Construction method of near-bit gamma machine learning sample

Abstract

Natural gamma logging is a method for measuring the intensity of natural gamma rays of a rock stratum along a well body, and the near-bit gamma logging technology and the cable logging technology are essentially the same. Reservoir section natural gamma logging reveals gamma values of a reservoir at different positions and the change trend of the gamma values. The simulation drill bit drills through the reservoir at different included angles, and gamma values of the drill bit at different positions of the reservoir, the measurement depth, the drill bit-stratum included angle and the vertical distance from the drill bit to the top of the reservoir are stored. And according to the length of the machine learning sample and the sampling interval, continuously sampling logging values to obtain a machine learning training sample.

Description

1. Technical Field

[0001] Horizontal well drilling technology is increasingly used in energy development. In order to develop resources quickly and efficiently, the drill bit is required to drill through as many reservoirs as possible. The use of artificial intelligence technology can help geologists determine the relationship between the drill bit and the reservoir, thereby improving the drilling rate. 2. Background Technology

[0002] Natural gamma cable logging reveals the size and variation of the gamma value in the reservoir section. The near-bit natural gamma logging instrument determines the relationship between the drill bit and the reservoir by measuring the values ​​of the upper and lower gamma curves at the tail of the drill bit and the changes in the curve shape caused by the changes in the values. III. Summary of the invention

[0003] The simulated drill bit drills through the reservoir at a fixed angle, and records the drill bit's travel distance in the reservoir, gamma value, height from the top plate, and angle between the drill bit and the reservoir. Taking into account the adjustment of the drill bit posture in the horizontal section of the reservoir, the simulation requirements must include drilling from the top of the reservoir to the bottom of the reservoir and from the bottom of the reservoir to the top of the reservoir, so as to ensure the adequacy of the training samples.

[0004] The near-bit natural gamma logging instrument records the gamma values ​​of a 360-degree scan along the drill bit. The while-drilling analysis only requires the values ​​of the upper and lower quadrants. The gamma values ​​combined with the continuously recorded curve shape can be used to analyze the position of the drill bit in the reservoir and the angle between the drill bit and the reservoir.

[0005] Simulate the drill bit to drill through the reservoir at different angles to build a machine learning library. The following parameters are required: The sample measurement length is l. The larger l is, the higher the anti-interference ability is and the lower the sensitivity is. The angle i between the drill bit and the reservoir is positive when the drill bit goes down through the reservoir and negative when it goes up through the reservoir; Reservoir thickness H, reservoir thickness refers to the true thickness of the reservoir. True thickness correction must be done after drilling cable logging. Cable logging sampling interval s; The drill bit diameter D should maintain consistency in each length unit. The natural gamma log GRi is obtained by simulating drilling through the reservoir section at an angle i. The length of the log curve is Li = H / sin(i), that is, when the drill bit drills vertically through the reservoir, the length of the log curve is equal to the reservoir thickness. When the drill bit-reservoir angle is 0, the log curve is infinitely long. Therefore, the number of sampling points for drilling through the reservoir at an angle i is H / sin(i) / s; When a drill bit with a diameter of D drills through a reservoir at an angle i, the reservoir thickness d caused by the drill bit diameter is d = D*cos(i), that is, d = 0 when the drill bit is perpendicular to the reservoir, and d = D when the drill bit is parallel to the reservoir. For natural gamma sampling, the difference in the number of sample points brought by the upper and lower gamma sampling positions brought by the drill bit is Sdi = D*cot(i) / s. The sample length is l, and the number of sampling points of each sample is l / s; When drilling through the formation at angle i to establish a sample, the upper gamma is GRi[up], and the lower gamma GRi[down] = GRi[up+Sdi]. The lower gamma is the difference in the number of sample points caused by the upper gamma extending the drill bit diameter; IV. Description of the drawings

[0006] Figure 1 It is the gamma curve characteristic of the reservoir section;

[0007] Figure 2 The drill bit drills through the formation at different angles;

[0008] Figure 3 It is to simulate the process of establishing near-bit GR logging samples;

[0009] Figure 4 It is the result of calculation using actual drilling data;

[0010] Figure 5 It is the way machine learning databases are organized;

[0011] (1) Typical gamma curve of the reservoir, (2) Reservoir drilling trajectory at a certain angle, (3) Training sample interception mode, (4) Upper gamma trajectory, (5) Lower gamma trajectory, (6) Upper gamma curve, (7) Lower gamma curve, (8) Upper edge of the reservoir, (9) Drilling trajectory, (10) Lower edge of the reservoir, (11) Drill bit-reservoir angle, (12) Actual drilling near-drill bit gamma curve; V. Specific implementation methods

[0012] Obtain the vertical well cable logging gamma log curve GR and interpret the reservoir section thickness H;

[0013] Calculate the well logging curve GRi when drilling through the reservoir at angle i, the length of the natural gamma logging curve Li, and the number of sample points Li / s;

[0014] Calculate the difference Sdi of the upper and lower gamma sampling points caused by the drill bit size when drilling through the reservoir at angle i;

[0015] According to the natural gamma logging curve GRi calculated by

[0013] , combined with Sdi obtained by

[0014] , the upper / lower gamma curve of the near-bit instrument is calculated;

[0016] Using the GRi[upper] and GRi[lower] logging curves combined with the sample length l, the training samples when drilling through the reservoir at angle i are calculated;

[0017] according to Figure 5 The X features of the training samples are organized in the format, the distance between the drill bit and the reservoir top and the angle between the drill bit and the reservoir are used as the training features y, and the relevant machine learning algorithm training model model is introduced;

[0018] According to the sample length l, the upper and lower natural gamma logging curves of the drilling well are intercepted and organized into real drilling samples for calculating the drill bit posture. The real drilling sample data is sent to the model trained by

[0017] to calculate the position of the drill bit in the reservoir and the angle between the drill bit and the reservoir.

Claims

1. A method for constructing near-bit gamma machine learning samples, characterized in that: Step 1: Collect the natural gamma logging curves of the completed drilling cable logging reservoir and interpret the reservoir thickness H; Step 2, calculate the gamma logging curve GRi when drilling through the reservoir at an angle i, the length of the natural gamma logging curve Li, and the number of sample points Li / s; Step 3, calculate the difference Sdi of the upper and lower gamma sampling points caused by the drill bit size when drilling through the reservoir at an angle i; Step 4: Calculate the upper / lower gamma curve of the near-bit instrument based on the natural gamma logging curve GRi calculated in step 2 and Sdi obtained by combining [0014]; Step 5: Calculate the training sample when the angle i penetrates the reservoir by using the GRi[upper] and GRi[lower] logging curves combined with the sample length 1; Step 6: Organize the X features of the training samples according to Table 1, take the distance between the drill bit and the reservoir top and the angle between the drill bit and the reservoir as the training features y, and introduce the relevant machine learning algorithm training model mode1; Step 7: According to the sample length 1, the upper and lower natural gamma logging curves of the drilling well are intercepted and organized into samples to be calculated. The actual drilling sample data is sent to the training model mode1 to calculate the position of the drill bit in the reservoir and the angle between the drill bit and the reservoir.

Images