Horizontal well real-time earthquake steering method

Abstract

The invention discloses a real-time seismic steering method for a horizontal well, which comprises the following steps of: fitting an impedance curve of a forward drilling horizontal well in real time by utilizing a machine learning method, calibrating the forward drilling horizontal well and a newly drilled straight fixed well, adding a new horizon and then updating a space-variant velocity field model; and completing updating of the shaft impedance inversion model constrained by real-time drilling information one by one along the horizontal well trajectory, and carrying out high-resolution reservoir inversion to realize dynamic prediction of the reservoir and dynamic adjustment of the drilling trajectory. According to the horizontal well real-time seismic steering method, the problem that real-time while-drilling prediction seismic steering cannot be achieved through an existing traditional seismic steering technology is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of oil physical exploration and development methods, and particularly relates to a real-time seismic guidance method for horizontal wells. Background Art

[0002] For unconventional oil and gas resources, in recent years, a development model of controlling reservoirs with vertical wells and increasing production with horizontal wells has been explored. Large-scale implementation of horizontal wells can improve the degree of reserve utilization, increase the production of single wells, and contribute to the scale-efficient development of oil and gas fields. During the deployment and implementation of horizontal wells, in order to obtain information such as the development degree and structure of underground reservoirs, 3D seismic guidance is required. However, at the same time, there are also problems such as thin reservoirs and rapid structural changes, and a high degree of dependence on traditional manual single-well guidance. Therefore, improving the prediction accuracy of reservoirs and their structures, and improving the quality, accuracy, and efficiency of horizontal well guidance are the keys to the large-scale and effective development of unconventional resources such as shale oil in the future.

[0003] Traditional seismic guidance often uses the impedance and velocity of completed vertical wells for reservoir prediction and trajectory adjustment, which is a static prediction. The dynamic effective information of the actual drilling of horizontal wells cannot be involved, and it is difficult to make trajectory adjustment decisions when the lithology change encountered during drilling does not match the inversion result; while the VSP while drilling requires stopping drilling for acquisition, with high costs and low efficiency, and it is not a real-time prediction while drilling; the bit seismic technology while drilling requires improvement in both the bit drilling equipment and the processing technology. Summary of the Invention

[0004] The purpose of the present invention is to provide a real-time seismic guidance method for horizontal wells, which solves the problem that the existing traditional seismic guidance technology cannot achieve real-time seismic guidance with prediction while drilling.

[0005] The technical solution adopted by the present invention is as follows: a real-time seismic guidance method for horizontal wells, which uses a lithology-based machine learning method to fit the impedance curve of a horizontal well in real time, uses this horizontal well and a new well for calibration, and after adding a new horizon, updates the velocity field of the current horizontal well platform in real time. At the same time, after adding the information of the new well, updates the impedance model of the current platform reservoir inversion. On the basis of the velocity model, completes the update of the wellbore dynamic inversion model with real-time drilling information constraints channel by channel along the horizontal well trajectory, and conducts high-resolution reservoir inversion to achieve dynamic prediction of the reservoir and dynamic adjustment of the drilling trajectory.

[0006] The characteristics of the technical solution adopted by the present invention also lie in:

[0007] Further, the real-time seismic guidance method for horizontal wells is specifically implemented according to the following steps:

[0008] Step 1: Obtain the drilling data of a standard horizontal well, and calculate the standard impedance curve;

[0009] Step 2: Use the standard impedance curve obtained in Step 1 as the target curve for machine learning. Input the real-time data while drilling the horizontal well, and use the support vector machine method of machine learning to fit the impedance curve of the horizontal well being drilled in real time according to the sensitive curve by lithology.

[0010] Step 3: Calibrate the horizontal well being drilled and the newly completed vertical well based on the impedance curve of the horizontal well being drilled obtained in Step 2, add new interpreted horizons, and update the spatially variable velocity field model.

[0011] Step 4: Update the impedance model inverted most recently by adding new wells and new horizons to obtain a new inverted impedance model. Use the spatially variable velocity field model updated in Step 3 to convert the new inverted impedance model into an impedance model in the depth domain.

[0012] Step 5: According to the pattern of the horizontal well trajectory being drilled in the seismic trace interval, update the wellbore impedance depth domain model with horizontal section information constraints for each trace in the depth domain impedance model in Step 4.

[0013] Step 6: Use the spatially variable velocity field model in Step 3 again to perform a depth-time conversion on the depth domain model updated in Step 5 to obtain a new impedance model in the time domain.

[0014] Step 7: Use the impedance model in the time domain in Step 6 to carry out high-resolution reservoir inversion to obtain the reservoir dynamic prediction results.

[0015] Step 8: Based on the reservoir dynamic prediction results, conduct sub-layer interpretation with the sandstone group as the target, complete the prediction of the location of the dominant sub-layers, and dynamically adjust the drilling trajectory of the horizontal well being drilled based on the location of the dominant sub-layers.

[0016] Step 9: Dynamically repeat Steps 1 - 8 as the drilling progresses until the horizontal well being drilled is completed.

[0017] Furthermore, the specific operation in Step 1 is as follows:

[0018] Select a completed horizontal well in the work area that has drilled through both sandstone and mudstone and has measured acoustic transit time, density, and logging-while-drilling curves in the horizontal section as the standard horizontal well. Calculate the standard impedance curve P through the acoustic transit time and density curves, as shown in formula (1) specifically:

[0019]

[0020] where DEN is the density curve and AC is the acoustic transit time curve.

[0021] Furthermore, the specific operation in Step 2 is as follows:

[0022] Step 2.1: Take the standard impedance curve P obtained in Step 1 as the target curve for machine learning; take the logging-while-drilling gas measurement curve QC, the logging-while-drilling gamma ray curve GR, and the logging-while-drilling shale content curve V sh as the input sample curves for machine learning; among them, the logging-while-drilling gas measurement curve QC and the logging-while-drilling gamma ray curve GR are obtained through measurement, and the logging-while-drilling shale content curve V sh is calculated by formula (2):

[0023] V sh =(GR - GR min ) / (GR max - GR min ) * 100 (2)

[0024] where GR min is the gamma ray value of the pure sandstone section of the standard well, and GR max is the gamma ray value of the pure shale section of the standard well;

[0025] Step 2.2: Classify the input samples according to the lithology of sandstone and shale. For shale, use the logging-while-drilling gamma ray curve GR and the logging-while-drilling shale content curve V sh as samples. For sandstone, select the logging-while-drilling gas measurement curve QC and the logging-while-drilling gamma ray curve GR as samples, and conduct support vector machine machine learning for sandstone and shale respectively to train prediction models for different lithologies;

[0026] Step 2.3: Obtain the logging-while-drilling data of the horizontal well being drilled in real time, calculate the logging-while-drilling shale content curve V sh , conduct machine learning by lithology, and fit the impedance curve of the horizontal well in real time.

[0027] Furthermore, the specific operations in Step 3 are as follows:

[0028] Step 3.1: Based on the impedance curve of the horizontal well being drilled obtained in Step 2, calibrate the vertical section of the horizontal well being drilled; if there are measured acoustic travel time and density curve data in the vertical section of the horizontal well being drilled, use the measured curves for calibration;

[0029] Step 3.2: Detect the newly completed vertical section of the well and conduct well-seismic calibration;

[0030] Step 3.3: Detect the latest interpreted horizons, add new horizons for spatial constraint, and at the same time add the time-depth relationship of the newly completed vertical section of the well and the horizontal well being drilled to update the spatially variable velocity field model.

[0031] Furthermore, the specific operations in Step 4 are as follows:

[0032] Step 4.1: If this method is carried out for the first time, the impedance model obtained from the most recent inversion is the impedance inversion model adopted in the horizontal well design stage. On this basis, newly drilled vertical wells, pilot holes, and newly interpreted horizons are added to obtain a new impedance inversion model;

[0033] If it is not the first time, newly drilled vertical wells, pilot holes, and newly interpreted horizons are added to the impedance model obtained from the previous iterative inversion to obtain a new impedance inversion model;

[0034] Step 4.2: Use the spatially-varying velocity field model in Step 3 to convert the new impedance inversion model obtained in Step 4.1 into a depth-domain impedance model.

[0035] Furthermore, in Step 5, the specific operation is as follows:

[0036] Taking the depth-domain impedance model in Step 4 as the initial model, for 4 different trajectory patterns of the actual horizontal well trajectory within the seismic trace interval, different methods are adopted to update the wellbore model of real-time drilling information channel by channel during drilling, specifically as follows;

[0037] When the actual horizontal well trajectory is horizontal between the nth and (n + 1)th seismic traces, the method of taking the end value is adopted to update the model according to formula (3):

[0038] Y = P imp (3)

[0039] where Y is the impedance value of the model at the intersection of the impedance model of the (n + 1)th trace and the horizontal well, and P imp is the impedance curve value of the horizontal well at the intersection of the horizontal well and the (n + 1)th seismic trace;

[0040] When the actual horizontal well trajectory is monotonically increasing between the nth and (n + 1)th seismic traces, the method of projecting the impedance curve of the horizontal section to the end trace is adopted. Starting from the vertical depth d n of the nth trace, the model is updated point by point along the horizontal well trajectory according to formula (4) to d n+1 :

[0041] Y _d = P imp_d (4)

[0042] where d is the vertical depth, and its value range is from the vertical depth dn of the nth trace to the vertical depth d n+1 of the (n + 1)th trace; P imp_d is the impedance curve value when the vertical depth of the horizontal well is d, and Y _d is the impedance value of the model when the vertical depth of the (n + 1)th seismic trace is d;

[0043] When the actual horizontal well trajectory is in a monotonically decreasing pattern between the nth and (n + 1)th seismic traces, the method of projecting to the end trace is adopted. Starting from the vertical depth d of the nth tracen Start, update the model point by point according to formula (4) until d n+1 ;

[0044] When the seismic trace interval of the actual horizontal well trajectory is a non-monotonic pattern, use the method of splitting, encrypting and projecting. Split the horizontal well section trajectory within the trace interval into monotonic well sections in sequence, encrypt the model at the end of each well section, and then update the encrypted model according to the method of projecting towards the end.

[0045] Furthermore, in step 7, the specific operation is as follows:

[0046] Utilize the impedance model in the time domain of step 6 to carry out geostatistics, waveform indication, and machine learning high-resolution reservoir inversion to obtain the impedance result of reservoir dynamic prediction.

[0047] Furthermore, in step 8, the specific operation is as follows:

[0048] Step 8.1: Based on the reservoir dynamic prediction result, within the well inclination range of 88 - 92 degrees that can be implemented in the drilling engineering, interpret the target small layer horizons respectively according to the level of geological sand groups;

[0049] Set the impedance threshold of the effective reservoir. After subtracting this threshold from the dynamic prediction result, convert the point at the impedance threshold to a 0 value point, and then use the method of automatically tracking the 0 value inflection point to quickly interpret the small layer horizons;

[0050] Step 8.2: Calculate the thickness of each small layer reservoir, estimate the length of the remaining horizontal well trajectory in the favorable reservoir of each small layer respectively, and use the method of bubble sorting of reservoir length to dynamically select the optimal drilling trajectory;

[0051] Step 8.3: Compare the real-time reservoir section lengths for other deployed horizontal wells to be drilled, complete the dynamic implementation sequence sorting, and provide a reference for the horizontal well drilling construction sequence.

[0052] Furthermore, in step 9, according to the following situation, dynamically repeat steps 1 - 8:

[0053] Update in real time at intervals of the length of the completed seismic trace distance in the horizontal section, repeat steps 1 - 8, and plan the optimal drilling suggestions until the horizontal well is completed.

[0054] The beneficial effects of the present invention are:

[0055] The present invention provides a real-time seismic guidance method for horizontal wells. By using the idea of lithology-based machine learning to fit impedance curves, calibration of the horizontal well being drilled is carried out, the velocity is continuously updated, a spatially variable velocity with higher precision is constructed, and the accuracy of microstructural interpretation is improved. Based on the spatially variable velocity model, the wellbore model of real-time drilling information is updated channel by channel along the horizontal well trajectory, realizing dynamic prediction of the reservoir and dynamic optimization of the trajectory. The seismic guidance accuracy and efficiency of the horizontal wells carried out by this method are higher. In the practical application of this method, the seismic prediction coincidence rate is 93%, and the average reservoir drilling encounter rate reaches more than 85%, effectively supporting the seismic guidance work of horizontal wells and filling the gap in this technology. This method can be widely applied to the horizontal well development of conventional and unconventional oil and gas, providing a solid technical guarantee for improving quality and efficiency, having a broad application prospect and considerable economic benefits. Description of the Drawings

[0056] Figure 1 is the flowchart of the method of the present invention;

[0057] Figure 2 is the comparison diagram of the measured impedance and the machine learning impedance in Example 1 of the present invention;

[0058] Figure 3a is the velocity field diagram before the real-time velocity constraint of the horizontal well H2 in Example 1 of the present invention;

[0059] Figure 3b is the velocity field diagram after the real-time velocity constraint of the horizontal well H2 in Example 1 of the present invention;

[0060] Figure 4a is the impedance inversion profile diagram before (left) and after (right) the dynamic inversion of the 6th target point constrained by the real-time information of the horizontal well in Example 1 of the present invention;

[0061] Figure 4b is the impedance inversion profile diagram before (left) and after (right) the dynamic inversion of the 13th target point constrained by the real-time information of the horizontal well in Example 1 of the present invention;

[0062] Figure 4c is the impedance inversion profile diagram before (left) and after (right) the dynamic inversion of the 18th target point constrained by the real-time information of the horizontal well in Example 1 of the present invention;

[0063] Figure 5 is the plan view of the reservoir thickness of the small layer in Example 1 of the present invention;

[0064] Figure 6a is the schematic diagram of the inversion model (left) and the inversion impedance profile (right) of Well H2 deployed in Example 1 of the present invention;

[0065] Figure 6b is the schematic diagram of the inversion model (left) and the inversion impedance profile (right) of Well H2 constrained by real-time information in Example 1 of the present invention;

[0066] Figure 7 In Embodiment 2 of the present invention, the impedance profile of Well H3 is obtained by real-time information-constrained dynamic inversion;

[0067] Figure 8 In Embodiment 3 of the present invention, the impedance profile of Well H4 is obtained by real-time information-constrained dynamic inversion;

[0068] Figure 9 In Embodiment 4 of the present invention, the impedance profile of Well H5 is obtained by real-time information-constrained dynamic inversion. Specific Embodiment

[0069] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0070] The present invention discloses a real-time seismic guidance method for horizontal wells, which uses a machine learning method to real-time fit the impedance curve of the horizontally drilled well being drilled, calibrate the horizontally drilled well being drilled and the newly completed vertical well, update the spatially variant velocity field model after adding new horizons, and on the basis of the spatially variant velocity field model, update the borehole impedance inversion model with real-time drilling information constraint for each trace along the horizontal well trajectory, and conduct high-resolution reservoir inversion to achieve dynamic prediction of the reservoir and dynamic adjustment of the drilling trajectory.

[0071] As Figure 1 shown, it is specifically implemented according to the following steps:

[0072] Step 1: Obtain the standard horizontal well drilling data and calculate the standard impedance curve;

[0073] Select the completed horizontal well in the work area that has drilled through both sandstone and mudstone and has measured acoustic transit time, density and logging-while-drilling curves in the horizontal section as the standard horizontal well, and calculate the standard impedance curve P through the acoustic transit time and density curves, as specifically shown in formula (1):

[0074] P = DEN / AC (1)

[0075] where DEN is the density curve and AC is the acoustic transit time curve;

[0076] Step 2: Use the standard impedance curve obtained in Step 1 as the target curve of machine learning, input the logging-while-drilling data of the horizontally drilled well being drilled, and use the machine learning support vector machine method to real-time fit the impedance curve of the horizontally drilled well being drilled according to the sensitive curves by lithology;

[0077] The specific operation is as follows:

[0078] Step 2.1: Use the standard impedance curve P obtained in Step 1 as the target curve of machine learning; use the gas logging curve while drilling QC, the gamma ray curve while drilling GR, and the shale content curve while drilling Vsh As the input sample curve for machine learning; among them, the logging-while-drilling gas measurement curve QC and the logging-while-drilling gamma curve GR are obtained through measurement, and the logging-while-drilling shale content curve V sh is calculated by formula (2):

[0079] V sh =(GR - GR min ) / (GR max - GR min )*100 (2)

[0080] wherein, GR min is the gamma value of the pure sandstone section of the standard well, and GR max is the gamma value of the pure shale section of the standard well;

[0081] Step 2.2: Classify the input samples according to the lithology of sandstone and shale. For shale, use the logging-while-drilling gamma curve GR and the logging-while-drilling shale content curve V sh as samples, and for sandstone, select the logging-while-drilling gas measurement curve QC and the logging-while-drilling gamma curve GR as samples, and respectively carry out support vector machine machine learning for sandstone and shale to train prediction models of different lithologies;

[0082] Step 2.3: Obtain the logging-while-drilling data of the horizontal well being drilled in real time, calculate the logging-while-drilling shale content curve V sh , carry out machine learning by lithology, and fit the impedance curve of the horizontal well in real time;

[0083] Step 3: Based on the impedance curve of the horizontal well being drilled obtained in Step 2, calibrate the horizontal well being drilled and the newly completed vertical well, and add new interpreted horizons to update the variable velocity field model;

[0084] The specific operation is as follows:

[0085] Step 3.1: Based on the impedance curve of the horizontal well being drilled obtained in Step 2, calibrate the vertical well section of the horizontal well being drilled; if there are measured acoustic travel time and density curve data in the vertical well section of the horizontal well being drilled, use the measured curves for calibration;

[0086] Step 3.2: Detect the newly completed vertical well and carry out well-seismic calibration;

[0087] Step 3.3: Detect the latest interpreted horizons, add new horizons for spatial constraint, and at the same time add the time-depth relationship of the newly completed vertical well and the horizontal well being drilled, and then update the variable velocity field model;

[0088] Step 4: Add the newly completed vertical well and new horizons to update the impedance model of the last inversion to obtain a new inversion impedance model, and use the variable velocity field model updated in Step 3 to convert the new inversion impedance model into a depth-domain impedance model;

[0089] The specific operation is as follows:

[0090] Step 4.1: If this method is carried out for the first time, add the newly drilled vertical wells, pilot holes and newly interpreted horizons to the impedance inversion model used in the horizontal well design stage to obtain a new impedance inversion model;

[0091] If it is not the first time, add the newly drilled vertical wells, pilot holes and newly interpreted horizons to the impedance model obtained from the most recent inversion to obtain a new impedance inversion model;

[0092] Step 4.2: Use the spatially variable velocity field model in Step 3 to convert the new impedance inversion model obtained in Step 4.1 into an impedance model in the depth domain;

[0093] Step 5: According to the pattern of the horizontal well trajectory being drilled in the seismic trace interval, update the wellbore impedance depth domain model with horizontal section information constraints for each trace in the depth domain impedance model in Step 4;

[0094] The specific operation is as follows:

[0095] Using the depth domain impedance model in Step 4 as the initial model, for the 4 different trajectory patterns of the horizontal well trajectory being drilled within the seismic trace interval, adopt different methods to complete the real-time wellbore model update of the drilling information for each trace while drilling, specifically as follows;

[0096] When the horizontal well trajectory being drilled is horizontal between the nth and (n + 1)th seismic traces, adopt the method of taking the end value and update the model according to formula (3):

[0097] Y = P imp (3)

[0098] Where Y is the impedance value of the model at the intersection of the impedance model of the (n + 1)th trace and the horizontal well, and P imp is the impedance curve value of the horizontal well at the intersection of the horizontal well and the (n + 1)th seismic trace;

[0099] When the horizontal well trajectory being drilled is monotonically increasing between the nth and (n + 1)th seismic traces, adopt the method of projecting the horizontal section impedance curve to the end trace. Starting from the vertical depth d n of the nth trace, update the model point by point along the horizontal well trajectory according to formula (4) to d n+1 :

[0100] Y _d = P imp_d (4)

[0101] Where d is the vertical depth, and the value range is from the vertical depth dn of the nth trace to the vertical depth d n+1 of the (n + 1)th trace; P imp_d is the impedance curve value when the vertical depth of the horizontal well is d, and Y _dis the impedance value of the model when the vertical depth of the (n + 1)-th seismic trace is d;

[0102] When the actual horizontal well trajectory between the n-th and (n + 1)-th seismic traces is a monotonically decreasing pattern, the method of projecting to the end trace is adopted. Starting from the vertical depth d of the n-th trace, the model is updated point by point according to formula (4) until d n ; n+1 ;

[0103] When the seismic trace spacing of the actual horizontal well trajectory is a non-monotonic pattern, the method of splitting, densifying, and projecting is adopted. The horizontal well section trajectory within the trace spacing is successively split into monotonic well sections, the model is densified at the end of each well section, and then the densified model is updated according to the method of projecting to the end;

[0104] Step 6: Using the spatially variant velocity field model in Step 3 again, perform depth-time conversion on the depth-domain model updated in Step 5 to obtain a new time-domain impedance model;

[0105] Step 7: Using the time-domain impedance model in Step 6, conduct geostatistics, waveform indication, and machine learning high-resolution reservoir inversion to obtain the reservoir dynamic prediction impedance result;

[0106] Step 8: Based on the reservoir dynamic prediction result, conduct sub-layer interpretation with the sand bed group as the target, complete the prediction of the advantageous sub-layer position, and dynamically adjust the drilling trajectory of the horizontal well being drilled based on the advantageous sub-layer position;

[0107] The specific operation is as follows:

[0108] Step 8.1: Based on the reservoir dynamic prediction result, within the well inclination range of 88 - 92 degrees that can be implemented in the drilling engineering, interpret the target sub-layer horizons according to the level of the geological sand bed group respectively;

[0109] Set the impedance threshold of the effective reservoir. After subtracting the threshold from the dynamic prediction result, convert the point at the impedance threshold to a 0 value point, and then use the method of automatically tracking the 0 value inflection point to quickly interpret the sub-layer horizons;

[0110] Step 8.2: Calculate the reservoir thickness of each sub-layer, estimate the length of the remaining horizontal well trajectory in the favorable reservoir of each sub-layer respectively, and use the method of reservoir length bubble sorting to dynamically optimize the best drilling trajectory;

[0111] Step 8.3: Compare the real-time reservoir section lengths for other deployed horizontal wells to be drilled, complete the dynamic implementation sequence sorting, and provide a reference for the horizontal well drilling construction sequence.

[0112] Step 9: When each seismic trace is completed, update in real time, repeat Steps 1 - 8, plan the optimal drilling suggestions until the horizontal well is completed.

[0113] The technical solution of the present invention will be further described below through embodiments and the accompanying drawings.

[0114] Embodiment 1

[0115] Step 1: Use machine learning to fit the impedance curve of the horizontal well in real time. Obtain the logging-while-drilling data of the horizontal well in real time, select the standard wells in the work area, and use the method of machine learning by lithology to carry out the fitting of the impedance curve. For mudstone, the combination of gamma and shale content is used, and for sandstone, the combination of gas logging and gamma is selected; from Figure 2 It can be seen from the comparison chart of the measured impedance and the machine learning impedance that the similarity between the measured impedance and the machine learning impedance curves is relatively high, and the correlation coefficient is relatively high, which can be used as the input curve for calibration and inversion.

[0116] Step 2: Based on the newly drilled vertical wells and pilot holes, calibrate the horizontal well being drilled, continuously update and correct it, and construct a high-precision variable velocity field; as shown in Figure 3, use the real-time velocity constraint technology of Well H2 being drilled to carry out velocity modeling while drilling. Compared with the original velocity Figure 3a , after adding the constraint of this well Figure 3b the velocity around the well increases, and there are slight differences in the velocity field before and after the constraint, which helps to improve the accuracy of micro-structure interpretation and can guide the azimuth adjustment of horizontal well drilling.

[0117] Step 3: Dynamically predict the distribution of wellbore sweet spots by real-time drilling information constraint; on the basis of updating the inversion model in the newly drilled wells, pilot holes, and the horizontal straight well section being drilled, use the variable velocity to obtain the depth-domain model. For the 4 different styles of seismic trace spacing of the actual horizontal well trajectory, 4 corresponding different formulas are used, as shown in Table 1, and the wellbore model update of real-time drilling information is completed for each trace while drilling;

[0118] Table 1 4 model update formulas corresponding to 4 kinds of inter-channel trajectories

[0119]

[0120] Note: Y is the impedance value of the model at the intersection of the impedance model of the (n + 1)-th trace and the horizontal well; P imp is the impedance curve value of the horizontal well at the intersection of the horizontal well and the (n + 1)-th seismic trace; d is the vertical depth, and the value range is from the vertical depth dn of the n-th trace to the vertical depth dn+1 of the (n + 1)-th trace; P imp_d is the impedance curve value of the horizontal well when the vertical depth is d, and Y _d is the impedance value of the model when the vertical depth of the (n + 1)-th seismic trace is d; the method of splitting, encrypting and projecting is used, that is, the horizontal well section trajectory within the trace spacing is successively split into monotonic well sections, and the model is encrypted at the end of each well section.

[0121] Step 4: After dynamically updating the model in the depth domain, use the variable velocity to perform deep-time conversion to obtain a new high-precision time-domain model;

[0122] Step 5: Using the dynamic new model, conduct dynamic high-resolution reservoir inversion again, as Figure 4a shown on the right; in Well H2, the lithology changed from sandstone to mudstone at the 6th target point, and the gamma increased, while the inversion predicted sandstone, as Figure 4a shown on the left. The prediction did not match the actual drilling results and could not effectively assist in adjusting the drilling direction. After adopting high-precision dynamic inversion constrained by real-time information, as Figure 4a shown on the right, the prediction results were consistent with the measured gamma curve while drilling of the horizontal well, and could guide the real-time adjustment of the horizontal well drilling trajectory;

[0123] Step 6: Dynamically adjust the horizontal well trajectory based on the dominant sub-layers; the high-precision dynamic inversion results constrained by real-time information at the 6th target point, as Figure 4a shown on the right, predicted that the length of the dominant reservoir was the longest when the remaining horizontal section was drilled in the reservoir of the upper sub-layer. As Figure 5 shown, the subsequent target well drilling trajectory was optimized accordingly;

[0124] Step 7: When each seismic trace is drilled, dynamically repeat Steps 1-6 until the horizontal well is completed. From the comparison effect of dynamic inversion before and after real-time information constraint at the 13th and 18th target points of the well being drilled Figure 4c 、 4b it was analyzed that at the position of lithology mutation, this method provided key seismic guidance suggestions for real-time adjustment of the horizontal well trajectory, ensuring that the final horizontal section length of Well H2 reached 1520 m and the oil layer drilling encounter rate reached 88%.

[0125] Example 2

[0126] Step 1: Use machine learning to fit the impedance curve of horizontal well H3 in real time. Real-time obtain the logging data while drilling the horizontal well, select the standard wells in the work area, and adopt the method of machine learning by lithology to conduct impedance curve fitting. For mudstone, use the combination of gamma and shale content, and for sandstone, select the combination of gas logging and gamma to obtain the impedance curve of Well H3.

[0127] Step 2: Based on the newly drilled vertical wells and pilot wells, calibrate the horizontal well being drilled and continuously update to construct a high-precision spatially variable velocity field.

[0128] Step 3: Update the wellbore impedance model with real-time drilling information constraint. On the basis of updating the inversion model in the vertical sections of the newly drilled wells, pilot wells, and horizontal well being drilled, use the spatially variable velocity to obtain the depth-domain model. For the 4 different patterns of seismic trace spacing of the actual drilling horizontal well trajectory of Well H3, adopt different update methods (Table 1) to complete the real-time drilling information-based wellbore model update trace by trace while drilling.

[0129] Step 4: After dynamically updating the model in the depth domain, use the spatially variable velocity to perform deep-time conversion to obtain a high-precision time-domain new model.

[0130] Step 5: Use the dynamic new model to carry out dynamic high-resolution reservoir inversion again.

[0131] Step 6: Dynamically adjust the horizontal well trajectory based on the dominant sub-layers.

[0132] Step 7: Dynamically repeat Steps 1-6 according to the drilling situation until the horizontal well is completed. As Figure 7 shown, the final horizontal section length of Well H3 is 1505 m, and the oil reservoir encounter rate reaches 96%.

[0133] Example 3

[0134] Step 1: Use machine learning to fit the impedance curve of horizontal well H4 in real time. Obtain the logging-while-drilling data of the horizontal well in real time, select the standard wells in the work area, and use the method of machine learning by lithology to carry out impedance curve fitting. For mudstone, use the combination of gamma and shale content, and for sandstone, select the combination of gas logging and gamma to obtain the impedance curve of Well H4.

[0135] Step 2: Based on the newly drilled vertical wells and pilot wells, calibrate the horizontal well being drilled and continuously update to construct a high-precision variable velocity field.

[0136] Step 3: Update the wellbore impedance model with real-time drilling information constraints. On the basis of updating the inversion model in the vertical sections of the newly drilled wells, pilot wells, and horizontal well being drilled, use the variable velocity to obtain the depth-domain model. For the 4 different patterns of seismic trace spacing of the actual drilled horizontal well trajectory of Well H4, adopt different update methods (Table 1) to complete the real-time update of the wellbore model of real-time drilling information trace by trace while drilling.

[0137] Step 4: After dynamically updating the model in the depth domain, use the variable velocity for depth-time conversion to obtain a high-precision new model in the time domain.

[0138] Step 5: Use the dynamic new model to carry out dynamic high-resolution reservoir inversion again.

[0139] Step 6: Dynamically adjust the horizontal well trajectory based on the dominant sub-layers.

[0140] Step 7: Dynamically repeat Steps 1-6 according to the drilling situation until the horizontal well is completed. As Figure 8 shown, the final horizontal section length of Well H4 is 1535 m, and the oil reservoir encounter rate reaches 91%.

[0141] Example 4

[0142] Step 1: Use machine learning to fit the impedance curve of horizontal well H5 in real time. Obtain the logging-while-drilling data of the horizontal well in real time, select the standard wells in the work area, and use the method of machine learning by lithology to carry out impedance curve fitting. For mudstone, use the combination of gamma and shale content, and for sandstone, select the combination of gas logging and gamma to obtain the impedance curve of Well H5.

[0143] Step 2: Based on the newly completed vertical wells and pilot holes, calibrate the horizontal well being drilled, continuously update, and construct a high-precision variable velocity field in space.

[0144] Step 3: Update the wellbore impedance model with real-time drilling information. On the basis of updating the inversion model in the vertical sections of the newly completed wells, pilot holes, and horizontal wells being drilled, use the variable velocity in space to obtain the depth-domain model. For the 4 different patterns of seismic trace spacing of the actual horizontal well trajectory of Well H5, adopt different update methods (Table 1), and complete the wellbore model update of real-time drilling information channel by channel while drilling.

[0145] Step 4: After dynamically updating the model in the depth domain, use the variable velocity in space for depth-time conversion to obtain a new high-precision time-domain model.

[0146] Step 5: Use the dynamic new model to carry out dynamic high-resolution reservoir inversion again.

[0147] Step 6: Dynamically adjust the horizontal well trajectory based on the dominant sub-layers.

[0148] Step 7: Dynamically repeat Steps 1 - 6 according to the drilling situation until the horizontal well is completed. As Figure 9 shown, the final horizontal section length of Well H5 is 1192 m, and the oil-bearing formation encounter rate reaches 96%.

[0149] Comparative Example 1

[0150] Comparison of the models and inversion profiles before and after being constrained by the real-time drilling information of Well H2 Figure 6a 、 6b Analysis shows that: after using this method in Well H2, the reservoir prediction compliance rate has increased by 19%, helping the oil-bearing formation encounter rate of this well reach 88%. The real-time and efficient horizontal well guidance method has significant effects, and the final inversion result can provide a reliable basis for the next-step horizontal well test fracturing design, with broad application prospects.

Claims

1. A real-time seismic guidance method for horizontal wells, characterized in that, Utilize the lithology-based machine learning method to fit the impedance curve of the horizontal well in real time. Use this horizontal well and the new well for calibration. After adding a new horizon, update the velocity field of the current horizontal well platform in real time. At the same time, after adding the new well information, update the impedance model of the current platform reservoir inversion. Based on the velocity model, complete the update of the wellbore dynamic inversion model with real-time drilling information constraints for each trace along the horizontal well trajectory, and conduct high-resolution reservoir inversion to achieve dynamic prediction of the reservoir and dynamic adjustment of the drilling trajectory.

2. The real-time seismic guidance method for horizontal wells according to claim 1, characterized in that, The specific implementation steps are as follows: Step 1: Obtain the drilling data of the standard horizontal well and calculate the standard impedance curve. Step 2: Use the standard impedance curve obtained in Step 1 as the target curve of machine learning, input the data while drilling the horizontal well, and use the machine learning support vector machine method to fit the impedance curve of the horizontal well being drilled in real time according to the sensitive curve by lithology. Step 3: Based on the impedance curve of the horizontal well being drilled obtained in Step 2, calibrate the horizontal well being drilled and the newly completed vertical well, and add the newly interpreted horizon to update the spatially variable velocity field model. Step 4: Add the new well and the new horizon to update the impedance model of the last inversion to obtain a new inverted impedance model. Use the spatially variable velocity field model updated in Step 3 to convert the new inverted impedance model into a depth-domain impedance model. Step 5: According to the pattern of the horizontal well trajectory in the seismic trace interval, complete the update of the wellbore impedance depth-domain model with horizontal section information constraints for each trace in the depth-domain impedance model in Step 4. Step 6: Use the spatially variable velocity field model in Step 3 again to perform a depth-time conversion on the depth-domain model updated in Step 5 to obtain a new time-domain impedance model. Step 7: Use the time-domain impedance model in Step 6 to conduct high-resolution reservoir inversion to obtain the dynamic prediction results of the reservoir. Step 8: Based on the dynamic prediction results of the reservoir, conduct sub-layer interpretation with the sandstone group as the target, complete the prediction of the location of the dominant sub-layers, and dynamically adjust the drilling trajectory of the horizontal well being drilled based on the location of the dominant sub-layers. Step 9: As the drilling progresses, dynamically repeat Steps 1-8 until the horizontal well being drilled is completed.

3. The real-time seismic guidance method for horizontal wells according to claim 2, wherein The specific operation in Step 1 is as follows: Select a completed horizontal well in the work area that has drilled through both sandstone and mudstone and has measured acoustic travel time, density, and logging-while-drilling curves in the horizontal section as the standard horizontal well. Calculate the standard impedance curve P through the acoustic travel time and density curves, as shown in formula (1): P = DEN / AC (1) Where DEN is the density curve and AC is the acoustic travel time curve.

4. The real-time seismic guidance method for horizontal wells according to claim 2, wherein The specific operation in Step 2 is as follows: Step 2.1: Take the standard impedance curve P obtained in Step 1 as the target curve for machine learning; take the gas logging while drilling curve QC, the gamma ray while drilling curve GR, and the shale content while drilling curve V sh as the input sample curves for machine learning; among them, the gas logging while drilling curve QC and the gamma ray while drilling curve GR are obtained by measurement, and the shale content while drilling curve V sh is calculated by formula (2): V sh = (GR - GR min ) / (GR max - GR min ) * 100 (2) Among them, GR min is the gamma value of the pure sandstone section of the standard well, and GR max is the gamma value of the pure shale section of the standard well; Step 2.2: Classify the input samples according to the lithology of sandstone and mudstone. For mudstone, use the gamma ray curve GR while drilling and the shale content curve V while drilling sh as samples. For sandstone, select the gas logging curve QC while drilling and the gamma ray curve GR while drilling as samples, and carry out support vector machine machine learning for sandstone and mudstone respectively to train prediction models for different lithologies; Step 2.3: Obtain the logging-while-drilling data of the horizontal well being drilled in real time, and calculate the shale content curve V while drilling. sh Conduct machine learning by lithology and fit the impedance curve of the horizontal well in real time.

5. The real-time seismic guidance method for horizontal wells according to claim 2, characterized in that The specific operation in Step 3 is as follows: Step 3.1: Based on the impedance curve of the horizontal well being drilled obtained in Step 2, calibrate the vertical section of the horizontal well being drilled. If there are measured acoustic travel time and density curve data in the vertical section of the horizontal well being drilled, use the measured curves for calibration. Step 3.2: Detect the newly completed vertical well and conduct well-seismic calibration. Step 3.3: Detect the latest interpreted horizon, add the new horizon for spatial constraint, and update the spatially variable velocity field model after adding the time-depth relationship of the newly completed vertical well and the horizontal well being drilled.

6. The real-time seismic guidance method for horizontal wells according to claim 2, characterized in that, The specific operation in Step 4 is as follows: Step 4.1: If this method is carried out for the first time, the impedance model obtained from the most recent inversion is the impedance inversion model adopted in the horizontal well design stage. On this basis, newly drilled vertical wells, pilot holes and newly interpreted horizons are added to obtain a new impedance inversion model; If it is not the first time, newly drilled vertical wells, pilot holes and newly interpreted horizons are added to the impedance model obtained from the previous iterative inversion to obtain a new impedance inversion model; Step 4.2: The new impedance inversion model obtained in Step 4.1 is converted into a depth-domain impedance model by using the spatially-variant velocity field model in Step 3.

7. The real-time seismic orientation method for horizontal wells according to claim 2, characterized in that, In the said Step 5, the specific operation is as follows: Taking the depth-domain impedance model in Step 4 as the initial model, for 4 different trajectory patterns of the actual horizontal well trajectory within the seismic trace interval, different methods are adopted to complete the update of the wellbore model of real-time drilling information channel by channel while drilling, specifically as follows; When the actual horizontal well trajectory is horizontal between the nth and (n + 1)th seismic traces, the method of taking the end value is adopted to update the model according to Formula (3): Y = P imp (3) Among them, Y is the impedance value of the model at the intersection of the n+1 impedance model and the horizontal well, and P imp is the impedance curve value of the horizontal well at the intersection of the horizontal well and the n+1 seismic traces; When the actual horizontal well trajectory is monotonically increasing between the nth and (n + 1)th seismic traces, the method of projecting the horizontal section impedance curve to the end trace is adopted. Starting from the vertical depth d of the nth trace n and updating the model point by point along the horizontal well trajectory according to formula (4) to d n+1 : Y _d = P imp_d (4) Among them, d is the vertical depth, and the value range is from the vertical depth dn of the nth vertical depth to the vertical depth d of the n + 1th vertical depth n+1 ; P imp_d is the impedance curve value when the vertical depth of the horizontal well is d, and Y _d is the impedance value of the model when the vertical depth of the (n + 1)th seismic trace is d; When the actual horizontal well trajectory is in a monotonically decreasing pattern between the nth and (n + 1)th seismic traces, the method of projecting to the end trace is adopted. Starting from the vertical depth d of the nth trace n and updating the model point by point according to formula (4) until d n+1 ; When the seismic trace interval of the actual horizontal well trajectory is a non-monotonic pattern, the method of splitting, densifying and projecting is adopted. The horizontal well section trajectory within the trace interval is successively split into monotonic well sections, the model is densified at the end of each well section, and then the densified model is updated according to the method of projecting towards the end.

8. The real-time seismic guidance method for horizontal wells according to claim 2, characterized in that In the said Step 7, the specific operation is as follows: Using the time-domain impedance model in Step 6, geostatistical, waveform-indicating and machine learning high-resolution reservoir inversion are carried out to obtain the impedance result of reservoir dynamic prediction.

9. The real-time seismic guidance method for horizontal wells according to claim 2, wherein In the said Step 8, the specific operation is as follows: Step 8.1: On the basis of the reservoir dynamic prediction result, within the well inclination range of 88 - 92 degrees that can be implemented in the drilling engineering, the target sub-layer horizons are interpreted respectively according to the level of geological sand groups; Set the impedance threshold of the effective reservoir. After subtracting this threshold from the dynamic prediction result, the impedance threshold is converted into a 0-value point, and then the method of automatically tracking the 0-value inflection point is adopted to quickly interpret the sub-layer horizons; Step 8.2: Calculate the reservoir thickness of each sub-layer, estimate respectively the lengths of the remaining horizontal well trajectories in the favorable reservoirs of each sub-layer, and adopt the method of bubble sorting of reservoir lengths to dynamically select the optimal drilling trajectory; Step 8.3: Compare the lengths of the real-time reservoir sections for other planned horizontal wells to complete the sorting of the dynamic implementation sequence, providing a reference for the construction sequence of horizontal well drilling.

10. The real-time seismic guidance method for horizontal wells according to claim 2, wherein In the said Step 9, according to the following situation, Steps 1 - 8 are dynamically repeated: Update in real time at intervals of the length of the completed seismic trace interval of the horizontal section, repeat Steps 1 - 8, and plan the optimal drilling suggestions until the horizontal well is completed.

Images