Intelligent early warning method for early overflow of well drilling

Abstract

The invention discloses an intelligent early warning method for early overflow of well drilling. The intelligent early warning method comprises the following steps: (1) acquiring logging data related to pressure change of a vertical pipe; (2) obtaining an IMF component through logging data obtained through EMD decomposition; (3) extracting IMF components mainly influencing the pressure of the riser through KPCA dimensionality reduction; (4) using an INFORMER deep learning algorithm and combining the actually measured riser pressure to carry out training learning on the main component to obtain a deep learning time sequence prediction model of the riser pressure under the normal and risk-free condition; (5) predicting the pressure change of the vertical pipe under the drilling working condition of the next stage by using the current deep learning parameters; and (6) identifying early overflow by comparing the deviation between the predicted value and the measured value of the riser pressure, further intercepting a curve different from the predicted riser pressure, and performing quantitative analysis on the overflow risk by using the early overflow riser pressure simulated by the multiphase flow model. The method provided by the invention provides technical means and theoretical support for early overflow risk early warning.

Description

Technical Field

[0001] The invention relates to the technical field of oil drilling well control, in particular to an intelligent early warning method for overflow in drilling. Background Art

[0002] As my country's oil and gas field development continues to increase, the scope of development is also expanding. The development environment is also more severe, the formation conditions are more complex, and overflows occur more frequently during drilling. If overflows are not discovered and controlled in time, they may further develop into well kicks or even blowouts, causing serious safety accidents. In the early stages of overflows, timely detection of overflows through intelligent technology can greatly improve drilling safety and further ensure safe and efficient development of oil and gas.

[0003] When an early overflow occurs, the formation pressure is greater than the bottom hole pressure, and the formation gas enters the wellbore. Generally, the riser pressure suddenly increases and then gradually decreases. However, other characterization parameters are delayed and will not respond at the same time. This problem of unclear change patterns will affect the ground personnel's reasoning and judgment of the overflow. At the same time, the riser pressure is affected by factors such as eccentric annulus, downhole power drilling tools, cuttings return, drilling pressure torque, and drilling displacement. It is always in a fluctuating state and has complex time-frequency signal characteristics, which makes it difficult for field personnel to identify the abnormal riser pressure caused by early overflow, greatly increasing the risk of drilling operations. In addition, since the downhole gas invasion parameters cannot be determined, the overflow warning model based on pure data drive ignores the characterization law of the multiphase flow gas invasion mechanism model, which is prone to false alarms. Summary of the invention

[0004] In view of the problem that the existing technology cannot accurately and timely detect drilling overflow, the present invention provides an intelligent early warning method for drilling overflow, which provides a new idea for early intelligent warning of overflow.

[0005] The intelligent early warning method for drilling overflow provided by the present invention comprises the following steps:

[0006] S1. Acquisition of data: The data are logging data related to the riser pressure change, including drilling pressure, torque, hook load, drilling time, rotation speed, pump stroke, inlet flow rate, outlet flow rate and other parameter data.

[0007] S2. Data decomposition: Decompose the logging data obtained in step S1 by using the EMD decomposition algorithm to obtain the IMF components of each parameter.

[0008] S3. KPCA dimensionality reduction is used to extract the IMF components that mainly affect the riser pressure.

[0009] S4. Using the INFORMER deep learning algorithm and combining it with the measured riser pressure, the IMF components are trained and learned to obtain a deep learning time series prediction model for the riser pressure under normal risk-free conditions.

[0010] S5. Use the current deep learning parameters to predict the riser pressure changes under the next stage of drilling conditions. Select the real-time data of normal drilling data, decompose each parameter into multiple IMF components through the EMD decomposition algorithm, and then use KPCA to achieve dimensionality reduction, retaining the IMF components with a cumulative contribution of more than 90%, and achieve accurate prediction of the riser pressure.

[0011] S6. Compare the deviation between the predicted and measured riser pressures. If the deviation between the predicted and measured riser pressures is continuously greater than 0.5 MPa for more than 2 minutes, it is considered that overflow has occurred and an alarm is issued. Intercept the curve that is different from the predicted riser pressure and use the early overflow riser pressure simulated by the multiphase flow model to perform quantitative analysis of the overflow risk.

[0012] The method for quantitative analysis of overflow risk is: setting different gas invasion rates, simulating different riser pressure change curves through multiphase flow model, and comparing the similarity between abnormal riser pressure curve and simulated riser pressure curve. If the similarity exceeds 80%, the gas invasion rate is considered to be the real gas invasion rate at the bottom of the well; using the gas invasion rate to simulate the specific situation and trend of overflow through multiphase flow model, and conduct quantitative analysis.

[0013] For the riser pressure simulation, the calculation formula is as follows:

[0014] P bottom =ρ l gH1+ρ m g(H-H1)

[0015] P pump =P bottom -ρ l HkDJ

[0016] Where: P bottom is the bottom hole pressure, MPa; P pump is the riser pressure, MPa; ρ l is the drilling fluid density, kg / m 3 ρ m is the density of the annular air-liquid mixture, kg / m 3 ; H is the vertical depth of the wellbore, m; H1 is the depth of the annular air-liquid mixing section, m; g is the gravitational acceleration, m / s 2 .

[0017] The multiphase flow model is as follows:

[0018] The gas mass conservation equation is as follows:

[0019]

[0020] When located in the overflow layer, the right side of the equation is q g ;

[0021] The liquid mass conservation equation is as follows:

[0022]

[0023] When it is at the drill bit, the right side of the equation is q l ;

[0024] The momentum conservation equation for gas-liquid two phases is:

[0025]

[0026] Where: q g is the gas intrusion rate, m 3 / s; A is the cross-sectional area of ​​the annulus, m 2 ; f is the volume fraction, dimensionless; ρ is the density, kg / m 3 ; v is the true velocity, m / s; θ is the well inclination, rad; x is the spatial step, m; t is the time step, s; q is the flow rate, m 3 / s; P is pressure, MPa; F is friction, MPa; subscript g represents gas phase, subscript l represents drilling fluid, and subscript m represents miscible phase.

[0027] The specific algorithm of the multiphase flow model is as follows:

[0028] The calculation steps of the dynamic process of gas invasion between any two nodes j and j+1 in the annulus from time t to time t+1 are as follows:

[0029] (1) Assume that the bottom hole pressure at the bottom hole node at time t+1 is

[0030] (2) Based on The gas invasion rate equation is used to calculate the bottom hole gas invasion rate;

[0031] (3) Randomly set the pressure of node j at time t+1

[0032] (4) Calculate density, viscosity and coefficient of deviation based on the real gas state equation;

[0033] (5) Assume that the gas holding rate of node j at time t+1 is

[0034] (6) Calculate the gas-liquid phase velocity based on the difference result of the continuity equation and

[0035] (7) Calculate the gas holdup rate according to the gas holdup rate equation like Continue to the next step of calculation; otherwise, return to step (5) to readjust the gas holdup rate Until satisfied conditions;

[0036] (8) Substitute the difference result into the motion equation and calculate if The calculation of node j is completed. Continue to use the parameters calculated at node j as the known conditions of point j+1. Otherwise, return to step (3) to adjust the pressure. Recalculate until all calculation conditions are met;

[0037] (9) When calculating to the wellhead, determine whether the calculation result meets the wellhead boundary conditions. If so, proceed to the next moment calculation; otherwise, return to step (1) and recalculate until the calculation conditions are met.

[0038] Compared with the prior art, the present invention is beneficial in that:

[0039] The present invention determines the main overflow warning parameters for the main working conditions where overflow occurs, and based on a large amount of normal drilling working condition data, uses the nonlinear fitting ability of the deep learning algorithm to establish a data-driven model for deep learning time series prediction. By accurately predicting the drilling riser pressure, the abnormal nodes of the riser pressure of early overflow can be discovered in time, and the alarm can be issued in time. The mathematical model of transient multiphase drift flow in the wellbore is used to simulate the change law of the riser pressure affected by gas invasion under different formation parameters, and quantitatively evaluate the early overflow risk. By constructing a hybrid-driven early warning method of deep learning time series and multiphase flow gas invasion model, technical means and theoretical support are provided for early overflow risk warning.

[0040] Other advantages, objectives and features of the present invention will be embodied in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flow chart of the intelligent early warning method for drilling overflow provided by the present invention.

[0042] Figure 2 Decomposition of the main parameters affecting the riser pressure in Well C (taking the hook load as an example).

[0043] Figure 3 The main IMF components affecting the standpipe pressure extracted by EMD-KPCA for Well C.

[0044] Figure 4Flowchart for deep learning time series forecasting in INFORMER.

[0045] Figure 5 Comparison between the predicted and measured results of the riser pressure in Well C.

[0046] Figure 6 Identify the abnormal decrease in riser pressure in Well C.

[0047] Figure 7 The simulation results of the riser pressure change under different gas invasion velocities in Well C. DETAILED DESCRIPTION

[0048] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0049] like Figure 1-7 As shown, the intelligent early warning method for well overflow provided by the present invention takes Well C as the research object and introduces the method steps in detail:

[0050] S1. Get data

[0051] Obtain logging data related to riser pressure changes, including drilling pressure, torque, hook load, drilling time, speed, pump stroke, inlet flow, outlet flow and other parameter data.

[0052] S2. Data decomposition

[0053] The logging data related to the standpipe pressure were decomposed by EMD to obtain the IMF components of each parameter.

[0054] During the decomposition process, if h1(t) satisfies the IMF condition, it is the first IMF component of the original signal x(t); if it does not satisfy the IMF condition, h1(t) is taken as the original signal and the following steps (1) to (3) are repeated until the IMF condition is satisfied. The first IMF component h1 of x(t) can be obtained. 1k (t), denoted as c1(t).

[0055] Step (1): Use the standard deviation S of the last two iterations to calculate the evaluation decomposition result. The threshold S0 is usually set to 0.2-0.3.

[0056] If S is less than the threshold S0, then h is considered to be satisfied. 1k (t) is an IMF, otherwise continue the iteration.

[0057]

[0058] Step (2): Separate the first IMF component from the original signal x(t) to obtain the residual component r1(t);

[0059] r1(t)=x(t)-c1(t)

[0060] Step (3): Take the residual component r1(t) as the new original signal and repeat steps (1)-(4) to obtain the second IMF component c2(t). Repeat the cycle n times until the last participating component r n (t) is a constant or a monotonic function, and n IMF components of the signal x(t) are obtained.

[0061]

[0062] Step (4): The original signal x(t) is:

[0063]

[0064] Taking the hook load as an example, the main parameters affecting the riser pressure of Well C are decomposed as follows: Figure 2 .

[0065] S3. Extract the IMF components that mainly affect the riser pressure:

[0066] KPCA is used to reduce the dimension and extract the IMF components that mainly affect the riser pressure. Although the IMF components obtained by decomposing the original data can fully reflect the characteristics of the original data, there is still a lot of redundant information, which seriously affects the subsequent prediction effect. KPCA is widely used in data dimensionality reduction operations. It can reduce the data dimension and eliminate the influence of redundant elements, while extracting key elements from massive data and retaining the key features of the data set.

[0067] KPCA first maps nonlinear data into feature space, and then performs principal component analysis in the feature space to extract the main components. In the present invention, the Ploynomial Kernel function is used to reduce the dimension of the EMD component:

[0068] k(x, y) = (1 +<x,y> ) d

[0069] The specific steps are as follows:

[0070] Step (1): Obtain IMF components. Assume that the original data is X = [x1, x2, x3, ···x n ]∈R m×n .

[0071] Step (2): Polynomial kernel calculation, through the mapping function Mapping the data to the feature space, the covariance matrix of the feature space data is:

[0072]

[0073] Step (3): Calculate the kernel matrix. From the matrix eigenvalue decomposition relationship, we know that the eigenvalues ​​and eigenvectors of the covariance matrix must satisfy:

[0074] λV=CV

[0075] Step (4): Centering the kernel matrix. Since C is unknown, the eigenvalues ​​and eigenvectors in the above formula cannot be directly calculated. This can be solved by introducing a nonlinear transformation. Multiply both sides of the above formula by as follows:

[0076]

[0077] Where V is the eigenvector of all λ≠0, and all exist in the feature space. Therefore, there is a coefficient a i (i=1, 2, ..., n) such that:

[0078]

[0079] The known kernel matrix can be further expressed as:

[0080]

[0081] Step (5): Eigenvalue decomposition and normalization. Convert the eigenvalue and eigenvector of C into the eigenvalue and eigenvector of the kernel matrix.

[0082] nλa=Kα

[0083] Where a is the feature vector. After calculation, the required data mapping result is as follows:

[0084]

[0085] Where: P is the number of selected eigenvalues.

[0086] Step (6): Select the principal component IMF component based on the cumulative contribution. The value of P can be determined based on the contribution rate, T 2 Statistics and SPE statistics, the specific calculation process is as follows:

[0087] T 2 =t T A -1 t

[0088] Where: t is the eigenvector; A is the matrix composed of eigenvalues ​​selected according to the sorting.

[0089]

[0090] Where n0 is the number of non-zero eigenvalues.

[0091] The main IMF components affecting the standpipe pressure extracted by EMD-KPCA in Well C are as follows: Figure 3 shown.

[0092] S4. Establish a riser pressure prediction model

[0093] The Informer deep learning algorithm is used to train and learn the main components in combination with the measured riser pressure, and a deep learning time series prediction model for the riser pressure under normal risk-free conditions is obtained. Figure 4 As shown in the figure, the core of the Informer model is the self-attention mechanism, which calculates the importance of each element in the sequence by weighted summation of each element in the sequence. Unlike the traditional self-attention mechanism, the Informer model uses an adaptive length attention mechanism, which can dynamically adjust the attention range according to the sequence length. In addition, the Informer model also uses a multi-scale time encoder and decoder, allowing the model to capture information on different time scales at the same time. The steps for the Informer model to predict time series are as follows:

[0094] Step (1): Data preprocessing: Preprocess the input time series data, including missing value processing, normalization and other operations.

[0095] Step (2): Multi-scale time encoding: Input the preprocessed data into the multi-scale time encoder to obtain feature representations at different time scales. The Informer encoder is as follows:

[0096]

[0097] Step (3): Adaptive length attention calculation: The output of the multi-scale time encoder is used as the input of the self-attention mechanism, and the importance of each element is calculated through the adaptive length attention mechanism. The sparse self-attention (ProSparse Self-attention) of the Informer model filters out the most important query and reduces the complexity of the calculation. It is necessary to input three matrices (Query, Key, Value), and then perform standardized point multiplication, that is:

[0098]

[0099] Step (4): Decoding and prediction: Input the output of the adaptive length attention mechanism into the decoder to obtain the final prediction result. The decoder of the Informer model is as follows:

[0100]

[0101] In the formula, is a marked start sequence, is the target prediction sequence, with an initial value of 0.

[0102] S5. Predict riser pressure changes

[0103] The current deep learning parameters are used to predict the riser pressure changes under the next stage of drilling conditions. The real-time data of normal drilling data, such as drilling pressure, torque, hook load, outlet flow, inlet flow, pump strokes 1, 2 and turntable speed, are selected. Each parameter is decomposed into multiple IMF components through the EMD decomposition algorithm, and then the dimension is reduced through KPCA, retaining the IMF components with a cumulative contribution of more than 90%, achieving accurate prediction of the riser pressure.

[0104] S6. Quantitative analysis of overflow risk

[0105] Early overflow identification: Compare the deviation between the predicted and measured riser pressures. If the deviation between the predicted and measured riser pressures is continuously greater than 0.5 MPa and lasts for more than 2 minutes, it is considered that early overflow has occurred and an alarm is issued. Figure 5 The comparison between the predicted result and the measured result of the riser pressure of Well C. The example shows that the early warning time of the well using the EMD-KPCA-INFORMER deep learning time series algorithm combined with the multiphase gas invasion model was 18:10, which was 23 minutes earlier than the 18:33 on site.

[0106] Quantitative analysis of early overflow: intercept abnormal riser pressure curves that are different from the predicted riser pressure, and use the early overflow riser pressure simulated by the multiphase flow model to perform quantitative analysis of overflow risk. Set different gas invasion rates, and use the multiphase flow model to simulate different riser pressure change curves. By comparing the similarity of the abnormal riser pressure curve and the simulated riser pressure curve, if the similarity exceeds 80%, it is considered that the gas invasion rate is the real gas invasion rate at the bottom of the well. Figure 6 This is the abnormal reduction in the riser pressure of Well C. The gas invasion rate can be used to simulate the specific situation and trend of the overflow through the multiphase flow model and conduct quantitative analysis.

[0107] For the riser pressure simulation: For the extensive pressure model of the entire wellbore, the wellhead casing pressure during the early overflow period is atmospheric pressure, and there are pure drilling fluid sections and gas-liquid mixed sections in the wellbore annulus. bottom The calculation formula is as follows:

[0108] P bottom =ρ l gH1+ρ m g(H-H1)

[0109] Since the drill pipe is in pure liquid phase during overflow, the riser pressure can be further solved from the bottom hole pressure. The riser pressure calculation formula is as follows:

[0110] P pump =P bottom -ρ l gH=ρ l gH1+ρ m g(H-H1)-ρ l HkDJ

[0111] Where: P bottom is the bottom hole pressure, MPa; P pump is the riser pressure, MPa; ρ l is the drilling fluid density, kg / m 3 ρ m is the density of the annular air-liquid mixture, kg / m 3 ; H is the vertical depth of the wellbore, m; H1 is the depth of the annular air-liquid mixing section, m; g is the gravitational acceleration, m / s 2 .

[0112] The simulation results of the riser pressure change under different gas invasion velocities in Well C are shown in Figure 7 .

[0113] Since the wellbore pressure changes transiently, in order to accurately calculate the pressure, it is necessary to combine the multiphase flow model to describe the transient changes. The multiphase flow model is as follows:

[0114] The gas mass conservation equation is as follows (the right side of the equation is q when it is in the overflow layer g ):

[0115]

[0116] The liquid mass conservation equation is as follows (the right side of the equation is q when it is at the drill bit l ):

[0117]

[0118] The momentum conservation equation for gas-liquid two phases is:

[0119]

[0120] Where: q g is the gas intrusion rate, m 3 / s; A is the cross-sectional area of ​​the annulus, m 2 ; f is the volume fraction, dimensionless; ρ is the density, kg / m 3 ; v is the true velocity, m / s; θ is the well inclination, rad; x is the spatial step, m; t is the time step, s; q is the flow rate, m 3 / s; P is pressure, MPa; F is friction, MPa; subscript g represents gas phase, subscript l represents drilling fluid, and subscript m represents miscible phase.

[0121] Gas will slip toward the wellhead relative to the liquid in the wellbore, which is described by the drift flow model. The true velocity of the gas in the drift flow model is calculated by the following formula:

[0122] v g =C0v m +v d

[0123] The gas-liquid mixing velocity is calculated as follows:

[0124] v m =v g f g +v l f l

[0125] Gas distribution coefficient:

[0126]

[0127] Gas slip velocity:

[0128]

[0129] Where: C0 is the gas phase distribution coefficient; v m is the mixing speed, m / s; v d is the drift velocity, m / s; B is the profile parameter, which is 1.2; γ is the profile parameter restriction item, note that γ should be between 0 and 1; K(f g ) is the critical Kutateladze number; v c is the characteristic velocity, m / s; m(φ) represents the effect of well inclination on drift velocity.

[0130] The specific algorithm of the multiphase flow model is as follows:

[0131] The calculation steps of the dynamic process of gas invasion between any two nodes j and j+1 in the annulus from time t to time t+1 are as follows (the parameters of nodes j and j+1 at time n are known):

[0132] (1) Assume that the bottom hole pressure at the bottom hole node at time t+1 is

[0133] (2) Based on The gas invasion rate equation is used to calculate the bottom hole gas invasion rate;

[0134] (3) Randomly set the pressure of node j at time t+1

[0135] (4) Calculate density, viscosity and coefficient of deviation based on the real gas state equation;

[0136] (5) Assume that the gas holding rate of node j at time t+1 is

[0137] (6) Calculate the gas-liquid phase velocity based on the difference result of the continuity equation and

[0138] (7) Calculate the gas holdup rate according to the gas holdup rate equation like Continue to the next step of calculation; otherwise, return to step (5) to readjust the gas holdup rate Until satisfied ε is the error requirement, which is 0.01 in this embodiment;

[0139] (8) Substitute the difference result into the motion equation and calculate if The calculation of node j is completed. Continue to use the parameters calculated at node j as the known conditions of point j+1. Otherwise, return to step (3) to adjust the pressure. Recalculate until all calculation conditions are met;

[0140] (9) When calculating to the wellhead, determine whether the calculation result meets the wellhead boundary conditions. If so, proceed to the next moment calculation; otherwise, return to step (1) and recalculate until the calculation conditions are met.

[0141] The above description is only a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment as above, it is not intended to limit the present invention. Any technician familiar with the profession can make some changes or modifications to equivalent embodiments of equivalent changes using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of the technical solution of the present invention.

Claims

1. An intelligent early warning method for drilling overflow, characterized in that: The following steps are involved: S1. Acquiring data: The data are logging data related to the riser pressure change, including drilling pressure, torque, hook load, drilling time, rotation speed, pump stroke, inlet flow rate and outlet flow rate; S2, data decomposition: decomposing the logging data obtained in step S1 by using the EMD decomposition algorithm to obtain the IMF components of each parameter; S3, extract the IMF components that mainly affect the standpipe pressure through KPCA dimensionality reduction; S4. Use the INFORMER deep learning algorithm and the measured riser pressure to train and learn the IMF components, and obtain a deep learning time series prediction model for the riser pressure under normal risk-free conditions; S5. Use the current deep learning parameters to predict the riser pressure change under the next stage of drilling conditions; S6. Compare the deviation between the predicted and measured riser pressures. If the deviation between the predicted and measured riser pressures is continuously greater than 0.5 MPa for more than 2 minutes, it is considered that overflow has occurred and an alarm is issued. Intercept a curve different from the predicted riser pressure and use the early overflow riser pressure simulated by the multiphase flow model to conduct a quantitative analysis of the overflow risk.

2. The intelligent early warning method for drilling overflow according to claim 1, characterized in that: In step S5, real-time data of normal drilling data is selected, and each parameter is decomposed into multiple IMF components by using the EMD decomposition algorithm. Then, dimension reduction is achieved by KPCA, and the IMF components with cumulative contributions of more than 90% are retained for accurate prediction of the standpipe pressure.

3. The intelligent early warning method for drilling overflow according to claim 1, characterized in that: In step S6, the method for quantitative analysis of overflow risk is: setting different gas invasion rates, simulating different riser pressure change curves through the multiphase flow model, and comparing the similarity between the abnormal riser pressure curve and the simulated riser pressure curve. If the similarity exceeds 80%, it is considered that the gas invasion rate is the actual gas invasion rate at the bottom of the well; using the gas invasion rate to simulate the specific situation and trend of overflow through the multiphase flow model, and perform quantitative analysis.

4. The intelligent early warning method for drilling overflow according to claim 3, characterized in that: In step S6, the riser pressure simulation calculation formula is as follows: P bottom =ρ l gH1+p m g(H-H1) P pump =P bottom -ρ l gH Where: P bottom is the bottom hole pressure, MPa; P pump is the riser pressure, MPa; ρ l is the drilling fluid density, kg / m 3 ; ρ m is the density of the annular air-liquid mixture, kg / m 3 ; H is the vertical depth of the wellbore, m; H1 is the depth of the annular air-liquid mixing section, m; g is the gravitational acceleration, m / s 2 .

5. The intelligent early warning method for drilling overflow according to claim 4, characterized in that: The multiphase flow model is as follows: The gas mass conservation equation is as follows: When located in the overflow layer, the right side of the equation is q g ; The liquid mass conservation equation is as follows: When it is at the drill bit, the right side of the equation is q l ; The gas-liquid two-phase momentum conservation equation is: Where: q g is the gas intrusion rate, m 3 / s; A is the cross-sectional area of ​​the annulus, m 2 ; f is the volume fraction, dimensionless; ρ is the density, kg / m 3 ; v is the true speed, m / s; θ is the well inclination angle, rad; x is the spatial step length, m; t is the time step, s; q is the flow rate, m 3 / s; P is pressure, MPa; F is friction, MPa; subscript g represents gas phase, subscript l represents drilling fluid, and subscript m represents miscible phase.

6. The intelligent early warning method for drilling overflow according to claim 5, characterized in that: The specific algorithm of the multiphase flow model is as follows: The calculation steps of the dynamic process of gas invasion between any two nodes j and j+1 in the annulus from time t to time t+1 are as follows: (1) Assume that the bottom hole pressure at the bottom hole node at time t+1 is (2) Based on The gas invasion rate equation is used to calculate the bottom hole gas invasion rate; (3) Randomly set the pressure of node j at time t+1 (4) Calculate density, viscosity and coefficient of deviation based on the real gas state equation; (5) Assume that the gas holding rate of node j at time t+1 is (6) Calculate the gas-liquid phase velocity based on the difference result of the continuity equation and (7) Calculate the gas holdup rate according to the gas holdup rate equation like Continue to the next step of calculation; otherwise, return to step (5) to readjust the gas holdup rate Until satisfied conditions; (8) Substitute the difference result into the motion equation and calculate if The calculation of node j is completed. Continue to use the parameters calculated at node j as the known conditions of point j+1. Otherwise, return to step (3) to adjust the pressure. Recalculate until all calculation conditions are met; (9) When calculating to the wellhead, determine whether the calculation result meets the wellhead boundary conditions. If so, proceed to the next moment calculation; Otherwise, return to step (1) and recalculate until the calculation condition is met.

Images