Monitoring-based deepwater drilling riser system state prediction method, medium and system

Abstract

The invention provides a monitoring-based deepwater drilling riser system state prediction method, medium and system, and belongs to the technical field of offshore oil and gas development.The monitoring-based deepwater drilling riser system state prediction method, medium and system comprises the following steps that firstly, the state of a deepwater drilling riser system needing to be monitored is determined; establishing a deepwater drilling riser system mechanical simulation model, establishing a database by using the simulation model, performing initial parameter optimization on the gated circulation network through an ant colony algorithm, training the gated circulation network by using the database to establish a deepwater drilling riser system prediction model, and constructing a test set; verifying a deepwater drilling riser system state prediction model by using the determined monitoring state, and optimizing the prediction model based on a verification result; the prediction precision can be improved, and the adaptability of the prediction method is improved by using the prediction algorithm.

Description

Technical Field

[0001] The present invention belongs to the technical field of marine oil and gas development, and in particular relates to a monitoring-based deepwater drilling riser system state prediction method, medium and system. Background Art

[0002] Deepwater drilling riser systems, a core component of offshore oil and gas production, play a crucial role in connecting offshore platforms to wellheads thousands of meters below the seabed. They must not only withstand immense weight and top tension, but also withstand the complex loads of the marine environment, such as strong winds, massive waves, and currents. The system's high height-to-diameter ratio (large aspect ratio) makes it highly susceptible to external interference during operation, leading to frequent safety incidents.

[0003] Deepwater drilling riser systems are critical equipment connecting offshore drilling platforms to the subsea wellhead. They primarily serve to isolate seawater, guide drilling tools, circulate drilling fluid, and operate and lower subsea blowout preventer stacks. Deepwater drilling riser systems must withstand the combined effects of their own weight, top tension, external marine environmental loads, and internal drilling conditions such as the drill string and drilling fluid. Therefore, their design and manufacture are highly complex and require significant technical expertise.

[0004] Deepwater drilling riser systems are critical equipment connecting offshore platforms to subsea wellheads. Their large aspect ratio makes them susceptible to accidents under loads such as waves and currents. To mitigate these accidents, monitoring devices are used in engineering projects. However, full-scale status monitoring of riser systems in deepwater operations is difficult, and local status monitoring provides limited assurance of riser system safety. Predicting the status of deepwater drilling riser systems is an effective solution, but a prediction method for deepwater drilling riser systems based on local monitoring has yet to be proposed. Summary of the Invention

[0005] The present invention is achieved in that:

[0006] The present invention provides a method for predicting the status of a deepwater drilling riser system based on monitoring, which includes the following steps:

[0007] S10: Determine the status of the deepwater drilling riser system that needs to be monitored;

[0008] S20: Use the deepwater drilling riser system mechanical simulation model to establish a database;

[0009] S30: Establish a deepwater drilling riser system status prediction model based on ant colony algorithm and gated recurrent neural network;

[0010] S40: Testing the deepwater drilling riser system status prediction model;

[0011] S50: Optimizing the state prediction model for deepwater drilling riser systems.

[0012] On the basis of the above technical solution, the monitoring-based deepwater drilling riser system status prediction method of the present invention can also be improved as follows:

[0013] Wherein, the S30 specifically includes the following steps:

[0014] Preprocess the data in the simulation database for model training;

[0015] Establishing the input and output of the deepwater drilling riser system state prediction model, wherein the input layer is the data processed rotation angle, displacement, current velocity and wave height data, and the output is the deepwater drilling riser system state quantity;

[0016] Designing the structure of the deepwater drilling riser system state prediction model, including the number of network layers and nodes in each layer, and determining hyperparameters such as learning rate and hidden layer dimension;

[0017] Use the ant colony optimization algorithm to optimize the initial weight thresholds between each layer;

[0018] The model is trained using the AdamW optimizer, and after the model predicts the output, the model is evaluated based on the size of the loss function.

[0019] Determine whether the maximum number of iterations has been reached. If so, the model training ends. Otherwise, the loss function is updated through backpropagation, and the model training continues until the maximum number of iterations is reached.

[0020] After the model training is completed, the output is the predicted state of the deepwater drilling riser system.

[0021] Furthermore, the method of using the ant colony optimization algorithm to optimize the initial weight thresholds between the layers specifically includes the following steps:

[0022] Initialize the initial position of the ant colony;

[0023] Calculate the probability that the k-th ant selects the i-th subinterval to obtain a set of parameter values, which are used to represent a set of initial weight thresholds generated after the k-th ant traverses all intervals;

[0024] After the kth ant has walked through all nodes according to the probability interval, a set of feasible solutions of the gated recurrent network model is obtained, and the state prediction model of the deepwater drilling riser system is initialized with the feasible solutions;

[0025] The deepwater drilling riser system state prediction model is trained to obtain the model mean square error, and the pheromone is updated after all ants have walked through all parameters to be optimized;

[0026] Iterate the ant colony to find the optimal value, and determine whether the number of iterations has been reached. If so, output the optimal initial parameters. If not, return to the second step and repeat until the number of iterations has been reached to obtain the minimum mean square error and the optimal value of the initial parameters.

[0027] The optimal parameters are substituted into the entry control circulation model to construct a riser system status prediction model.

[0028] Furthermore, in said S10, the determined state of the deepwater drilling riser system includes the movement state of the drilling platform, the rotation angle state of the upper flexible joint, the state of the ocean current velocity profile and the state of the local riser acceleration.

[0029] Furthermore, the motion state of the drilling platform, the rotation angle state of the upper flexible joint, and the current velocity profile state are solved by establishing a mechanical simulation model of the deep-water drilling riser system to obtain state parameter values; the acceleration state of the local riser is obtained through a monitoring device.

[0030] Furthermore, in S20, the mechanical simulation model of the deepwater drilling riser system includes: a platform motion model, a tensioner model, a riser mechanical model, and an environmental load model.

[0031] Furthermore, in S20, the database is established by using the mechanical simulation model of the deep-water drilling riser system to establish the correspondence between parameters such as ocean current, wave amplitude and wave period and the response of the riser system, and the parameters such as ocean current, wave amplitude and wave period are divided using a slider time window.

[0032] Furthermore, the S40 specifically includes the following steps:

[0033] Selecting a test set from the database, wherein the test set only retains the drilling platform motion state, the upper flexible joint rotation angle, the ocean current velocity, and the riser system response;

[0034] Inputting the movement state of the drilling platform, the rotation angle of the upper flexible joint, and the ocean current velocity into the deepwater drilling riser system state prediction model;

[0035] Compare the riser system acceleration prediction result output by the deepwater drilling riser system status prediction model with the riser system local acceleration monitoring response.

[0036] Furthermore, in S50, the deepwater drilling riser system state prediction model is optimized by using an ant colony algorithm.

[0037] Compared with the existing technology, the beneficial effects of the deep-water drilling riser system status prediction method, medium and system provided by the present invention based on monitoring are: the present invention provides a deep-water drilling riser system status prediction method based on monitoring, which combines the ant colony algorithm and the gated recurrent neural network to establish a prediction model for training, thereby improving the training accuracy of the prediction model. After the prediction model training is completed, the established prediction model is evaluated and verified in the testing process, and the ant colony algorithm is used again for optimization based on the evaluation and verification results to ensure that the prediction model has better adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0039] Figure 1 The flowchart of a method for predicting the status of a deepwater drilling riser system based on monitoring is shown;

[0040] Figure 2 This is a flow chart for optimizing the initial parameters of a gated loop network for a monitoring-based deepwater drilling riser system status prediction method.

[0041] Figure 3 The present invention is a prediction flow chart of a deepwater drilling riser system status prediction system based on monitoring;

[0042] Figure 4 A schematic diagram of the external information of a model for a deepwater drilling riser system status prediction system based on monitoring;

[0043] Figure 5 Schematic diagram of a deepwater drilling riser system status prediction system based on monitoring;

[0044] In the accompanying drawings, the components represented by the reference numerals are as follows:

[0045] 1. Drilling platform; 10. Doppler current meter; 11. Ocean current; 2. Riser; 3. Upper flexible joint; 4. Lower flexible joint; 5. Bottom riser assembly; 6. Underwater blowout preventer; 7. Platform motion sensor; 8. Angle sensor; 9. Acceleration sensor. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0047] like Figure 1 FIG. 1 is a first embodiment of a method for predicting the state of a deepwater drilling riser system based on monitoring provided by the present invention. In this embodiment, the method includes the following steps:

[0048] S10: Determine the status of the deepwater drilling riser system that needs to be monitored;

[0049] S20: Use the deepwater drilling riser system mechanical simulation model to establish a database;

[0050] S30: Establish a deepwater drilling riser system status prediction model based on ant colony algorithm and gated recurrent neural network;

[0051] S40: Testing the deepwater drilling riser system status prediction model;

[0052] S50: Optimizing the state prediction model for deepwater drilling riser systems.

[0053] like Figure 3 As shown, in the above technical solution, S30 specifically includes the following steps:

[0054] Preprocess the data used for model training in the simulation database: Use the Min-Max normalization method to preprocess the data and normalize the data to the [0,1] interval. The normalization method formula is as follows:

[0055]

[0056] In the formula, y is the normalized data; x is the unprocessed original data; max and min are the maximum and minimum values ​​of the data respectively;

[0057] The input layer is the processed data of rotation angle, displacement, current velocity and wave height, and the output is the state of the deepwater drilling riser system;

[0058] Design the model structure, including the number of network layers and nodes in each layer, and determine hyperparameters such as learning rate and hidden layer dimension;

[0059] Use the ant colony optimization algorithm to optimize the initial weight thresholds between each layer;

[0060] The model is trained using the AdamW optimizer, and after the model predicts the output, the model is evaluated based on the size of the loss function.

[0061] Determine whether the maximum number of iterations has been reached. If so, the model training ends. Otherwise, the loss function is updated through backpropagation, and the model training continues until the maximum number of iterations is reached.

[0062] After the model training is completed, the output is the predicted state of the deepwater drilling riser system.

[0063] like Figure 2 As shown, further, in the above technical solution, using the ant colony optimization algorithm to optimize the initial weight thresholds between each layer specifically includes the following steps:

[0064] Initialize ant colony parameters and initial position of ant colony;

[0065] Calculation of the probability of ant selection interval: Each ant starts from the first parameter and chooses the subinterval to reach the next parameter according to the probability. In the tth cycle, the probability of ant k choosing the subinterval at parameter i is:

[0066]

[0067] Where, the relative importance of α and β reaction pheromones and heuristic information, n ij Represents the heuristic function of the jth interval of the i-th parameter, the value is d mn To predict the distance between the parameters m and n to be optimized in the model, there is no actual distance between the parameters in this model, and a random number between (0,1] is selected to represent the heuristic function;

[0068] The initial parameters of the gated recurrent network include the weights between the input and hidden layers, the weights between hidden layers, and the threshold of the output layer. These parameters to be optimized are considered the number of city nodes corresponding to the city. After an ant has completed all nodes according to the probability interval, a set of feasible solutions is formed. This is used to initialize the riser system state prediction model, which is trained to obtain the model mean square error. After all ants have completed all parameters to be optimized, the pheromone is updated.

[0069] Pheromone Update:

[0070] τ ij (t+1)=(1-ρ)*τ ij (t)+Δτ ij (t);

[0071]

[0072] Where 1-ρ represents the persistence coefficient of pheromone, Δτ ij represents the pheromone increment, is the mean square error of the watertight pipe system state prediction model constructed by the kth ant in the tth cycle, Q is the pheromone intensity which is a constant value;

[0073] Iterate the ant colony to find the optimal value, and determine whether the number of iterations has been reached. If so, output the optimal initial parameters. If not, return to the second step and repeat until the number of iterations has been reached to obtain the minimum mean square error and the optimal value of the initial parameters.

[0074] The optimal parameters are substituted into the entry control circulation model to construct a riser system status prediction model.

[0075] Furthermore, in the above technical solution, in S10, the determined state of the deepwater drilling riser system includes the movement state of the drilling platform, the rotation angle state of the upper flexible joint, the current velocity profile state and the local riser acceleration state.

[0076] Furthermore, in the above technical solution, in S20, the mechanical simulation model of the deepwater drilling riser system includes: a platform motion model, a tensioner model, a riser mechanical model, and an environmental load model.

[0077] Furthermore, in the above technical solution, in S20, the database is established by using a mechanical simulation model of a deep-water drilling riser system to establish a correspondence between parameters such as ocean currents, wave amplitudes, and wave periods and the response of the riser system. Parameters such as ocean currents, wave amplitudes, and wave periods are divided using a slider time window.

[0078] Furthermore, in the above technical solution, S40 specifically includes the following steps:

[0079] A test set is selected from the database, which only retains the drilling platform motion state, upper flexible joint rotation angle, ocean current velocity, and riser system response;

[0080] The drilling platform motion state, upper flexible joint rotation angle, and ocean current velocity are input into the deepwater drilling riser system state prediction model.

[0081] Comparison of deepwater drilling riser system status prediction model output and riser system response.

[0082] Furthermore, in the above technical solution, in S50, the deepwater drilling riser system state prediction model is optimized by using an ant colony algorithm.

[0083] The following are the meanings of the symbols in the formulas of the present invention:

[0084] Table 1 Meaning of symbols in the invention formula

[0085]

[0086]

[0087] A second aspect of the present invention provides a computer-readable storage medium comprising a monitoring-based deepwater drilling riser system status prediction method according to any one of claims 1 to 8.

[0088] A third aspect of the present invention provides a monitoring-based deepwater drilling riser system state prediction system, comprising: an angle sensor 8, an acceleration sensor 9, a Doppler current meter 10, a motion reference unit, a data acquisition card, and a computer;

[0089] The rotation angle sensor 8 and the acceleration sensor 9 are used to monitor the motion state of the watertight pipe 2, the Doppler current meter 10 is used to detect the ocean current velocity, the motion reference unit is used to monitor the motion state of the platform, and the data acquisition card is used to monitor the transmission of data and communicate with the computer. The computer stores the computer-readable storage medium as claimed in claim 9.

[0090] It is used to store the received data, and at the same time, calculate the received signals according to the prediction algorithm to output the predicted watertight pipe system status.

[0091] like Figure 4 The figure shows a schematic diagram of a deepwater drilling riser system, including a drilling platform 1, a riser 2, an upper flexible joint 3, a lower flexible joint 4, a bottom riser assembly 5, an underwater blowout preventer 6, and in addition, platform motion sensors 7, angle sensors 8, riser local acceleration sensors 9, current velocity meters 10, and ocean currents 11 in the marine environment.

[0092] like Figure 5 As shown, the physical structure of the riser system includes a drilling platform 1, a riser 2, an upper flexible joint 3, a lower flexible joint 4, a bottom riser assembly 5, and a subsea blowout preventer 6. The platform is connected to the riser 2 via the upper and lower flexible joints 3 and 4, which reduce the bending stiffness of the connection. Waves and currents 11 impose loads on the riser 2. To ensure the safe and efficient operation of the riser system, monitoring sensors are used to monitor its structure, and prediction methods are used to obtain predicted values ​​for the riser system's status.

[0093] Specifically, the principles of the present invention are: determining the status of the deep-water drilling riser system that needs to be monitored; using the mechanical simulation model of the deep-water drilling riser system to establish a database; establishing a deep-water drilling riser system status prediction model based on the ant colony algorithm and the gated recurrent neural network; testing the deep-water drilling riser system status prediction model; and optimizing the deep-water drilling riser system status prediction model.

Claims

1. A monitoring-based deepwater drilling riser system status prediction method, characterized in that: The following steps are involved: S10: Determine the status of the deepwater drilling riser system that needs to be monitored; S20: Use the deepwater drilling riser system mechanical simulation model to establish a database; S30: Establish a deepwater drilling riser system status prediction model based on ant colony algorithm and gated recurrent neural network; S40: Testing the deepwater drilling riser system status prediction model; S50: Optimizing the state prediction model for deepwater drilling riser systems.

2. The method for predicting the status of a deepwater drilling riser system based on monitoring according to claim 1, characterized in that: The S30 specifically includes the following steps: Preprocess the data in the simulation database for model training; Establishing the input and output of the deepwater drilling riser system state prediction model, wherein the input layer is the data processed rotation angle, displacement, current velocity and wave height data, and the output is the deepwater drilling riser system state quantity; Designing the structure of the deepwater drilling riser system state prediction model, including the number of network layers and nodes in each layer, and determining hyperparameters such as learning rate and hidden layer dimension; Use the ant colony optimization algorithm to optimize the initial weight thresholds between each layer; The model is trained using the AdamW optimizer, and after the model predicts the output, the model is evaluated based on the size of the loss function. Determine whether the maximum number of iterations has been reached. If so, the model training ends. Otherwise, the loss function is updated through backpropagation, and the model training continues until the maximum number of iterations is reached. After the model training is completed, the output is the predicted state of the deepwater drilling riser system.

3. The method for predicting the status of a deepwater drilling riser system based on monitoring according to claim 2, characterized in that: The method of using the ant colony optimization algorithm to select the optimal initial weight threshold between each layer specifically includes the following steps: Initialize the initial position of the ant colony; Calculate the probability that the k-th ant selects the i-th subinterval to obtain a set of parameter values, which are used to represent a set of initial weight thresholds generated after the k-th ant traverses all intervals, where k represents the ant number and i represents the number of intervals; After the kth ant has walked through all nodes according to the probability interval, a set of feasible solutions of the gated recurrent network model is obtained, and the state prediction model of the deepwater drilling riser system is initialized with the feasible solutions; The deepwater drilling riser system state prediction model is trained to obtain the model mean square error, and the pheromone is updated after all ants have walked through all parameters to be optimized; Iterate the ant colony to find the optimal value, and determine whether the number of iterations has been reached. If so, output the optimal initial parameters. If not, return to the second step and repeat until the number of iterations has been reached to obtain the minimum mean square error and the optimal value of the initial parameters. The optimal parameters are substituted into the entry control circulation model to construct a riser system status prediction model.

4. The method for predicting the status of a deepwater drilling riser system based on monitoring according to claim 3, characterized in that: In said S10, the determined deepwater drilling riser system state includes the drilling platform movement state, the upper flexible joint rotation angle state, the ocean current velocity profile state and the local riser acceleration state.

5. The method for predicting the status of a deepwater drilling riser system based on monitoring according to claim 4, characterized in that: In the above S20, the mechanical simulation model of the deepwater drilling riser system includes: a platform motion model, a tensioner model, a riser mechanical model, and an environmental load model.

6. The method for predicting the status of a deepwater drilling riser system based on monitoring according to claim 5, characterized in that: In S20, the database is established by using the deep-water drilling riser system mechanical simulation model to establish the correspondence between parameters such as ocean current, wave amplitude and wave period and the response of the riser system. The parameters such as ocean current, wave amplitude and wave period are divided using a slider time window.

7. The method for predicting the status of a deepwater drilling riser system based on monitoring according to claim 6, characterized in that: The S40 specifically includes the following steps: Selecting a test set from the database, wherein the test set only retains the drilling platform motion state, the upper flexible joint rotation angle, the ocean current velocity, and the riser system response; The movement state of the drilling platform, the rotation angle of the upper flexible joint, and the ocean current velocity are input into the deepwater drilling riser system state prediction model. Compare the deepwater drilling riser system state prediction model output with the riser system response.

8. The method for predicting the status of a deepwater drilling riser system based on monitoring according to claim 7, characterized in that: In S50, the deepwater drilling riser system state prediction model is optimized by using an ant colony algorithm.

9. A computer-readable storage medium, characterized in that It comprises a deepwater drilling riser system status prediction method based on monitoring as described in any one of claims 1-8.

10. A monitoring-based deepwater drilling riser system status prediction system, comprising: a rotation angle sensor, an acceleration sensor, a Doppler flowmeter, a motion reference unit, a data acquisition card, and a computer; The rotation angle sensor and the acceleration sensor are used to monitor the motion state of the watertight pipe, the Doppler current meter is used to detect the ocean current velocity, the motion reference unit is used to monitor the motion state of the platform, the data acquisition card is used to monitor data transmission and communicate with the computer, and the computer stores the computer-readable storage medium according to claim 9. It is used to store the received data, and at the same time, calculate the received signals according to the prediction algorithm to output the predicted watertight pipe system status.

Images