Productivity splitting method based on machine learning and petroleum fingerprint screening
Through machine learning and petroleum fingerprint screening methods, the problems of high costs and long cycles in the existing technology are solved, and the rapid and accurate capacity splitting of combined production wells is achieved, which improves production efficiency and market competitiveness.
Patent Information
- Application Number
- CN202510489705.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-18
- Publication Date
- 2025-08-05
AI Technical Summary
The existing technology relies on separate terminal well modeling for combined production well capacity splitting, resulting in high costs and long cycles, making it difficult to meet the demand for rapid capacity splitting.
Using machine learning and petroleum fingerprint screening methods, the chromatographic fingerprints of the end element wells and co-production wells of each oil group in the reservoir were obtained, and the compound peak calibration, alignment and feature extraction were performed. The contribution ratio of each oil group was calculated by combining the least squares method to achieve fast and low-cost capacity splitting.
It improves the accuracy and efficiency of production capacity allocation of co-production wells, reduces production costs, and meets the rapid and economical capacity splitting needs.
Smart Images
Figure CN120432046A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of oil and gas exploration and development, and in particular to a production capacity splitting method based on machine learning and oil fingerprint screening. Background Art
[0002] In multi-layer commingled production reservoirs, accurate oil well production splitting is the basis for dynamic analysis of each layer, development effect evaluation and residual oil research.
[0003] The significance of commingled oil production capacity splitting is mainly reflected in the following aspects:
[0004] (1) Optimize resource utilization: Through reasonable splitting, production equipment and human resources can be fully utilized, production efficiency can be improved, and resource waste can be reduced;
[0005] (2) Meeting different market demands: By splitting production capacity, we can meet the requirements of different markets for different product specifications and quality, thereby improving market competitiveness;
[0006] (3) Reduce production costs: Allocating funds based on the production costs and market pricing of different products will help reduce overall production costs and increase corporate profit margins.
[0007] The study found that existing technologies rely on modeling separate end-member wells and then performing data fitting to achieve capacity splitting of combined production wells, which is costly and time-consuming; it is difficult to meet the demand for rapid capacity splitting. Summary of the Invention
[0008] The present invention provides a productivity splitting method based on machine learning and oil fingerprint screening, which can achieve simple, efficient, rapid and low-cost productivity splitting of commingled production wells.
[0009] To achieve the above object, the present invention adopts the following technical solutions:
[0010] In a first aspect, the present application provides a production capacity splitting method based on machine learning and oil fingerprint screening, comprising:
[0011] S1, obtaining the end-member oil chromatographic fingerprint of the end-member wells of each oil group in the oil reservoir and the mixed oil chromatographic fingerprint of the combined production wells, and calibrating the compound peaks therein;
[0012] S2, aligning the end-member oil chromatographic fingerprint and the mixed oil chromatographic fingerprint based on the calibrated common compound peaks;
[0013] S3, using machine learning methods to extract chromatographic fingerprint features from the chromatographic fingerprints of each end-member oil;
[0014] S4, based on the chromatographic fingerprint features of each end-member oil chromatographic fingerprint map, combined with the mixed oil chromatographic fingerprint map of each commingled production well, calculate the production capacity splitting of each oil group of the commingled production well.
[0015] In one implementation, the step S1 includes:
[0016] The automatic integration method is used to integrate the compound peak areas of the total hydrocarbon chromatograms of each end-member oil sample and each commingled oil sample in the reservoir to obtain the calibrated compound peaks.
[0017] In one implementation, in S2, the dynamic time warping method is used to align the common compound peaks in each end-member oil chromatographic fingerprint and the mixed oil chromatographic fingerprint to obtain time series data with the same peak time.
[0018] In one implementation, the S3 includes:
[0019] For multiple sets of time series data with the same peak time, the peak area of the compound is normalized;
[0020] The EDA analysis method is used to analyze the chromatographic fingerprint characteristics of each end-member oil well in each oil group in the normalized data to determine the representative end-member well of each oil group;
[0021] Based on the characteristic peak ratios with the largest differences in the end-member oils of representative end-member wells in each oil group, a characteristic peak ratio set is established.
[0022] In one implementation, the step S4 includes: calculating the contribution ratio of each end-member oil in each commingled oil using the least squares method based on a preset productivity splitting mathematical model.
[0023] In one implementation, in the productivity splitting mathematical model, there are n end-member oil layers in combined production, and there are p elements in the characteristic peak ratio set. Then, the mixed crude oil chromatographic fingerprint peak ratio satisfies:
[0024]
[0025] or in matrix form
[0026] Y=AX
[0027] Among them, a m is the contribution ratio of the mth end member oil layer in the mixed crude oil, X mi is the value of the i-th characteristic peak ratio in the m-th end member oil layer;
[0028] The characteristic peak ratios of more than the number of end-member oils are selected and brought into the model for calculation, that is, p>n; the characteristic peak ratios in the end-member oil and the mixed oil are known, that is, X and Y are known; a least squares model is established, and the optimal solution of the matrix equation is fitted to obtain the optimal production capacity ratio A.
[0029] The technical solution of the present invention establishes a process for peak selection, integration, and selection of characteristic peak area ratios for calculating the production capacity of combined production wells using crude oil chromatographic fingerprinting through machine learning, which can improve the accuracy and efficiency of the calculation results of the production capacity allocation of combined production wells. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] Figure 1 A diagram of a method for extracting characteristic peaks, calculating characteristic peak area ratios, determining characteristic peak ratio sets, and performing commingled well productivity splitting using a machine learning method in a specific implementation method.
[0031] Figure 2 For example, the distribution map of the end wells and combined production wells in the oil field is studied
[0032] Figure 3 The similarity analysis results of each end member oil of II oil group by EDA method are shown in Figure 2.
[0033] Figure 4 This is a schematic diagram of the results of machine learning in a specific implementation method for selecting characteristic peaks and determining characteristic ratios in the co-produced oil P17 and the end-member oils of oil groups I and II. DETAILED DESCRIPTION
[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the described embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field are within the scope of protection of the present invention.
[0035] Aiming at the defects and problems of the existing technology, based on the attached Figure 1 This application provides a production capacity splitting method based on machine learning and oil fingerprint screening, including:
[0036] S1, obtaining the end-member oil chromatographic fingerprint of the end-member wells of each oil group in the oil reservoir and the mixed oil chromatographic fingerprint of the combined production wells, and calibrating the compound peaks therein;
[0037] S2, aligning the end-member oil chromatographic fingerprint and the mixed oil chromatographic fingerprint based on the calibrated common compound peaks;
[0038] S3, using machine learning methods to extract chromatographic fingerprint features from the chromatographic fingerprints of each end-member oil;
[0039] S4, based on the chromatographic fingerprint features of each end-member oil chromatographic fingerprint map, combined with the mixed oil chromatographic fingerprint map of each commingled production well, calculate the production capacity splitting of each oil group of the commingled production well.
[0040] The above method is described below in a more detailed embodiment with reference to more drawings.
[0041] In one embodiment, the implementation steps and advantages of this method are described in detail using Reservoir A as an example. The end-member wells of Reservoir A include horizontal wells P6, P28H, and P33H in the single-production II oil group, horizontal wells P1H, W1H, and P12H in the single-production III oil group, and horizontal wells P5H and P27H in the single-production I oil group. Wells P8 and P15 are wells for the production of oil in the combination of I, II, and III oil, and P17 is a well for the production of oil in the combination of I and II oil. Figure 2 .
[0042] The method provided in this embodiment specifically includes the following steps:
[0043] Step 1) Calibrate the crude oil chromatographic fingerprint
[0044] S1.1: Compound peak area integration was performed on the total hydrocarbon chromatograms of eight end-member oil samples and three commingled oil samples from Oilfield A using the automated integration method in Thermo Xcalibur 2.2.42 and Qual Browser (Thermo Fisher Scientific Inc.). The software automatically identified 3821 to 5413 peaks and calculated peak areas.
[0045] S1.2: The laboratory issued batch testing and compound identification reports for 11 end-member oils and commingled oils, detecting a total of 86 common peaks (Table 1). These common peaks were calibrated in the software's automated peak integration list (Table 2). Table 2 lists the partial calibration results for end-member oils P27H and P28H and commingled oil P17.
[0046] Table 1. Compound identification report issued by the laboratory (partial)
[0047]
[0048]
[0049] Table 2. Calibration results of end-use oils P27H, P28H and combined production oil P17 (partial)
[0050]
[0051]
[0052] Step 2) Align the chromatographic fingerprints of the end-member oil and the mixed oil with the labeled peak
[0053] Using the dynamic time warping (DTW) method, the chromatographic fingerprints of the 11 calibrated end-member oils and the combined oil were aligned using the calibration peaks. After alignment, the chromatographic fingerprints of all 11 samples had the same time sequence (Table 3).
[0054] Table 3. Aligned sample chromatogram (partial)
[0055]
[0056]
[0057]
[0058] Step 3) Use machine learning exploratory data analysis (EDA) to extract the chromatographic fingerprint characteristics of each end member oil
[0059] S3.1: Normalize the peak areas of the common peaks from the aligned 11 samples. Set the lower limit of the normalized peak area for machine-selected feature peaks to 0.1 to remove the influence of noise peaks and prepare for feature extraction.
[0060] S3.2: Use EDA analysis method to analyze the chromatographic fingerprint of each end-member oil sample belonging to oil groups I, II, and III, and analyze the similarity of different end-member oils in the same oil group (Appendix Figure 3 Of the three endmember oil samples in Group II, P28H, P6, and P33H, P28H had the largest similarity interval with P6 and P33H. Therefore, P28H was selected as the representative endmember oil for Group II. The same method was used to select representative endmember oil samples, P27H and P12H, for Groups I and III, respectively.
[0061] S3.3: Feature extraction of oil groups I, II, and III. Set rules and use machine language to randomly select adjacent or similar characteristic peaks in P27H, P28H, and P12H to calculate ratios. Create a set containing all characteristic peak ratios. From this set, select the ratio with the largest difference among the three end-member oils as the final characteristic peak ratio set for the capacity allocation of oil field A (see Appendix). Figure 4 ).
[0062] Step 4) Calculate the contribution ratio of each end member oil in each combined oil production
[0063] By using the selected characteristic peak ratio set (Table 4) through the least partial squares method, the contribution ratios of oil groups I, II, and III in the commingled oil P8, P15, and P17 were calculated to be 21:45:34, 20, 36, 44, and 96:4, respectively.
[0064] Table 4. Set of characteristic peak ratios of 11 end-member oils and commingled oils selected by machine learning
[0065]
[0066] In the several embodiments provided by the present invention, it should be understood that the disclosed methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the above units is merely a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interface, indirect coupling or communication connection of devices or units, which may be electrical, mechanical or other forms.
[0067] The above are only preferred embodiments of the present invention and are 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 in the scope of protection of the present invention.
Claims
1. A production capacity splitting method based on machine learning and oil fingerprint screening, characterized in that: include: S1, obtaining the end-member oil chromatographic fingerprint of the end-member wells of each oil group in the oil reservoir and the mixed oil chromatographic fingerprint of the combined production wells, and calibrating the compound peaks therein; S2, aligning the end-member oil chromatographic fingerprint and the mixed oil chromatographic fingerprint based on the calibrated common compound peaks; S3, using machine learning methods to extract chromatographic fingerprint features from the chromatographic fingerprints of each end-member oil; S4, based on the chromatographic fingerprint features of each end-member oil chromatographic fingerprint map, combined with the mixed oil chromatographic fingerprint map of each commingled production well, calculate the production capacity splitting of each oil group of the commingled production well.
2. The production capacity splitting method based on machine learning and oil fingerprint screening according to claim 1 is characterized in that: Said S1 includes: The automatic integration method is used to integrate the compound peak areas of the total hydrocarbon chromatograms of each end-member oil sample and each commingled oil sample in the reservoir to obtain the calibrated compound peaks.
3. The production capacity splitting method based on machine learning and oil fingerprint screening according to claim 2 is characterized in that: In the S2, the dynamic time warping method is used to align the common compound peaks in each end-member oil chromatographic fingerprint and the mixed oil chromatographic fingerprint to obtain time series data with the same peak time.
4. The production capacity splitting method based on machine learning and oil fingerprint screening according to claim 3 is characterized in that: The S3 includes: For multiple sets of time series data with the same peak time, the peak area of the compound is normalized; The EDA analysis method is used to analyze the chromatographic fingerprint characteristics of each end-member oil well in each oil group in the normalized data to determine the representative end-member well of each oil group; Based on the characteristic peak ratios with the largest differences in the end-member oils of representative end-member wells in each oil group, a characteristic peak ratio set is established.
5. The production capacity splitting method based on machine learning and oil fingerprint screening according to claim 4 is characterized in that: The step S4 includes: calculating the contribution ratio of each end-member oil in each commingled oil by using the least square method based on a preset productivity splitting mathematical model.
6. The production capacity splitting method based on machine learning and oil fingerprint screening according to claim 5 is characterized in that: In the productivity splitting mathematical model, there are n end-member oil layers produced together, and there are p elements in the characteristic peak ratio set. Then the chromatographic fingerprint peak ratio of the mixed crude oil satisfies: or in matrix form Y=AX Among them, a m is the contribution ratio of the mth end member oil layer in the mixed crude oil, X mi is the value of the i-th characteristic peak ratio in the m-th end member oil layer; The characteristic peak ratios of more than the number of end-member oils are selected and brought into the model for calculation, that is, p>n; the characteristic peak ratios in the end-member oil and the mixed oil are known, that is, X and Y are known; a least squares model is established, and the optimal solution of the matrix equation is fitted to obtain the optimal production capacity ratio A.