A multi-sensor redundant combination fault-tolerant dynamic measurement model and its establishment method

Abstract

The present invention discloses a multi-sensor redundant combination fault-tolerant dynamic measurement model and a method for establishing the same, belonging to the technical field of measurement-while-drilling systems, and comprising the following steps: Step S1: Using a fluxgate, an accelerometer, and a gyroscope to form a multi-sensor residual error model under redundant information for the calculation methods of well inclination angle, azimuth angle, and tool face angle; Step S2: By analyzing the error characteristics and failure laws of each sensor, constructing a multi-sensor fault-tolerant judgment model based on deep learning. Step S3: By constructing a first-order prediction state equation, constructing a state equation under multi-sensor redundant combination, and further realizing the construction of an attitude fault-tolerant dynamic measurement model for a near-bit measurement-while-drilling device. Through the above method, the present invention can further suppress the measurement error of an inertial measurement-while-drilling system and improve the positioning accuracy of a multi-sensor redundant combination positioning system.

Description

Technical Field

[0001] The present invention relates to the technical field of measurement-while-drilling systems, and particularly to a multi-sensor redundant combination fault-tolerant dynamic measurement model and a method for establishing the same. Background Art

[0002] The exploitation and utilization of coalbed methane is of great significance in ensuring the safe production of coal mines, optimizing the energy industrial structure, protecting the ecological environment, etc. In the current situation of coalbed methane drilling technology and equipment, improving the recovery rate of coalbed methane and reducing the drilling cost are of great significance for the vigorous development of coalbed methane. In order to realize the industrialization and large-scale development of coalbed methane, it is necessary to vigorously develop coalbed methane exploitation drilling technology and equipment, and gradually form a multi-branch cluster well drilling technology with directional drilling as the core.

[0003] Guiding the drill bit to reach the coal seam area along a predetermined three-dimensional trajectory and maintaining drilling in the coal seam is the core problem in the exploitation of multi-branch horizontal wells of coalbed methane. In downhole measurement-while-drilling inclinometers, the fluxgate is easily affected by factors such as drill string magnetization and magnetic interference of the drill string assembly, and the accelerometer and gyroscope are easily affected by factors such as complex vibrations of the drill string and friction deformation of the drill string. Therefore, it is necessary to reduce the error influence of the sensors.

[0004] Based on this, the present invention designs a multi-sensor redundant combination fault-tolerant dynamic measurement model and a method for establishing the same to solve the above problems. Summary of the Invention

[0005] In view of the above-mentioned drawbacks existing in the prior art, the present invention provides a multi-sensor redundant combination fault-tolerant dynamic measurement model and a method for establishing the same.

[0006] To achieve the above object, the present invention is realized through the following technical solutions:

[0007] A method for establishing a multi-sensor redundant combination fault-tolerant dynamic measurement model includes the following steps:

[0008] Step S1: Using the calculation methods of the well inclination angle, azimuth angle, and tool face angle by the fluxgate, accelerometer, and gyroscope to form a multi-sensor residual error model under redundant information;

[0009] Step S2: By analyzing the error characteristics and failure laws of each sensor, constructing a multi-sensor fault-tolerant judgment model based on deep learning.

[0010] Step S3: By constructing a primary prediction state equation, constructing a state equation under multi-sensor redundancy combination, and further realizing the construction of an attitude fault-tolerant dynamic measurement model for a near-bit measurement-while-drilling device.

[0011] Furthermore, step S1 specifically includes:

[0012] According to the measurement method of the three-dimensional attitude of the drill string by the near-bit measurement-while-drilling system under the combination of a three-axis fluxgate and a three-axis accelerometer, a real-time calculation method for the well inclination angle, azimuth angle, and tool face angle using Euler angle theory is constructed;

[0013] According to the measurement method of the three-dimensional attitude of the drill string by the near-bit measurement-while-drilling system under the combination of a three-axis gyroscope and a three-axis accelerometer, a real-time calculation method for the well inclination angle, azimuth angle, and tool face angle using quaternion theory is constructed;

[0014] For the two three-dimensional attitude calculation methods of the drill string, a single residual error model for the calculation method is established respectively, and then combined with the Kalman filter algorithm to construct a multi-sensor residual error model under redundant information.

[0015] Furthermore, in step S2, according to the influence of drill string magnetization error and drill string assembly magnetic interference on the magnetic field, and the influence of factors such as complex vibration of the drill string and friction deformation of the drill string on the accelerometer and gyroscope, the error characteristics and failure laws of measurement parameters under multiple influencing factors are constructed, and then a measurement error model of multi-sensors while drilling is established; the intelligent recognition strategy of measurement error and failure law of multi-parameters under abnormal conditions is studied using fuzzy clustering analysis theory, and then the construction of a multi-sensor redundant combination fault-tolerant dynamic measurement model of the measurement-while-drilling system during the exploitation process of coalbed methane multi-branch directional wells is realized.

[0016] Furthermore, the constraint conditions of the objective function set of the fuzzy clustering algorithm are as follows:

[0017]

[0018] Among them, U is the fuzzy C-partition matrix, V is the vector of C fault clustering centers, C is the number of clustering centers, n is the number of samples, Ci represents the i-th clustering center, xj represents the j-th sample, is the membership degree of the identification sample xj to the clustering center Ci, and m is the number of clusters of the clustering;

[0019] The iterative formula of the above algorithm is obtained by using the Lagrange multiplier method:

[0020]

[0021] Combining the above measurement error model of multi-sensors while drilling, fuzzy clustering analysis, and the multi-sensor residual error model in step S1, a multi-sensor fault-tolerant judgment model based on deep learning is established.

[0022] Furthermore, in step S2, a gated recurrent neural network is used to construct a multi-sensor fault-tolerant judgment model based on deep learning, including a GUR layer and a fully connected layer;

[0023] The GUR layer includes an update gate, a reset gate, a candidate state, and a hidden state;

[0024] The internal expression of the GUR hidden layer is as follows:

[0025] Z t = σ(W Z [s t-1 , x t + b z (3)

[0026] r t = σ(W r [s t-1 , x t + b r (4)

[0027]

[0028]

[0029] Among them, Z t represents the update gate activation vector, s t-1 is the hidden state vector at the previous moment, x t is the input vector at the current moment, W Z is the weight matrix of the update gate, b z is the bias vector of the update gate, σ represents the Sigmoid function, which is used to convert data into values in the range of 0 to 1; r t is the activation vector of the reset gate, W r is the weight matrix of the reset gate, b r is the bias vector of the reset gate, h t is the candidate state vector, tanh is the activation function, represents the Hadamard product, which is the product of the corresponding elements in the operation matrix, W h is the weight matrix for controlling the candidate state information, b n is the bias vector for controlling the candidate state information, s t is the hidden state vector at the current moment.

[0030] Furthermore, step S3 specifically includes:

[0031] Using the attitude solution model for the near-bit measurement-while-drilling system of coalbed methane multi-branch directional wells obtained in step S1, setting reasonable state vectors, constructing the primary prediction state equation of the near-bit measurement-while-drilling system, and realizing the construction of the state equation under the multi-sensor redundant combination through the multi-sensor dynamic transfer function model of the near-bit measurement-while-drilling system;

[0032] Combined with the fault tolerance judgment model based on deep learning described in step S2, analyze the influence law under abnormal parameter observations of the fluxgate, accelerometer, and gyroscope, establish the observation equation of the fault tolerance combined positioning system based on deep learning, realize the construction of the state space equation of the inertial measurement-while-drilling combined positioning system, and further realize the attitude fault tolerance dynamic measurement of the near-bit measurement-while-drilling device.

[0033] The present invention also provides a multi-sensor redundant combined fault tolerance dynamic measurement model constructed by the above construction method.

[0034] Beneficial effects

[0035] The present invention can realize the establishment of the attitude fault tolerance dynamic measurement model of the measurement-while-drilling system under multi-sensor redundancy near the bit, so as to further suppress the measurement error of the inertial measurement-while-drilling system and improve the positioning accuracy of the multi-sensor redundant combined positioning system. Description of the drawings

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0037] Figure 1 It is a flow chart of the method for establishing a multi-sensor redundant combined fault tolerance dynamic measurement model of the present invention;

[0038] Figure 2 It is a comparison diagram of the actual trajectory and the preset trajectory of the drill bit in the embodiment of the multi-sensor redundant combined fault tolerance dynamic measurement model of the present invention. Detailed implementation manners

[0039] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0040] The following further describes the present invention with reference to the embodiments.

[0041] Embodiment 1

[0042] Please refer to the attached specification Figure 1, a method for establishing a multi-sensor redundant combination fault-tolerant dynamic measurement model of a measurement-while-drilling system, comprising the following steps:

[0043] Step S1: Use the calculation methods of a fluxgate, an accelerometer, and a gyroscope for well inclination angle, azimuth angle, and tool face angle to form a multi-sensor residual error model under redundant information.

[0044] Optionally, according to the method for measuring the three-dimensional attitude of a drill string in a near-bit measurement-while-drilling system under the combination of a three-axis fluxgate and a three-axis accelerometer, construct a real-time calculation method for well inclination angle, azimuth angle, and tool face angle using Euler angle theory.

[0045] Optionally, according to the method for measuring the three-dimensional attitude of a drill string in a near-bit measurement-while-drilling system under the combination of a three-axis gyroscope and a three-axis accelerometer, construct a real-time calculation method for well inclination angle, azimuth angle, and tool face angle using quaternion theory.

[0046] Optionally, to increase the reliability of data and the accuracy of data under multiple sensors, establish a single residual error model for each of the two drill string three-dimensional attitude calculation methods, and then combine the Kalman filter algorithm to construct a multi-sensor residual error model under redundant information.

[0047] Step S2: By analyzing the error characteristics and failure laws of each sensor, construct a multi-sensor fault-tolerant judgment model based on deep learning.

[0048] Optionally, to prevent the sensors from being affected by factors such as magnetic fields and vibrations during the drilling process as the drill bit penetrates deeper, resulting in errors or sensor failures, in this embodiment, according to the influence of drill string magnetization error and drill string assembly magnetic interference on the magnetic field, and the influence of complex drill string vibrations and drill string friction deformation on accelerometers and gyroscopes, construct the error characteristics and failure laws of measurement parameters under multiple influencing factors, and then establish a measurement-while-drilling multi-sensor measurement error model.

[0049] Optionally, for the above problems, in this embodiment, use the fuzzy clustering analysis theory to study the intelligent recognition strategy of measurement errors and failure laws of multi-parameters under abnormal conditions, and then realize the construction of a multi-sensor redundant combination fault-tolerant dynamic measurement model of a measurement-while-drilling system during the exploitation of coalbed methane multi-branch directional wells.

[0050] The constraint conditions of the objective function set of the fuzzy clustering algorithm are as follows:

[0051]

[0052] Among them, U is the fuzzy C-partition matrix, V is the C fault clustering center vectors, C is the number of clustering centers, n is the number of samples, Ci represents the i-th clustering center, xj represents the j-th sample, is the membership degree of the identification sample xj to the clustering center Ci, and m is the number of clusters of the clustering;

[0053] The iterative formula of the above algorithm obtained by Lagrange multipliers is as follows:

[0054]

[0055] Combined with the above-mentioned measurement error model of multi-sensors while drilling, fuzzy clustering analysis, and the multi-sensor residual error model in step S1, a multi-sensor fault-tolerant judgment model based on deep learning is established.

[0056] The gated recurrent neural network has the characteristic of memory, can better handle the problem of data fault tolerance, and can better capture the dependencies with a large time step distance in the time series. The gated recurrent neural network is used to establish a multi-sensor fault-tolerant judgment model based on deep learning, including a GUR layer and a fully connected layer.

[0057] The GUR layer includes an update gate, a reset gate, a candidate state, and a hidden state.

[0058] The internal expression of the GUR hidden layer is as follows:

[0059] Z t = σ(W Z [s t-1 , x t +b z (3)

[0060] r t = σ(W r [s t-1 , x t +b r (4)

[0061]

[0062]

[0063] Among them, Z t represents the update gate activation vector, s t-1 is the hidden state vector of the previous moment, x t is the input vector of the current moment, W Z is the weight matrix of the update gate, b z is the bias vector of the update gate, σ represents the Sigmoid function, which is used to convert data into values in the range of 0 to 1; r t is the activation vector of the reset gate, W r is the weight matrix of the reset gate, b r is the bias vector of the reset gate, h t is the candidate state vector, and tanh is the activation function. denotes the Hadamard product, which is the product of the corresponding elements in the operation matrix, W h is the weight matrix for controlling the weight of candidate state information, b n is the bias vector for controlling the candidate state information, s t is the hidden state vector at the current moment.

[0064] Step S3: By constructing a single-step prediction state equation, the state equation under the multi-sensor redundant combination is constructed, and then the attitude fault-tolerant dynamic measurement model of the near-bit measurement-while-drilling device is realized.

[0065] Optionally, using the attitude solution model of the near-bit measurement-while-drilling system for coalbed methane multi-branch directional wells obtained in Step S1, a reasonable state vector is set, a single-step prediction state equation of the near-bit measurement-while-drilling system is constructed, and through the multi-sensor dynamic transfer function model of the near-bit measurement-while-drilling system, the construction of the state equation under the multi-sensor redundant combination is realized.

[0066] Optionally, combining the fault-tolerant judgment model based on deep learning described in Step S2, analyzing the influence law under abnormal parameter observations of the fluxgate, accelerometer, and gyroscope, establishing an observation equation of the fault-tolerant combined positioning system based on deep learning, realizing the construction of the state space equation of the inertial measurement-while-drilling combined positioning system, and further realizing the attitude fault-tolerant dynamic measurement of the near-bit measurement-while-drilling device.

[0067] Using the Kalman filter algorithm, the establishment of the attitude fault-tolerant dynamic measurement model of the measurement-while-drilling system with multi-sensor redundancy near the bit is realized, so as to further suppress the measurement error of the inertial measurement-while-drilling system.

[0068] Figure 2 This is a comparison diagram of the actual trajectory and the preset trajectory of the drill bit in the embodiment of the multi-sensor redundant combination fault-tolerant dynamic measurement model of the present invention. It can be seen that compared with the drill bit trajectory without applying the model of the present invention, the actual trajectory of the drill bit of the present invention basically coincides with the preset trajectory, and the error is smaller.

[0069] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements will not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for establishing a multi-sensor redundant combined fault-tolerant dynamic measurement model, characterized in that, Including the following steps: Step S1: A multi-sensor residual error model under redundant information is formed by using the calculation methods of well inclination angle, azimuth angle, and tool face angle with a fluxgate, an accelerometer, and a gyroscope. Step S2: By analyzing the error characteristics and failure laws of each sensor, a multi-sensor fault tolerance judgment model based on deep learning is constructed. In Step S2, according to the influence of drill string magnetization error and drill string assembly magnetic interference on the magnetic field, and the influence of complex drill string vibration and drill string friction deformation factors on the accelerometer and gyroscope, the error characteristics and failure laws of measurement parameters under multiple influencing factors are constructed, and then a measurement error model of multi-sensors while drilling is established. The intelligent recognition strategy of measurement error and failure law of multi-parameters under abnormal conditions is studied by using the fuzzy clustering analysis theory, and then the construction of a multi-sensor redundant combination fault tolerance dynamic measurement model of the measurement system while drilling in the process of coalbed methane multi-branch directional well exploitation is realized. Step S3: By constructing a primary prediction state equation, a state equation under multi-sensor redundant combination is constructed, and then the construction of an attitude fault tolerance dynamic measurement model of the near-bit measurement-while-drilling device is realized. Step S3 specifically includes: Using the attitude calculation model of the near-bit measurement-while-drilling system for coalbed methane multi-branch directional wells obtained in Step S1, setting reasonable state vectors, constructing a primary prediction state equation of the near-bit measurement-while-drilling system, and realizing the construction of the state equation under multi-sensor redundant combination through the multi-sensor dynamic transfer function model of the near-bit measurement-while-drilling system. Combined with the fault tolerance judgment model based on deep learning described in Step S2, analyzing the influence law under abnormal observations of fluxgate, accelerometer, and gyroscope parameters, establishing an observation equation of the fault tolerance combination positioning system based on deep learning, realizing the construction of the state space equation of the inertial measurement-while-drilling combination positioning system, and then realizing the attitude fault tolerance dynamic measurement of the near-bit measurement-while-drilling device.

2. The method for establishing a multi-sensor redundant combination fault-tolerant dynamic measurement model according to claim 1, characterized in that Step S1 specifically includes: According to the drill string three-dimensional attitude measurement method of the near-bit measurement-while-drilling system under the combination of three-axis fluxgate and three-axis accelerometer, a real-time calculation method of well inclination angle, azimuth angle, and tool face angle using Euler angle theory is constructed. According to the drill string three-dimensional attitude measurement method of the near-bit measurement-while-drilling system under the combination of three-axis gyroscope and three-axis accelerometer, a real-time calculation method of well inclination angle, azimuth angle, and tool face angle using quaternion theory is constructed. For the two drill string three-dimensional attitude calculation methods, a single residual error model for the calculation method is established respectively, and then a multi-sensor residual error model under redundant information is constructed by combining the Kalman filter algorithm.

3. The method for establishing a multi-sensor redundant combination fault-tolerant dynamic measurement model according to claim 2, characterized in that, The constraint conditions of the objective function set of the fuzzy clustering algorithm are as follows: ； Among them, U is the fuzzy C-partition matrix, V is the vector of C fault clustering centers, C is the number of clustering centers, n is the number of samples, c i represents the i-th category, represents the j-th category c i is the clustering center of, is the membership function for the clustering center c i is; m is the number of clusters of the clustering; The iterative formula of the above algorithm obtained by the Lagrange multiplier method is as follows: ; Combining the above measurement error model of multi-sensors while drilling, fuzzy clustering analysis, and the multi-sensor residual error model in Step S1, a multi-sensor fault tolerance judgment model based on deep learning is established.

4. The method for establishing a multi-sensor redundant combination fault-tolerant dynamic measurement model according to claim 3, characterized in that In Step S2, a gated recurrent neural network is used to construct a multi-sensor fault tolerance judgment model based on deep learning, including a GUR layer and a fully connected layer. The GUR layer includes an update gate, a reset gate, a candidate state, and a hidden state. The internal expression of the GUR hidden layer is as follows: ; Among them, represents the update gate activation vector, is the hidden state vector at the previous moment, is the input vector at the current moment, is the weight matrix of the update gate, is the bias vector of the update gate, represents the Sigmoid function, which is used to convert data into values in the range of 0 to 1; is the activation vector of the reset gate, is the weight matrix of the reset gate, is the bias vector of the reset gate, is the candidate state vector, and tanh is the activation function, represents the Hadamard product, which is the product of the corresponding elements in the operation matrix, is the weight matrix for controlling candidate state information, is the bias vector for controlling candidate state information, is the hidden state vector at the current moment.

5. A multi-sensor redundant combined fault-tolerant dynamic measurement model constructed by the method for establishing a multi-sensor redundant combined fault-tolerant dynamic measurement model according to any one of claims 1 to 4.

Images