes production-geology deviation, production-environment deviation and geology-environment deviation; Extract the actual change curve of the first type of dynamic timing segment in the deviation candidate combination, and extract the actual change curve of the second type of dynamic timing segment in the deviation candidate combination, the horizontal coordinate of the actual change curve is the time in the segment, and the vertical coordinate of the actual change curve is the data value; According to the standard trend consistency range in the standard coupling relationship, a standard trend curve range corresponding to the first type of dynamic timing segment is generated, and a standard trend curve range corresponding to the second type of dynamic timing segment is generated, the standard trend curve range includes an upper limit trend curve and a lower limit trend curve, and the standard trend curve range includes an upper limit trend curve and a lower limit trend curve; Calculate the difference area of the actual change curve of the first type of dynamic timing segment and the upper limit of the standard trend curve range, calculate the difference area of the actual change curve of the first type of dynamic timing segment and the lower limit of the standard trend curve range, and take the maximum value of the two difference areas as the first type of segment trend deviation area; Calculate the difference value area between the actual change curve of the second type of dynamic timing segment and the upper limit of the standard trend curve range, calculate the difference value area between the actual change curve of the second type of dynamic timing segment and the lower limit of the standard trend curve range, and take the maximum of the two difference value areas as the trend deviation area of the second type of segment; According to the standard phase difference cooperativity range in the standard coupling relationship, calculate the absolute value of the difference between the actual phase difference and the upper limit of the standard phase difference cooperativity range in the deviation candidate combination, and the actual phase difference is the difference value at the inflection point, calculate the absolute value of the difference between the actual phase difference and the lower limit of the standard phase difference cooperativity range, and take the maximum of the two difference absolute values as the phase difference deviation value; According to the standard amplitude cooperativity range in the standard coupling relationship, calculate the absolute value of the difference between the actual amplitude ratio and the upper limit of the standard amplitude cooperativity range in the deviation candidate combination, and the actual amplitude ratio is the ratio of the average change amplitude, calculate the absolute value of the difference between the actual amplitude ratio and the lower limit of the standard amplitude cooperativity range, and take the maximum of the two difference absolute values as the amplitude cooperativity deviation value; Set the trend deviation weight, phase difference deviation weight and amplitude cooperativity deviation weight, and the sum of the trend deviation weight, phase difference deviation weight and amplitude cooperativity deviation weight is 1; Respectively, the average value of the first type of segment trend deviation area, the second type of segment trend deviation area, the phase difference deviation value and the amplitude cooperativity deviation value are normalized, the normalized trend deviation area score value is multiplied by the trend deviation weight, the normalized phase difference deviation score value is added to the phase difference deviation weight, and the normalized amplitude cooperativity deviation score value is added to the amplitude cooperativity deviation weight. Get the deviation value corresponding to the deviation candidate combination. 6.The AI timing prediction-based coalbed gas well production performance analysis method according to claim 1, characterized in that, According to the ternary timing coupling deviation result, combined with forward conduction positioning and reverse trace verification, the deviation conduction closed loop path affecting production dynamics is determined, and a deviation conduction analysis report containing deviation conduction direction, correlation strength and trace verification result is generated, including: From the ternary timing coupling deviation result, all deviation candidate combinations and corresponding deviation values and deviation evolution trend are extracted, deviation candidate combinations with expanding trend of deviation evolution trend are screened out, and the screened deviation candidate combinations are sorted in descending order of deviation value; Select the top preset number of deviation candidate combinations as key deviation combinations, extract the timestamp information of the two types of dynamic timing segments in each key deviation combination, and determine the time sequence of the two types of dynamic timing segments; According to the time sequence, the first appearing dynamic timing segment is marked as the starting dynamic timing segment, and the later appearing dynamic timing segment is marked as the affected dynamic timing segment, and the forward conduction direction of the deviation from the starting dynamic timing segment to the affected dynamic timing segment is preliminarily determined; extracting the link correlation strength between the node corresponding to the starting dynamic time sequence segment and the node corresponding to the affected dynamic time sequence segment from the ternary time sequence dynamic correlation network, and simultaneously extracting the link correlation strength between the node corresponding to the starting dynamic time sequence segment and other related nodes, the other related nodes being nodes that exist in correlation with the affected dynamic time sequence segment, and extracting the link correlation strength between the node corresponding to the affected dynamic time sequence segment and the other related nodes; constructing a forward conduction strength matrix, a row of the forward conduction strength matrix representing a starting node and a related node, a column of the forward conduction strength matrix representing an affected node and a related node, and an element of the forward conduction strength matrix being a correlation strength between corresponding nodes; calculating a cumulative correlation strength of a forward conduction path, and screening a path with a cumulative correlation strength higher than a forward conduction threshold value as a candidate forward conduction path, the cumulative correlation strength being an average of correlation strengths between all nodes on the path; for each candidate forward conduction path, starting reverse tracing verification: taking the affected dynamic time sequence segment as a starting point and the starting dynamic time sequence segment as a terminal point, querying a reverse correlation path from the affected node to the starting node in the ternary time sequence dynamic correlation network, calculating a cumulative correlation strength of the reverse correlation path, the cumulative correlation strength being an average of correlation strengths between all nodes on the path; if the cumulative correlation strength of the reverse correlation path is higher than a reverse tracing threshold value, and the reverse correlation path and the candidate forward conduction path form a node-intercommunicating closed loop structure, then determining that the candidate forward conduction path is an effective deviation conduction closed loop path; if the cumulative correlation strength of the reverse correlation path is lower than the reverse tracing threshold value, or the reverse correlation path cannot form a closed loop structure, then eliminating the candidate forward conduction path; statistically analyzing a forward cumulative correlation strength, a reverse cumulative correlation strength, a number of involved nodes, and a dynamic time sequence segment type of each effective deviation conduction closed loop path, integrating a deviation conduction direction, a correlation strength, and a tracing verification result of the effective deviation conduction closed loop path, generating a deviation conduction analysis report, the deviation conduction direction being a forward conduction direction, the correlation strength being the forward cumulative correlation strength, and the tracing verification result including the reverse cumulative correlation strength and a closed loop verification result.

7. The AI timing prediction-based coalbed gas well production dynamic analysis method according to claim 6, characterized in that, The constructing of the forward conduction strength matrix comprises: determining a starting node number corresponding to the starting dynamic time sequence segment and an affected node number corresponding to the affected dynamic time sequence segment; from a node correlation list of the ternary time sequence dynamic correlation network, querying all nodes having a link correlation with the starting node as starting related nodes, and querying all nodes having a link correlation with the affected node as affected related nodes; integrating the starting node and the starting related nodes to form a matrix row node set, and arranging nodes in the matrix row node set in ascending order of node numbers; integrating the affected node and the affected related nodes to form a matrix column node set, and arranging nodes in the matrix column node set in ascending order of node numbers; constructing a blank forward conduction strength matrix, the number of rows of the blank forward conduction strength matrix being equal to the number of nodes in the matrix row node set, and the number of columns of the blank forward conduction strength matrix being equal to the number of nodes in the matrix column node set; For each node in the matrix row node set, the node is recorded as a row node, for each node in the matrix column node set, the node is recorded as a column node, and the modified correlation strength between the row node and the column node in the ternary time sequence dynamic correlation network is queried; If there is a connection correlation between the row node and the column node, the modified correlation strength is filled into the element position of the corresponding row of the row node and the corresponding column of the column node in the forward conduction strength matrix; If there is no connection correlation between the row node and the column node, the element value is set to 0 and filled into the element position of the corresponding row of the row node and the corresponding column of the column node in the forward conduction strength matrix, to obtain a complete forward conduction strength matrix. 8.The AI timing prediction-based coalbed gas well production performance analysis method according to claim 1, characterized in that, Based on the deviation conduction analysis report, a dynamic control strategy for the production parameters of the coalbed methane well is generated in combination with a production dynamic prediction threshold, including: Analyzing the effective deviation conduction closed loop path in the deviation conduction analysis report, determining the starting deviation type corresponding to each effective deviation conduction closed loop path, and the starting deviation type includes a geological condition deviation and an environmental factor deviation, extracting the forward cumulative correlation strength and the reverse cumulative correlation strength of each effective deviation conduction closed loop path; A pre-trained production dynamic prediction model is called to input the current production dynamic data sequence, the current geological condition data sequence, and the current environmental data sequence, to predict the gas production rate prediction value and the bottom hole pressure prediction value in a future preset time period as the production dynamic prediction result; A production dynamic prediction threshold is set, including a gas production rate prediction threshold range and a bottom hole pressure prediction threshold range, the gas production rate prediction threshold range including an upper threshold and a lower threshold, and the bottom hole pressure prediction threshold range including an upper threshold and a lower threshold; The gas production rate prediction value in the production dynamic prediction result is compared with the gas production rate prediction threshold range, and if the gas production rate prediction value is lower than the lower threshold or higher than the upper threshold, it is marked as a gas production rate anomaly; the bottom hole pressure prediction value in the production dynamic prediction result is compared with the bottom hole pressure prediction threshold range, and if the bottom hole pressure prediction value is lower than the lower threshold or higher than the upper threshold, it is marked as a bottom hole pressure anomaly; For the effective deviation conduction closed loop path with the starting deviation type of the geological condition deviation, if there is a gas production rate anomaly, a preset geological-gas production control rule library is queried to extract the production parameter adjustment direction corresponding to the geological condition deviation and the gas production rate anomaly type; if there is a bottom hole pressure anomaly, the production parameter adjustment direction corresponding to the geological condition deviation and the bottom hole pressure anomaly type is extracted; For the effective deviation conduction closed loop path with the starting deviation type of the environmental factor deviation, if there is a gas production rate anomaly, a preset control rule library is queried to extract the production parameter adjustment direction corresponding to the environmental factor deviation and the gas production rate anomaly type; if there is a bottom hole pressure anomaly, the production parameter adjustment direction corresponding to the environmental factor deviation and the bottom hole pressure anomaly type is extracted; The adjustment priority coefficient is determined according to the forward cumulative correlation strength of the effective deviation conduction closed loop path, and the higher the forward cumulative correlation strength, the greater the adjustment priority coefficient. The abnormal influence coefficient is determined in combination with the production dynamic anomaly degree, and the production dynamic anomaly degree is the absolute value of the difference between the predicted value of the gas production rate and the threshold value, the absolute value of the difference between the predicted value of the bottom hole pressure and the threshold value, the higher the abnormal degree, the greater the abnormal influence coefficient; The adjustment priority coefficient and the abnormal influence coefficient are normalized respectively, the normalized adjustment priority coefficient is added to the normalized abnormal influence coefficient, and a comprehensive priority score of each production parameter adjustment direction is obtained; The production parameter adjustment directions are sorted in descending order of the comprehensive priority score, and the effective deviation propagation closed loop path number, the correlation strength data and the predicted abnormal type corresponding to each production parameter adjustment direction are marked; The sorted production parameter adjustment directions, the marked information and the adjustment implementation period are integrated, the adjustment implementation period is determined based on the deviation evolution trend, the faster the deviation expands, the shorter the implementation period, and a dynamic control strategy including the adjustment order, the adjustment object, the adjustment direction, the implementation period and the correlation basis is generated, and the adjustment object is a specific production parameter. 9.The AI timing prediction-based coalbed gas well production performance analysis method according to claim 8, characterized in that, The pre-trained production dynamic prediction model is called to input the current production dynamic data sequence, the current geological condition data sequence and the current environment data sequence, to predict the predicted value of the gas production rate and the predicted value of the bottom hole pressure in a future preset time period as the production dynamic prediction result, which includes: The production dynamic prediction model training data set is constructed by collecting the production dynamic data sequence, the geological condition data sequence, the environment data sequence and the actual gas production rate data and the actual bottom hole pressure data in a future preset time period of a historical coalbed methane well; The production dynamic prediction model training data set is divided into a training set, a validation set and a test set, the training set is used for model parameter learning, the validation set is used for model hyperparameter adjustment, and the test set is used for model performance evaluation; The network structure of the production dynamic prediction model includes a time sequence feature coding layer, a multi-source feature fusion layer and a prediction output layer; the time sequence feature coding layer adopts a bidirectional long short-term memory network structure and is used for respectively performing time sequence feature extraction on the input production dynamic data sequence, the input geological condition data sequence and the input environment data sequence to output production dynamic time sequence coding features, geological condition time sequence coding features and environment time sequence coding features; the multi-source feature fusion layer adopts an attention mechanism fusion structure and is used for performing weighted fusion on the production dynamic time sequence coding features, the geological condition time sequence coding features and the environment time sequence coding features, the weights are dynamically adjusted based on the influence degree of each feature on the production dynamic prediction result, and multi-source fusion features are output; and the prediction output layer adopts a full connection network structure and is used for processing the multi-source fusion features to output the predicted value of the gas production rate and the predicted value of the bottom hole pressure at each time in a future preset time period; The production dynamic prediction model is trained using the training set, and after each preset training round, the prediction error of the production dynamic prediction model is calculated using the validation set, and the prediction error is the mean square error of the predicted value and the actual value of the gas production rate and the mean square error of the predicted value and the actual value of the bottom hole pressure; If the prediction error is higher than the preset error threshold, the number of hidden layer nodes of the bidirectional long short-term memory network, the weight calculation method of the attention mechanism, and the number of layers of the fully connected network are adjusted, and the production dynamic prediction model is retrained. If the prediction error is lower than the preset error threshold, the test set is used for final performance verification, and if the prediction error of the test set is also lower than the preset error threshold, a pre-trained production dynamic prediction model is obtained. The current production dynamic data sequence, the current geological condition data sequence, and the current environmental data sequence are input into the pre-trained production dynamic prediction model to extract current production dynamic time sequence coding features, current geological condition time sequence coding features, and current environmental time sequence coding features through a time sequence feature coding layer. The current production dynamic time sequence coding features, the current geological condition time sequence coding features, and the current environmental time sequence coding features are weighted and fused through a multi-source feature fusion layer to obtain current multi-source fusion features. The current multi-source fusion features are processed through a prediction output layer to obtain gas production rate prediction values and bottom hole pressure prediction values at each time point in a future preset time period as production dynamic prediction results.

10. A coalbed methane well production dynamic analysis system based on AI timing prediction, characterized in that, It comprises: a processor; a machine readable storage medium for storing machine executable instructions of the processor; wherein the processor is configured to execute the machine executable instructions to perform the AI time series prediction based coalbed methane well production dynamic analysis method of any one of claims 1 to 9.

Images