Method and system for predicting and evaluating working condition of deep sea drilling fluid lifting system

Abstract

The invention discloses a working condition prediction and evaluation method and system for a deep sea drilling fluid lifting system, and relates to the technical field of ocean engineering equipment and working condition prediction and evaluation, and the method comprises the steps: generating a working condition point sample database based on a physical model; a machine learning agent model is constructed and trained; performing real-time working condition prediction and parameter optimization, and outputting a real-time working condition prediction result, a frequency recommendation result and / or a capability boundary analysis result; the working condition prediction result comprises corresponding predicted flow and predicted lift under various power supply frequencies and pump configuration combinations; the frequency recommendation result comprises an optimal power supply frequency which accords with a preset operation target set by a user and meets system security constraints; the capacity boundary analysis result comprises the lifting capacity of the deep sea drilling fluid lifting system under the limiting condition. According to the method, the operation condition points (flow and lift) of the double-pump series / parallel configuration under different power supply frequencies can be quickly and accurately predicted based on the drilling fluid density changing in real time, and the optimal power supply frequency can be intelligently recommended.

Description

Technical Field

[0001] This application relates to the field of marine engineering equipment and operating condition prediction and evaluation technology, and in particular to a method and system for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system. Background Technology

[0002] In deep-sea oil and gas exploration and development, drilling fluid lift systems are core equipment for maintaining stable wellbore pressure and ensuring the safe and efficient return of drill cuttings. To cope with the complex and variable conditions in deepwater environments, drilling fluid lift systems, due to their ability to provide higher lift, greater flow rates, and system redundancy, have become a key component of advanced drilling technologies such as drilling fluid lift systems. Dual-pump lift units can operate in series or parallel to adapt to different operational needs.

[0003] Currently, determining the operating point (i.e., actual flow rate and head) of a dual-pump lifting unit on-site mainly relies on the traditional graphical method of "pump characteristic curve - pipeline characteristic curve". This method involves manually or semi-manually plotting the equivalent performance curve of the pump at a specific speed and the resistance curve of the pipeline system on a coordinate graph. The intersection of the two curves is the stable operating point of the system.

[0004] However, existing technologies have the following significant drawbacks:

[0005] (1) Response lag, unable to control in real time. During deep-sea drilling, the rheological parameters of drilling fluid, such as density and viscosity, change dynamically. After each change, the pipeline characteristic curve changes accordingly, requiring redrawing to find the intersection point. The calculation process is cumbersome and time-consuming, which cannot meet the on-site demand for "real-time prediction and control" of working conditions.

[0006] (2) The computational load is large and the coverage of operating conditions is limited. When considering both series / parallel configurations of dual pumps and frequency conversion regulation (usually the power supply frequency is continuously or discretely adjustable within a certain range), the number of operating condition combinations that need to be analyzed is enormous. Traditional graphical methods are difficult to quickly and comprehensively evaluate all potential operating schemes, which limits the system's operational optimization potential.

[0007] (3) Reliance on manual work and insufficient accuracy. Graphical methods rely on manual reading and interpretation, which can easily introduce subjective errors and make it difficult to handle complex nonlinear relationships, resulting in low prediction accuracy and affecting the reliability of operational decisions.

[0008] Therefore, there is an urgent need for a technical solution that can quickly, accurately, and intelligently predict the working status of drilling fluid lift systems under different drilling fluid densities, different operating configurations, and different power supply frequencies, and can provide optimization decision support. Summary of the Invention

[0009] The purpose of this application is to provide a method and system for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system. Based on the real-time changes in drilling fluid density, it can quickly and accurately predict the operating conditions (flow rate, head) of dual-pump series / parallel configurations under different power supply frequencies, and can intelligently recommend the optimal power supply frequency.

[0010] To achieve the above objectives, this application provides the following solution.

[0011] In a first aspect, this application provides a method for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system, which specifically includes the following steps.

[0012] Acquire historical operating data of deep-sea drilling fluid lift systems and generate a sample database of operating points based on a physical model.

[0013] A machine learning proxy model is constructed, and the machine learning proxy model is trained using the working condition sample database to obtain a trained machine learning proxy model. The machine learning proxy model is a model that takes drilling fluid density, power supply frequency and pump configuration as inputs, and predicts flow rate and predicted head under the corresponding working conditions as outputs, and is used to predict the working conditions of deep-sea drilling fluid lifting system.

[0014] The system collects the current drilling fluid density in real time and, based on this density, uses the trained machine learning proxy model to perform real-time operating condition prediction and parameter optimization, outputting real-time operating condition prediction results, frequency recommendation results, and / or capacity boundary analysis results. The operating condition prediction results include predicted flow rate and predicted head under various power supply frequencies and pump configuration combinations. The frequency recommendation results include the optimal power supply frequency that meets the user-defined preset operating objectives and satisfies system safety constraints. The capacity boundary analysis results include the lifting capacity of the deep-sea drilling fluid lift system under extreme conditions.

[0015] Secondly, this application provides a deep-sea drilling fluid lift system condition prediction and evaluation system, which applies the deep-sea drilling fluid lift system condition prediction and evaluation method as described in the first aspect, and includes the following functional modules.

[0016] The data generation module is used to acquire historical operating condition data of deep-sea drilling fluid lift systems and generate a sample database of operating points based on physical models.

[0017] The model training module is used to construct a machine learning proxy model and train the machine learning proxy model using the working condition sample database to obtain a trained machine learning proxy model. The machine learning proxy model is a model that takes drilling fluid density, power supply frequency and pump configuration as inputs and predicts flow rate and head under the corresponding working conditions as outputs, and is used to predict the working conditions of the deep-sea drilling fluid lift system.

[0018] The real-time prediction and optimization module is used to collect the current drilling fluid density in real time, and based on the current drilling fluid density, use the trained machine learning proxy model to perform real-time operating condition prediction and parameter optimization, outputting real-time operating condition prediction results, frequency recommendation results, and / or capacity boundary analysis results. The operating condition prediction results include predicted flow rate and predicted head under various power supply frequencies and pump configuration combinations; the frequency recommendation results include the optimal power supply frequency that meets the user-defined preset operating objectives and satisfies system safety constraints; and the capacity boundary analysis results include the lifting capacity of the deep-sea drilling fluid lift system under extreme conditions.

[0019] According to the specific embodiments provided in this application, this application has the following technical effects.

[0020] This application provides a method and system for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system. By acquiring historical operating condition data of the deep-sea drilling fluid lift system, a physical model-based database of operating condition samples can be generated. Then, a machine learning proxy model is trained using the operating condition samples in this database to obtain a trained machine learning proxy model. Finally, real-time operating condition prediction and parameter optimization are performed based on the trained machine learning proxy model, outputting real-time operating condition prediction results, frequency recommendation results, and / or capacity boundary analysis results. The operating condition prediction results include predicted flow rate and predicted head under various power supply frequencies and pump configuration combinations; the frequency recommendation results include the optimal power supply frequency that meets the user-defined preset operating objectives and satisfies system safety constraints; and the capacity boundary analysis results include the lifting capacity of the deep-sea drilling fluid lift system under extreme conditions. This application utilizes neural network technology and employs a machine learning proxy model to replace the traditional complex physical iterative calculation process, reducing the single operating condition prediction time from minutes to milliseconds. It can respond to real-time changes in drilling fluid density and provide immediate decision support on-site. Furthermore, since the machine learning proxy model is trained using sample data from a physical model-based working condition point sample database, the model training allows it to fully learn the correlations between working condition data. This results in highly accurate working condition predictions that fully conform to the fundamental principles of fluid mechanics, effectively improving the accuracy and efficiency of working condition predictions, ensuring the physical rationality of the predictions, and avoiding the absurd predictions that might occur with purely data-driven models. Additionally, this application can quickly evaluate system performance under a large number of power supply frequencies and pump configuration combinations, easily achieving optimization across all working conditions. The developed frequency recommendation and capacity boundary analysis functions can further obtain frequency recommendation results and / or capacity boundary analysis results based on the working condition prediction results, freeing engineers from tedious chart lookups and trial calculations, and improving the scientific nature of decision-making and the level of operational intelligence. Moreover, this method is easy to implement in engineering and can be integrated into existing drilling monitoring and data acquisition systems, significantly improving the safety, efficiency, and economy of deep-sea drilling operations. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is an application environment diagram of a method for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system, provided in an embodiment of this application.

[0023] Figure 2 This is a flowchart illustrating a method for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system, provided as an embodiment of this application.

[0024] Figure 3 This is a schematic diagram illustrating the working principle of a method for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system, provided in one embodiment of this application.

[0025] Figure 4 This is a schematic diagram illustrating the principle of solving the operating point of a pump-pipeline system for physical model calculation, provided in one embodiment of this application.

[0026] Figure 5 This is a functional block diagram of a real-time prediction and optimization application system provided in an embodiment of this application.

[0027] Figure 6 The equivalent performance curve of a dual-pump series configuration provided in one embodiment of this application is shown.

[0028] Figure 7 The equivalent performance curve of a dual-pump parallel configuration provided in an embodiment of this application is shown.

[0029] Figure 8 This is a schematic diagram of the available operating condition prediction interface provided in one embodiment of this application.

[0030] Figure 9 for Figure 8 Predicted 3D plots of series and parallel configurations.

[0031] Figure 10 This is a schematic diagram of a maximum capacity prediction interface provided in an embodiment of this application.

[0032] Figure 11 This is a schematic diagram of a deep-sea drilling fluid lift system operating condition prediction and evaluation system provided in an embodiment of this application. Detailed Implementation

[0033] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0034] The purpose of this application is to provide a method and system for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system. In essence, it is a method and system for real-time prediction of system operating conditions (flow rate, head) and optimization recommendation of operating parameters (power supply frequency) that combines a fluid dynamics physical model and a machine learning proxy model. It aims to quickly and accurately predict the operating conditions (flow rate, head) of a dual-pump series / parallel configuration under different power supply frequencies based on the real-time changing drilling fluid density, and to intelligently recommend the optimal operating frequency. This solves the technical problems of existing dual-pump lift system operating condition prediction methods, such as slow response, large computational load, low accuracy, and inability to meet the needs of real-time dynamic control.

[0035] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] The deep-sea drilling fluid lift system operating condition prediction and evaluation method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on other servers. Terminal 102 can send deep-sea drilling fluid lift system operating condition data to server 104. After receiving the deep-sea drilling fluid lift system operating condition data, server 104 generates a physical model-based operating point sample database; constructs and trains a machine learning proxy model; performs real-time operating condition prediction and parameter optimization, and outputs real-time operating condition prediction results, frequency recommendation results, and / or capacity boundary analysis results. The operating condition prediction results include the predicted flow rate and predicted head under various power supply frequencies and pump configuration combinations; the frequency recommendation results include the optimal power supply frequency that meets the user-defined preset operating targets and satisfies system safety constraints; and the capacity boundary analysis results include the lifting capacity of the deep-sea drilling fluid lift system under extreme conditions. Server 104 can feed back the obtained operating condition prediction results, frequency recommendation results, and / or capability boundary analysis results to terminal 102. Furthermore, in some embodiments, the deep-sea drilling fluid lift system operating condition prediction and evaluation method can also be implemented independently by server 104 or terminal 102. For example, terminal 102 can directly perform operating condition prediction and evaluation processing on the deep-sea drilling fluid lift system operating condition data, or server 104 can obtain historical operating condition data of the deep-sea drilling fluid lift system from the data storage system and perform operating condition prediction and evaluation processing on the deep-sea drilling fluid lift system operating condition data.

[0037] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, and IoT devices. The server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers, or it can be a cloud server.

[0038] In one exemplary embodiment, such as Figure 2 As shown, a method for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps S1 to S3.

[0039] S1: Acquire historical operating condition data of deep-sea drilling fluid lift systems and generate a sample database of operating point data based on a physical model.

[0040] S2: Construct a machine learning proxy model and train the machine learning proxy model using the working condition sample database to obtain a trained machine learning proxy model; the machine learning proxy model refers to a model that takes drilling fluid density, power supply frequency and pump configuration as inputs and predicts flow rate and predicted head under the corresponding working conditions as outputs, and is used to predict the working conditions of the deep-sea drilling fluid lifting system.

[0041] S3: Real-time acquisition of the current drilling fluid density, and based on the current drilling fluid density, using the trained machine learning proxy model to perform real-time operating condition prediction and parameter optimization, outputting real-time operating condition prediction results, frequency recommendation results, and / or capacity boundary analysis results; wherein, the operating condition prediction results include the predicted flow rate and predicted head corresponding to various power supply frequencies and pump configuration combinations; the frequency recommendation results include the optimal power supply frequency that meets the user-set preset operating objectives and satisfies system safety constraints; the capacity boundary analysis results include the lifting capacity of the deep-sea drilling fluid lifting system under extreme conditions.

[0042] In this embodiment, step S1 acquires historical operating condition data of the deep-sea drilling fluid lift system and generates a sample database of operating points based on a physical model, specifically including the following steps.

[0043] S11: Obtain historical operating data of the deep-sea drilling fluid lift system.

[0044] In this embodiment, the historical operating condition data of the deep-sea drilling fluid lift system mainly refers to the historical operating condition data of the deep-sea drilling fluid lift system within a certain period, such as the historical operating condition data of flow rate, drilling fluid density, power supply frequency, pump configuration and pump head within one year. This data is mainly used in step S1 to establish a sample database of operating condition points based on the physical model, and in step S2 to train the machine learning agent model.

[0045] S12: Based on the historical operating data of the deep-sea drilling fluid lift system, a hydraulic characteristic model of the pipeline system is constructed. The Darcy-Weisbach formula is used to calculate the frictional pressure loss along the pipeline, the Haaland formula is used to calculate the friction coefficient, and the effective static pressure head caused by the density difference between seawater and drilling fluid in the pipe is combined to obtain the total head demand curve of the pipeline.

[0046] In this embodiment, the expression for the total head demand curve of the pipeline is as follows.

[0047] H system (Q)=H f +H static (ρ);

[0048] Among them, H system H represents the total head demand curve for the pipeline. f H represents the frictional pressure loss along the friction path. static This represents the effective static head caused by the density difference between seawater and drilling fluid inside the pipe, where Q is the flow rate and ρ is the drilling fluid density.

[0049] S13: Based on the historical operating data of the deep-sea drilling fluid lifting system, construct a performance model of the dual-pump lifting unit, obtain the equivalent performance curves of a single pump under different power supply frequencies, and establish the equivalent performance curves of dual-pump series and parallel configurations according to the rules of superimposed head under the same flow rate in series and superimposed flow rate under the same head in parallel.

[0050] In this embodiment, the equivalent performance curves under the dual-pump series and parallel configurations are as follows.

[0051] Series configuration: Under the same flow rate, the head is superimposed, as shown in the following formula.

[0052] H series (Q)=H pump1 (Q)+H pump2 (Q);

[0053] Among them, H series (Q) represents the pump head in a dual-pump series configuration, H pump1 (Q) and H pump2 (Q) represents the head of the first and second pumps in a dual-pump series configuration, respectively.

[0054] Parallel configuration: Under the same head, the flow rates are superimposed, as shown in the following formula.

[0055] Q parallel (H)=Q pump1 (H)+Q pump2 (H);

[0056] Among them, Q parallel (H) represents the flow rate under the dual-pump parallel configuration, Q pump1 (H) and Q pump2 (H) represents the flow rates of the first and second pumps in the dual-pump parallel configuration, respectively.

[0057] S14: For each set of parameters in the preset parameter space, the intersection of the total head demand curve of the pipeline and the equivalent performance curve under the series and parallel configuration of the dual pumps is solved by numerical method to obtain the balanced flow rate and head under the corresponding working condition, and to generate a sample database of working condition points; the parameter combination includes drilling fluid density, power supply frequency and pump configuration.

[0058] In this embodiment, the intersection point of the total head demand curve of the pipeline and the equivalent performance curve under the series and parallel configuration of the dual pumps is calculated using the following formula.

[0059] H pump_effective (Q, f) freq ,Config)=H system (Q, ρ);

[0060] Among them, H pump_effective (Q, f) freq Config) is for head / flow supply under multi-pump collaborative operation, H system This represents the total head demand curve for the pipeline, where Q is the flow rate, ρ is the drilling fluid density, Config indicates the pump configuration, and f freq This refers to the power supply frequency.

[0061] In this embodiment, step S2 constructs a machine learning agent model and uses the working condition point samples in the working condition point sample database to train the machine learning agent model to obtain a trained machine learning agent model. Specifically, it includes the following steps.

[0062] S21: Divide the working condition sample database into a training set and a test set.

[0063] S22: Standardize the continuous numerical features in the training set and the test set respectively, and perform one-hot encoding on the categorical features in the training set and the test set to obtain the preprocessed training set and the preprocessed test set; wherein, the continuous numerical features include drilling fluid density and power supply frequency, and the categorical features include pump configuration.

[0064] S23: Construct a machine learning proxy model, and use the drilling fluid density, power supply frequency and pump configuration in the preprocessed training set as inputs, and the predicted flow rate and predicted head under the corresponding working conditions as outputs to train the machine learning proxy model to obtain the trained model.

[0065] S24: Use the preprocessed test set to evaluate the performance of the trained model, and use the model with the best performance as the trained machine learning proxy model.

[0066] In this embodiment, step S3 collects the current drilling fluid density in real time, and uses the trained machine learning proxy model to perform real-time operating condition prediction and parameter optimization based on the current drilling fluid density, and outputs the real-time operating condition prediction result, specifically including the following steps.

[0067] S31: Real-time acquisition of current drilling fluid density.

[0068] S32: Based on the current drilling fluid density, iterate through all preset power supply frequencies and two pump configuration methods: dual-pump series configuration and dual-pump parallel configuration. Input the current drilling fluid density, each power supply frequency, and each pump configuration method into the trained machine learning proxy model to obtain the predicted flow rate and predicted head under various power supply frequency and pump configuration combinations, which are used as the working condition prediction results.

[0069] In this embodiment, after step S3, which involves real-time acquisition of the current drilling fluid density and, based on the current drilling fluid density, using the trained machine learning proxy model to perform real-time operating condition prediction and parameter optimization, and outputting real-time operating condition prediction results, the deep-sea drilling fluid lift system operating condition prediction and evaluation method further includes step A, which specifically includes the following steps.

[0070] A1: Obtain the user-defined preset operation target; the preset operation target includes target flow rate and / or target head.

[0071] A2: Based on the preset work target, select the power supply frequency and pump configuration combination that meets the system safety constraints from the work condition prediction results; the system safety constraints include the pump's maximum speed, the pump's minimum speed, and the pump's maximum power.

[0072] A3: The power supply frequency that satisfies the system safety constraints and the power supply frequency corresponding to the pump configuration combination are taken as the optimal power supply frequency to obtain the frequency recommendation result.

[0073] A4: Recommend the optimal power supply frequency from the frequency recommendation results to the user.

[0074] In this embodiment, after step S3, which involves real-time acquisition of the current drilling fluid density and, based on the current drilling fluid density, using the trained machine learning proxy model to perform real-time operating condition prediction and parameter optimization, and outputting real-time operating condition prediction results, the deep-sea drilling fluid lift system operating condition prediction and evaluation method further includes step B, which specifically includes the following steps.

[0075] B1: With a fixed pump configuration and power supply frequency, the drilling fluid density is gradually adjusted, and the trained machine learning agent model is invoked. The fixed pump configuration, fixed power supply frequency, and various drilling fluid densities are input into the trained machine learning agent model to obtain the predicted flow rate and predicted head under the corresponding working conditions.

[0076] B2: Based on the predicted flow rate, the predicted head, the drilling fluid density, the pump configuration, the power supply frequency, and the pump physical limit parameters, determine the maximum drilling fluid density that the deep-sea drilling fluid lift system can handle under the corresponding configuration.

[0077] To make the technical solution of this application clearer, the specific implementation steps of the technical solution of this application will be explained in detail below with examples.

[0078] This embodiment provides a method for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system. Figure 3 The overall flowchart of the method is shown, demonstrating the complete technical path from physical modeling, data generation, model training to real-time prediction and optimization. It mainly includes the following steps:

[0079] Step 1: Obtain historical operating data of the deep-sea drilling fluid lift system and generate a sample database of operating points based on a physical model.

[0080] This step aims to obtain a large number of high-fidelity "(input parameters) -> (output operating points)" data pairs through precise physical calculations, which will serve as the basis for subsequent machine learning model training.

[0081] 1) Modeling the hydraulic characteristics of the pipeline system, that is, constructing a hydraulic characteristic model of the pipeline system.

[0082] In this embodiment, the pressure loss (head requirement) H of the pipeline system is established. lossThe functional relationship between the total pressure loss and the flow rate Q. The total pressure loss consists of the static pressure difference and the frictional pressure loss along the pipe. For a given pipeline geometry (length L, inner diameter D, relative roughness ∈ / D), the total pressure loss is mainly affected by the drilling fluid density ρ and the flow rate Q. Frictional pressure loss along the pipe H... f Calculated using the Darcy-Weisbach formula:

[0083]

[0084] Where v is the average velocity of the fluid inside the pipe (which can be calculated from the flow rate Q and the pipe diameter D), and g is the acceleration due to gravity. The key parameter, the friction coefficient f, is a function of the Reynolds number Re and the relative roughness ∈ / D. This embodiment uses the Haaland formula for accurate calculation, which is applicable to the entire flow regime and avoids the complexity of piecewise calculations.

[0085]

[0086] The Reynolds number Re is defined as:

[0087]

[0088] Where μ is the dynamic viscosity of the drilling fluid. In addition, the effective static head H caused by the density difference between seawater and the drilling fluid inside the pipe must also be considered. static Ultimately, the total head requirement curve for the pipeline is as follows:

[0089] H system (Q)=H f +H static (ρ);

[0090] Among them, H system H represents the total head demand curve for the pipeline. f H represents the frictional pressure loss along the friction path. static This represents the effective static head caused by the density difference between seawater and drilling fluid inside the pipe, where Q is the flow rate and ρ is the drilling fluid density.

[0091] 2) Performance modeling of the dual-pump lifting unit, i.e., constructing a performance model of the dual-pump lifting unit.

[0092] In this embodiment, the data of a single pump at different power supply frequencies f are obtained. freq Equivalent performance curve H pump (Q) (Typically provided by the pump manufacturer). Based on the single-pump curve, establish equivalent performance curves for dual-pump series and parallel configurations. Figure 6 The equivalent performance curves for a dual-pump series configuration are shown. Figure 7 The equivalent performance curves for a dual-pump parallel configuration are shown.

[0093] Series configuration: Under the same flow rate, the head is superimposed:

[0094] H series (Q)=H pump1 (Q)+H pump2 (Q);

[0095] Among them, H series (Q) represents the pump head in a dual-pump series configuration, H pump1 (Q) and H pump2 (Q) represents the head of the first and second pumps in a dual-pump series configuration, respectively.

[0096] Parallel configuration: Under the same head, the flow rate is superimposed.

[0097] Q parallel (H)=Q pump1 (H)+Q pump2 (H);

[0098] Among them, Q parallel (H) represents the flow rate under the dual-pump parallel configuration, Q pump1 (H) and Q pump2 (H) represents the flow rates of the first and second pumps in the dual-pump parallel configuration, respectively.

[0099] 3) Solving for operating points and generating data.

[0100] For a given operating condition (considering drilling fluid density ρ, pump configuration Config, and power supply frequency f), freq (Uniquely determined), the stable operating point of the system is the point where the pump's supply capacity balances the pipeline's demand, i.e., the intersection of the pump's equivalent performance curve and the pipeline characteristic curve. This intersection point is obtained by solving the following set of nonlinear equations:

[0101] H pump_effective (Q, f) freq ,Config)=H system (Q, ρ);

[0102] Among them, H pump_effective (Q,f freq ,Config) represents the head / flow rate supply under multi-pump collaborative operation, where Q is the flow rate, ρ is the drilling fluid density, Config indicates the pump configuration, and f freq This refers to the power supply frequency.

[0103] The above equations are solved using numerical methods (such as efficient root-finding algorithms like Brentq) to obtain the equilibrium flow rate Q under this operating condition. op He Yangcheng H op .

[0104] Iterate through the preset parameter space, such as the following parameters.

[0105] Drilling fluid density ρ: e.g., from 1000 kg / m³ 3 Up to 1500 kg / m 3 Step length 10kg / m 3 .

[0106] Power supply frequency f freq For example, from 35Hz to 50Hz, the step size is 1Hz.

[0107] Pump configuration: series or parallel.

[0108] For each parameter combination, the operating point solution is performed, ultimately generating a large structured dataset containing thousands or even tens of thousands of records. Each record contains the input (ρ, f) freq ,Config) and output (Q op H op ).

[0109] Figure 4 The principle of solving the operating point is illustrated, with one curve representing the head demand of the pipeline system (pipeline characteristic curve) and another curve representing the pump's supply capacity at a specific frequency (pump equivalent performance curve). The intersection of the curves is clearly marked as the "system stable operating point".

[0110] Step 2: Construction and training of machine learning agent model.

[0111] This step aims to train a lightweight, efficient mathematical model that can simulate the complex physical calculations in step one.

[0112] 1) Data preprocessing.

[0113] The dataset generated in step one is divided into a training set and a test set. The data is then standardized, including:

[0114] Normalize or standardize continuous numerical characteristics (density, frequency) to eliminate the influence of dimensions.

[0115] The categorical features (pump configuration: series / parallel) are processed by one-hot encoding and converted into a numerical form that can be processed by machine learning models.

[0116] 2) Model selection and training.

[0117] Select a machine learning model suitable for multi-input, multi-output regression tasks. Alternative models may include, but are not limited to: Random Forest Regressor, Support Vector Regressor (SVR), and Multilayer Perceptron (MLP, i.e., artificial neural network). Train the selected model using preprocessed parameters (density, frequency, configuration) as model input (X) and corresponding parameters (flow rate, head) as model output (y). The model aims to learn the nonlinear mapping relationship between input and output as follows.

[0118] X(ρ,f freq ,Config)→y(Q op H op ).

[0119] 3) Model evaluation and selection.

[0120] The performance of multiple trained models is evaluated using a test set, and the evaluation metric can be the coefficient of determination (R²). 2 Score, root mean square error (RMSE), etc. Choose the one with the best overall performance (e.g., average R-squared). 2 The model with the highest score is used as the final machine learning proxy model and is saved as a .pkl file.

[0121] Step 3: Real-time prediction and optimization application system based on machine learning agent model.

[0122] This embodiment deploys the trained machine learning agent model as an application system to provide decision support for field operations.

[0123] 1) Real-time operating condition prediction function.

[0124] In this embodiment, the system receives the drilling fluid density ρ monitored in real time on site. current As input, the system internally iterates through a series of power supply frequencies (e.g., [35, 40, 45, 50] Hz) and two pump configurations (series and parallel) of interest to the user. These combinations (ρ) are then processed. current f freq The Config is input into the loaded machine learning agent model. The model can instantly calculate the predicted flow Q for all combinations within milliseconds. pred and H pred And present it to the user in a table or graphical format.

[0125] 2) Intelligent frequency recommendation function.

[0126] In this embodiment, the system can determine the user's task objectives (such as target traffic Q). target Or target lift H targetThe system searches and sorts the prediction results to automatically recommend the optimal power supply frequency that is closest to the target and meets system safety constraints (such as the pump's maximum / minimum speed, maximum power, etc.).

[0127] 3) Capability boundary analysis function.

[0128] In this embodiment, the system can be used to analyze the lifting capacity of a dual-pump lifting unit under extreme conditions. For example, by continuously increasing the input density at a certain highest frequency (e.g., 50Hz) and calling the model for prediction, while determining whether the predicted operating point exceeds the physical limits of the pump (e.g., maximum flow rate or maximum head of a single pump), the maximum drilling fluid density that the system can handle in series and parallel configurations can be determined, providing a basis for risk assessment of unconventional operating conditions.

[0129] Figure 5 This diagram illustrates the functional block diagram of a real-time prediction and optimization application system, showcasing its main components, including a user input interface, a computational engine loaded with a machine learning proxy model, and a result output and display module. It also explains how the system operates in practical applications. Real-time user interaction refers to the software system's visual GUI interface, allowing users to directly select functional modules or input corresponding data within these modules to achieve the desired function (pressure loss calculation, operating point calculation, operating point prediction, etc.). The 3D prediction map originates from the operating point prediction module. This module loads a pre-trained optimal model, and users input data such as density / frequency of interest to generate a visual 3D prediction map. This 3D prediction map displays the operating points of the lifting system at a set density / frequency for both series and parallel configurations (the image below can be enlarged for direct viewing).

[0130] This application generates a dataset of "operating point" conditions covering a wide range of operating conditions through high-fidelity physical model calculations; then, it uses this dataset to train a machine learning proxy model to replace time-consuming physical calculations; finally, it deploys the machine learning proxy model in practical applications to achieve rapid prediction and intelligent optimization of operating conditions.

[0131] The following examples illustrate several implementation schemes.

[0132] Example: Prediction of operating conditions and frequency optimization of deep-sea RMR dual-pump lifting system.

[0133] 1. Establish a high-fidelity physical model and generate a database of operating point samples based on the physical model (refer to...). Figure 1 and Figure 2 ).

[0134] First, assume the total length of the piping in the dual-pump lifting system is L = 2000m, the inner diameter is D = 130mm, and the seawater density is ρ.sea =1030kg / m 3 The relative roughness of the pipe wall is ∈ / D=4.5×10. -5 The pump's equivalent performance curve data (flow rate-head correspondence at different frequencies) is known. First, a pipeline characteristic curve model is constructed. For any given drilling fluid density ρ and flow rate Q, the flow velocity v = 4Q / (πD) is calculated. 2 The Reynolds number Re = ρvD / μ is used to obtain the friction coefficient f in the Haaland formula, and then the frictional pressure loss H is calculated. f The total head requirement is H. system (Q)=H f +H static (ρ).

[0135] Secondly, a pump performance model is constructed. Based on the performance data of a single pump, the performance of two pumps at different frequencies (f) is obtained by superimposing the head (in series) or the flow rate (in parallel). freq Equivalent performance curve H below pump_effective (Q,f freq (Config). Then, perform iterative solutions. Set the parameter traversal range: density ρ from 1000 to 1500 kg / m³. 3 (Step length 10kg / m) 3 ), frequency f freq From 35 to 50 Hz (in 1 Hz increments), configured in both series and parallel. For each parameter combination (ρ, f) freq ,Config), solving equations using numerical methods Find the intersection point, i.e., the operating point (Q). op H op ).like Figure 2 As shown, this intersection point represents the point under which, under specific conditions, the pump's output capacity just overcomes the resistance of the pipeline system, and the system reaches stable operation. All calculated operating points together constitute a large sample database.

[0136] 2. Train and validate the machine learning agent model.

[0137] The generated database is divided into training and testing sets in an 8:2 ratio. The input feature "density" and "frequency" are standardized, and the "configuration" (serial / parallel) is one-hot encoded.

[0138] Three regression models—random forest, SVR, and MLP—were constructed respectively. The processed (density, frequency, configuration) data were used as input, and (flow rate, head) data were used as output; these models were then trained on the training set.

[0139] The performance of the three models was evaluated on the test set. For example, the random forest model was found to have better R-values ​​in flow and head prediction. 2All scores reached above 0.995, demonstrating the best overall performance. Therefore, this random forest model was selected as the final machine learning surrogate model, and it was saved as a pump_model.pkl file.

[0140] 3. Deploy a real-time prediction and optimization system (see reference) Figure 3 ).

[0141] Develop a graphical user interface (GUI) or web application as a real-time prediction and optimization system. This system loads pump_model.pkl upon startup.

[0142] Application Scenario 1: Real-time operating condition prediction.

[0143] The on-site SCADA (Drilling Monitoring and Data Acquisition) system detected that the current drilling fluid density is ρ. current =1250kg / m 3 The operator enters the value into the system interface. The system immediately combines this density with a preset list of frequencies (e.g., 35, 40, 45, 50 Hz) and series / parallel configurations, and passes this data to the machine learning agent model. Within milliseconds, the model returns predicted flow and head for all combinations, clearly displayed in a table. The operator can immediately see the impact of switching different frequencies or configurations on system performance at the current density. Figure 8 This is a schematic diagram of the interface for predicting available operating conditions.

[0144] Application Scenario 2: Goal-oriented frequency recommendation.

[0145] Assuming the current operation needs to maintain 80m 3 / h of return flow. The operator enters a target flow of 80m³ / h into the system. 3 / h and current density 1250kg / m³ 3 The system invokes a machine learning proxy model to predict the flow rates of series and parallel configurations at different frequencies (e.g., in 0.5Hz steps within the 35-50Hz range). Then, the system filters out the configurations closest to 80m. 3 The system provides several combinations of frequencies and configurations per hour, ranked by secondary indicators such as energy consumption or stability, and recommends the optimal solution to the operator. For example, it suggests "parallel configuration with the frequency adjusted to 42.5Hz, with an expected flow rate of 79.8m³ / h." 3 / h". Figure 9 Predicted 3D plots for series and parallel configurations.

[0146] Application Scenario 3: Maximum Lifting Capacity Analysis.

[0147] To assess the system's ability to withstand harsh operating conditions, the operator can use this function. The system's pump configuration (e.g., in series) and frequency (e.g., maximum frequency 50Hz) are fixed, then the pump is started at 1500 kg / m³. 3 Initially, the drilling fluid density is gradually increased, with a machine learning proxy model used to predict the operating point each time. Simultaneously, the system checks whether the predicted flow rate and head exceed the maximum permissible values ​​marked on the pump nameplate (e.g., a maximum flow rate of 120 m³ / h when connected in series). 3 / h, maximum lift 1240m). When the predicted value first exceeds the safe range, the corresponding density is considered to be the maximum liftable density under this configuration. Figure 10 This is a schematic diagram of the maximum capacity prediction interface.

[0148] This embodiment analyzes a dual-pump lifting system in drilling fluid densities of 1000-1800 kg / m³. 3 The analysis results show that the maximum lifting density of the system is 1660 kg / m³ in a series configuration. 3 Parallel configuration 1350kg / m 3 The main findings are that the series configuration is suitable for high-head, medium-flow conditions; the parallel configuration is suitable for high-flow, medium-head conditions; and the overall maximum lifting capacity of the system is 1660 kg / m³. 3 .

[0149] This embodiment focuses on operating point prediction, using a trained machine learning model to predict the performance of a dual-pump system under different densities. For pipeline pressure loss calculation, the Haaland formula is used to calculate pipeline pressure loss, including friction loss and effective static head. The following constraints are set for verification: Series configuration: Flow rate ≤ 120 m³ / h 3 / h, head ≤1240m; parallel configuration: flow rate ≤240m³ / h 3 / h, head ≤620m; pump head ≥ pipeline pressure loss.

[0150] In this embodiment, all operating points that meet the constraints are sorted and visualized, and the operating points are configured in series as shown in Table 1.

[0151] Table 1 Operating conditions for series configuration

[0152]

[0153] The operating points for parallel configurations are shown in Table 2.

[0154] Table 2 Parallel Configuration Operating Points

[0155]

[0156]

[0157] Finally, based on the above analysis, this embodiment draws the following conclusion: Maximum lift density for tandem configuration: 1660 kg / m³ 3 Maximum lift density in parallel configuration: 1350 kg / m³ 3 The final engineering recommendations are: (1) For densities greater than 1200 kg / m³ 3 (1) For drilling fluids, series configuration is preferred; (2) For working conditions with high flow requirements, parallel configuration is preferred; (3) During actual operation, a safety margin of 10-15% should be reserved on the basis of the maximum density; (4) The pump performance model should be calibrated regularly to ensure the accuracy of prediction.

[0158] In summary, the deep-sea drilling fluid lift system operating condition prediction and evaluation method and system proposed in this application, through the organic integration of physical models and machine learning, successfully overcomes the limitations of traditional methods, and is more efficient, accurate and intelligent, with the following specific advantages:

[0159] (1) High timeliness and real-time performance. The machine learning proxy model replaces the complex physical iterative calculation, reducing the prediction time of a single working condition from minutes to milliseconds. It can respond to real-time changes in drilling fluid density and provide immediate decision support on site.

[0160] (2) High accuracy and high reliability. Since the training data for the machine learning surrogate model comes from high-fidelity physical model calculation data (i.e., sample data in the working point sample database based on the physical model), its prediction results are not only highly accurate (simulation verification R...). 2 The score can reach above 0.99, and it fully conforms to the basic principles of fluid mechanics, ensuring the physical rationality of the prediction results and avoiding the absurd predictions that may occur with purely data-driven models.

[0161] (3) Wide operating condition coverage and intelligent operation. This application can quickly evaluate the system performance under a large number of frequency and configuration combinations, and easily achieve full-condition optimization. The developed frequency recommendation and capability boundary analysis functions free engineers from tedious chart looking and trial calculations, improving the scientific nature of decision-making and the level of intelligent operation.

[0162] (4) Good prospects for engineering applications. The method and system framework proposed in this application are clear, easy to implement in engineering, and can be integrated into existing SCADA systems, which can significantly improve the safety, efficiency and economy of deep-sea drilling operations.

[0163] In one exemplary embodiment, a deep-sea drilling fluid lift system condition prediction and evaluation system is provided, which mainly includes the following functional modules.

[0164] Data generation module: Runs offline, used to perform the physical calculations in step one and generate a database of working point samples.

[0165] Model training module: Runs offline and is used to perform step two, completing the training and saving of the machine learning agent model.

[0166] Real-time prediction and optimization module: runs online and includes the following content.

[0167] A user input interface is used to receive parameters such as real-time drilling fluid density.

[0168] A computational engine loaded with a pre-trained machine learning agent model.

[0169] An interface for displaying results and providing decision support, used to show predicted operating conditions, recommendation frequency, and capability boundary analysis results.

[0170] Based on the same inventive concept, this application also provides a deep-sea drilling fluid lift system condition prediction and evaluation system for implementing the aforementioned deep-sea drilling fluid lift system condition prediction and evaluation method. The solution provided by this system is similar to the implementation scheme described in the above method. Therefore, the specific limitations in the embodiments of the deep-sea drilling fluid lift system condition prediction and evaluation system provided below can be found in the limitations of the deep-sea drilling fluid lift system condition prediction and evaluation method described above, and will not be repeated here.

[0171] In one exemplary embodiment, such as Figure 11 As shown, a deep-sea drilling fluid lift system operating condition prediction and evaluation system is provided. The deep-sea drilling fluid lift system operating condition prediction and evaluation system applies the aforementioned deep-sea drilling fluid lift system operating condition prediction and evaluation method. The deep-sea drilling fluid lift system operating condition prediction and evaluation system includes the following functional modules.

[0172] The data generation module is used to acquire historical operating condition data of deep-sea drilling fluid lift systems and generate a sample database of operating points based on physical models.

[0173] The model training module is used to construct a machine learning proxy model and train the machine learning proxy model using the working condition sample database to obtain a trained machine learning proxy model. The machine learning proxy model is a model that takes drilling fluid density, power supply frequency and pump configuration as inputs and predicts flow rate and head under the corresponding working conditions as outputs, and is used to predict the working conditions of the deep-sea drilling fluid lift system.

[0174] The real-time prediction and optimization module is used to collect the current drilling fluid density in real time, and based on the current drilling fluid density, use the trained machine learning proxy model to perform real-time operating condition prediction and parameter optimization, outputting real-time operating condition prediction results, frequency recommendation results, and / or capacity boundary analysis results. The operating condition prediction results include predicted flow rate and predicted head under various power supply frequencies and pump configuration combinations; the frequency recommendation results include the optimal power supply frequency that meets the user-defined preset operating objectives and satisfies system safety constraints; and the capacity boundary analysis results include the lifting capacity of the deep-sea drilling fluid lift system under extreme conditions.

[0175] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0176] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system, characterized in that, The method for predicting and evaluating the operating conditions of deep-sea drilling fluid lift systems includes: Acquire historical operating condition data of deep-sea drilling fluid lift systems and generate a sample database of operating points based on physical models; A machine learning proxy model is constructed, and the machine learning proxy model is trained using the working condition sample database to obtain a trained machine learning proxy model. The machine learning proxy model is a model that takes drilling fluid density, power supply frequency and pump configuration as inputs and predicts flow rate and head under the corresponding working conditions as outputs to predict the working conditions of the deep-sea drilling fluid lift system. The system collects the current drilling fluid density in real time and, based on this density, uses the trained machine learning proxy model to perform real-time operating condition prediction and parameter optimization, outputting real-time operating condition prediction results, frequency recommendation results, and / or capacity boundary analysis results. The operating condition prediction results include predicted flow rate and predicted head under various power supply frequencies and pump configuration combinations. The frequency recommendation results include the optimal power supply frequency that meets the user-defined preset operating objectives and satisfies system safety constraints. The capacity boundary analysis results include the lifting capacity of the deep-sea drilling fluid lift system under extreme conditions.

2. The method for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system according to claim 1, characterized in that, Acquire historical operating condition data of deep-sea drilling fluid lift systems and generate a sample database of operating point data based on a physical model, specifically including: Acquire historical operating data of deep-sea drilling fluid lift systems; Based on the historical operating data of the deep-sea drilling fluid lift system, a hydraulic characteristic model of the pipeline system is constructed. The Darcy-Weisbach formula is used to calculate the frictional pressure loss along the pipeline, the Haaland formula is used to calculate the friction coefficient, and the effective static pressure head caused by the density difference between seawater and drilling fluid in the pipe is combined to obtain the total head demand curve of the pipeline. Based on the historical operating data of the deep-sea drilling fluid lifting system, a performance model of the dual-pump lifting unit is constructed, and the equivalent performance curves of a single pump under different power supply frequencies are obtained. According to the rules of superimposed head under the same flow rate when connected in series and superimposed flow rate under the same head when connected in parallel, the equivalent performance curves of dual-pump series and parallel configurations are established. For each set of parameters within the preset parameter space, the intersection point of the total pipeline head demand curve and the equivalent performance curve under the series and parallel configuration of the dual pumps is solved using numerical methods to obtain the balanced flow rate and head under the corresponding operating conditions, and to generate a sample database of operating conditions; the parameter combination includes drilling fluid density, power supply frequency and pump configuration.

3. The method for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system according to claim 2, characterized in that, The expression for the total head demand curve of the pipeline is: H system (Q)=H f +H static (ρ); Among them, H system H represents the total head demand curve for the pipeline. f H represents the frictional pressure loss along the friction path. static This represents the effective static head caused by the density difference between seawater and drilling fluid inside the pipe, where Q is the flow rate and ρ is the drilling fluid density.

4. The method for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system according to claim 2, characterized in that, The equivalent performance curves for the dual-pump series and parallel configurations are expressed as follows: Series configuration: Under the same flow rate, the head is superimposed, which is expressed as: H series (Q)=H pump1 (Q)+H pump2 (Q); Among them, H series (Q) represents the pump head in a dual-pump series configuration, H pump1 (Q) and H pump2 (Q) represents the head of the first and second pumps in a dual-pump series configuration, respectively. Parallel configuration: Under the same head, the flow rates are superimposed, as shown below: Q parallel (H)=Q pump1 (H)+Q pump2 (H); Among them, Q parallel (H) represents the flow rate under the dual-pump parallel configuration, Q pump1 (H) and Q pump2 (H) represents the flow rates of the first and second pumps in the dual-pump parallel configuration, respectively.

5. The method for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system according to claim 2, characterized in that, The intersection point of the total head demand curve of the pipeline and the equivalent performance curves under the series and parallel configurations of the two pumps is calculated using the following formula: H pump_effective (Q,f freq ,Config)=H system (Q,ρ); Among them, H pump_effective (Q, f) freq Config) is for head / flow supply under multi-pump collaborative operation, H system This represents the total head demand curve for the pipeline, where Q is the flow rate, ρ is the drilling fluid density, Config indicates the pump configuration, and f freq This refers to the power supply frequency.

6. The method for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system according to claim 1, characterized in that, Construct a machine learning agent model and train it using working condition point samples from the working condition point sample database to obtain a trained machine learning agent model, specifically including: The working condition sample database is divided into a training set and a test set; The continuous numerical features in the training set and the test set are standardized respectively, and the categorical features in the training set and the test set are encoded using one-hot encoding to obtain the preprocessed training set and the preprocessed test set; wherein, the continuous numerical features include drilling fluid density and power supply frequency, and the categorical features include pump configuration. A machine learning proxy model is constructed, and the drilling fluid density, power supply frequency and pump configuration in the preprocessed training set are used as inputs, and the predicted flow rate and predicted head under the corresponding working conditions are used as outputs to train the machine learning proxy model to obtain the trained model. The preprocessed test set is used to evaluate the performance of the trained model, and the model with the best performance is selected as the trained machine learning proxy model.

7. The method for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system according to claim 1, characterized in that, The system collects the current drilling fluid density in real time, and based on this density, uses the trained machine learning proxy model to perform real-time operating condition prediction and parameter optimization, outputting real-time operating condition prediction results, specifically including: Real-time acquisition of current drilling fluid density; Based on the current drilling fluid density, the system iterates through all preset power supply frequencies and two pump configurations: dual-pump series configuration and dual-pump parallel configuration. The current drilling fluid density, each power supply frequency, and each pump configuration are input into the trained machine learning proxy model to obtain the predicted flow rate and predicted head under various power supply frequency and pump configuration combinations, which are used as the operating condition prediction results.

8. The method for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system according to claim 7, characterized in that, After the steps of real-time acquisition of the current drilling fluid density, and based on the current drilling fluid density, using the trained machine learning proxy model to perform real-time operating condition prediction and parameter optimization, and outputting real-time operating condition prediction results, the deep-sea drilling fluid lift system operating condition prediction and evaluation method further includes: Obtain the user-defined preset operation target; the preset operation target includes target flow rate and / or target head; Based on the preset operational objectives, power supply frequency and pump configuration combinations that meet system safety constraints are selected from the operational condition prediction results; the system safety constraints include the pump's maximum speed, minimum speed, and maximum power. The optimal power supply frequency is obtained by taking the power supply frequency that satisfies the system safety constraints and the power supply frequency corresponding to the pump configuration combination as the optimal power supply frequency. The optimal power supply frequency from the frequency recommendation results is recommended to the user.

9. The method for predicting and evaluating the operating conditions of a deep-sea drilling fluid lift system according to claim 7, characterized in that, After the steps of real-time acquisition of the current drilling fluid density, and based on the current drilling fluid density, using the trained machine learning proxy model to perform real-time operating condition prediction and parameter optimization, and outputting real-time operating condition prediction results, the deep-sea drilling fluid lift system operating condition prediction and evaluation method further includes: With a fixed pump configuration and power supply frequency, the drilling fluid density is gradually adjusted, and the trained machine learning agent model is invoked. The fixed pump configuration, fixed power supply frequency, and various drilling fluid densities are input into the trained machine learning agent model to obtain the predicted flow rate and predicted head under the corresponding working conditions. Based on the predicted flow rate, the predicted head, the drilling fluid density, the pump configuration, the power supply frequency, and the pump's physical limit parameters, the maximum drilling fluid density that the deep-sea drilling fluid lift system can handle under the corresponding configuration is determined.

10. A predictive evaluation system for the operating conditions of a deep-sea drilling fluid lift system, characterized in that, The deep-sea drilling fluid lift system operating condition prediction and evaluation system applies the deep-sea drilling fluid lift system operating condition prediction and evaluation method as described in any one of claims 1-9, and the deep-sea drilling fluid lift system operating condition prediction and evaluation system includes: The data generation module is used to acquire historical operating condition data of the deep-sea drilling fluid lift system and generate a sample database of operating points based on a physical model. The model training module is used to construct a machine learning proxy model and train the machine learning proxy model using the working condition sample database to obtain a trained machine learning proxy model. The machine learning proxy model is a model that takes drilling fluid density, power supply frequency and pump configuration as inputs and predicts flow rate and head under the corresponding working conditions as outputs to predict the working conditions of the deep-sea drilling fluid lift system. The real-time prediction and optimization module is used to collect the current drilling fluid density in real time, and based on the current drilling fluid density, use the trained machine learning proxy model to perform real-time operating condition prediction and parameter optimization, outputting real-time operating condition prediction results, frequency recommendation results, and / or capacity boundary analysis results. The operating condition prediction results include predicted flow rate and predicted head under various power supply frequencies and pump configuration combinations; the frequency recommendation results include the optimal power supply frequency that meets the user-defined preset operating objectives and satisfies system safety constraints; and the capacity boundary analysis results include the lifting capacity of the deep-sea drilling fluid lift system under extreme conditions.

Images