A method for constructing a multi-coal seam combined coalbed methane well productivity splitting model

Abstract

The present application relates to the technical field of coal seam mining, and relates to a multi-coal seam combined coalbed methane well productivity splitting model construction method.The present application screens the historical combined production data and historical production capacity of each similar historical coalbed methane well similar to the static data of the reservoir, establishes a preliminary productivity splitting model of the target coalbed methane well, outputs the predicted production capacity of the target coalbed methane well in each unit period through the real-time combined production data sequence of each coal seam in each unit period, adjusts the preliminary productivity splitting model based on the deviation value of the predicted production capacity and the actual production capacity to obtain an optimized productivity splitting model, realizes the incremental training and iterative adjustment of the model, and predicts the predicted production capacity of each coal seam in a future set time length, identifies the main coal seam of the target coalbed methane well in the future set time length, helps to exert the productivity potential of the main coal seam, improves the overall mining efficiency of multi-coal seam combined mining, and reduces the cost of ineffective mining.

Description

Technical Field

[0001] This invention relates to the field of coal seam mining technology, and specifically to a method for constructing a production capacity splitting model for coalbed methane wells that mine multiple coal seams. Background Technology

[0002] With the deepening development of coalbed methane resources, multi-seam co-production has become an important technical means to improve the productivity and recovery rate of single wells. During multi-seam co-production, significant differences exist in the reservoir properties, gas content, and permeability of different coal seams, leading to dynamic changes in the contribution of each coal seam to the total production. Accurately predicting the productivity contribution of coal seams and identifying the dominant coal seams are crucial for optimizing drainage systems, formulating precise development strategies, and achieving long-term stable production of coalbed methane wells.

[0003] Existing technologies mostly use empirical models with fixed parameters or static prediction models trained once for capacity allocation, which mainly have the following problems: First, existing technologies usually establish static prediction models based on initial geological conditions, lacking dynamic optimization mechanisms during production, resulting in insufficient model timeliness. As the mining stage progresses, the model cannot adapt to the dynamic changes in the capacity of each coal seam, the prediction accuracy gradually decreases, and it is difficult to accurately identify the main coal seams at different times, affecting the timely adjustment of mining plans.

[0004] Second, existing technologies do not fully consider the similarity matching of coal seam reservoir characteristics when using historical data. They simply use all historical well data or random sampling, resulting in poor model applicability. This makes it difficult to adapt to the individual characteristics of coalbed methane wells, which in turn affects the accuracy of subsequent mining strategy formulation and leads to unreasonable allocation of mining resources. Summary of the Invention

[0005] The present invention aims to overcome the shortcomings of the prior art and provide a method for constructing a production capacity splitting model for coalbed methane wells in multi-coal-seam co-mining, thereby providing technical support for the efficient development of multi-coal-seam co-mining.

[0006] The technical solution adopted by the present invention to solve its technical problem is: a method for constructing a production capacity splitting model for coalbed methane wells with multiple coal seams, including: obtaining the reservoir static data of each coal seam in the target coalbed methane well, screening the historical co-production data and historical production capacity of each similar historical coalbed methane well with similar reservoir static data, and establishing a preliminary production capacity splitting model for the target coalbed methane well.

[0007] Collect real-time combined mining data of each coal seam in the target coalbed methane well, and generate a real-time combined mining data sequence of each coal seam in each unit time period.

[0008] The real-time coalbed methane data sequence is input into the established preliminary capacity splitting model, and the predicted production capacity of the target coalbed methane well in each unit time period is output. The deviation between the predicted production capacity and the actual production capacity is calculated.

[0009] Based on the deviation value analysis, it is determined whether the initial capacity splitting model needs to be optimized. If optimization is required, the initial capacity splitting model is adjusted to obtain the optimized capacity splitting model.

[0010] Based on the real-time combined mining data sequence of each coal seam in each unit time period, and combined with the capacity splitting optimization model, the predicted production capacity of each coal seam for a set time period is predicted, and the main coal seam of the target coalbed methane well for the set time period is identified and displayed.

[0011] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention screens the historical co-production data and historical production capacity of similar historical coalbed methane wells that are similar to the static data of the reservoir, and establishes a preliminary model of the production capacity of the target coalbed methane well, thereby improving the adaptability of the model and the accuracy of the initial prediction, providing effective data for the accuracy of the later mining strategy formulation, and improving the rationality of mining resource allocation.

[0012] (2) This invention outputs the predicted production capacity of the target coalbed methane well in each unit time period through the real-time combined mining data sequence of each coal seam in each unit time period. Based on the deviation between the predicted production capacity and the actual production capacity, the preliminary production capacity splitting model is adjusted to obtain the production capacity splitting optimization model, realizing the incremental training and iterative adjustment of the model, ensuring that the model can match the dynamic changes in the production stage in real time, and effectively improving the prediction accuracy of the model.

[0013] (3) Based on the capacity splitting optimization model, this invention predicts the production capacity of each coal seam for a set time in the future, identifies the main coal seam of the target coalbed methane well for a set time in the future, provides a basis for resource tilting configuration and parameter adjustment in the mining stage, helps to maximize the production capacity potential of the main coal seam, improve the overall mining efficiency of multi-coal seam mining, and reduce ineffective mining costs. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a schematic diagram of the method steps of the present invention.

[0016] Figure 2 This is a schematic diagram illustrating the steps involved in establishing a preliminary model for the production capacity of the target coalbed methane well in this invention.

[0017] Figure 3 This is a schematic diagram illustrating the steps involved in obtaining the capacity splitting optimization model in this invention.

[0018] Figure 4 This is a schematic diagram of the main coal seam identification step in this invention. Detailed Implementation

[0019] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention. Furthermore, it should be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale.

[0020] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use. Techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and apparatus should be considered part of the specification.

[0021] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0022] Please see Figure 1 As shown, the present invention provides a method for constructing a production capacity splitting model for coalbed methane wells with multiple coal seams, including: S1, obtaining the reservoir static data of each coal seam in the target coalbed methane well, screening the historical co-production data and historical production capacity of each similar historical coalbed methane well with similar reservoir static data, and establishing a preliminary production capacity splitting model for the target coalbed methane well.

[0023] Given the heterogeneity of static reservoir data in multi-coalbed reservoirs—such as differences in reservoir depth, coal seam thickness, gas content, permeability, porosity, and reservoir pressure—directly using all historical coalbed methane well data or random sampling would result in low data matching with the target coalbed methane well and poor model generalization ability. Therefore, static data similarity screening is necessary to ensure consistency between historical data and the basic reservoir characteristics of the target coalbed methane well, laying a data foundation for subsequent model accuracy.

[0024] Furthermore, considering that the production capacity contribution of multi-coal-seam co-mining is significantly affected by inter-seam interference, for example, the target coalbed methane well is a 3-seam co-mining well, while the historical coalbed methane well is a 2-seam co-mining well. The difference in inter-seam pressure interaction is huge. The historical coalbed methane well cannot reflect the inter-seam interaction mechanism of the target coalbed methane well. For example, in 3-seam co-mining, there is interference from water production in the upper layer suppressing gas production in the middle layer, while there is no such phenomenon in 2-seam co-mining. Directly using the historical coalbed methane well of 2-seam co-mining will lead to the model misjudging the production capacity of the middle layer.

[0025] Based on this, the process of screening historical coalbed methane wells with similar reservoir static data and their historical combined production data and historical production capacity is as follows: First, the reservoir static data of each coal seam in the target coalbed methane well is standardized to form a reservoir static data vector for each coal seam.

[0026] Secondly, retrieve the same number of historical coalbed methane wells as the coal seams from the coal seam mining database, and obtain the historical reservoir static data vectors of each coal seam in each historical coalbed methane well.

[0027] Next, by combining the reservoir static data vectors of each coal seam in the target coalbed methane well, the comprehensive similarity between the target coalbed methane well and each historical coalbed methane well is analyzed through similarity analysis. Historical coalbed methane wells with a comprehensive similarity greater than a preset similarity threshold are regarded as similar historical coalbed methane wells.

[0028] Finally, historical combined mining data and historical production capacity of similar historical coalbed methane wells were screened from the coal seam mining database.

[0029] In one specific example, the preset similarity threshold in this invention can be set to 0.7. The setting process is as follows: different similarity thresholds are selected to verify historical data. When the threshold is 0.7, the preliminary production capacity splitting model trained by the selected similar historical coalbed methane wells has the highest determination coefficient on the test set. If the threshold is lower than 0.7, historical coalbed methane wells with poor similarity will be introduced, and if it is higher than 0.8, the amount of data selected will be insufficient. Implementers can also adjust the preset similarity threshold themselves.

[0030] This invention employs Z-score normalization to process the static reservoir data of each coal seam in the target coalbed methane well. If the distribution of the static reservoir data is non-normal, min-max normalization can be used to eliminate dimensional interference between different static reservoir data, providing a calculable basic data format for subsequent similarity analysis.

[0031] It should be noted that the method for analyzing the comprehensive similarity between the target coalbed methane well and each historical coalbed methane well is as follows: First, each coal seam in the target coalbed methane well is paired with each coal seam in each historical coalbed methane well to obtain each coal seam group. Then, the similarity of the reservoir static data vector between the target coalbed methane well and each coal seam group in each historical coalbed methane well is calculated using cosine similarity.

[0032] The second step is to select the coal seam group with the highest similarity of the static data vector of the reservoir, and to record the coal seam in the target coalbed methane well corresponding to the coal seam group as the first coal seam, and to record the coal seam in each historical coalbed methane well corresponding to the first coal seam as the associated coal seam of each historical coalbed methane well.

[0033] The third step is to remove coal seam groups that include the first coal seam and its corresponding associated coal seams, and then filter the reservoir static data vector similarity of all remaining coal seam groups by the maximum value until the corresponding associated coal seams of each coal seam in the target coalbed methane well are determined.

[0034] The fourth step is to calculate the similarity of the static data vectors of the reservoirs of each coal seam in the target coalbed methane well with the corresponding associated coal seams in each historical coalbed methane well. The average value of the similarity of the static data vectors of the reservoirs of each coal seam with the corresponding associated coal seams is taken as the comprehensive similarity between the target coalbed methane well and each historical coalbed methane well.

[0035] This invention uses the static reservoir data vectors of the target coalbed methane well and each coal seam in each historical coalbed methane well to analyze the comprehensive similarity between the target coalbed methane well and each historical coalbed methane well. This quantifies the overall matching degree between the target coalbed methane well and the historical coalbed methane well, reduces noise interference caused by dissimilar data, provides a reliable training basis for the preliminary model of production capacity splitting, and directly improves the initial prediction accuracy.

[0036] In one specific embodiment, such as Figure 2 As shown, the preliminary model for the production capacity division of the target coalbed methane well is established as follows: S11. Based on the static reservoir data of each coal seam in the target coalbed methane well, the corresponding associated coal seams of each coal seam in the target coalbed methane well in each similar historical coalbed methane well are selected.

[0037] S12. Obtain the historical combined mining data and historical production capacity of each coal seam in the corresponding associated coal seams of each similar historical coal seam gas well, and extract the time series of historical combined mining data and historical production capacity of the associated coal seams in each drainage stage.

[0038] In one specific example, each drainage stage in this invention includes the initial stage of water production, the initial stage of gas production, the stable production stage, and the decline stage.

[0039] S13. Match the historical commutation data time series and the historical production capacity time series in terms of time dimension to form a dataset of commutation data and production capacity corresponding to each coal seam in each mining stage.

[0040] S14. Divide the dataset into a training set and a test set according to a set ratio. Use the combined data in the training set as the input feature and the production capacity as the output feature to train the model and obtain the initial preliminary model of production capacity splitting.

[0041] It should be noted that the preferred ratio in this invention is 7:3, but the implementer may also adjust the ratio.

[0042] S15. Predict the production capacity of the test set using the initial capacity splitting preliminary model, optimize the initial capacity splitting preliminary model using the accuracy, and output the optimized capacity splitting preliminary model.

[0043] In a specific example, the present invention preferably uses a gradient boosting decision tree or random forest model for training. The model is used to fit the mapping between the joint procurement data and the production capacity. Both the gradient boosting decision tree and random forest models are existing technologies and will not be described in detail here.

[0044] The accuracy rate represents the percentage of samples where the deviation between the model's predicted production capacity and the actual production capacity is within the allowable deviation range. When the accuracy rate is less than the set accuracy rate, the number of decision trees in the decision tree or random forest model is increased by optimizing the gradient, and the accuracy rate of the optimized model is re-evaluated. When the accuracy rate is greater than the set accuracy rate, the optimized preliminary model for capacity splitting is output.

[0045] S16. Statistically analyze the preliminary model of production capacity division for each coal seam at each drainage stage to form a preliminary model of production capacity division for the target coalbed methane well.

[0046] This invention filters historical synergistic production data and historical production capacity of similar historical coalbed methane wells that are similar to reservoir static data, and establishes a preliminary model for capacity partitioning of target coalbed methane wells. This improves the adaptability of the constructed model and the accuracy of initial prediction, provides effective data for the accuracy of subsequent mining strategy formulation, and improves the rationality of mining resource allocation.

[0047] S2. Collect real-time combined mining data of each coal seam in the target coalbed methane well and generate a real-time combined mining data sequence of each coal seam in each unit time period.

[0048] Specifically, the real-time combined mining data sequence of each coal seam in each unit time period is generated as follows: First, the real-time combined mining data of each coal seam is collected by the monitoring device deployed in each coal seam in the target coalbed gas well. The combined mining data includes fluid flow pressure, fluid flow velocity, fluid density and temperature.

[0049] Secondly, the real-time combined data is timestamped to obtain the real-time combined data for each time period within the current set duration.

[0050] Finally, outlier removal and interpolation are performed on the real-time combined mining data for each unit time period to obtain the processed real-time combined mining data, generating a real-time combined mining data sequence for each coal seam in each unit time period.

[0051] It should be noted that in this invention, the unit time period can be set to 1 day. The outlier removal and interpolation processing are existing technologies and will not be elaborated upon here. This invention effectively prevents erroneous data caused by sensor malfunctions or environmental interference from entering the model by removing outliers; simultaneously, through interpolation processing, it compensates for a small amount of missing data, ensuring the integrity of the data sequence.

[0052] In one specific embodiment, the monitoring device deployed in each coal seam includes a strain gauge pressure sensor, a flow velocity sensor, a density sensor, and a temperature sensor, which respectively collect the fluid flow pressure, fluid flow velocity, fluid density, and temperature of each coal seam.

[0053] Fluid flow pressure determines the production pressure difference between the coal seam and the wellbore, and is the core driving force for fluid to flow from the reservoir to the wellbore. The higher the fluid flow pressure, the greater the production pressure difference and the stronger the gas production potential.

[0054] Fluid velocity is the amount of fluid produced per unit time, which is positively correlated with production capacity. The higher the velocity, the more fluid is produced per unit time, distinguishing between gas-dominated and water-dominated production states.

[0055] Fluid density can identify the type of fluid produced by a single coal seam. The density increases sharply during water intrusion, providing a basis for judging factors that inhibit production capacity.

[0056] Temperature affects the adsorption of coalbed methane. Higher temperatures reduce the adsorption capacity of coalbed for methane, promoting gas production, while lower temperatures inhibit desorption.

[0057] S3. Input the real-time coalbed methane data sequence into the established preliminary capacity splitting model, output the predicted production capacity of the target coalbed methane well in each unit time period, and calculate the deviation between the predicted production capacity and the actual production capacity.

[0058] Specifically, the predicted production capacity of the target coalbed methane well in each unit time period is as follows: based on the current drainage stage corresponding to the target coalbed methane well, the preliminary production capacity splitting model of each coal seam in the current drainage stage is selected from the preliminary production capacity splitting model of the target coalbed methane well.

[0059] Input the real-time combined mining data sequence of each coal seam in each unit time period into the corresponding preliminary capacity splitting model, and output the predicted production capacity of each coal seam in each unit time period.

[0060] The predicted production capacity of each coal seam in each unit time period is accumulated to obtain the predicted production capacity of the target coalbed methane well in each unit time period.

[0061] It should be noted that the calculation of the deviation between the predicted production capacity and the actual production capacity specifically involves: obtaining the actual production capacity of the target coalbed methane well in each unit time period, comparing the absolute difference between the actual production capacity in each unit time period and the predicted production capacity in the corresponding unit time period, and obtaining the deviation between the predicted production capacity and the actual production capacity.

[0062] This invention effectively eliminates the impact of production capacity differences by calculating the deviation between predicted and actual production capacity, reflecting the degree of fit between the constructed preliminary capacity splitting model and the actual production capacity, and making the predicted production capacity fit the current production status.

[0063] S4. Analyze whether the preliminary capacity splitting model needs to be optimized based on the deviation value. If optimization is needed, adjust the preliminary capacity splitting model to obtain the optimized capacity splitting model.

[0064] Specifically, the method for analyzing whether the preliminary model for capacity splitting needs optimization is as follows: Step 1: Based on the current drainage stage corresponding to the target coalbed methane well, extract the historical predicted production capacity and historical actual production capacity of each similar historical coalbed methane well in the corresponding drainage stage from the coal seam mining database, and compare them to obtain the corresponding historical deviation value.

[0065] Step 2: Remove outliers from the historical deviation values ​​of each similar historical coalbed methane well in the corresponding drainage stage for each unit time period, form a set of historical deviation values ​​from the remaining historical deviation values, and select the largest historical deviation value as the set deviation value threshold.

[0066] Step 3: If the deviation value of a certain unit time period is greater than the set deviation value threshold, the preliminary capacity splitting model needs to be optimized; otherwise, the preliminary capacity splitting model does not need to be optimized.

[0067] It should be noted that the outlier removal mentioned above uses the 3σ principle of normal distribution for outlier removal. The 3σ principle is an existing technology and will not be elaborated further.

[0068] The set deviation threshold represents the maximum reasonable deviation that the model may have during normal operation when similar historical coalbed methane wells are in the same drainage stage. If the deviation value of the current target well does not exceed the threshold, it means that the model prediction accuracy is within a reasonable range; if it exceeds the threshold, it means that the current model accuracy is lower than the historical best level and optimization needs to be initiated.

[0069] In one specific embodiment of the present invention, such as Figure 3 As shown, the capacity splitting optimization model is obtained as follows: First, each unit time period with a deviation value less than a set deviation value threshold is selected, and the real-time combined mining data sequence and predicted production capacity of each coal seam in each selected unit time period are statistically analyzed and supplemented into the dataset corresponding to each coal seam in the current mining stage.

[0070] Secondly, incremental training is conducted on the preliminary capacity splitting model for each coal seam in the current mining stage to update the preliminary capacity splitting model.

[0071] Next, based on the updated preliminary capacity splitting model, the predicted production capacity of the target coalbed methane well in each unit time period is re-predicted, and the updated preliminary capacity splitting model is analyzed to determine whether it needs to be optimized.

[0072] Finally, if no optimization is needed, the updated preliminary capacity allocation model will be used as the optimized capacity allocation model. If optimization is needed, the updated preliminary capacity allocation model will be iteratively adjusted until the iteratively adjusted model no longer requires optimization.

[0073] It should be noted that selecting unit time periods with a deviation value less than the set deviation value threshold indicates that the predicted production capacity of that unit time period is accurate and effective, and can reflect the latest production patterns in the current scheduling stage. Supplementing these units to the original dataset can enrich the training samples of the model, enabling the model to learn the latest dynamic changes. Conversely, if data with a deviation value greater than the threshold is added, it will cause the model to learn incorrect mapping relationships, further reducing the prediction accuracy.

[0074] This invention outputs the predicted production capacity of the target coalbed methane well in each unit time period by using the real-time combined mining data sequence of each coal seam in each unit time period. Based on the deviation between the predicted production capacity and the actual production capacity, the preliminary production capacity splitting model is adjusted to obtain the optimized production capacity splitting model. This enables incremental training and iterative adjustment of the model, ensuring that the model can match the dynamic changes in the drainage stage in real time and effectively improve the prediction accuracy of the model.

[0075] S5. Based on the real-time combined mining data sequence of each coal seam in each unit time period, and combined with the capacity splitting optimization model, predict the predicted production capacity of each coal seam for a set time period in the future, identify the main coal seam of the target coalbed methane well for a set time period in the future and display it.

[0076] Given the strong correlation between the production capacity of multi-coal-seam mining and the mining data, the changing trend of the mining data directly determines the direction of future production capacity changes, and the rate of change reflects the speed of trend evolution. If current data is used directly to replace future data without trend analysis, the dynamic changes in the reservoir will be ignored, leading to distorted future production capacity predictions.

[0077] Based on this, such as Figure 4 As shown, the main coal seam identification method is as follows: S51, based on the real-time combined mining data sequence of each coal seam in each unit time period, analyze the changing trend and rate of change of each coal seam corresponding to each combined mining data.

[0078] One specific example is to select a suitable trend model based on the distribution characteristics of real-time combined data for each unit time period. For example, if the combined data changes approximately linearly over time, linear regression is used for fitting. If the slope of the fluid flow pressure fitting equation is negative, it is determined to be a linear downward trend, and the slope is used as its rate of change.

[0079] If the combined production data shows exponential growth or decline, such as an exponential increase in fluid velocity with the increase in desorption during the initial stage of gas production, an exponential smoothing method is used for fitting. When the fluid velocity fitting equation is a growth coefficient, it is determined to be an exponential upward trend, and the rate of change is calculated using a month-on-month comparison. The calculation formula is as follows: .

[0080] In the formula, For the rate of change, This represents the combined data for the t-th time period. This is the combined collection data for the (t-1)th time period. , This represents the total number of items per time period.

[0081] If the combined procurement data remains unchanged, it is determined to be a stable trend, and its rate of change is 0.

[0082] S52. Based on the changing trends and rates of change of each coal seam mining data, determine the coal seam mining data for each unit time period within the future set time period, thus forming the coal seam mining data sequence for each unit time period within the future set time period.

[0083] It should be noted that in this invention, the future set duration is set according to the remaining duration of the current gas production stage. For example, if the current gas production stage is the initial stage of gas production and the total duration of the initial stage of gas production is 30 days, and the current time is the 10th day of the initial stage of gas production, then the future set duration is 20 days.

[0084] S53. Substitute the combined mining data sequence of each unit time period within the future set time period of each coal seam into the corresponding capacity splitting optimization model, output the predicted production capacity of each unit time period, and determine the predicted production capacity of each coal seam within the future set time period.

[0085] It should be noted that the predicted production capacity of each coal seam for the specified future time period is the sum of the predicted production capacity of each unit time period within the specified future time period.

[0086] S54. Based on the predicted production capacity of each coal seam for a set future time, analyze the production capacity contribution rate of each coal seam, and identify the main coal seam of the target coalbed methane well for a set future time based on the production capacity contribution rate.

[0087] In one specific embodiment, the capacity contribution rate analysis method of each coal seam is as follows: the predicted production capacity of each coal seam for a set future time is accumulated to obtain the total predicted production capacity for a set future time, and the proportion of the predicted production capacity of each coal seam for a set future time to the total predicted production capacity is calculated, and this proportion is used as the capacity contribution rate of each coal seam.

[0088] This invention predicts the production capacity of each coal seam over a set time period based on a capacity splitting optimization model, identifies the main coal seam of the target coalbed methane well over the set time period, provides a basis for resource allocation and parameter adjustment during the mining stage, helps to maximize the production potential of the main coal seam, improve the overall mining efficiency of multi-coal seam co-mining, and reduce ineffective mining costs.

[0089] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0090] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0091] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0092] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0093] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for constructing a production capacity partitioning model for multi-coal-seam commingled coalbed methane wells, characterized in that, include: Obtain the reservoir static data of each coal seam in the target coalbed methane well, screen the historical synergistic production data and historical production capacity of similar historical coalbed methane wells with similar reservoir static data, and establish a preliminary model for capacity splitting of the target coalbed methane well. Collect real-time combined production data of each coal seam in the target coalbed methane well, and generate a real-time combined production data sequence of each coal seam in each unit time period; The combined data includes fluid flow pressure, fluid flow velocity, fluid density, and temperature; The real-time coalbed methane data sequence is input into the established preliminary capacity splitting model, and the predicted production capacity of the target coalbed methane well in each unit time period is output. The deviation value between the predicted production capacity and the actual production capacity is calculated. Specifically, the deviation value between the predicted production capacity and the actual production capacity is calculated by: obtaining the actual production capacity of the target coalbed methane well in each unit time period, comparing the absolute difference between the actual production capacity in each unit time period and the predicted production capacity in the corresponding unit time period, and obtaining the deviation value between the predicted production capacity and the actual production capacity. Based on the deviation value analysis, it is determined whether the initial capacity splitting model needs to be optimized. If optimization is required, the initial capacity splitting model is adjusted to obtain the optimized capacity splitting model. Analyzing whether the preliminary production capacity splitting model needs optimization includes setting a deviation threshold based on the historical deviation values ​​of each similar historical coalbed methane well in each unit time period during the corresponding drainage stage; If the deviation value of a certain unit time period is greater than the set deviation value threshold, the preliminary capacity allocation model needs to be optimized; otherwise, the preliminary capacity allocation model does not need to be optimized. When optimizing the initial capacity splitting model, real-time combined mining data sequences and predicted production capacity of each coal seam within each unit time period with deviation values ​​less than the set deviation value threshold are selected and added to the dataset corresponding to each coal seam in the current mining stage. Incremental training is performed on the initial capacity splitting model of each coal seam in the current mining stage to update the initial capacity splitting model. Based on the real-time combined mining data sequence of each coal seam in each unit time period, and combined with the capacity splitting optimization model, the predicted production capacity of each coal seam for a set time period is predicted, and the main coal seam of the target coalbed methane well for the set time period is identified and displayed.

2. The method for constructing a production capacity partitioning model for multi-coal-seam commingled coalbed methane wells according to claim 1, characterized in that: The process of filtering historical synergistic production data and historical production capacity of similar historical coalbed methane wells that are similar to reservoir static data is as follows: The reservoir static data of each coal seam in the target coalbed methane well are standardized to form a reservoir static data vector for each coal seam. Retrieve historical coalbed methane wells with the same number of coal seams from the coal seam mining database, and obtain the historical reservoir static data vector of each coal seam in each historical coalbed methane well. By combining the reservoir static data vectors of each coal seam in the target coalbed methane well, the comprehensive similarity between the target coalbed methane well and each historical coalbed methane well is analyzed through similarity analysis. Historical coalbed methane wells with a comprehensive similarity greater than a preset similarity threshold are regarded as similar historical coalbed methane wells. Historical combined mining data and historical production capacity of similar historical coalbed methane wells were screened from the coal seam mining database.

3. The method for constructing a production capacity partitioning model for multi-coal-seam commingled coalbed methane wells according to claim 1, characterized in that: The preliminary model for dividing the production capacity of the target coalbed methane well is established as follows: Based on the reservoir static data of each coal seam in the target coalbed methane well, the corresponding associated coal seams of each coal seam in the target coalbed methane well in each similar historical coalbed methane well are screened. Obtain historical synergistic mining data and historical production capacity of each coal seam in the corresponding associated coal seams of each similar historical coal seam gas well, and extract the time series of historical synergistic mining data and historical production capacity of the associated coal seams in each drainage stage. By matching the time series of historical joint mining data with the time series of historical production capacity, a dataset of joint mining data and production capacity corresponding to each coal seam at each mining stage is formed. The dataset is divided into a training set and a test set according to a set ratio. The combined data in the training set is used as the input feature and the production capacity is used as the output feature to train the model and obtain the initial preliminary model of capacity splitting. The production capacity of the test set is predicted by the initial capacity splitting preliminary model. The initial capacity splitting preliminary model is optimized by the accuracy and the optimized capacity splitting preliminary model is output. A preliminary model of capacity splitting for each coal seam at each drainage stage was developed to form a preliminary model of capacity splitting for the target coalbed methane well.

4. The method for constructing a production capacity partitioning model for multi-coal-seam commingled coalbed methane wells according to claim 1, characterized in that: The method for generating the real-time combined mining data sequence of each coal seam in each unit time period is as follows: Real-time combined mining data of each coal seam is collected by the monitoring devices deployed in each coal seam of the target coalbed methane well. Timestamp alignment is performed on the real-time aggregated data to obtain the real-time aggregated data for each time period within the current set duration; Outlier removal and interpolation are performed on the real-time combined mining data for each unit time period to obtain the processed real-time combined mining data, and a real-time combined mining data sequence for each coal seam in each unit time period is generated.

5. The method for constructing a production capacity partitioning model for multi-coal-seam commingled coalbed methane wells according to claim 3, characterized in that: The predicted production capacity of the target coalbed methane well in each unit time period is as follows: Based on the current production stage of the target coalbed methane well, the preliminary production capacity splitting model for each coal seam at the current production stage is selected from the preliminary production capacity splitting model of the target coalbed methane well. Input the real-time combined mining data sequence of each coal seam in each unit time period into the corresponding preliminary capacity splitting model, and output the predicted production capacity of each coal seam in each unit time period; The predicted production capacity of each coal seam in each unit time period is accumulated to obtain the predicted production capacity of the target coalbed methane well in each unit time period.

6. The method for constructing a production capacity partitioning model for multi-coal-seam commingled coalbed methane wells according to claim 3, characterized in that: The methods for analyzing whether the preliminary capacity allocation model needs optimization also include: Based on the current drainage stage of the target coalbed methane well, extract the historical predicted production capacity and historical actual production capacity of each similar historical coalbed methane well in the corresponding drainage stage from the coal seam mining database, and compare them to obtain the corresponding historical deviation value. Outliers are removed from the historical deviation values ​​of each unit time period in the corresponding drainage stage of each similar historical coalbed methane well. The remaining historical deviation values ​​are then used to form a set of historical deviation values. The largest historical deviation value is selected and used as the set deviation value threshold.

7. The method for constructing a production capacity partitioning model for multi-coal-seam commingled coalbed methane wells according to claim 6, characterized in that: The methods for obtaining the capacity splitting optimization model also include: Based on the updated preliminary capacity splitting model, the predicted production capacity of the target coalbed methane well in each unit time period is re-predicted. Similarly, it is analyzed whether the updated preliminary capacity splitting model needs to be optimized. If no optimization is needed, the updated preliminary capacity allocation model will be used as the optimized capacity allocation model. If optimization is needed, the updated preliminary capacity allocation model will be iteratively adjusted until the iteratively adjusted model no longer requires optimization.

8. The method for constructing a production capacity partitioning model for multi-coal-seam commingled coalbed methane wells according to claim 1, characterized in that: The method for identifying the main coal seam is as follows: Based on the real-time combined mining data sequence of each coal seam in each unit time period, analyze the changing trend and rate of change of each coal seam corresponding to each combined mining data; Based on the changing trends and rates of change of each coal seam mining data, the mining data for each unit time period within the future set time period are determined, thus forming the mining data sequence for each unit time period within the future set time period for each coal seam. Substitute the combined mining data sequence of each unit time period within the future set time period of each coal seam into the corresponding capacity splitting optimization model, output the predicted production capacity of each unit time period, and determine the predicted production capacity of each coal seam within the future set time period. Based on the projected production capacity of each coal seam over a set future time period, the capacity contribution rate of each coal seam is analyzed, and the main coal seam of the target coalbed methane well over the set future time period is identified based on the capacity contribution rate.

9. The method for constructing a production capacity partitioning model for multi-coal-seam commingled coalbed methane wells according to claim 8, characterized in that: The method for analyzing the production capacity contribution rate of each coal seam is as follows: The predicted production capacity of each coal seam over a set future time period is summed to obtain the total predicted production capacity over the set future time period. The proportion of the predicted production capacity of each coal seam over the set future time period to the total predicted production capacity is calculated, and this proportion is used as the production capacity contribution rate of each coal seam.

Images