A wellbore cleaning evaluation method incorporating a cuttings weighing device

Abstract

A wellbore cleanliness evaluation method incorporating a cuttings weighing device, relating to the field of drilling engineering technology, includes the following steps: acquiring real-time cuttings weight data online using a cuttings weighing device, performing anomaly detection, and performing purification and completion processing. The cuttings weighing device is equipped with multiple sets of weight sensors distributed in a rectangular array for online collection of cuttings weight, classifying wellbore cleanliness status levels and numerical ranges. Based on the processed cuttings weight data, core parameters related to wellbore cleanliness status are calculated. These core parameters, combined with drilling engineering parameters, are used as feature vectors input to a random forest model for training, resulting in a wellbore cleanliness status judgment model. This model is used for online evaluation and optimization of wellbore cleanliness status during production. This invention, through a novel weighing bucket, combined with a weighted average algorithm using multiple distributed sensors and outlier removal logic, achieves accurate and reliable assessment of wellbore status in drilling scenarios.

Description

Technical Field

[0001] This invention relates to the field of drilling engineering technology, and more specifically to a wellbore cleanliness evaluation method combined with a cuttings weighing device. Background Technology

[0002] In oil and gas drilling, real-time and accurate weighing is crucial for assessing wellbore cleanliness, cuttings return rate, downhole cuttings bed accumulation, and formation lithology changes. Therefore, cuttings weighing directly impacts drilling safety, efficiency, and cost. Currently, the industry's main cuttings weighing solutions utilize sensors, employing weighing trays and pressure sensors installed at the end of a vibrating screen or on a conveyor belt to achieve continuous measurement. For example, patent CN217179733U discloses a cuttings weighing device with an automatic tilting function, using a combination of electromagnetic valves and cylinders to control the hopper for weighing cuttings. Patent CN202111204099.3 discloses a wellbore cleanliness quantitative evaluation method based on cuttings return, which compares the amount of cuttings returned to the surface during drilling with the theoretically generated amount at the well bottom to determine the cuttings accumulation at the bottom, providing a basis for cuttings bed morphology identification and guiding surface parameter optimization. However, such systems or methods are often susceptible to interference from factors such as drilling fluid humidity, particle adhesion, equipment vibration, and ambient temperature. Single-point measurements lead to insufficient accuracy and a lack of automatic calibration capabilities. Data distortion is prone to occur in complex conditions such as horizontal wells and highly deviated wells. For example, the aforementioned cuttings weighing device only measures and weighs cuttings through hardware, resulting in limited data acquisition and difficulty in eliminating errors. Furthermore, the weighing systems involved in the aforementioned evaluation methods have been found to have the following problems: poor real-time performance, weak anti-interference capabilities, and a lack of intelligent decision-making capabilities, making it difficult to meet the demands of modern drilling for high efficiency, safety, and intelligence. Therefore, there is an urgent need for a high-efficiency online cuttings weighing system for accurately assessing wellbore cleanliness. Summary of the Invention

[0003] In view of this, the present invention proposes a wellbore cleanliness evaluation method that combines a cuttings weighing device. Through hardware structure innovation and deep integration with machine learning algorithms, a high-precision real-time weighing system is constructed. The hopper structure is optimized to reduce cuttings accumulation, and distributed sensors are designed to avoid abnormal fluctuations in weighing data, ensuring the integrity and reliability of weighing data. Combined with machine learning methods, a real-time and reliable assessment of the wellbore status in drilling scenarios is achieved.

[0004] To solve at least one of the above-mentioned technical problems, the present invention provides a method for evaluating wellbore cleanliness by combining a cuttings weighing device, comprising the following steps: Step S1: Obtain real-time rock cutting weight data online through the rock cutting weighing device, perform anomaly detection, and perform purification and completion processing. The rock cutting weighing device is equipped with multiple sets of weight sensors distributed in a rectangular array for online collection of rock cutting weight. Step S2: Divide the wellbore cleanliness status level and numerical range, and calculate the core parameters related to the wellbore cleanliness status by combining the processed cuttings weight data; Step S3: Combine the core parameters with drilling engineering parameters and use them as feature vectors to train the random forest model to obtain a wellbore cleanliness judgment model, which is used for online evaluation and control optimization of wellbore cleanliness during production.

[0005] The technical effects achieved by this invention are: (1) This invention uses a novel weighing bucket, combined with a weighted average algorithm and outlier removal logic of multiple distributed sensors, to effectively avoid interference from drilling fluid humidity, equipment vibration and other factors. It can efficiently control the cuttings weighing error, provide a reliable data basis for subsequent wellbore cleaning evaluation, and realize accurate and reliable assessment of wellbore status in drilling scenarios.

[0006] (2) The washing system in the weighing hopper of this invention dynamically adjusts the water pressure and washing strategy according to the rock cuttings adhesion characteristics, with ΔM≤5g as the cleaning standard to ensure that there is no residual measurement in the weighing hopper; the isolated forest + LSTM fusion model accurately identifies nonlinear abnormal data and completes the time-series interrupted data, adapting to the uneven distribution of rock cuttings and data interruption problems under complex working conditions such as horizontal wells and highly deviated wells, and its robustness far exceeds that of traditional interpolation methods and 3σ anomaly detection schemes.

[0007] (3) The random forest model in this invention is based on multiple dimensions such as ECD and cuttings return rate to achieve accurate evaluation of the three-level status of "safety-early warning-danger" and outputs standardized signals including abnormal indicators and confidence level; the gradient boosting model takes over the early warning results and dynamically outputs the optimal drilling fluid discharge and cuttings removal frequency. The hardware is driven by the PLC control system to perform regulation and control, forming a closed loop of "data acquisition-anomaly handling-status evaluation-parameter regulation-feedback optimization".

[0008] (4) This invention can effectively reduce the incidence of major accidents such as stuck drill and wellbore collapse by providing early warning of risks such as cuttings bed collapse and wellbore instability; the intelligent control strategy optimizes the drilling fluid’s cuttings carrying capacity and cuttings removal efficiency, while reducing drilling fluid loss and equipment energy consumption, thus achieving a balance between safety and economy. Attached Figure Description

[0009] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0010] Figure 1 This is an overall flowchart of the present invention; Figure 2 This is a top view of the weighing bucket of the present invention; Figure 3 This is a flowchart illustrating the collaborative workflow of the LSTM model and the isolated forest model in this invention. Figure 4 This is a flowchart illustrating the overall structure of the isolated forest model in this invention. Figure 5 This is a flowchart illustrating the overall structure of the LSTM model in this invention. Figure 6 This is a flowchart of the overall model coordinated control process in this invention; Figure 7 This is a schematic diagram illustrating the process of the random forest model outputting early warning classification results in this invention; In the diagram, 1-weighing hopper, 2-weight sensor, 3-direct nozzle, 4-fan-shaped nozzle, 5-waste liquid guide channel, and 6-guide rib. Detailed Implementation

[0011] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings.

[0012] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention.

[0013] A wellbore cleanliness evaluation method incorporating a cuttings weighing device includes the following steps: Step S1: Obtain real-time rock cutting weight data online through the rock cutting weighing device, perform anomaly detection, and perform purification and completion processing. The rock cutting weighing device is equipped with multiple sets of weight sensors distributed in a rectangular array for online collection of rock cutting weight. See Figure 2 The rock cuttings weighing device includes a weighing hopper 1, which is used to receive and weigh the rock cuttings after they have been processed by a common rock cuttings vibration device during construction. The weighing hopper 1 is an inverted trapezoid, and its bottom surface is equipped with 6 sets of pressure plate type weight sensors 2 arranged in a rectangular array of 2 rows and 3 columns. Each weighing is performed by transferring the evenly dispersed rock cuttings from the vibrating screen to the weighing hopper 1 for weighing.

[0014] The bottom surface of the weighing hopper 1 is also provided with a waste liquid guide channel 5 and a direct spray nozzle 3. The waste liquid guide channel 5 is distributed close to the side of the weighing hopper 1 and is connected to an external waste liquid equipment. Four sets of direct-fire nozzles 3 connected to an external liquid source are equally spaced outside the rectangular array. Their spray range covers the gap between the three rows of weight sensors 2, as well as the gap between the rectangular array and the side of the weighing hopper 1.

[0015] On the side of the weighing hopper 1, there are also multiple sets of fan-shaped nozzles 4 and guide ribs 6 at intervals. The fan-shaped nozzles 4 connected to the external liquid source spray downward along the side, and the spray range covers the entire side where they are located. The guide rib 6 is set downward along its side, located below the fan-shaped nozzle 4, and the bottom end of the guide rib 6 is in contact with the bottom surface of the weighing hopper 1.

[0016] In some embodiments, the main body of the weighing hopper 1 can adopt a composite irregular profile of "wide upper body and narrow lower body arc shape + flat bottom + lateral flow guidance structure", with an outward-folding baffle added to avoid rock cuttings loss by design. At the same time, this irregular weighing hopper can greatly prevent rock cuttings from moving within the weighing hopper, improving the integrity of rock cuttings collection (especially spherical rock cuttings). Specifically, the main structural states are as follows: The main body is wider at the top and narrower at the bottom, with an arc-shaped structure. The longitudinal section adopts an inverted trapezoidal + arc-shaped waist structure, with the upper opening being 30% wider than the lower opening. The side wall is inclined at a 15° angle to the vertical direction. To adapt to the mainstream 1500mm wide vibrating screen in domestic drilling sites, the total length of the long side (axis of rotation) is 1500mm. The longitudinal section (perpendicular to the axis of rotation) has a total upper opening width of 1200mm and a lower opening width of 923mm. The total vertical height of the weighing hopper is 900mm. An arc-shaped transition section is set in the middle of the side wall, with a total arc length of 480mm and a corresponding chord length of 462mm, completely covering the arc and curvature of the core area in the middle of the side wall. The radius of the arc is 800mm, and the central angle is 36°. The position is limited: the upper end of the arc section is 210mm vertically from the upper opening of the weighing hopper, and the lower end is 210mm vertically from the lower opening of the weighing hopper, strictly located in the middle of the side wall. The tilt angle ensures that rock chips slide down naturally under gravity, and the arc transition avoids the "right-angle dead angle" of the straight-walled weighing hopper, structurally eliminating rock chip residue. An outward-curving baffle is installed at the top of the transverse section, tilted 30° outwards to intercept rock chips splashed during receiving. The baffle does not affect the trajectory of the weighing hopper during tilting, enhancing its anti-splash capability.

[0017] Bottom plate: The bottom design structure of the weighing hopper 1 is a plate + reinforcing rib + waste liquid guide channel 5. The bottom of the flat body is an absolute plane. The mounting holes of the 6 sets of weight sensors 2 are preset according to a "3 columns × 2 rows" matrix. The columns of the matrix are strictly distributed along the long side (rotation axis direction) of the weighing hopper.

[0018] Based on a 1500mm wide main vibrating screen, the center-to-center distance between two adjacent columns is 460mm, and the center-to-center distance between two adjacent rows is 4600mm. The column direction installation edge distance is 60mm from the left inner wall of the long side of the weighing hopper in the first column and 60mm from the right inner wall of the long side of the weighing hopper in the third column. The row direction installation edge distance is 61.5mm from the front inner wall of the short side of the weighing hopper in the first row and 61.5mm from the rear inner wall of the short side of the weighing hopper in the second row (avoiding the installation positions of the bottom guide channel and low-pressure direct spray nozzle to avoid structural interference). After the sensor is embedded, the top is flush with the inner wall of the flat bottom of the weighing hopper (after the weight sensor 2 is embedded and installed, its weighing force top surface is completely flush with the inner wall of the rock debris contacting the flat bottom of the weighing hopper, without any protrusions or depressions). The weight sensor 2 is connected to the PLC input module through a double-shielded twisted pair cable extending along the preset wiring groove on the side wall of the weighing hopper. This design is suitable for complex scenarios such as highly deviated wells and horizontal wells, where rock cuttings tend to be heavier on the left and lighter on the right, slide along the wall, or accumulate in the middle within the weighing hopper. The "multi-area full coverage" design allows for real-time capture of weight data from all areas at the bottom of the weighing hopper (left, middle, right, front, and back), preventing "local weight omissions" caused by uneven rock cuttings distribution. Furthermore, by applying a "weighted averaging algorithm" to multiple sets of sensor data, outliers caused by instantaneous impacts (such as concentrated rock cuttings falling from a particular spot) can be eliminated, ensuring that the weight of every rock cutting is accurately captured, providing a reliable basis for subsequent volume conversion.

[0019] Lateral flow guide structure: Three longitudinal flow guide ribs 6 are added to the inner wall of the weighing hopper (at the bottom 1 / 3 height, parallel to the axis of rotation). The effective length of the flow guide ribs 6 is 595mm, the vertical projection height is 390mm, the horizontal projection width is 105mm, and the center distance between two adjacent flow guide ribs 6 is 375mm (equally distributed). The maximum protrusion height is ≤15mm to ensure unobstructed spraying from the nozzle. Attention should be paid to the positional relationship with the high-pressure fan-shaped nozzle 4 (at the top 1 / 2 height of the inner wall of the weighing hopper). Each flow guide rib 6 extends along the arc contour of the side wall and smoothly connects with the flat bottom, guiding the rock debris to gather in the center area of ​​the bottom, avoiding local weight concentration caused by the rock debris "sliding against the wall", which would further deviate the sensor weighing data.

[0020] High-pressure fan-shaped flushing zone on the side wall: Two sets of high-pressure fan-shaped nozzles 4 are symmetrically installed on the inner wall of the weighing hopper 1 (at about 1 / 2 of the height from the top opening, parallel to the axis of rotation). The nozzles are angled downwards at 45°, covering the arc transition section and the surface of the guide ribs. The water pressure is adjustable (0.3~0.8MPa). For rocks with weak rock fragment adhesion, such as various sandstones (quartz sandstone, feldspar sandstone, lithic sandstone), carbonate rocks (limestone, dolomite), and igneous / metamorphic rocks (granite, basalt, gneiss): the initial water pressure is 0.3 MPa, which can be increased to 0.5 MPa when ΔM>5g. For rocks with strong rock fragment adhesion, such as mudstone, shale (including carbonaceous shale, oil shale, and mudstone with high montmorillonite / illite content), silty mudstone, silty mudstone, coal, soft mudstone, and claystone with strong hydration swelling properties: the initial water pressure is 0.5 MPa, which can be increased to 0.8 MPa when ΔM>5g. This structure effectively removes residual rock fragments from the sidewalls and guide ribs, preventing residual rock fragments from sliding down and affecting the bottom metering.

[0021] Bottom low-pressure direct flushing area: Four sets of low-pressure direct-injection nozzles are installed on the outer side of the bottom of weighing hopper 1. These four nozzles are distributed along the long side, with the nozzle direction inclined at 30° to the inner wall of the flat bottom of the weighing hopper. The nozzles spray upwards towards the inner wall of the flat bottom, rinsing only the bottom plate. The "directional water flow" of the direct-injection nozzles allows for precise control of the rinsing range, avoiding direct contact with the sensor. This effectively removes residual rock debris while preventing water flow interference with the sensor, ensuring accurate measurement. Furthermore, the rinsing strategy (including water pressure, rinsing duration, and number of rinsing cycles) for the high-pressure fan-shaped rinsing zone on the side walls can be adjusted based on the structural differences of the rinsing area, the adhesion characteristics of rock debris in different mining environments, and real-time rinsing effect evaluation, to achieve the core requirements of "precise cleaning with no residue and no interference with measurement."

[0022] Based on this, the method for online acquisition of rock cuttings weight data by a rock cuttings weighing device includes the following steps: Step S111: After the rock cuttings are vibrated evenly by the vibrating screen, they are transferred into the weighing hopper. The weight of the rock cuttings is collected by 6 sets of weight sensors, and anomaly detection is performed. The rock debris accumulation threshold inside the weighing hopper can be adjusted according to the actual situation, usually set at 50% to 80% of the rated capacity of the weighing hopper. For weakly cohesive rock debris (sandstone, carbonate rock, etc.), the basic dumping threshold can be increased to a maximum of 80%; while for strongly cohesive rock debris (mudstone, shale, hydrated expansive claystone, etc.), the basic dumping threshold is usually reduced to 50% to 60% to reduce the amount of rock debris accumulated at one time and reduce sidewall adhesion.

[0023] Step S112: Clean and complete the rock cutting weight data collected by each group of sensors after anomaly detection, and calculate the total weight of rock cuttings; Step S113: Tilt the weighing hopper to discharge the weighed rock cuttings, and flush the inside of the weighing hopper with a direct jet nozzle and a fan-shaped nozzle until the residual value is flushed out. If the residual value after rinsing does not meet the requirements, it is necessary to rinse again and correct the cuttings data. The total weight of rock cuttings was collected using a weighted average method, as shown in the following formula: in, The weighted average total weight of the rock cuttings is calculated by weighting and averaging the weight data of all rock cuttings collected by multiple sets of sensors in a single weighing cycle after purification and completion processing. The weighting coefficients for the weight sensors in the i-th column and j-th row are given, and the weighting coefficients for the two sets of weight sensors in the second column are given. , The weighting coefficient for each of the three weight sensors is 0.22, while the weighting coefficient for the other weight sensors is 0.14. The weight data collected online by the weight sensor in the i-th column and j-th row; For each meter of drilling, the cuttings are carried from the bottom of the well to the surface by the drilling fluid, and there is a fixed circulation time. Therefore, all the cuttings belonging to this well section are: all the cuttings collected within the time period of the sum of the drilling time and the circulation time of the cuttings in this well section returning to the surface.

[0024] Rinse residual value The calculation method is shown in the following formula: In the formula, This represents the total unloaded weight of the weighing hopper collected by the weight sensor after rinsing. This represents the initial unloaded total weight after the weighing bucket has been calibrated before weighing.

[0025] Due to the highly complex conditions during on-site construction, nonlinear anomalies caused by concentrated falling rock cuttings, high-frequency vibration of the vibrating screen, and mud splashing can easily be obtained during the transfer of rock cuttings to the weighing hopper after uniform vibration in the vibrating screen. Therefore, this invention employs a 3σ criterion-based method to determine whether the data are outliers, including the following steps: Step S121: Calculate the arithmetic mean and standard deviation of the instantaneous weight data of the 6 sets of weight sensors at a certain moment; Step S122: Use the 3σ criterion to determine the abnormality of the weight data of each group of sensors at this time, that is, the abnormality of a certain sensor data. satisfy If so, it is considered an outlier; Step S123: If more than two sets of abnormal values ​​are found in three consecutive sampling cycles, the system will issue a sensor interference alarm, prompting on-site inspection of the weighing hopper for blockage.

[0026] The sampling cycle consists of six sets of sensors simultaneously acquiring data → data calculation → anomaly detection. This anomaly detection is performed solely by the PLC, and the data is instantaneously acquired and used only to determine the condition of the weighing hopper. In some implementations, the sampling frequency can be set to 1Hz, meaning one cycle is one second. The sampling frequency can also be adjusted according to the actual drilling conditions; for example, 0.5Hz corresponds to one cycle = 2 seconds.

[0027] To address the issues of nonlinear anomalies caused by concentrated falling cuttings, high-frequency vibration of vibrating screens, and mud splashing at drilling sites, as well as the interruption of time-series data due to the tipping of weighing hoppers, this invention integrates the advantages of unsupervised anomaly identification of the isolated forest model and the advantages of time-series fitting and completion of the LSTM model. This achieves accurate purification and continuous completion of cuttings data, solving the problems of the traditional 3σ principle, which can only identify Gaussian distribution anomalies and the poor continuity of data completion by polynomial fitting.

[0028] This fusion model adopts a "clean-up then complete" working logic. First, an isolated forest model is used to remove outlier data. Then, the clean time-series data is input into the LSTM model to complete the interruption repair. The training process and working mechanism of the two models are as follows: Figure 3 As shown, the specific steps include: Step S131: Normalize the real-time rock cutting weight data acquired online by each set of weight sensors and input it into the isolated forest model to identify and remove abnormal data, and retain the pure time series rock cutting weight data. Step S132: Input the clean time series rock cuttings weight data into the LSTM model to complete the interruption, and output the final data after secondary verification by the isolated forest model; The features input to the isolated forest model include the main features consisting of real-time rock cutting weight data collected by six sets of pellet-type weight sensors, and auxiliary features consisting of sampling time, vibrating screen vibration frequency, and weighing hopper tilting state, totaling nine features, as shown in Table 1: Table 1. Data characteristics of isolated forests To address the nonlinear and non-Gaussian distribution characteristics of drilling site data, the parameters of the isolated forest model were specifically optimized, with the number of trees set to 200 to improve the model's detection stability under complex data distributions. The anomaly ratio was set to 0.06, matching the approximately 6% anomaly interference probability at the drilling site. Furthermore, multi-threaded parallel computing was employed to ensure real-time monitoring requirements.

[0029] Compared to the unidirectional serial structure of existing isolated forest + LSTM models, the isolated forest model in this invention sets up a pre-filtering branch (using the weighing hopper flipping state as the priority splitting condition for the root node); when splitting a node, six sets of sensor weighing data are used as the core splitting features, and auxiliary working condition features such as the vibrating screen vibration frequency and sampling time are only introduced when the core features cannot complete the splitting. The overall structure of the isolated forest model is as follows: Figure 4 As shown, the anomaly detection feature matrix constructed by this invention can accurately identify and eliminate abnormal values ​​in the real-time weight data of 6 sets of distributed weighing sensors. The model introduces 3 working condition auxiliary features, namely sampling time, vibration frequency of vibrating screen, and weighing hopper flipping state (this feature is obtained through the weighing hopper's own hardware detection, such as reading magnetic switch data), which can be used to help determine the working condition attributes of weight data fluctuations and distinguish between normal working condition fluctuations and real abnormal data.

[0030] To address the temporal characteristics of rock cuttings weighing data and the data interruption caused by the weighing hopper overturning and discharging chips, the LSTM model performs continuous temporal completion on discrete outliers and continuous chip discharge interruptions, rather than replacing isolated single points, thus ensuring the continuity of the data temporal sequence. The overall structure and flow of the LSTM model are as follows: Figure 5 As shown, the LSTM model adopts a three-layer structure of "input layer, hidden layer, and output layer" connected in sequence. The input layer has a dimension of 9, the number of hidden layers is 2, each layer contains 128 neurons, and the output layer has a dimension of 6. The model time step is set to 60, the batch size is 32, the number of training epochs is 50, the optimizer is Adam, the learning rate is set to 0.001, and the loss function is mean squared error to minimize the deviation between the imputed value and the true value.

[0031] The weight sensor in this invention is an industrial-grade sensor, typically with a measuring range of 0~100kg. Therefore, the mean square error between the compensated value and the true value is... Mean absolute error kg, ensuring that the compensation error does not exceed the measurement accuracy of the sensor itself, while allowing for small and reasonable fluctuations at individual sampling points.

[0032] Step S2: Divide the wellbore cleanliness status level and numerical range, and calculate the core parameters related to the wellbore cleanliness status by combining the processed cuttings weight data; The core parameters mainly include the weight data of rock cuttings after purification and replenishment from 6 sets of compressed weight sensors, the rock cuttings return rate K, and the annular rock cuttings volume concentration. Annular equivalent circulating density (ECD), late arrival well depth Rock cuttings volume deviation rate This will be used as a feature input later, and its calculation method is as follows: The cuttings return rate K is calculated using the following formula: In the formula, This represents the actual volume of cuttings per meter of well section; This represents the theoretical volume of cuttings per meter of well section; Actual rock cuttings volume The calculation method is shown in the following formula: In the formula, This represents the total weight of the rock cuttings. Density of rock fragments; Annular rock fragment volume concentration The calculation method is shown in the following formula: In the formula, This refers to the empty volume corresponding to each meter of well section; The calculation method for the annular equivalent cyclic density (ECD) is shown in the following formula: In the formula, H represents the drilling fluid density; H represents the well depth. For annular pressure loss; Late arrival at the well The calculation method is shown in the following formula: In the formula, The time for performing ground weighing; The drilling time for completing a certain section of drilling indicates the moment when the drill bit breaks through the rock cuttings in that section and completes the drilling of the corresponding well section; The drilling time is the pure drilling time required to drill 1 meter of well section, and its reciprocal 1 / t is the mechanical drilling rate; it can be seen that... This indicates the length of the newly drilled section within the cuttings circulation period, while the late arrival depth... This indicates the current drilling depth minus the length of the newly drilled section within the cuttings circulation period, which is the original bottom depth of the well corresponding to the cuttings returned in that section.

[0033] Rock cuttings volume deviation rate The calculation method is shown in the following formula: ; Step S3: Combine the core parameters with drilling engineering parameters and use them as feature vectors to train the random forest model to obtain a wellbore cleanliness judgment model, which is used for online evaluation and control optimization of wellbore cleanliness during production.

[0034] The input features of the random forest model include the weight data of rock cuttings from 6 sets of compressed weight sensors after purification and completion, the vibration frequency of the vibrating screen, the overturning state of the weighing hopper, the rock cuttings return rate K, and the annular rock cuttings volume concentration. Annular equivalent circulating density (ECD), flushing residual value Late arrival at the well Rock cuttings volume deviation rate ; The output of the random forest model includes a classification of wellbore cleanliness status, operational recommendations, and data reliability. The classification results are categorized into three levels: safe, warning, and dangerous. The classification primarily uses the ECD threshold level, combined with the cuttings return rate K and flushing residue value. Classify them.

[0035] Based on this, the output warning includes the following states, specifically as follows: Figure 7 As shown: when , , At this point, if the ECD is within the safe window and the flushing is satisfactory, it indicates that the cuttings return rate is normal, the drilling fluid's cuttings carrying capacity matches the drilling rate, there is no cuttings retention or wellbore spalling, no cuttings accumulation in the annulus, the wellbore is under stable stress, and the output wellbore is in a safe clean state. The formation equilibrium density is obtained from well logging during the drilling design phase.

[0036] when and , At that time, the ECD increased due to the increase in rock fragment concentration (annular rock fragment volume concentration). Increased drilling fluid flow rate (DF) leads to decreased K due to cuttings retention, requiring prevention of cuttings bed development. The output wellbore cleanliness status is a warning sign, specifically the cuttings accumulation status. Output construction recommendations: appropriately increase drilling fluid flow rate, reduce drilling speed, and prevent cuttings bed development.

[0037] when and , At this time, the K value is relatively high due to slight rock cuttings falling off the wellbore (additional cuttings are included). The low ECD is due to insufficient annular pressure and weak wellbore support. The output wellbore cleanliness status is a warning sign, specifically a state of slight wellbore instability. Output construction recommendations: appropriately increase drilling fluid density to increase annular support and closely monitor the proportion of returned cuttings.

[0038] when Furthermore, when the K-value deviation is greater than 10%, residual rock debris in the weighing bucket leads to... The calculation deviation needs to be eliminated and re-evaluated. The output wellbore cleanliness status is a warning, and the specific status type is metering interference. The output construction suggestion is to start the zone adaptive flushing system to flush the weighing bucket again, eliminate the interference of rock cuttings residue, and then re-collect data for evaluation.

[0039] when and , At this time, the large amount of rock cuttings retained leads to a reduction in the effective flow area of ​​the annulus, and a sharp increase in ECD, which can easily cause pump stalling or rock cuttings bed collapse and stuck pipe. The output wellbore clean status is dangerous, and the specific status type is rock cuttings bed collapse risk. Output construction recommendation: immediately reduce the drilling rate, increase the drilling fluid discharge rate, start circulation well flushing, and prevent rock cuttings bed collapse and stuck pipe.

[0040] when , Furthermore, when the proportion of broken rock fragments in the returned cuttings exceeds 30%, the wellbore is severely fragmented (K is significantly high), the ECD is too low to support the wellbore, and the accumulation of broken rock fragments can easily clog the annulus. The output wellbore cleanliness is dangerous, specifically the risk of wellbore collapse. Output construction recommendation: Stop drilling immediately, increase the drilling fluid density to a safe range, and continuously observe the status of the returned cuttings.

[0041] when Furthermore, if the ECD calculation deviation exceeds 10 kg / m³, the weighing hopper will have significant residue, resulting in completely distorted cuttings data. The ECD calculation will then lack a basis, making it impossible to determine the downhole condition. The output wellbore cleanliness status will be deemed dangerous, specifically a metering failure risk. The recommended operation is to immediately cease wellbore cleanliness assessment, conduct a comprehensive inspection of the zoned adaptive flushing system and distributed sensors, and resume operations only after ruling out equipment malfunctions.

[0042] In addition, see Figure 6 The model in this invention can coordinate and control the entire process to form a complete closed loop, and achieves intelligent adaptive control of drilling conditions through five core logic steps, including the following steps: Step S31: Aggregate the wellbore cleaning status and the data input stage obtained in this step using multi-source hardware; Step S32: The PLC control system parses the instructions; Step S33: Drive the hardware module to perform control actions; Step S34: The data after adjustment is fed back by the sensor and used as the input for the next round to achieve continuous optimization and ensure that the system dynamically adapts to changes in drilling conditions.

[0043] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the embodiments of the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for evaluating wellbore cleanliness combined with a cuttings weighing device, characterized in that, Includes the following steps: Step S1: Obtain real-time rock cutting weight data online through the rock cutting weighing device, perform anomaly detection, and perform purification and completion processing. The rock cutting weighing device is equipped with multiple sets of weight sensors distributed in a rectangular array for online collection of rock cutting weight. Step S2: Divide the wellbore cleanliness status level and numerical range, and calculate the core parameters related to the wellbore cleanliness status by combining the processed cuttings weight data; Step S3: Combine the core parameters with drilling engineering parameters and use them as feature vectors to train the random forest model to obtain a wellbore cleanliness judgment model, which is used for online evaluation and control optimization of wellbore cleanliness during production.

2. The wellbore cleanliness evaluation method combined with a cuttings weighing device according to claim 1, characterized in that: The rock cuttings weighing device described in step S1 includes a weighing hopper (1), wherein the weighing hopper (1) is an inverted trapezoid and its bottom surface is provided with 6 sets of press-type weight sensors (2) arranged in a rectangular array of 2 rows and 3 columns. Waste liquid guide channel (5) and direct spray nozzle (3) are also provided on the bottom surface of the weighing hopper (1). The waste liquid guide channel (5) is distributed close to the side of the weighing hopper (1) and is connected to external waste liquid equipment. Four sets of direct-fire nozzles (3) connected to an external liquid source are equally spaced outside the rectangular array. Their spray range covers the gap between the three columns of weight sensors (2) and the gap between the rectangular array and the side of the weighing hopper (1).

3. The wellbore cleanliness evaluation method combined with a cuttings weighing device according to claim 2, characterized in that: The weighing hopper (1) is also provided with multiple sets of fan-shaped nozzles (4) and guide ribs (6) at intervals on its side. The fan-shaped nozzles (4) connected to the external liquid source spray downward along the side and the spray range covers the entire side where they are located. The guide rib (6) is set downward along its side, located below the fan-shaped nozzle (4), and the bottom end of the guide rib (6) is in contact with the bottom surface of the weighing bucket (1).

4. A wellbore cleanliness evaluation method combined with a cuttings weighing device according to claim 3, characterized in that: The method for acquiring rock cuttings weight data online using the rock cuttings weighing device described in step S1 includes the following steps: Step S111: After the rock cuttings are vibrated evenly by the vibrating screen, they are transferred into the weighing hopper. The weight of the rock cuttings is collected by 6 sets of weight sensors, and anomaly detection is performed. Step S112: Clean and complete the rock cutting weight data collected by each group of sensors after anomaly detection, and calculate the total weight of rock cuttings; Step S113: Tilt the weighing hopper to discharge the weighed rock cuttings, and flush the inside of the weighing hopper with a direct jet nozzle and a fan-shaped nozzle until the residual value is flushed out. ; The total weight of rock cuttings is calculated using a weighted average method, as shown in the following formula: in, This represents the total weight of the rock cuttings after weighted averaging. The weighting coefficients for the weight sensors in the i-th column and j-th row are given, and the weighting coefficients for the two sets of weight sensors in the second column are given. , The weighting coefficient for each of the three weight sensors is 0.22, while the weighting coefficient for the other weight sensors is 0.

14. The weight data collected online by the weight sensor in the i-th column and j-th row; Rinse residual value The calculation method is shown in the following formula: In the formula, This represents the total unloaded weight of the weighing hopper collected by the weight sensor after rinsing. This represents the initial unloaded total weight after the weighing bucket has been calibrated before weighing.

5. A wellbore cleanliness evaluation method combined with a cuttings weighing device according to claim 2, characterized in that: The anomaly determination method described in step S1 includes the following steps: Step S121: Calculate the arithmetic mean and standard deviation of the instantaneous weight data of the 6 sets of weight sensors at a certain moment; Step S122: Use the 3σ criterion to determine any abnormalities in the weight data of each group of sensors at this time; Step S123: If there are more than two sets of abnormal values ​​in three consecutive sampling cycles, the system will issue a sensor interference alarm, prompting on-site inspection of the weighing hopper for blockage.

6. A wellbore cleanliness evaluation method combined with a cuttings weighing device according to claim 2, characterized in that: The purification and completion method described in step S1 includes the following steps: Step S131: Normalize the real-time rock cutting weight data acquired online by each set of weight sensors and input it into the isolated forest model to identify and remove abnormal data, and retain the pure time series rock cutting weight data. Step S132: Input the clean time series rock cuttings weight data into the LSTM model to complete the interruption, and output the final data after secondary verification by the isolated forest model; The features of the input isolated forest model include the main features consisting of real-time rock cutting weight data collected by six sets of press-type weight sensors, as well as auxiliary features consisting of sampling time, vibration frequency of the vibrating screen, and the overturning state of the weighing hopper. The isolated forest model has 200 trees, an anomaly ratio of 0.06, and uses multi-threaded parallel computing.

7. A wellbore cleanliness evaluation method combined with a cuttings weighing device according to claim 6, characterized in that: The isolated forest model uses the weighing bucket flipping state as the priority splitting condition for the root node. When the main features cannot complete the node splitting, auxiliary working condition features, including the vibration frequency of the vibrating screen and the sampling time, are used to guide the splitting.

8. A wellbore cleanliness evaluation method combined with a cuttings weighing device according to claim 6, characterized in that: The LSTM model performs continuous time-series completion on discrete outliers and continuous chip removal interruptions. The mean square error between the completed value and the true value is... Mean absolute error kg; The LSTM model is a three-layer structure consisting of an input layer, a hidden layer, and an output layer connected in sequence. The input layer has a dimension of 9, the hidden layer has 2 layers, each containing 128 neurons, and the output layer has a dimension of 6. The model time step was set to 60, the batch size to 32, the number of training epochs to 50, the optimizer to Adam, the learning rate to 0.001, and the loss function to mean squared error.

9. A wellbore cleanliness evaluation method combined with a cuttings weighing device according to claim 4, characterized in that: The input features of the random forest model in step S3 include the weight data of cuttings after purification and completion of 6 sets of pellet-type weight sensors, vibration frequency of the vibrating screen, overturning state of the weighing hopper, cuttings return rate, annular cuttings volume concentration, annular equivalent circulation density, flushing residue value, late well depth, and cuttings volume deviation rate. The calculation method for the cuttings return rate K is shown in the following formula: In the formula, This represents the actual volume of cuttings per meter of well section; This represents the theoretical volume of cuttings per meter of well section; Actual rock cuttings volume The calculation method is shown in the following formula: In the formula, This represents the total weight of the rock cuttings. Density of rock fragments; Annular rock fragment volume concentration The calculation method is shown in the following formula: In the formula, This refers to the empty volume corresponding to each meter of well section; The calculation method for the annular equivalent cyclic density (ECD) is shown in the following formula: In the formula, H represents the drilling fluid density; H represents the well depth. For annular pressure loss; Late arrival at the well The calculation method is shown in the following formula: In the formula, The time for performing ground weighing; The time when a certain section of drilling is completed indicates the moment when the drill bit breaks through the rock cuttings in that section and completes the drilling of the corresponding section; Drilling time represents the pure drilling time required to drill 1 meter of well section. Rock cuttings volume deviation rate The calculation method is shown in the following formula: ； The output of the random forest model includes a classification of wellbore cleanliness status, construction recommendations, and data reliability. The classification results are categorized into three classes: safe, warning, and dangerous, including the following states: when , , At that time, the wellbore cleanliness status is considered safe, among which, This represents the equilibrium density of the formation. when and , At that time, the output wellbore cleanliness status is an early warning, and the specific status type is cuttings accumulation; when and , When the output wellbore cleanliness status is set as an early warning, the specific status type is wellbore micro-instability; when Furthermore, when the K-value deviation is greater than 10%, the output wellbore cleanliness status is an early warning, and the specific status type is metering interference; when and , At that time, the output wellbore cleanliness status is dangerous, specifically the risk of cuttings bed collapse; when , Furthermore, when the proportion of broken pieces in the returned cuttings is greater than 30%, the cleanliness of the output wellbore is considered dangerous, specifically the risk of wellbore collapse. when Furthermore, if the ECD calculation deviation is greater than 10 kg / m³, the output wellbore cleanliness status is dangerous, specifically the status type is metering failure risk.

10. A wellbore cleanliness evaluation method combined with a cuttings weighing device according to claim 1, characterized in that: The regulation optimization described in step S3 is feedback regulation, which includes the following steps: Step S31: Aggregate the wellbore cleaning status and the data input stage obtained in this step using multi-source hardware; Step S32: The PLC control system parses the instructions; Step S33: Drive the hardware module to perform control actions; Step S34: The data after adjustment is fed back by the sensor and used as the input for the next round to achieve continuous optimization and ensure that the system dynamically adapts to changes in drilling conditions.

Images