Marine riser-free drilling double-pump system working condition intelligent switching fluctuation minimization method and system

Abstract

The invention discloses a marine riser-free drilling double-pump system working condition intelligent switching fluctuation minimization method. The method comprises the steps of obtaining real-time working condition operation data of a marine riser drilling double-pump system; analyzing the working condition operation data through a preset fusion decision model, making a decision, and selecting an operation working condition; searching an optimal solution set according to the selected operation condition, finding a matched optimal solution, and setting corresponding process parameters according to the optimal solution to minimize the condition switching fluctuation; wherein the acquisition process of the optimal solution set is as follows: a prediction model of each working condition is pre-constructed, the prediction model specifically predicts flow fluctuation and pressure fluctuation of the current working condition in the process from starting to stopping according to the collected rotating speed of the double pumps and the opening speed and the opening degree of the multiple valves under the corresponding working condition, and a prediction data set of each working condition is obtained; and calculating an optimal solution set in each working condition prediction data set through a multi-objective optimization algorithm. According to the method, intelligent working condition switching with fluctuation minimization can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control of deep-sea drilling equipment, and in particular to a method and system for minimizing fluctuations in intelligent switching of working conditions of a dual-pump system for riser-less drilling. Background Art

[0002] The control system of traditional dual-pump riserless drilling relies on manual intervention or fixed rules. Under sudden operating conditions such as single pump failure and load overload, it is unable to automatically and accurately determine and switch to series / parallel mode in real time. This leads to problems such as sudden pressure changes and flow fluctuations during the switching process, threatening drilling safety (such as equipment fatigue and blowout risks). Existing methods lack quantitative modeling of flow and pressure fluctuations and multi-objective collaborative optimization, making it difficult to quickly find the optimal control parameter combination that takes into account both fluctuation indicators in complex nonlinear systems. Furthermore, they have not formed a closed-loop control link of "mode decision-parameter optimization-execution feedback", and are unable to meet the high real-time and reliability requirements of deep-sea drilling. Domestic and foreign research mostly remains at the level of theoretical simulation or single algorithm application, lacking systematic solutions. Especially in emergency conditions (such as single pump failure, sudden load changes, etc.), it is impossible to achieve intelligent identification of operating modes and dynamic adaptive adjustment of control parameters. Summary of the Invention

[0003] The main purpose of the present invention is to provide a method for automatically selecting the optimal working condition based on the real-time working condition operation data of the watertight pipe drilling dual pump system, under which the flow fluctuation and pressure fluctuation can be minimized.

[0004] The technical solution adopted in the present invention is: A method for minimizing fluctuations in intelligent switching of operating conditions of a dual-pump system for riserless drilling is provided, comprising the following steps: Obtain real-time operating data of the dual-pump system for riser drilling; Analyze the operating data through the preset fusion decision model, make a decision, and select an operating condition, including single pump operation, dual pump parallel operation, or dual pump series operation; Search for the optimal solution set based on the selected operating conditions, find the matching optimal solution, and set the corresponding process parameters based on the optimal solution to minimize the fluctuation of operating condition switching; The process of obtaining the optimal solution set is as follows: pre-constructing a prediction model for each working condition. The prediction model predicts the flow fluctuation and pressure fluctuation during the start-up to stop process of the current working condition based on the collected speed of the dual pumps, the opening speed and opening degree of multiple valves under the corresponding working conditions, and obtains the prediction data set for each working condition. Then, the optimal solution set in the prediction data set for each working condition is calculated through a multi-objective optimization algorithm.

[0005] According to the above technical solution, the operating data include the inlet and outlet pressures, inlet and outlet flow rates, total trunk flow rate and total trunk pressure of the dual pumps.

[0006] Following the above technical solution, the preset fusion decision model selects a working condition based on rule-driven and data-driven respectively, and selects the working condition with higher confidence as the final decision.

[0007] Following the above technical solution, rule-driven is based on a rule base, which includes multiple rules, each of which includes a triggering condition, an execution action and a priority.

[0008] Following the above technical solution, data is driven based on a pre-trained neural network classification decision model, which is trained based on the data collected from each working condition and the corresponding working condition labels.

[0009] Following the above technical solution, the prediction model of each working condition is constructed using a three-layer feedforward neural network based on the nonlinear characteristics of the dual-pump system for riserless drilling, including an input layer, a hidden layer, and an output layer, where the number of nodes in the input layer is equal to the dimension of the control parameters.

[0010] Following the above technical solution, the multi-objective optimization algorithm is the Rime multi-objective optimization algorithm or the NSGA-II algorithm.

[0011] Following the above technical solution, when searching for the optimal matching solution, the solution that minimizes pressure fluctuation is preferably selected.

[0012] The present invention also provides a system for minimizing fluctuations in intelligent switching of working conditions of a dual-pump system for riser-less drilling, comprising: Operation data acquisition module, used to obtain real-time operating data of the dual-pump system in riser drilling; The fusion decision module is used to analyze the operating data of the working condition through the preset fusion decision model and make a decision to select an operating condition, including a single pump operating condition, a dual pump parallel operating condition, or a dual pump series operating condition; The process parameter setting module is used to find the optimal solution set according to the selected operating conditions, find the matching optimal solution, and set the corresponding process parameters according to the optimal solution to minimize the fluctuation of the operating condition switching; Among them, the optimal solution set is obtained through the optimal solution set construction module, which is specifically used to pre-construct a prediction model for each working condition. The prediction model predicts the flow fluctuation and pressure fluctuation of the current working condition from start to stop based on the collected speed of the dual pumps, the opening speed and opening degree of multiple valves under the corresponding working conditions, and obtains the prediction data set of each working condition. Then, the optimal solution set in the prediction data set of each working condition is calculated through a multi-objective optimization algorithm.

[0013] The present invention also provides a computer storage medium storing a computer program executable by a processor, wherein the computer program executes the method for minimizing fluctuations in intelligent switching of working conditions of a dual-pump system for watertight drilling without a riser as described in the above technical solution.

[0014] The beneficial effects of the present invention are as follows: the present invention realizes intelligent decision-making and emergency switching control of the operating mode of the dual-pump pumping system in the watertight pipeless drilling system. Based on the real-time operating data of the dual-pump system, the process parameters under the optimal working condition are found by integrating the decision model and combining the prediction model of each working condition to switch the working condition, thereby achieving the minimum control of flow and pressure fluctuations during the working condition switching process, thereby forming a complete intelligent link from working condition identification to precise control. In addition, based on dual-index modeling, the present invention takes flow and pressure fluctuation amplitude as the joint optimization target for the first time, accurately characterizing the stability of the system switching process. Compared with traditional single-index control (such as controlling only pressure or flow), it significantly improves the comprehensive control performance under complex working conditions.

[0015] Furthermore, the fusion decision model integrates rule-driven and data-driven decision-making mechanisms, and combines the prediction models of various working conditions with multi-objective optimization algorithms. It can solve the problems of adaptive switching of dual-pump system working modes in complex deep-sea environments and minimizing flow and pressure fluctuations during switching. It is suitable for emergency collaborative control scenarios of deepwater oil and gas drilling operations.

[0016] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are 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 work.

[0018] Figure 1A Flowchart 1 of a method for minimizing fluctuations in intelligent switching of operating conditions of a dual-pump system for riserless drilling according to an embodiment of the present invention; Figure 1B This is a process for minimizing fluctuations in intelligent switching of dual-pump systems for riser-less drilling according to an embodiment of the present invention. Figure 2 ; Figure 2 This is a flow chart of fusion decision making according to an embodiment of the present invention; Figure 3 Schematic diagram of a feedforward neural network structure according to an embodiment of the present invention; Figure 4 This is a basic flow chart of a multi-objective optimization algorithm according to an embodiment of the present invention; Figure 5 This is a schematic structural diagram of a dual-pump system for riserless drilling according to an embodiment of the present invention; Figure 6 yes Figure 5 Schematic diagram of the control piping structure on both sides of the dual pump system. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0020] It should be noted that the illustrations provided in the embodiments of the present invention are only schematic illustrations of the basic concept of the present invention. Therefore, the drawings only show components related to the present invention and are not drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component can be changed at will, and the component layout type may also be more complicated.

[0021] In the present invention, it should also be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer" and the like are used to indicate positions or locations based on those shown in the accompanying drawings. These terms are intended solely to facilitate the description of the present application and to simplify the description. They are not intended to indicate or imply that the devices or components referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present application. Furthermore, the terms "first" and "second" are used solely for descriptive and distinguishing purposes and should not be construed as indicating or implying relative importance.

[0022] In addition, it should be noted that the features of the various embodiments of the present invention may be combined or coupled in part or in whole, and, as will be appreciated by those skilled in the art, may interact and operate in different ways. Each embodiment may be implemented independently of one another or in an associated relationship.

[0023] like Figure 1A As shown, the method for minimizing fluctuations in intelligent switching of dual-pump systems for riserless drilling according to an embodiment of the present invention includes the following steps: S1. Obtain real-time operating data of the dual-pump system for riser drilling; S2. Analyze the operating data using a preset fusion decision model and make a decision to select an operating condition, including a single-pump operating condition, a dual-pump parallel operating condition, or a dual-pump series operating condition; S3. Searching for an optimal solution set based on the selected operating conditions, finding a matching optimal solution, and setting corresponding process parameters based on the optimal solution to minimize operating condition switching fluctuations; Among them, the process of obtaining the optimal solution set in step S3 is: pre-constructing a prediction model for each working condition, the prediction model specifically predicts the flow fluctuation and pressure fluctuation during the current working condition from start to stop based on the collected speed of the dual pumps, the opening speed and opening degree of multiple valves under the corresponding working conditions, and obtains the prediction data set of each working condition, and then calculates the optimal solution set in the prediction data set of each working condition through a multi-objective optimization algorithm.

[0024] Among them, this embodiment is mainly based on a dual pump system for riser-less drilling, which is used for ten-stage seabed mud lifting. Its structural diagram is shown in Figure 5 , which mainly includes ten-stage centrifugal pumps, drill pipe joints, straightening blocks and other units. The ten-stage centrifugal pump is designed as a series structure. Through the principle of multi-stage energy superposition, the mud obtains progressively stronger braking force when flowing through each boosting unit, effectively responding to the transportation needs of deep-sea high-pressure conditions. The key parts of the pump body are made of wear-resistant materials to adapt to the environment of a large amount of sand and mineral particles in the seabed mud, and are suitable for complex seabed environments. The drill pipe joint is used to connect the pump and the drill bit to ensure a stable physical connection. The ball valve adjusts the mud flow to control the in and out of the fluid. The straightening block helps support the pump body, reduces vibration and ensures the smooth operation of the pump. The new four-way provides flexible path switching for the fluid and optimizes the flow control of the pump. The skid assembly and skid facilitate the installation and disassembly of the pump components and improve maintenance efficiency. The trolley adapter ensures the compatibility of the pump with the mobile platform, facilitates the movement and arrangement of the pump body, and the clamping mechanism provides additional assembly stability to reduce the risk of failure. Pipes are connected at both ends of the dual pump system, such as Figure 6 As shown, it mainly includes valves, 5 pressure gauges and 3 flow meters.

[0025] This implementation, based on dual-index modeling, uses flow and pressure fluctuation amplitude as joint optimization targets, accurately characterizing the stability of the system switching process. Compared to traditional single-index control (such as controlling only pressure or flow), it significantly improves comprehensive control performance under complex operating conditions. It can solve the problem of adaptively switching the operating modes of dual-pump systems in complex deepwater environments and minimizing flow and pressure fluctuations during switching. It is suitable for emergency collaborative control scenarios in deepwater oil and gas drilling operations.

[0026] Furthermore, in step S1, the real-time operating data collected includes the inlet and outlet pressures, inlet and outlet flow rates, total trunk flow rate, and total trunk pressure of the dual pumps. Specifically, it includes the inlet and outlet pressures of pump 1, the inlet and outlet pressures of pump 2, the inlet and outlet flow rates of pump 1, the inlet and outlet flow rates of pump 2, the total trunk flow rate, and the total trunk pressure. The operating conditions (i.e., working modes) of the dual-pump system for riserless drilling primarily include single pump, dual-pump series, and dual-pump parallel operation. This invention addresses the emergency coordinated control needs of dual-pump systems for deep-sea drilling, primarily focusing on adaptive selection of these three operating modes (single pump / series / parallel) and minimizing flow and pressure fluctuations during switching.

[0027] like Figure 1B As shown, operating mode decision-making primarily relies on a rule-data fusion decision-making model. A dynamic confidence selection mechanism leverages the complementary strengths of the two approaches, addressing the challenge of mode switching decisions under complex operating conditions. This fusion decision-making model primarily determines the optimal operating mode based on real-time monitoring data (flow rate, pressure, pump status, etc.). For the selected mode (single pump / series / parallel), the optimal control parameter combination is generated by constructing fluctuation characteristic indicators, predictive models for each operating condition, and a multi-objective optimization algorithm with the goal of minimizing flow and pressure fluctuations.

[0028] The fusion decision model primarily utilizes a dynamic decision-making mechanism, two mathematical models, and a dynamic confidence selection mechanism to leverage the complementary strengths of two approaches (rule-driven and data-driven). This dynamic confidence selection mechanism sets initial confidence levels (reflecting the method's prior reliability) and judgment accuracy (historical performance statistics) for both rules and algorithms. The system compares the current confidence levels of the two approaches in real time, prioritizing the method with the higher confidence level for decision making. High-confidence rule-based approaches ensure interpretability and security, while high-confidence algorithmic approaches meet the complex data-driven needs of pattern recognition.

[0029] like Figure 2 As shown in the figure, the real-time data collected by the dual-pump system includes: pump 1 inlet and outlet pressure, pump 2 inlet and outlet pressure, pump 1 inlet and outlet flow, pump 2 inlet and outlet flow, total trunk flow, and total trunk pressure; the fusion decision model uses the above data to determine the operating conditions and select the corresponding mode (single pump operation, dual pumps in parallel, dual pumps in series). The rules of the rule-driven part of the fusion decision model are shown in Table 1 below: Table 1 Rule-driven

[0030] The algorithm-driven part (i.e., the data-driven part) uses differential evolution support vector machine (ADE-SVM) for decision classification. Here, first, through orthogonal experiments, the data collected from each working condition (including single pump operation, two pumps in series, and two pumps in parallel) are obtained in advance. The data collected by the system is used as input and manually labeled. Three working conditions, single pump operation, two pumps in series, and two pumps in parallel, are specified and trained. The obtained training model is used for data-driven classification decision-making.

[0031] Initial confidence: Initial confidence of rule-driven method , algorithm-driven method confidence , with a value range of [0,1]. The confidence of the rule-driven method is given by expert experience, while the confidence of the algorithm-driven method is determined by the training effect.

[0032] Initial judgment accuracy: initial accuracy of rule-driven method , the initial accuracy of the algorithm-driven method , preset based on historical data or expert experience; Decision selection logic of the fusion decision model: When data x is input, the system compares the current confidence level and selects the method with higher confidence to perform the judgment:

[0033] in is the rule determination function (operating conditions are selected according to the rule-driven method), For algorithm decision functions (such as machine learning models), operating conditions are selected based on a data-driven approach.

[0034] To prevent the system from becoming dependent on a single method, this embodiment introduces a forced exploration strategy. This strategy sets an exploration trigger period (e.g., every 50 data runs). Upon reaching this period, the decision method is randomly selected (with a 50% probability for each of the rule and algorithm), rather than relying on confidence levels. This mechanism ensures that both methods receive continuous validation opportunities, avoiding decision blind spots caused by shifting data distribution or lags in the rule base. It is particularly useful for identifying rare fault conditions.

[0035] Specifically, set the exploration trigger cycle n. After processing k data times, the knth forced random selection judgment method is used, and the probability distribution is uniform distribution: The probability of choosing rule-driven approach = the probability of choosing algorithm-driven approach = 0.5 The forced exploration strategy mechanism introduces randomness to ensure that the two methods continue to obtain verification opportunities. The formula is expressed as: Explore trigger conditions Random selection Where count is the cumulative number of times data is processed. To prevent abnormal fluctuations in confidence or confidence, boundary constraints and system stability are introduced. The following boundary constraints are set: Confidence Range: ,in It is a very small positive number to avoid decision failure caused by numerical overflow or extreme values, such as limiting the confidence range to [0.5, 0.9].

[0036] And the confidence adopts a dynamic adjustment strategy, Specifically, the judgment result processing and confidence update of the fusion decision model are as follows: The basic confidence adjustment mechanism dynamically adjusts the confidence of the corresponding method based on the judgment results.

[0037] If the rule method is successful (the result is consistent with the true label), then

[0038] If the judgment fails,

[0039] The same applies to algorithm-driven approaches, where As the basic adjustment step, is the accuracy rate of the current rule method.

[0040] The specific rewards for success and penalties for failure are: Successful methods receive increased confidence (reflecting their current effectiveness), while failed methods receive decreased confidence (reducing reliance on inefficient methods). The magnitude of the confidence adjustment is linked to the method's historical accuracy. For example, methods with a high historical accuracy rate receive a larger confidence increase upon success, strengthening the preference for stable and reliable methods.

[0041] A continuous success negative feedback compensation mechanism is employed. For methods that consistently succeed, this mechanism reduces the attenuation factor of their confidence gain, limiting the rapid growth of confidence and preventing the rigidification of decisions caused by the "stronger, stronger" mentality. For example, after a method succeeds three times in a row, its confidence gain is automatically attenuated by 30%, preserving competitive opportunities for other methods. In this embodiment, in order to suppress the excessive dominance of a single method, when the number of consecutive successes of a method m>N, its confidence gain attenuation factor is reduced. :

[0042] At this time, the confidence gain is corrected to:

[0043] The confidence growth rate of the advantage method is limited by nonlinear adjustment of the attenuation factor.

[0044] By adopting continuous failure compensation, when a method fails continuously, the incremental coefficient of its confidence adjustment is increased, so that it can obtain more "exploration opportunities" in subsequent judgments and avoid being suppressed for a long time due to accidental errors.

[0045] In this embodiment, when a method fails several times continuously When the increment coefficient is increased To enhance exploration:

[0046] The confidence adjustment formula for failure is modified to avoid the long-term suppression of weak methods due to occasional failures

[0047] A dynamic attenuation mechanism and threshold balancing strategy are adopted. The dynamic attenuation mechanism is specifically as follows: the confidence of rules and algorithms automatically decays over time, simulating the impact of equipment performance changes or environmental parameter drift in reality, ensuring that the impact of historical performance decreases over time, and the system can respond to changes in current working conditions in a timely manner.

[0048] The specific purpose of the difference threshold-triggered random selection is: when the confidence difference between the rule and the algorithm is less than the preset threshold (such as 0.02), it is considered that the capabilities of the two are close. At this time, probabilistic selection (non-deterministic selection) is adopted to avoid decision bias caused by small differences and maintain the system's sensitivity to fluctuations in method performance. The time decay factor is introduced to simulate the timeliness of the method, and the confidence decays exponentially with the time step t:

[0049] in, They are the decay rates of rules and algorithmic methods, respectively. This mechanism ensures that the impact of historical data performance decreases over time and adapts to environmental changes.

[0050] Use confidence equalization and random selection strategy to set the difference threshold , when the confidence difference between the two methods is less than the threshold ( ), and the two are considered to have equal capabilities. In this case, probability selection is used instead of deterministic selection:

[0051] Avoid decision rigidification through probabilistic selection and keep the system sensitive to fluctuations in method performance.

[0052] After obtaining the prediction data set of each working condition, specifically, the optimal solution set in the prediction data set of each working condition is calculated through multi-objective optimization problem modeling.

[0053] Introduction Total Variation (TV) As a quantitative indicator of flow and pressure fluctuations, it reflects the overall change amplitude of the signal in the time domain: Traffic fluctuation indicators:

[0054] Pressure fluctuation index:

[0055] in, 、 、 、 are the flow and pressure monitoring values ​​at time t and t+1, and T is the sampling period. The optimization goal is to minimize the flow fluctuation index and pressure fluctuation index.

[0056] In one embodiment of the present invention, the prediction model of each working condition can adopt a feedforward neural network agent model. In view of the nonlinear characteristics of the dual pump system, a three-layer feedforward neural network (input layer - hidden layer - output layer) is used to establish the mapping relationship between the control parameters and the fluctuation index. The network structure is shown in Figure 3 : First, through orthogonal experiments, the data collected for each working condition (including single pump operation, dual pumps in series, dual pumps in parallel, single pump in series, single pump in parallel, parallel-to-series, and series-to-parallel) are obtained in advance. Here, the relevant parameter data are set as input (including pump 1 speed, pump 2 speed, valve 1-5 opening speed, valve 1-5 opening), and the flow fluctuation and pressure fluctuation during the start-up and stop of each working condition are used as output, and prediction is performed through the neural network. Input layer: The number of nodes is equal to the control parameter dimension m; Hidden layer: uses ReLU activation function, and the number of nodes is determined by trial and error (e.g. 2m+1) Output layer: The number of nodes is 2 (corresponding to There is no activation function. The training data is generated through system measurement and simulation, and the loss function is the mean square error:

[0057] in, (u k ) is the network prediction value, N is the number of training samples, F (u k ) is the weighted objective function, u k is the control parameter vector.

[0058] Specifically, the multi-objective optimization algorithm may be the Rime Multi-Objective Optimization Algorithm or the NSGA-II algorithm. This embodiment specifically adopts the Rime Multi-Objective Optimization Algorithm (MS-MOA).

[0059] Multi-objective optimization algorithm for rime based on surrogate model , the improved algorithm is used to search for the Pareto frontier. The basic process of the algorithm is shown in Figure 4 , its core steps are as follows: (1) Population initialization Generate population ,Include individuals, each of which corresponds to a set of control parameters ,satisfy .

[0060] (2) Fitness evaluation Calculate the non-dominated ranking of each individual and congestion distance .

[0061] Non-dominated sorting: If individual A is not worse than individual B in all objectives and is better in at least one objective, then A dominates B, and the lower the rank, the better; Crowding distance: The density estimation of individuals of the same level in the target space, used to maintain population diversity.

[0062] (3) Rime Iteration Operator Simulate the growth process of rime crystals and generate offspring populations through the following operations

[0063] A. Local search: Apply Gaussian perturbation to the current individual. is the standard deviation of the Gaussian perturbation, is a standard normal distribution random number, and u is a control parameter vector.

[0064]

[0065] B. Crossover fusion: Select parent individuals for parameter crossover, such as arithmetic crossover:

[0066] is the cross coefficient.

[0067] C. Elite retention: merge the parent and child generations, and select the best one according to the non-dominated sequence and congestion calculation. (population size) individuals enter the next generation (4) Optimal solution generation: Select the optimal solution from the Pareto front according to the engineering requirements. If the priority is to control flow fluctuations, select The smallest solution. If you are pursuing performance, choose a weighted objective function. The smallest solution.

[0068] When the fusion decision system determines that it is necessary to switch from single pump mode to series / parallel mode (for example, if a single pump is detected to be overloaded or with insufficient flow), the multi-objective optimization process is automatically triggered.

[0069] When selecting the optimal process parameters, the optimal solution sets for optimizing each operating condition are organized into an optimization database in advance. After the system makes a fusion decision based on real-time data, the solution sets in the optimization database are matched according to the changes in the system's operating conditions (for example, if switching from a single pump to parallel operation, the optimization results of switching from a single pump to parallel operation are matched). The solution set that minimizes the system flow and pressure fluctuations is selected to obtain the optimal input parameters (pump 1 speed, pump 2 speed, valve 1-5 opening speed, valve 1-5 opening) for adjustment. When making the selection here, priority is given to parameters that minimize pressure fluctuations (to prevent blowouts, rock chip sedimentation, etc.), but other optimal solutions can also be selected.

[0070] In summary, the steps for intelligent switching of the dual-pump system operating conditions in riserless drilling are roughly as follows: the sensor collects flow Q, pressure P and pump status data in real time; the fusion decision model selects the current operating mode, searches for the optimal solution set based on the selected operating conditions, and finds the matching optimal solution; if the mode is switched, the control parameters are found according to the optimal solution, and the actuator adjusts the pump speed, valve opening, f valve opening speed and other parameters according to the control parameters.

[0071] In order to implement the above method embodiment, the present invention further provides a riser-less drilling dual-pump system operating condition intelligent switching fluctuation minimization system, comprising: Operation data acquisition module, used to obtain real-time operating data of the dual-pump system in riser drilling; The fusion decision module is used to analyze the operating data of the working condition through the preset fusion decision model and make a decision to select an operating condition, including a single pump operating condition, a dual pump parallel operating condition, or a dual pump series operating condition; The process parameter setting module is used to find the optimal solution set according to the selected operating conditions, find the matching optimal solution, and set the corresponding process parameters according to the optimal solution to minimize the fluctuation of the operating condition switching; Among them, the optimal solution set is obtained through the optimal solution set construction module, which is specifically used to pre-construct a prediction model for each working condition. The prediction model predicts the flow fluctuation and pressure fluctuation of the current working condition from start to stop based on the collected speed of the dual pumps, the opening speed and opening degree of multiple valves under the corresponding working conditions, and obtains the prediction data set of each working condition. Then, the optimal solution set in the prediction data set of each working condition is calculated through a multi-objective optimization algorithm.

[0072] The operating data includes the inlet and outlet pressures, inlet and outlet flow rates, total trunk flow rate and total trunk pressure of the dual pumps.

[0073] Specifically, the fusion decision model preset in the fusion decision module selects a working condition based on rule-driven and data-driven respectively, calculates the confidence of the two working conditions, and selects the working condition with higher confidence as the final decision.

[0074] In the optimal solution set construction module, rule-driven systems are based on a rule base containing multiple rules, each of which includes a trigger condition, an execution action, and a priority. Data-driven systems are based on a pre-trained neural network classification decision model, which is trained based on data collected from each operating condition and the corresponding operating condition labels.

[0075] In the optimal solution set construction module, the prediction model for each operating condition is constructed using a three-layer feedforward neural network based on the nonlinear characteristics of the dual-pump system for riserless drilling. The model consists of an input layer, a hidden layer, and an output layer. The number of nodes in the input layer is equal to the dimensionality of the control parameters. The multi-objective optimization algorithm used is either the Rime Multi-Objective Optimization Algorithm or the NSGA-II algorithm.

[0076] Furthermore, when the process parameter setting module searches for the optimal matching solution, it gives priority to a solution that minimizes pressure fluctuations.

[0077] Each module is mainly used to implement each step of the above method embodiment, which will not be described here one by one.

[0078] The present invention also provides a computer-readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic storage device, a magnetic disk, an optical disk, a server, an app store, etc., storing a computer program that, when executed by a processor, implements a corresponding function. The computer-readable storage medium of this embodiment, when executed by a processor, implements the method for minimizing fluctuations in intelligent switching of operating conditions in a dual-pump system for riserless drilling, as described in the method embodiment.

[0079] It should be pointed out that, according to the needs of implementation, the various steps / components described in this application can be split into more steps / components, or two or more steps / components or partial operations of steps / components can be combined into new steps / components to achieve the purpose of the present invention.

[0080] The size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0081] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. A method for minimizing fluctuations in intelligent switching of dual-pump systems in riser-less drilling, characterized in that: The following steps are involved: Obtain real-time operating data of the dual-pump system for riser drilling; Analyze the operating data through the preset fusion decision model, make a decision, and select an operating condition, including single pump operation, dual pump parallel operation, or dual pump series operation; Search for the optimal solution set based on the selected operating conditions, find the matching optimal solution, and set the corresponding process parameters based on the optimal solution to minimize the fluctuation of operating condition switching; The process of obtaining the optimal solution set is as follows: pre-constructing a prediction model for each working condition. The prediction model predicts the flow fluctuation and pressure fluctuation during the start-up to stop process of the current working condition based on the collected speed of the dual pumps, the opening speed and opening degree of multiple valves under the corresponding working conditions, and obtains the prediction data set for each working condition. Then, the optimal solution set in the prediction data set for each working condition is calculated through a multi-objective optimization algorithm.

2. The method for minimizing fluctuations in intelligent switching of dual-pump systems for riserless drilling according to claim 1 is characterized in that: The operating data includes the inlet and outlet pressures, inlet and outlet flow rates, total trunk flow rate and total trunk pressure of the dual pumps.

3. The method for minimizing fluctuations in intelligent switching of dual-pump systems for riserless drilling according to claim 1 is characterized in that: The preset fusion decision model selects a working condition based on rule-driven and data-driven respectively, and selects the working condition with higher confidence as the final decision.

4. The method for minimizing fluctuations in intelligent switching of dual-pump systems for riserless drilling according to claim 3 is characterized in that: Rule-driven is based on a rule base, which includes multiple rules. Each rule includes a trigger condition, execution action and priority.

5. The method for minimizing fluctuations in intelligent switching of dual-pump systems for riserless drilling according to claim 3 is characterized in that: Data-driven is based on a pre-trained neural network classification decision model, which is trained based on the data collected from each working condition and the corresponding working condition labels.

6. The method for minimizing fluctuations in intelligent switching of dual-pump systems for riserless drilling according to claim 1, characterized in that: The prediction model for each working condition is constructed using a three-layer feedforward neural network based on the nonlinear characteristics of the dual-pump system for riserless drilling. The model includes an input layer, a hidden layer, and an output layer. The number of nodes in the input layer is equal to the dimension of the control parameters.

7. The method for minimizing fluctuations in intelligent switching of dual-pump systems for riserless drilling according to any one of claims 1 to 6, characterized in that: The multi-objective optimization algorithm is the Rime multi-objective optimization algorithm or the NSGA-II algorithm.

8. The method for minimizing fluctuations in intelligent switching of dual-pump systems for riserless drilling according to any one of claims 1 to 6, characterized in that: When searching for the optimal matching solution, the solution that minimizes pressure fluctuation is preferred.

9. A system for minimizing fluctuations in intelligent switching of dual-pump systems for riser-less drilling, characterized in that: include: Operation data acquisition module, used to obtain real-time operating data of the dual-pump system in riser drilling; The fusion decision module is used to analyze the operating data of the working condition through the preset fusion decision model and make a decision to select an operating condition, including a single pump operating condition, a dual pump parallel operating condition, or a dual pump series operating condition; The process parameter setting module is used to find the optimal solution set according to the selected operating conditions, find the matching optimal solution, and set the corresponding process parameters according to the optimal solution to minimize the fluctuation of the operating condition switching; Among them, the optimal solution set is obtained through the optimal solution set construction module, which is specifically used to pre-construct a prediction model for each working condition. The prediction model predicts the flow fluctuation and pressure fluctuation of the current working condition from start to stop based on the collected speed of the dual pumps, the opening speed and opening degree of multiple valves under the corresponding working conditions, and obtains the prediction data set of each working condition. Then, the optimal solution set in the prediction data set of each working condition is calculated through a multi-objective optimization algorithm.

10. A computer storage medium, characterized in that A computer program executable by a processor is stored therein, and the computer program executes the method for minimizing fluctuations in intelligent switching of working conditions of a dual-pump system for watertight drilling without a riser as described in any one of claims 1-8.

Images