Multi-working-condition well bottom pressure prediction method and device, computer device and storage medium

Abstract

The present application relates to the field of oil and gas exploration, and particularly relates to a multi-working-condition well bottom pressure prediction method and device, computer equipment and storage medium. The method comprises the following steps: acquiring a real-time working condition according to real-time logging data; determining a well bottom pressure prediction model of the real-time working condition, wherein the well bottom pressure prediction model of each working condition is trained based on historical input feature data related to pressure of each working condition, historical well bottom pressure data of each working condition and constraint conditions of each working condition, the constraint conditions of each working condition are established according to the correlation between the historical input feature data and the historical well bottom pressure data of each working condition and / or annular gas-liquid-solid three-phase flow mechanism; determining input feature data of the real-time working condition according to the real-time logging data; inputting the input feature data of the real-time working condition into the well bottom pressure prediction model of the real-time working condition to obtain the well bottom pressure.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of oil and gas exploration, and particularly to a multi-working-condition well bottom pressure prediction method and device, computer equipment and a storage medium. BACKGROUND

[0002] With the deepening of oil and gas exploration and development, oil and gas exploration and development gradually extend to deep, ultra-deep and low-permeability unconventional oil and gas resources.

[0003] The existing wellbore pressure prediction is based on an annular multiphase flow mechanism model. The model has high complexity, long calculation time and insufficient accuracy under complex geological and engineering conditions. On the other hand, in the high temperature and high pressure working environment, the well bottom pressure is measured by a downhole measuring device, and the measurement cost is high.

[0004] The existing wellbore pressure intelligent prediction model is mostly a machine learning or deep learning model established by directly combining intelligent algorithms with a large amount of geological and engineering data. Such a model is greatly affected by the quality of training data. Although it can improve the average accuracy of well bottom pressure prediction to some extent, the prediction result has large fluctuation and many abnormal values, and the stability and generalization performance of the model still need to be improved.

[0005] In view of the problems of large fluctuation of wellbore pressure prediction result and high measurement cost in the prior art, there is an urgent need to research a multi-working-condition well bottom pressure prediction method. SUMMARY

[0006] To solve the above problems of the prior art, the embodiments of the present application provide a multi-working-condition well bottom pressure prediction method. By adding mechanism constraints, the deep integration of algorithms and mechanisms can be realized, the prediction result can be made more consistent with the change mechanism of the prediction target while ensuring the prediction accuracy of the neural network model, the abnormal values and fluctuation values of the prediction result can be reduced, and the stability and generalization ability of the neural network model can be effectively improved. It provides technical support and important reference for efficient prediction and fine control of well bottom pressure under deep high temperature and high pressure drilling gas invasion conditions.

[0007] The embodiments of the present application provide a multi-working-condition well bottom pressure prediction method, comprising: acquiring a real-time working condition according to real-time logging data; determining a well bottom pressure prediction model of the real-time working condition, wherein the well bottom pressure prediction model of each working condition is trained based on historical input feature data related to pressure of each working condition, historical well bottom pressure data of each working condition and constraint conditions of each working condition, the constraint conditions of each working condition are established according to the correlation of historical input feature data and historical well bottom pressure data of each working condition and / or annular gas-liquid-solid three-phase flow mechanism; determining input feature data of the real-time working condition according to the real-time logging data; inputting the input feature data of the real-time working condition into the well bottom pressure prediction model of the real-time working condition to predict the well bottom pressure.

[0008] According to an aspect of the embodiments herein, the training process of the bottom hole pressure prediction model for each condition comprises: obtaining a historical sample data set for each condition, the historical sample data set for each condition comprising a plurality of sample data, each sample data comprising historical input feature data and historical bottom hole pressure data; establishing an equality constraint condition for each condition according to the annulus gas-liquid-solid three-phase flow mechanism; establishing an inequality constraint condition for each condition according to the correlation between the historical input feature data and the corresponding historical bottom hole pressure data for each condition; constructing a loss function for each condition according to the historical sample data set for each condition, the equality constraint condition and the inequality constraint condition for each condition; training the parameters in the neural network model using the loss function for each condition, and taking the trained neural network model for each condition as the bottom hole pressure prediction model for each condition.

[0009] According to an aspect of the embodiments herein, the establishing of the inequality constraint condition for each condition according to the correlation between the historical input feature data and the corresponding historical bottom hole pressure data for each condition comprises: for each condition, calculating the correlation between each input feature data in the historical input feature data and the historical bottom hole pressure data for the corresponding condition; taking the input feature data with a correlation lower than a first preset threshold as a non-sensitive parameter for the condition, and taking the input feature data with a correlation higher than the first preset threshold as a sensitive parameter for the condition; and establishing the inequality constraint condition for the condition according to the sensitive parameter and the non-sensitive parameter for the condition.

[0010] According to an aspect of the embodiments herein, for each condition, after the sensitive parameters for the condition are determined, the method further comprises: taking any two sensitive parameters for the condition as a sensitive parameter group, and calculating the correlation between the sensitive parameters in each sensitive parameter group; screening out the sensitive parameter groups with a correlation greater than a second preset threshold; and deleting one sensitive parameter from each sensitive parameter group screened out.

[0011] According to an aspect of the embodiments herein, after the historical bottom hole pressure data set for each condition is obtained, the method further comprises: estimating a bottom hole pressure range for each condition according to an annulus pressure calculation formula; and deleting the sample data in the historical sample data set for each condition that does not satisfy the bottom hole pressure range.

[0012] According to an aspect of the embodiments herein, the establishing of the inequality constraint condition for each condition according to the sensitive parameter and the non-sensitive parameter for each condition comprises:

[0013] The inequality constraint condition is established using the following formula:

[0014]

[0015] Wherein, x represents sensitive parameters and non-sensitive parameters of each working condition, P and F(x, w, b) both represent wellbore pressure expression output by the neural network model, w represents weights of each layer in the neural network, and b represents bias of each layer in the neural network.

[0016] According to an aspect of the embodiment, the equation constraint conditions of each working condition are established according to the annulus air-liquid-solid three-phase flow mechanism, which includes:

[0017] The equation constraint conditions are established by using the following formula:

[0018]

[0019]

[0020] So that A = A theory ; wherein A theory is a momentum conservation equation in the annulus air-liquid-solid three-phase flow mechanism, z represents the fixed vertical depth of each working condition, P and F(x, w, b) both represent wellbore pressure expression predicted by the neural network model; t represents time; m represents one of the three components of gas, liquid and solid; p m represents density; a m represents average flow velocity; V m represents volume fraction of gas, liquid and solid; cosθ represents inclination angle; g represents gravitational acceleration; F f represents friction between the annulus and the drilling fluid, w represents weights of each layer in the neural network, and b represents bias of each layer in the neural network.

[0021] The embodiment also provides a wellbore pressure prediction device for multiple working conditions, which includes: a working condition acquisition unit, configured to acquire a real-time working condition according to real-time logging data;

[0022] a wellbore pressure prediction model determination unit, configured to determine a wellbore pressure prediction model of the real-time working condition, wherein the wellbore pressure prediction model of each working condition is trained based on historical input feature data related to pressure of each working condition, historical wellbore pressure data of each working condition and constraint conditions of each working condition, the constraint conditions of each working condition are established according to correlation between the historical input feature data and the historical wellbore pressure data of each working condition and / or an annulus air-liquid-solid three-phase flow mechanism; and a wellbore pressure prediction unit, configured to input input features of the real-time working condition into the wellbore pressure prediction model of the real-time working condition, and predict wellbore pressure.

[0023] The embodiment also provides a computer device, which includes a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the above method when executing the computer program.

[0024] The embodiments of the present application also provide a computer readable storage medium, which stores computer instructions, and the computer instructions are executed by a processor to implement the method described above.

[0025] With the embodiments of the present application, the deep integration of the algorithm and the mechanism can be realized by adding the mechanism constraint method, the prediction result is more in line with the change mechanism of the prediction target while ensuring the prediction accuracy of the neural network model, the abnormal value and the fluctuation value of the prediction result are reduced, and the stability and the generalization ability of the neural network model are effectively improved. Technical support and important reference are provided for efficient prediction and fine regulation of bottom hole pressure under deep high temperature and high pressure gas invasion condition. BRIEF DESCRIPTION OF DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0027] Figure 1 A flow chart of a multi-condition bottom hole pressure prediction method is shown;

[0028] Figure 2 A flow chart of a multi-condition bottom hole pressure prediction method is shown;

[0029] Figure 3 A flow chart of a multi-condition bottom hole pressure prediction method is shown;

[0030] Figure 4 A flow chart of a multi-condition bottom hole pressure prediction method is shown;

[0031] Figure 5 A structure schematic diagram of a multi-condition bottom hole pressure prediction device is shown;

[0032] Figure 6 A structure schematic diagram of a multi-condition bottom hole pressure prediction device is shown;

[0033] Figure 7 A structure schematic diagram of a multi-condition bottom hole pressure prediction device is shown;

[0034] Figure 8 A structure schematic diagram of a multi-condition bottom hole pressure prediction device is shown;

[0035] BRIEF DESCRIPTION OF DRAWINGS

[0036] 501, condition acquisition unit;

[0037] 5011, working condition recognition rule determination module;

[0038] 502, well bottom pressure prediction model determination unit;

[0039] 5021, historical sample data set acquisition module;

[0040] 5022, historical sample data set screening module;

[0041] 5023, equality constraint establishment module;

[0042] 5024, inequality constraint establishment module;

[0043] 5025, loss function establishment module;

[0044] 503, input feature data determination unit;

[0045] 5031, correlation calculation module;

[0046] 5032, sensitive parameter determination module;

[0047] 5033, non-sensitive parameter determination module;

[0048] 5034, sensitive parameter group screening module;

[0049] 5035, sensitive parameter deletion module;

[0050] 504, well bottom pressure prediction unit;

[0051] 802, computer device;

[0052] 804, processor;

[0053] 806, memory;

[0054] 808, drive mechanism;

[0055] 810, input / output module;

[0056] 812, input device;

[0057] 814, output device;

[0058] 816, presentation device;

[0059] 818, graphical user interface;

[0060] 820, network interface;

[0061] 822, communication link;

[0062] 824, communication bus. DETAILED DESCRIPTION

[0063] In order to make the technical personnel in the art better understand the technical solutions in the specification, the technical solutions in the specification will be clearly and completely described below in combination with the drawings in the embodiments. Obviously, the described embodiments are only part of the embodiments of the specification, not all. Based on the embodiments herein, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection.

[0064] It should be noted that the terms "first", "second" and the like in the specification and claims of the specification and the above drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a chronological order. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the specification described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, device, product or equipment including a series of steps or units need not be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or equipment.

[0065] The specification provides method operation steps as described in the embodiments or flowcharts, but can include more or less operation steps based on routine or non-creative labor. The order of steps listed in the embodiments is only one of the many step execution orders, and does not represent the only execution order. In actual system or device product execution, the method order shown in the embodiments or drawings can be executed in sequence or in parallel.

[0066] It should be noted that the multi-working-condition well bottom pressure prediction method and device herein can be used in the field of oil and gas exploration, and the application field of the multi-working-condition well bottom pressure prediction method and device herein is not limited.

[0067] As shown in the flowchart of the multi-working-condition well bottom pressure prediction method of the embodiments herein, the method specifically includes the following steps: Figure 1 As shown in the flowchart of the multi-working-condition well bottom pressure prediction method of the embodiments herein, the method specifically includes the following steps:

[0068] Step 101, obtaining real-time working conditions according to real-time logging data.

[0069] In some embodiments of the present specification, the mud logging data includes various types of parameters such as well depth, vertical depth, bit depth, bit vertical depth, drilling time, drilling pressure, hook load, rotary speed, torque, gauge, hook position, hook speed, standpipe pressure, casing pressure, pump stroke 1, pump stroke 2, pump stroke 3, total pit volume arrival time, mud overflow, inlet / outlet flow, inlet / outlet density, inlet / temperature, total hydrocarbon, H2S content, C1 / C2 content, displacement volume, uphole depth, etc. Among them, the real-time mud logging data is measured and obtained by field mud logging instruments, downhole sensors, broken drill measuring devices and other equipment. Different real-time mud logging data corresponds to different real-time working conditions. The mud logging data of different working conditions may present different change trends with the change of time or the progress of drilling process. For example, it presents an increasing trend with time, a decreasing trend with time, or tends to be stable with time without obvious change. Therefore, the real-time working condition can be determined according to the change of real-time mud logging data. Specifically, according to theoretical analysis and actual field experience, the change law of wellhead pressure, pipe column movement, hook load, drill disc torque and other parameters under different working conditions is summarized to form the identification rule of drilling working condition, and further to form the working condition identification program. The working condition identification program can be described in Table 1.

[0070] Table 1: Drilling working condition identification rule

[0071]

[0072] Table 1 is an identification rule of a drilling working condition according to an embodiment of the present text. Table 1 shows the corresponding relationship between the working condition determined according to the change trend of the input feature data and the input feature data.

[0073] Seven working conditions are shown in Table 1, including tripping in, tripping out, reaming, reverse reaming, circulation, drilling, and stopping drilling, and the corresponding input feature data includes hook load, drilling pressure, rotary speed, torque, pump pressure, bit depth variable, well depth variable, etc. Among them, the numbers > 0, < 0 in the table represent the change trend of the input feature data. For example, in the drilling working condition, the hook load > 0 indicates that the hook load presents an increasing trend in this working condition. For example, in the reverse reaming working condition, the data corresponding to the bit depth variable is < 0, indicating that the bit depth variable presents a decreasing trend in this working condition. For example, in the stopping drilling working condition, the data corresponding to the drilling pressure is 0, indicating that the drilling pressure is substantially unchanged in the stopping drilling working condition.

[0074] Based on this and the change trend of a large amount of input feature data under different working conditions, the identification rule of each working condition can be formed. According to the change trend of multiple input feature data, the working condition corresponding to the current input feature data can be judged, so as to realize the working condition identification.

[0075] In practical applications, real-time mud logging data obtained by real-time measurement in the field is input into the determined working condition identification program to determine the corresponding real-time working condition. In some embodiments of the present specification, the real-time working condition corresponding to the above-mentioned multiple mud logging data can be determined as the drilling working condition according to the increasing trend of multiple real-time mud logging data such as hook load, drilling pressure, rotating speed, torque, pump pressure, drill bit depth variable, well depth variable, etc. with time.

[0076] Step 102, determining the bottom hole pressure prediction model of the real-time working condition, wherein the bottom hole pressure prediction model of each working condition is trained based on the historical input feature data related to pressure of each working condition, the historical bottom hole pressure data of each working condition and the constraint condition of each working condition, and the constraint condition of each working condition is established according to the correlation between the historical input feature data and the historical bottom hole pressure data of each working condition and / or the annular air-liquid-solid three-phase flow mechanism.

[0077] In some embodiments of the present specification, each working condition corresponds to a respective bottom hole pressure prediction model, and the bottom hole pressure prediction model of each working condition is stored in a database. When step 102 is implemented, the bottom hole pressure prediction model corresponding to the real-time working condition is found from the database according to the working condition determined in step 101.

[0078] The historical input feature data related to pressure and the historical bottom hole pressure data in the mud logging data of each working condition are screened, and then the bottom hole pressure prediction model of each working condition is trained according to the constraint condition of each working condition. The initial model of the bottom hole pressure prediction model is a neural network model.

[0079] The constraint condition of each working condition includes an equality constraint condition and an inequality constraint condition, and the parameters of the model are constrained during the model training process according to the change rule of the mud logging data during the drilling process. The inequality constraint is established according to the correlation between the historical input feature data and the historical bottom hole pressure data of each working condition, and the equality constraint is established according to the annular air-liquid-solid three-phase flow mechanism.

[0080] Step 103, determining the input feature data of the real-time working condition according to the real-time mud logging data.

[0081] In this step, the real-time mud logging data obtained by the downhole sensor and other measuring devices in real time may have noise, i.e., the real-time mud logging data collected in each working condition may have errors. As described in step 201, the historical input feature data related to pressure in each working condition can be determined after screening the historical mud logging data. Similarly, the historical input feature related to pressure in the real-time working condition can be determined after screening the implementation mud logging data. The screening of the real-time mud logging data can be referred to step 201.

[0082] Step 104, input the input feature data of the real-time working condition to the bottom hole pressure prediction model corresponding to the real-time working condition, and predict the bottom hole pressure. Each real-time working condition corresponds to a respective bottom hole pressure prediction model, and the training method of the bottom hole pressure prediction model is described in detail in Figure 2 The embodiment can realize deep integration of the algorithm and the mechanism by adding the mechanism constraint method, ensure the prediction accuracy of the neural network model, make the prediction result more consistent with the change mechanism of the prediction target, reduce the abnormal value and fluctuation value of the prediction result, and effectively improve the stability and generalization ability of the neural network model.

[0083] As shown in Figure 2 Fig. 1 is a flow chart of a bottom hole pressure prediction model training method for each working condition according to an embodiment of the present application. Specifically, the method comprises the following steps:

[0084] Step 201, obtain a historical sample data set for each working condition, and each historical sample data set for a working condition comprises a plurality of sample data, and each sample data comprises historical input feature data and historical bottom hole pressure data. In this step, each working condition has a respective historical sample data set, which includes a plurality of sample data, and the historical input feature data and the historical bottom hole pressure data of each working condition correspond to each other. In the model training process, the historical bottom hole pressure data corresponds to the label of the historical input feature data. The historical sample data set of each working condition is the historical data obtained by using downhole sensors and other measuring devices in historical drilling exploration tests.

[0085] Taking the drilling working condition as an example, the historical sample data set of the drilling working condition includes a plurality of historical input feature data and corresponding historical bottom hole pressure data. The historical input feature data includes fixed-point vertical depth, drilling fluid discharge, drilling fluid density, back pressure pump flow, funnel viscosity, drill bit depth, rotational speed, standpipe pressure, outlet flow, outlet density, sand content, etc. The historical bottom hole pressure data of the drilling working condition is the bottom hole pressure value corresponding to the plurality of historical input feature data.

[0086] In some other embodiments of the present application, the working conditions include, but are not limited to, sand washing, overflow pressure control, reverse reaming, drilling stop, drilling start, drilling start, and the like.

[0087] As described above, the historical bottom hole pressure data obtained by using the downhole sensor and other measuring devices may have noise, i.e., the historical bottom hole pressure data of each working condition may have errors. Therefore, in some other embodiments of the present specification, after obtaining the historical bottom hole pressure data set of each working condition, the following steps are further included: estimating the bottom hole pressure range of each working condition according to the annular pressure calculation formula; and deleting the sample data in the historical sample data set of each working condition that does not meet the bottom hole pressure range. The sample data that does not meet the bottom hole pressure range includes the historical bottom hole pressure data in the historical input feature data set. This step can be regarded as data processing of the historical bottom hole pressure data set. After estimating the approximate bottom hole pressure range of each working condition according to the annular pressure calculation formula, the historical bottom hole pressure data in the historical sample data set that does not meet the bottom hole pressure range estimated according to the annular pressure calculation formula can be deleted, or the historical input feature data that does not meet the bottom hole pressure range can be deleted.

[0088] In step 202, the equation constraint condition of each working condition is established according to the annular gas-liquid-solid three-phase flow mechanism.

[0089] In this step, the annular gas-liquid-solid three-phase flow mechanism represents the flow rule of gas, liquid and solid in the wellbore annulus. In the drilling scene, the downhole is in a gas-liquid-solid coexistence state. The gas includes but is not limited to air, natural gas of the formation, etc.; the liquid includes but is not limited to formation fluid, oil, drilling fluid, etc.; and the solid includes but is not limited to drilling fluid solid particles, rock debris, etc. Further, the annular gas-liquid-solid three-phase flow mechanism includes mass conservation equation and other equations. The momentum conservation equation has the partial derivative of the bottom hole pressure to the input feature data of the fixed point vertical depth, so according to the formula in the annular gas-liquid-solid three-phase flow mechanism, the equation of each working condition is established. Specifically, the partial derivative equation of the neural network predicted bottom hole pressure and the well depth is obtained by using the neural network to predict the bottom hole pressure, and then the equation of the well depth partial derivative in the momentum conservation equation in the annular gas-liquid-solid three-phase flow mechanism is used. The two equations are the same, so the equation constraint of each working condition is realized.

[0090] Specifically, the equation constraint condition is established by using the following formula:

[0091]

[0092]

[0093] So that A = A theory ; wherein A theoryP and F(x, w, b) represent the bottom hole pressure expression predicted by the neural network; t represents time; m represents one of the three components of gas, liquid, and solid; p m represents density; a m represents average flow velocity; V m represents the volume fraction of gas, liquid, and solid; cos θ represents the inclination angle; g represents the acceleration of gravity; F f represents the friction between the annulus and the drilling fluid, w represents the weight of each layer in the neural network, and b represents the bias of each layer in the neural network.

[0094] As described above, P is the bottom hole pressure expression predicted by the neural network. By using the momentum conservation equation A theory in the annulus gas-liquid-solid three-phase flow mechanism, the partial derivative of the bottom hole pressure expression predicted by the neural network with respect to the well depth is subjected to an equality constraint (formula A), and formula A is equal to A theory . Further, the equality constraint condition of each working condition is established. The above equation is further transformed to obtain:

[0095]

[0096] In step 203, according to the correlation between the historical input feature data of each working condition and the corresponding historical bottom hole pressure data, the inequality constraint condition of each working condition is established. In some embodiments of the present specification, the historical input feature data of each working condition has a certain correlation with its historical bottom hole pressure data. Specifically, when a part of the historical input feature data changes, the bottom hole pressure also changes, so the historical input feature data has a certain correlation with its corresponding historical bottom hole pressure data. Different historical input feature data has different correlation with historical bottom hole pressure data. For example, the change of the historical input feature data of drilling fluid discharge in the drilling working condition will cause a larger change of the bottom hole pressure, and the change of the historical input feature data of the drill bit depth will cause a smaller change of the bottom hole pressure.

[0097] The change relationship between the historical feature input data and the historical bottom hole pressure data can be expressed as the partial derivative of the bottom hole pressure with respect to the historical input feature data being greater than 0 in mathematics. The description of establishing the inequality constraint condition according to the correlation between the historical input feature data and the historical bottom hole pressure data of each working condition will be described below, and details are described in Figure 3 .

[0098] In step 204, according to the historical sample data set of each working condition, the equality constraint condition and the inequality constraint condition of each working condition, the loss function of each working condition is constructed.

[0099] In some embodiments, the loss function of each working condition is used to evaluate the difference between the bottom hole pressure prediction value of the bottom hole pressure prediction model being trained and the actual bottom hole pressure value actually measured by the measuring device. The bottom hole pressure prediction value of the bottom hole pressure prediction model is the bottom hole pressure prediction value obtained by inputting the historical input feature data of each working condition into the neural network. When the loss function of the trained bottom hole pressure prediction model is minimized, the model training is completed.

[0100] In this step, the loss function of each working condition includes the difference between the predicted value of the neural network model for the bottom hole pressure and the actual value of the bottom hole pressure, the difference between the predicted value of the neural network model for the bottom hole pressure and the momentum conservation equation, and the inequality constraint condition. The loss function is related to the weights, biases of the neural network and the penalty factors of the inequality constraint condition. Through model training, the weights, biases and penalty factors of the loss function are continuously adjusted until the loss function reaches a minimum value.

[0101] In some embodiments of the present specification, when the equality constraint based on the annular air-liquid-solid three-phase flow mechanism and the inequality constraint based on the correlation between the historical feature input data and the bottom hole pressure data are applied to the model at the same time, the bottom hole pressure prediction model is an equality and inequality constrained model. Using the penalty function method, the objective function and the constraint condition are constructed into the loss function, so as to convert the constrained nonlinear programming problem into an unconstrained nonlinear programming problem. The following formula is the expression of the difference between the predicted value of the neural network for the bottom hole pressure and the actual value of the bottom hole pressure: Where P pre represents the predicted value of the bottom hole pressure of the neural network model, P true represents the actual measured value of the bottom hole pressure. For example, 1 sensitive parameter and 3 non-sensitive parameters are used to establish the inequality constraint, so the inequality constraint condition includes 4 inequality constraint equations. Considering that the effects of the 4 inequality constraints on the final bottom hole pressure prediction result are different, in addition to using the particle swarm algorithm to calculate the weights and biases of the neural network, the 4 penalty factors of the inequality constraint also need to be solved, so that the model has different responses when facing different degrees of constraints. The corresponding function is as follows:

[0102]

[0103] Where λ1, λ2, λ3, λ4 are the 4 penalty factors of the inequality constraint condition. The particle swarm algorithm is used to solve the weights, biases of the neural network and the penalty factors λ of the constraint condition, to further determine the loss function.

[0104] Step 205, using the loss function of each working condition, training the parameters in the neural network model, and taking the trained neural network model of each working condition as the well bottom pressure prediction model of each working condition. After the weights, biases and penalty factors in the objective function of the neural network are solved, the intelligent prediction model of the well bottom pressure is successfully solved and established, and the intelligent prediction method of the well bottom pressure under the mechanism constraint is formed.

[0105] In some embodiments of the present specification, real-time mud logging data is judged in real time according to the working condition identification rule, the drilling working conditions corresponding to the real-time mud logging data are distinguished, different constraint conditions are selected for different drilling working conditions according to the constraint condition, and the constrained neural network is established and solved, so that the well bottom pressure prediction model of different working conditions can be obtained.

[0106] Figure 3 The method flow chart for establishing the inequality constraint condition is shown. Specifically, it includes:

[0107] Step 301, for each working condition, the correlation between each input feature data in the historical input feature data under the working condition and the historical well bottom pressure data under the corresponding working condition is calculated. In this step, the correlation between each input feature data in the historical input feature data under each working condition and the corresponding historical well bottom pressure data can be calculated by a correlation calculation formula. The correlation calculation includes but is not limited to: cosine similarity formula, Pearson correlation coefficient, Jaccard similarity coefficient, Tanimoto coefficient, etc. The method for calculating the correlation between the inequality constraint of the working condition and the historical well bottom pressure data is not limited in the present application. Taking the drilling working condition as an example, the correlation between the drilling fluid density and the well bottom pressure data is 0.98, the correlation between the drilling fluid discharge and the well bottom pressure data is 0.92, the correlation between the drill bit depth and the well bottom pressure data is 0.6, the correlation between the sand content and the well bottom pressure data is 0.47, the correlation between the rotation speed and the well bottom pressure data is 0.4, and the correlation between the stand pressure and the well bottom pressure data is 0.2. It can be seen that the correlations between different historical feature input data and well bottom pressure data under the drilling working condition are different.

[0108] Step 302, taking the input feature data with a correlation lower than a first preset threshold as a non-sensitive parameter of the working condition, and taking the input feature data with a correlation higher than the first preset threshold as a sensitive parameter of the working condition. In this step, the sensitive parameter represents a parameter closely related to the well bottom pressure; and the non-sensitive parameter represents a parameter with a small correlation with the well bottom pressure.

[0109] For example, taking drilling operation as an example, the first preset threshold is set as 0.5, the input feature data (such as rotation speed, sand content, standpipe pressure, etc.) with a correlation lower than 0.5 is taken as a non-sensitive parameter of the drilling operation, and the input feature data (such as drilling fluid density, drilling fluid displacement, drill bit depth, etc.) with a correlation higher than 0.5 is taken as a sensitive parameter of the drilling operation. The first preset threshold can be pre-set or adjusted on site according to real-time logging data and real-time operation conditions. The value of the first preset threshold is not limited in the application.

[0110] In step 303, inequality constraints of the operation are established according to the sensitive parameters and the non-sensitive parameters of the operation. The sensitive parameters have a greater constraint effect on the accuracy of the model, and the non-sensitive parameters have a greater constraint effect on the abnormal value of the model. Therefore, in some embodiments of the present application, the combination of the sensitive parameters and the non-sensitive parameters can be used to establish the inequality constraints of the corresponding operation. In some other embodiments of the present application, only the sensitive parameters can be used to establish the inequality constraints of the corresponding operation. The present application does not limit the use of the sensitive parameters and / or the non-sensitive parameters to establish the inequality constraints of the corresponding operation.

[0111] In some embodiments of the present application, the establishment of the inequality constraints of each operation includes establishing the inequality constraints by using the following formula:

[0112] wherein x represents the sensitive parameters and the non-sensitive parameters of each operation, P and F(x, w, b) represent the expression of the bottom hole pressure output by the neural network model, w represents the weight of each layer in the neural network, and b represents the bias of each layer in the neural network. The change of the sensitive parameters and the non-sensitive parameters will cause the change of the bottom hole pressure, and the inequality constraints are established according to the change rule. The inequality constraints can be established by using only one sensitive parameter, or the combination of the sensitive parameters and the non-sensitive parameters.

[0113] In this step, in some embodiments of the present application, F(x, w, b) can be represented as: wherein f1 is the first activation function in the neural network model, f2 is the second activation function in the neural network model, f3 is the third activation function in the neural network model, w1 represents the weight of the first layer of the neural network, w2 represents the weight of the second layer of the neural network, w3 represents the weight of the third layer of the neural network, b1 represents the bias of the first layer of the neural network, b2 represents the bias of the second layer of the neural network, and b3 represents the bias of the third layer of the neural network. [1] [2] [3] [1] [2] [3] ​​​​​Bias of the third layer of the neural network. In order to ensure the regression performance of the neural network model, the embodiment of the present specification sets a 4-layer neural network, including: 1 input layer, 2 hidden layers and 1 output layer. Among them, the activation function between the input layer and the first hidden layer is f1(Relu), the activation function between the first hidden layer and the second hidden layer is f2(Relu), and the activation function between the second hidden layer and the output layer is f3(Linear). As shown in Figure 7 Fig. 1 shows a structure diagram of a neural network model according to an embodiment of the present specification. The number of hidden layer neurons is reasonably set by grid search method to ensure the prediction accuracy of the model and avoid overfitting of the model, and a neural network model with a topology of 12-32-16-1 is established. In the figure, w ij [k] represents the weight coefficient between the i-th neuron of the k-th layer and the j-th neuron of the previous layer; Z i [k] represents the value of the i-th neuron of the k-th layer without activation function; f i (x) represents an activation function; A i [k] represents the value of the i-th neuron of the k-th layer after activation function.

[0114] Through the calculation of the neural network, the neural network mathematical expression P of the bottom hole pressure can be obtained, F(x, w, b) = f3(w [3] × f2(w [2] × f1(w [1] × X + b [1] ) + b [2] ) + b [3] ). Therefore, the training process of the neural network becomes a constrained nonlinear programming problem, that is, to find the optimal network weight and bias, so that the objective function is minimized. In some other embodiments of the present specification, the number of layers of the neural network can be 3, 5 or any other number, and the number of activation functions in the neural network can be 2, 4 or any other number, and the present application does not limit the number of layers of the neural network and the number of activation functions. Therefore, the specific representation of F(x, w, b) is not limited.

[0115] Figure 4 Fig. 2 shows a flow chart of a sensitive parameter processing method according to an embodiment of the present specification.

[0116] Step 401: Take any two sensitive parameters under the working condition as a sensitive parameter group, and calculate the correlation of the sensitive parameters in each sensitive parameter group.

[0117] As described in step 302, at least one sensitive parameter can be determined according to the correlation between the historical input feature data and the bottom hole pressure. When more than two sensitive parameters are obtained under a certain working condition, any two sensitive parameters are taken as a group, and the correlation between the two sensitive parameters in the group is calculated. For example, step 302 obtains sensitive parameters under drilling conditions, including drilling fluid density, drilling fluid displacement, and drill bit depth, etc. The drilling fluid density and the drilling fluid displacement can be taken as a group, and the correlation between the drilling fluid density and the drilling fluid displacement is calculated.

[0118] Step 402, screening out the sensitive parameter group with a correlation greater than a second preset threshold.

[0119] In this step, by screening the sensitive parameters in the group composed of two sensitive parameters, if the correlation between the two sensitive parameters is greater than the second preset threshold, it can be considered that the influence of the two sensitive parameters on the bottom hole pressure may be similar. For example, the second preset threshold is set to 0.9, and the correlation between the drilling fluid density and the drilling fluid displacement in the group composed of sensitive parameters calculated in step 401 is 0.95, which exceeds the second preset threshold. Therefore, it is considered that the drilling fluid density and the drilling fluid displacement have the same influence on the bottom hole pressure. For another example, the correlation between the drilling fluid density and the drill bit depth calculated in step 401 is 0.6, and it is considered that the drilling fluid density and the drill bit depth have different influences on the bottom hole pressure. The second preset threshold can be pre-set or adjusted according to real-time field conditions.

[0120] Step 403, deleting one sensitive parameter from each sensitive parameter group screened out.

[0121] When the correlation between the two sensitive parameters in each sensitive parameter group is greater than the second preset threshold, that is, the two sensitive parameters have relatively similar influences on the bottom hole pressure. Then, any one sensitive parameter in the sensitive parameter group is deleted, and the remaining one sensitive parameter is retained. Based on this, in the subsequent step, the remaining one sensitive parameter is used to establish an inequality constraint for further model training, which can reduce redundant data, reduce the calculation amount of model training, and improve the efficiency of model training.

[0122] As Figure 5 Fig. 1 shows a structural schematic diagram of a multi-condition bottom hole pressure prediction device according to an embodiment of the present application. The basic structure of the multi-condition bottom hole pressure prediction device is described in the figure, and the functional units and modules therein can be implemented in a software manner, or can be implemented by using general-purpose chips or special-purpose chips. The device specifically comprises:

[0123] The working condition acquisition unit 501 is configured to acquire a real-time working condition according to real-time logging data.

[0124] The well bottom pressure prediction model determination unit 502 is configured to determine a well bottom pressure prediction model of the real-time working condition, wherein the well bottom pressure prediction model of each working condition is trained based on historical input feature data related to pressure of each working condition, historical well bottom pressure data of each working condition, and constraint conditions of each working condition, and the constraint conditions of each working condition are established according to the correlation between the historical input feature data and the historical well bottom pressure data of each working condition and / or the annular air-liquid-solid three-phase flow mechanism;

[0125] The input feature data determination unit 503 is configured to determine input feature data of the real-time working condition according to the real-time mud logging data.

[0126] The well bottom pressure prediction unit 504 is configured to input the input feature of the real-time working condition into the well bottom pressure prediction model of the real-time working condition to predict the well bottom pressure.

[0127] The device can realize deep integration of algorithm and mechanism by adding mechanism constraints, ensure the prediction accuracy of the neural network model, make the prediction result more consistent with the change mechanism of the prediction target, reduce the abnormal value and fluctuation value of the prediction result, and effectively improve the stability and generalization ability of the neural network model. It provides technical support and important reference for efficient prediction and fine control of well bottom pressure under deep high temperature and high pressure drilling gas invasion conditions.

[0128] As an embodiment of the present document, reference can also be made to Figure 6 FIG. 5 shows a specific structural schematic diagram of a multi-working condition well bottom pressure prediction device, and the working condition acquisition unit 501 is further configured to determine a working condition identification rule.

[0129] As an embodiment of the present document, the working condition acquisition unit 501 further comprises:

[0130] The working condition identification rule determination module 5011 is configured to determine a working condition identification rule according to the mud logging data.

[0131] As an embodiment of the present document, the well bottom pressure prediction model determination unit 502 is further configured to obtain historical sample data sets of each working condition, establish equation constraint conditions of each working condition according to the annular air-liquid-solid three-phase flow mechanism, and establish inequality constraint conditions of each working condition and construct a loss function according to the correlation between historical input feature data and corresponding historical well bottom pressure data of each working condition. The well bottom pressure prediction model determination unit 502 further comprises:

[0132] The historical sample data set acquisition module 5021 is configured to obtain historical sample data sets of each working condition.

[0133] The historical sample data set screening module 5022 is configured to delete part of the sample data according to the annular pressure calculation formula.

[0134] The equality constraint establishment module 5023 is used to establish equal transport constraint conditions based on the annular air-liquid-solid three-phase flow mechanism.

[0135] The inequality constraint establishment module 5024 is used to establish inequality constraints based on the correlation between historical input characteristic data and historical bottom hole pressure data for each working condition.

[0136] The loss function building module 5025 is used to construct the loss function.

[0137] As one embodiment of this document, the input feature data determination unit 503 further includes:

[0138] The correlation calculation module 5031 is used to calculate the correlation between historical input feature data and bottom hole pressure data;

[0139] Sensitive parameter determination module 5032 is used to use data with a correlation greater than a first preset threshold as sensitive parameters;

[0140] The non-sensitive parameter determination module 5033 is used to use data with a correlation less than a first preset threshold as non-sensitive parameters;

[0141] The sensitive parameter group filtering module 5034 is used to filter out sensitive parameter groups with a correlation greater than the second preset threshold.

[0142] The sensitive parameter deletion module 5035 is used to delete one of the sensitive parameters from each group of filtered sensitive parameters.

[0143] like Figure 8 As shown in this embodiment, a computer device 802 may include one or more processors 804, such as one or more central processing units (CPUs), each of which can implement one or more hardware threads. The computer device 802 may also include any memory 806 for storing information of any kind, such as code, settings, data, etc. Non-limitingly, for example, the memory 806 may include any type of RAM, any type of ROM, flash memory, hard disk, optical disk, etc. More generally, any memory can use any technology to store information. Further, any memory can provide volatile or non-volatile retention of information. Further, any memory may represent a fixed or removable component of the computer device 802. In one case, when the processor 804 executes associated instructions stored in any memory or combination of memories, the computer device 802 can perform any operation of the associated instructions. The computer device 802 also includes one or more drive mechanisms 808 for interacting with any memory, such as hard disk drive mechanisms, optical disk drive mechanisms, etc.

[0144] The computer device 802 can also include input / output module(s) 810 (I / O) for receiving various input (via input device(s) 812) and for providing various output (via output device(s) 814). One particular output mechanism can include a presentation device 816 and associated graphical user interface (GUI) 818. In other embodiments, the input / output module(s) 810 (I / O), input device(s) 812, and output device(s) 814 can not be included, and the computer device 802 can be only a computer device in a network. The computer device 802 can also include one or more network interfaces 820 for exchanging data with other devices via one or more communication links 822. One or more communication buses 824 couple the above-described components so that each component can communicate with each other component.

[0145] The communication links 822 can be implemented in any manner, such as through a local area network, a wide area network (e.g., the Internet), a point-to-point connection, etc., or any combination thereof. The communication links 822 can include any combination of hardwired links, wireless links, routers, gateway functionality, name servers, etc., governed by any protocol or combination of protocols.

[0146] Corresponding to the method in Figures 1-4 The embodiments herein also provide a computer readable storage medium, having stored thereon a computer program, which, when executed by a processor, performs the steps of the above-mentioned method.

[0147] The embodiments herein also provide a computer readable instruction, wherein the program in the computer readable instruction, when executed by a processor, causes the processor to perform the method as shown in Figures 1 to 4

[0148] It should be understood that the size of the serial number of the above-mentioned processes in the various embodiments herein does not mean the order of execution, and the execution order of the processes should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments herein.

[0149] It should also be understood that in the embodiments herein, the term "and / or" is only a description of the association relationship of the associated objects, which means that there can be three relationships. For example, A and / or B can represent the following three cases: A exists alone, A and B exist together, and B exists alone. In addition, the character " / " in this paper generally represents that the front and rear associated objects are in an "or" relationship.

[0150] ​Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0151] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0152] In several embodiments provided herein, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are merely schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interfaces, devices or units, and can also be electrical, mechanical or other form of connection.

[0153] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiments of the present application.

[0154] In addition, each functional unit in each embodiment herein can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0155] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions herein or the entire or part of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments herein. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0156] The principles and implementation manners of the present application are described in the embodiments herein, and the above descriptions of the embodiments are only used to help understand the methods and core ideas thereof; meanwhile, for those skilled in the art, the specific implementation manners and application ranges can be changed according to the ideas herein, and the above descriptions should not be understood as limitations on the present application.

Claims

1. A multi-condition bottom hole pressure prediction method, characterized in that, The method comprises: According to real-time mud logging data, the real-time working condition is obtained; Determine the bottom hole pressure prediction model of the real-time working condition, wherein the bottom hole pressure prediction model of each working condition is trained based on the historical input feature data related to pressure of each working condition, the historical bottom hole pressure data of each working condition and the constraint condition of each working condition, and the constraint condition of each working condition is established according to the correlation between the historical input feature data and the historical bottom hole pressure data of each working condition and / or the annular air-liquid-solid three-phase flow mechanism; According to the real-time mud logging data, the input feature data of the real-time working condition is determined; The input feature data of the real-time working condition is input into the bottom hole pressure prediction model of the real-time working condition, and the bottom hole pressure is predicted.

2. The multi-condition bottom hole pressure prediction method of claim 1, wherein, The bottom hole pressure prediction model training process of each working condition comprises: Obtain the historical sample data set of each working condition, and each historical sample data set of a working condition comprises a plurality of sample data, and each sample data comprises historical input feature data and historical bottom hole pressure data; According to the annular air-liquid-solid three-phase flow mechanism, the equality constraint condition of each working condition is established; According to the correlation between the historical input feature data and the corresponding historical bottom hole pressure data of each working condition, the inequality constraint condition of each working condition is established; According to the historical sample data set of each working condition, the equality constraint condition and the inequality constraint condition of each working condition, the loss function of each working condition is constructed; The parameters in the neural network model are trained by using the loss function of each working condition, and the trained neural network model of each working condition is used as the bottom hole pressure prediction model of each working condition.

3. The multi-condition bottom hole pressure prediction method of claim 2, wherein, According to the correlation between the historical input feature data and the corresponding historical bottom hole pressure data of each working condition, the inequality constraint condition of each working condition is established, which comprises: For each working condition, the correlation between each input feature data in the historical input feature data and the historical bottom hole pressure data under the corresponding working condition is calculated; The input feature data with a correlation lower than a first preset threshold is used as a non-sensitive parameter of the working condition, and the input feature data with a correlation higher than the first preset threshold is used as a sensitive parameter of the working condition; According to the sensitive parameters and the non-sensitive parameters of the working condition, the inequality constraint condition of the working condition is established.

4. The multi-condition bottom hole pressure prediction method of claim 3, wherein, For each working condition, after determining the sensitive parameters under the working condition, the method further comprises: Any two sensitive parameters under the working condition are used as a sensitive parameter group, and the correlation of the sensitive parameters in each sensitive parameter group is calculated; Sensitive parameter groups with a correlation greater than a second preset threshold are screened out; One sensitive parameter is deleted from each sensitive parameter group screened out.

5. The multi-condition bottom hole pressure prediction method of claim 2, wherein, After obtaining the historical sample data set of each working condition, the method further comprises: The bottom hole pressure range of each working condition is estimated according to the annular pressure calculation formula; The sample data in the historical sample data set of each working condition which does not satisfy the bottom hole pressure range is deleted.

6. The multi-condition bottom hole pressure prediction method of claim 3, wherein, According to the sensitive parameters and the non-sensitive parameters of each working condition, the inequality constraint condition of each working condition is established, which comprises: The inequality constraint condition is established by using the following formula: Wherein, x represents sensitive parameters and non-sensitive parameters of each working condition, P and F(x, w, b) both represent the expression of bottom hole pressure output by the neural network model, w represents the weight of each layer in the neural network, and b represents the bias of each layer in the neural network.

7. The multi-condition bottom hole pressure prediction method of claim 2, wherein, According to the annular air-liquid-solid three-phase flow mechanism, the equation constraint conditions of each working condition include: The equation constraint conditions are established by using the following formula: Let A = A theory ; where A theory is the momentum conservation equation in the annulus air-liquid-solid three-phase flow mechanism, z represents the fixed point vertical depth of each working condition, P and F(x, w, b) represent the bottom hole pressure expression predicted by the neural network; t represents time; m represents one of the three components of gas, liquid and solid; p m represents density; a m represents average flow velocity; V m represents the volume fraction of gas, liquid and solid; cosθ represents the inclination angle; g represents the acceleration of gravity; F f represents the friction between the annulus and the drilling fluid, w represents the weight of each layer in the neural network, and b represents the bias of each layer in the neural network.

8. A multi-condition bottom hole pressure prediction device, characterized in that, The device comprises: A working condition acquisition unit configured to acquire a real-time working condition according to real-time logging data; A bottom hole pressure prediction model determination unit configured to determine a bottom hole pressure prediction model of the real-time working condition, wherein the bottom hole pressure prediction model of each working condition is trained based on historical input feature data related to pressure of each working condition, historical bottom hole pressure data of each working condition and constraint conditions of each working condition, the constraint conditions of each working condition are established according to a response relationship between the historical input feature data and the historical bottom hole pressure data of each working condition and / or an annular air-liquid-solid three-phase flow mechanism; An input feature data determination unit configured to determine input feature data of the real-time working condition according to the real-time logging data; A bottom hole pressure prediction unit configured to input the input feature of the real-time working condition into the bottom hole pressure prediction model of the real-time working condition to predict a bottom hole pressure.

9. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the method of any one of claims 1-7 when executing the computer program.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the method of any one of claims 1-7.

Images