Multi-system combined self-regulation method for offshore oilfield

Abstract

The invention discloses a multi-system combination self-control method for an offshore oilfield. The method comprises the steps that a drilling track is optimized through multi-sensor data and a geologic model, oil reservoir data are monitored in real time in combination with a seabed sensing network, and injection-production parameters and production balance are optimized through the technologies of a genetic algorithm, a neural network and the like. Meanwhile, a resource allocation scheme is generated based on an intelligent scheduling algorithm, a risk signal is handled, the digital twinborn model is updated, an oil reservoir management strategy and a production deployment scheme are finally generated, and intelligent oil field management is achieved. According to the invention, intelligent management of the whole life cycle of the submarine oil field is realized, the production efficiency and safety are improved, and a new technical scheme is provided for oil and gas development in a complex environment.

Description

Technical Field

[0001] The present invention belongs to the field of oilfield exploitation, and in particular relates to a multi-system combined self-regulating method for offshore oilfields. Background Art

[0002] Offshore oilfield development is a core area of ​​global energy strategy. Its efficient, safe, and sustainable extraction is directly linked to energy supply and economic benefits. With the increasing demand for deepwater oil and gas resource development, the complex geological environment, harsh ocean conditions, and high operating costs of offshore oilfields place extremely high demands on production technologies. However, existing offshore oilfield development methods have significant limitations in coping with complex dynamic environments. Traditional technologies often rely on a single system or static control strategies, making it difficult to achieve multi-dimensional, real-time, and refined production management. This leads to low resource utilization, insufficient emergency response capabilities, and limited operational efficiency. Against this backdrop, the core challenges facing offshore oilfield development lie in three technical factors: multi-system collaboration, data-driven decision-making, and refined control. First, the lack of multi-system collaboration hinders the efficient integration of subsystems such as drilling, communications, and scheduling, impacting overall production efficiency. Second, data-driven decision-making is limited by insufficient real-time data acquisition and processing capabilities, making it difficult to construct dynamically updated digital twin models and thus limiting the level of intelligent management. Finally, the limited resolution of refined control makes it difficult for existing methods to implement differentiated control for different reservoir blocks, resulting in difficulties in timely detection and resolution of local anomalies. Unresolved technical issues like these make it difficult to optimize production and mitigate risks in complex offshore oilfield environments. Therefore, integrating multiple systems technologies, building a digital twin model driven by real-time data, and implementing high-resolution, differentiated control to enhance intelligent management and production efficiency in offshore oilfields has become a critical issue that needs to be addressed. Summary of the Invention

[0003] To solve the above technical problems, the present invention provides a multi-system combined self-regulation method for offshore oil fields, comprising:

[0004] Acquire multi-sensor data from drilling equipment and calculate drilling trajectory deviations based on pre-established geological models;

[0005] If the drilling trajectory deviation exceeds a preset threshold, the geological parameters in the digital twin model are updated based on the multi-sensor data, and a genetic algorithm is used to optimize the drilling angle and speed to obtain the optimal drilling path;

[0006] Update the reservoir geological model according to the optimal drilling path, obtain pressure, temperature and flow data from the seabed sensor network, and transmit it to the edge computing node through the data communication system;

[0007] Analyze the pressure, temperature, and flow data at the edge computing node, calculate the production status of each control unit, and extract abnormal fluctuation characteristics;

[0008] If the abnormal fluctuation characteristics exceed the preset range, a neural network is used to analyze the time series regularity of the abnormal fluctuation characteristics to generate a differentiated water injection and oil production parameter adjustment strategy;

[0009] Based on the differentiated water injection and oil production parameter adjustment strategy, combined with the oilfield spatial layout data, a numerical simulation method is used to evaluate the production efficiency of each control unit, and a particle swarm optimization algorithm is used to coordinate the production balance between units and generate global optimized control parameters;

[0010] Based on the global optimization control parameters, combined with weather forecasts, equipment status and production plan data, an intelligent scheduling algorithm is used to calculate resource allocation plans and generate dynamic scheduling instructions for maintenance operations and logistics supplies;

[0011] If the submarine sensor network detects a typhoon or equipment failure risk signal, the emergency plan management system matches the preset risk model, generates production parameter adjustment or shutdown instructions, and determines safety protection measures;

[0012] According to the security protection measures, real-time production data is obtained from the edge computing node and transmitted to the land data center, and combined with historical production data, long-term trends are mined using a long-short-term memory network to update the digital twin model and generate a reservoir management strategy;

[0013] Based on the reservoir management strategy, an agent model is used to simulate well location and pipeline network adjustment plans. Combined with the oilfield spatial layout data, layout optimization suggestions are generated and the final production deployment plan is determined.

[0014] Preferably, the process of acquiring multi-sensor data of the drilling equipment and calculating the drilling trajectory deviation in combination with a pre-established geological model includes:

[0015] Acquiring multi-sensor data based on drilling equipment; wherein the multi-sensor data includes drill bit position and orientation data and geological parameters, and storing the data as a raw data set;

[0016] The data preprocessing method is used to denoise and standardize the original data set to obtain the processed sensor data set;

[0017] The processed sensor data set is fused through the Kalman filter algorithm to generate drill bit position and direction data and determine the fused positioning data;

[0018] Generating a real-time drilling trajectory using a geometric calculation method based on the fused positioning data and a preset geological model to obtain drilling trajectory data;

[0019] If there is a deviation between the drilling trajectory data and the expected trajectory of the preset geological model, the trajectory deviation is calculated by the least square method to obtain a deviation value;

[0020] Based on the deviation value, the PID control algorithm is used to adjust the drilling equipment parameters and generate a control instruction data set;

[0021] The drilling equipment is driven by the control instruction data set, the drill bit position and direction are updated, and new sensor data are obtained.

[0022] Preferably, the process of obtaining the optimal drilling path includes:

[0023] If the drilling trajectory deviation exceeds a preset threshold, real-time drilling parameters are obtained from the multi-sensor data, and the sensor data is processed using a filtering algorithm to obtain denoised drilling parameters;

[0024] Based on the denoised drilling parameters, the geological parameters in the digital twin model are updated, and the spatial distribution of the geological parameters is calculated using an interpolation algorithm to determine the updated geological model.

[0025] If the deviation between the updated geological model and the preset geological model exceeds a threshold, a genetic algorithm is used to optimize the drilling angle and speed based on the drilling parameters and the geological model to obtain a preliminary optimized drilling path;

[0026] Calculating the deviation between the drilling trajectory and the target path using the preliminary optimized drilling path, fitting the trajectory deviation using the least squares method, and determining trajectory adjustment parameters;

[0027] Adjusting parameters according to the trajectory, updating the drilling path in the digital twin model, and verifying the stability of the adjusted path using a simulation algorithm to obtain a verified drilling path;

[0028] If the stability of the verified drilling path meets a preset threshold, extracting drilling control parameters from the verified path to determine a final drilling path;

[0029] According to the final drilling path, a drilling equipment control instruction is generated and sent to the drilling equipment using a real-time transmission protocol to obtain the drilling status after execution.

[0030] Preferably, the process of updating the reservoir geological model according to the optimal drilling path, acquiring pressure, temperature and flow data from the seabed sensor network, and transmitting the data to the edge computing node through the data communication system includes:

[0031] Using a preset sampling frequency, pressure data, temperature data, and flow data are collected from the submarine sensor network in real time to obtain the original sensor data set;

[0032] The original sensor data set is encoded and compressed using a differential encoding technique through a data communication system to obtain a compressed data packet;

[0033] If the integrity check of the compressed data packet passes, the compressed data packet is transmitted to the edge computing node through a low-latency communication protocol to obtain a transmission data stream;

[0034] Decoding and preprocessing the transmission data stream at the edge computing node, smoothing data fluctuations using a sliding window averaging method to obtain a smoothed data sequence;

[0035] updating the reservoir geological model using a particle filter algorithm based on the smoothed data sequence and preset reservoir geological parameters to obtain an updated geological model;

[0036] Based on the updated geological model, a genetic algorithm is used to optimize the drilling path and determine the optimal drilling trajectory;

[0037] If the deviation of the optimal drilling trajectory is less than a preset threshold, the optimal drilling trajectory is output to the drilling control system to obtain a path control instruction.

[0038] Preferably, the process of calculating the production status of each control unit and extracting abnormal fluctuation characteristics includes:

[0039] Obtain pressure, temperature, and flow data collected by edge computing nodes, clean the data using preset standardization rules, and obtain a standardized data set;

[0040] Using a time series analysis method, the normalized data set is segmented, the production state parameters of each control unit are calculated, and the state characteristic vector is determined;

[0041] If the fluctuation amplitude of the state feature vector exceeds a preset threshold, the abnormal fluctuation frequency feature is extracted by Fourier transform to obtain an abnormal fluctuation feature set;

[0042] According to the abnormal fluctuation feature set, a cluster analysis algorithm is used to group the features and determine the abnormal state category of each control unit;

[0043] Extracting high-risk categories from the abnormal state categories, generating abnormal fluctuation feature descriptions of the control unit through preset mapping rules, and determining a distribution pattern of the abnormal fluctuation features;

[0044] According to the distribution pattern of the abnormal fluctuation characteristics, a decision tree algorithm is used to generate an optimized control strategy to obtain adjustment parameters for each control unit;

[0045] The adjustment parameters are sent to each control unit through the edge computing node, the production status is updated in real time, and closed-loop processing of abnormal fluctuations is completed.

[0046] Preferably, if the abnormal fluctuation characteristics exceed a preset range, a process of using a neural network to analyze the time series regularity of the abnormal fluctuation characteristics and generating a differentiated water injection and oil production parameter adjustment strategy includes:

[0047] If the abnormal fluctuation characteristics exceed the preset range, time series data is extracted from the collected data, and the time series data is denoised using a data processing method to obtain a smoothed time series;

[0048] Performing regularity analysis on the smoothed time series through a neural network, extracting periodic features in the smoothed time series using a long short-term memory network, and obtaining fluctuation regularity features;

[0049] If the fluctuation regularity matches the historical fluctuation pattern, the preset water injection and oil production parameter adjustment rules are queried based on the matching results to determine the preliminary adjustment parameters;

[0050] Based on the preliminary adjustment parameters, a genetic algorithm is used to optimize the water injection and oil production parameters to generate a differentiated adjustment strategy;

[0051] Extracting water injection parameters and oil production parameters based on the differentiated adjustment strategy, determining whether the parameter adjustment meets the production constraints through parameter comparison and analysis, and obtaining a parameter set that meets the constraints;

[0052] If the parameter set meets the constraints, the stability of the adjusted parameters is verified through real-time data, and the fluctuation trend of the parameters after adjustment is determined by time series prediction method to obtain the final adjustment strategy;

[0053] According to the final adjustment strategy, control instructions for water injection and oil production equipment are generated, and the control instructions are transmitted through the equipment interface to complete the parameter adjustment.

[0054] Preferably, the process of generating the global optimization control parameters includes:

[0055] Using numerical simulation methods, the geological and production data of each control unit are obtained from the spatial layout data of the oil field, and the production efficiency of the control unit is calculated;

[0056] If the production efficiency of the control unit is lower than a preset threshold, the water injection parameters and the oil production parameters are adjusted based on the differentiation strategy to generate preliminary adjustment parameters;

[0057] Through the particle swarm optimization algorithm, the inter-unit production data is obtained from the preliminary adjustment parameters, and the inter-unit production balance index is calculated;

[0058] If the inter-unit production balance index does not reach the preset threshold, the water injection parameters and oil production parameters are iteratively adjusted to generate optimized adjustment parameters;

[0059] According to the optimized adjustment parameters, dynamic production data is extracted from the oil field spatial layout data, and the updated control unit production efficiency is recalculated;

[0060] By comparing the updated control unit production efficiency with the initial efficiency, the convergence of the global optimization parameters is judged and the final global optimization control parameters are generated;

[0061] Based on the final global optimization control parameters and combined with the oilfield spatial layout data, the water injection and oil production execution plans of each control unit are generated.

[0062] Preferably, the process of generating dynamic scheduling instructions for maintenance operations and logistical supplies includes:

[0063] Obtain weather forecast data, equipment status information, and production plan data, and generate a comprehensive data set through data fusion processing;

[0064] If the comprehensive data set is complete, the control parameters are calculated using a particle swarm optimization algorithm to obtain an optimized parameter set;

[0065] Based on the optimized parameter set, a constraint satisfaction algorithm is used to generate a resource allocation plan and determine the resource allocation result;

[0066] If the resource allocation result satisfies the production plan data constraints, a maintenance operation instruction is generated to obtain a maintenance scheduling sequence;

[0067] Generate logistics supply instructions and determine the supply scheduling sequence according to the maintenance scheduling sequence and logistics supply requirements;

[0068] Fusion of the maintenance scheduling sequence and the supply scheduling sequence to generate dynamic scheduling instructions;

[0069] Get real-time information to update comprehensive data sets, and loop execution to achieve continuously optimized dynamic scheduling instructions.

[0070] Preferably, the process of generating a production parameter adjustment or production stoppage instruction and determining safety protection measures includes:

[0071] Obtain typhoon or equipment failure risk signals through submarine sensor networks and determine the risk signal type;

[0072] If the risk signal type is typhoon, the wind speed and wave height are calculated using the preset risk model to obtain the risk level;

[0073] According to the risk level, a corresponding preset risk model is matched from the emergency plan management system to generate a production parameter adjustment instruction;

[0074] If the risk level exceeds the preset threshold, a production suspension instruction is generated and the suspension time range is determined;

[0075] The management system matches the safety protection measures database to obtain the protection measures plan corresponding to the production suspension order;

[0076] Use machine learning classification algorithms to optimize the protection measures and obtain the final security protection measures;

[0077] Execution parameters are extracted from the final security protection measures to generate device control instructions.

[0078] Preferably, the process of generating a reservoir management strategy includes:

[0079] Collect real-time production data from edge computing nodes and transmit it to the land data center using a preset encryption protocol to obtain an encrypted real-time data set;

[0080] If the received encrypted real-time data set is complete, the data is decrypted in the land data center, combined with the stored historical production data, and data cleaning methods are used to remove noise to obtain the integrated production data set;

[0081] For the integrated production data set, long short-term memory network is used to perform time series analysis, extract long-term trend features, and obtain trend feature sets;

[0082] According to the trend feature set, the parameters of the digital twin model are updated, and the model accuracy is judged using a preset model verification method to obtain an updated digital twin model;

[0083] The updated digital twin model is used to simulate the reservoir operation status, and a rule engine is used to generate preliminary reservoir management strategies to obtain a candidate set of strategies.

[0084] If the strategies in the strategy candidate set meet the preset constraints, the optimal strategy is selected from them to generate the final reservoir management strategy and obtain the management strategy output;

[0085] According to the management policy output, the collection parameters of the edge computing nodes are adjusted to optimize the collection process of real-time production data and obtain the optimized collection configuration.

[0086] Compared with the prior art, the present invention has the following advantages and technical effects:

[0087] The present invention discloses an intelligent submarine oilfield production management method, which realizes drilling trajectory optimization and reservoir geological model update through multi-sensor data acquisition and digital twin model. Combined with submarine sensor network and edge computing, production anomalies are analyzed and differentiated adjustment strategies are generated. Intelligent algorithms are used to coordinate production balance, optimize resource allocation and formulate dynamic scheduling instructions. For sudden risks, emergency plans are matched and safety measures are taken. Long-term trends are mined using long-term and short-term memory networks, digital twin models are updated and reservoir management strategies are generated. Finally, through agent model simulation, well locations and pipeline network layouts are optimized. The present invention realizes intelligent management of the entire life cycle of submarine oilfields, improves production efficiency and safety, and provides a new technical solution for oil and gas development in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0089] Figure 1 Schematic diagram of a method flow in an embodiment of the present invention. DETAILED DESCRIPTION

[0090] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0091] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0092] like Figure 1 As shown, this embodiment provides a multi-system combined self-regulation method for offshore oil fields, including:

[0093] Acquire multi-sensor data from drilling equipment and calculate drilling trajectory deviations based on pre-established geological models;

[0094] If the drilling trajectory deviation exceeds the preset threshold, the geological parameters in the digital twin model are updated based on multi-sensor data, and a genetic algorithm is used to optimize the drilling angle and speed to obtain the optimal drilling path;

[0095] Update the reservoir geological model based on the optimal drilling path, obtain pressure, temperature, and flow data from the seabed sensor network, and transmit it to the edge computing node through the data communication system;

[0096] Analyze pressure, temperature, and flow data at the edge computing node, calculate the production status of each control unit, and extract abnormal fluctuation characteristics;

[0097] If the abnormal fluctuation characteristics exceed the preset range, a neural network is used to analyze the time series regularity of the abnormal fluctuation characteristics and generate differentiated water injection and oil production parameter adjustment strategies;

[0098] Based on differentiated water injection and oil production parameter adjustment strategies and combined with oilfield spatial layout data, a numerical simulation method is used to evaluate the production efficiency of each control unit. A particle swarm optimization algorithm is used to coordinate the production balance between units and generate global optimal control parameters.

[0099] Based on global optimization control parameters, combined with weather forecasts, equipment status, and production plan data, an intelligent scheduling algorithm is used to calculate resource allocation plans and generate dynamic scheduling instructions for maintenance operations and logistics supplies.

[0100] If the submarine sensor network detects a typhoon or equipment failure risk signal, the emergency plan management system matches the preset risk model, generates production parameter adjustment or shutdown instructions, and determines safety protection measures;

[0101] Based on security measures, real-time production data is obtained from edge computing nodes and transmitted to land data centers. Long-short-term memory networks are used to mine long-term trends in combination with historical production data to update digital twin models and generate reservoir management strategies.

[0102] Based on reservoir management strategies, a proxy model is used to simulate well location and pipeline network adjustment plans. Combined with oilfield spatial layout data, layout optimization suggestions are generated and the final production deployment plan is determined.

[0103] Furthermore, the process of obtaining multi-sensor data from the drilling equipment and calculating the drilling trajectory deviation in combination with a pre-established geological model includes:

[0104] Acquire multi-sensor data based on drilling equipment; wherein the multi-sensor data includes drill bit position and orientation data and geological parameters, and is stored as a raw data set;

[0105] The data preprocessing method is used to denoise and standardize the original data set to obtain the processed sensor data set;

[0106] The processed sensor data set is fused through the Kalman filter algorithm to generate drill bit position and direction data and determine the fused positioning data;

[0107] Based on the fused positioning data and the preset geological model, a geometric calculation method is used to generate a real-time drilling trajectory to obtain drilling trajectory data;

[0108] If there is a deviation between the drilling trajectory data and the expected trajectory of the preset geological model, the trajectory deviation is calculated using the least squares method to obtain the deviation value;

[0109] Based on the deviation value, the PID control algorithm is used to adjust the drilling equipment parameters and generate a control instruction data set;

[0110] The drilling equipment is driven by the control instruction data set, the drill bit position and direction are updated, and new sensor data is obtained.

[0111] For example, during the drilling process, a multi-sensor system installed on the drill bit collects real-time drill bit position data (e.g., latitude and longitude coordinates 116.404°E, 39.915°N, vertical depth 1524.36 meters), directional parameters (inclination 12.5°, azimuth 45.8°), and geological parameters (gamma value 78 API, resistivity 15Ω·m). This data is transmitted to a data processing system at a frequency of 10 Hz. The raw data is denoised using a Kalman filter algorithm, where the process noise covariance matrix Q is set to [0.10; 0.01] and the observation noise covariance matrix R is set to [10; 0.1].

[0112] The system compares and analyzes the real-time data with the geological model, which is constructed using the Kriging interpolation method with a grid accuracy of 5 meters × 5 meters × 0.5 meters, including the formation density (2.65g / cm 3 ), porosity (12%) and other parameters. When it is detected that the deviation between the actual drill bit position and the planned trajectory exceeds the threshold (horizontal deviation 0.75 meters or vertical deviation 0.3 meters), the system automatically triggers the trajectory correction calculation module and uses the minimum curvature method to generate a correction plan, in which the inclination rate is 3° / 30 meters and the tool face angle is adjusted to 215°. At the same time, the geological model is dynamically adjusted through real-time updated geological parameters (such as the sudden drop in the resistivity of the sandstone layer to 8Ω·m), and the Bayesian probability model (prior probability 0.3, likelihood function value 0.85) is used to re-evaluate the type of formation encountered. The corrected trajectory parameters are then transmitted to the automatic control system via the Modbus protocol to perform correction operations.

[0113] Furthermore, the process of obtaining the optimal drilling path includes:

[0114] If the drilling trajectory deviation exceeds the preset threshold, real-time drilling parameters are obtained from the multi-sensor data, and the sensor data is processed using a filtering algorithm to obtain the denoised drilling parameters;

[0115] Based on the denoised drilling parameters, the geological parameters in the digital twin model are updated, and the spatial distribution of the geological parameters is calculated using an interpolation algorithm to determine the updated geological model.

[0116] If the deviation between the updated geological model and the preset geological model exceeds a threshold, a genetic algorithm is used to optimize the drilling angle and speed based on the drilling parameters and the geological model to obtain a preliminary optimized drilling path;

[0117] Through the preliminary optimized drilling path, the deviation between the drilling trajectory and the target path is calculated, and the trajectory deviation is fitted using the least squares method to determine the trajectory adjustment parameters;

[0118] Adjust parameters based on the trajectory, update the drilling path in the digital twin model, and use simulation algorithms to verify the stability of the adjusted path to obtain the verified drilling path;

[0119] If the stability of the verified drilling path meets the preset threshold, the drilling control parameters are extracted from the verified path to determine the final drilling path;

[0120] Based on the final drilling path, drilling equipment control instructions are generated and sent to the drilling equipment using a real-time transmission protocol to obtain the drilling status after execution.

[0121] For example, the multi-sensor system can collect drill bit position, inclination, azimuth, and geological parameters, such as the drill bit's position coordinates of 116.405°E, 39.916°N, vertical depth of 1525.2 meters, inclination of 13.2°, and azimuth of 46.5°. Geological parameters include a gamma value of 80 API and a resistivity of 14 Ω·m. This data is transmitted to the processing system at a 15 Hz frequency.

[0122] Specifically, filtering algorithms often use extended Kalman filtering to process sensor data. Raw data can contain noise due to downhole vibrations or electromagnetic interference. The extended Kalman filter integrates historical data with current observations through prediction and update steps to generate smoothed drilling parameters.

[0123] For example, the processed inclination data stabilized from a fluctuating value of 13.0°-13.4° to 13.2°, improving the accuracy of subsequent calculations.

[0124] Preferably, the digital twin model updates the geological information based on the denoising parameters. Geological parameters such as porosity 13%, formation density 2.64g / cm 3 The spatial distribution is calculated using the Kriging interpolation algorithm, generating a geological grid with a resolution of 4 m × 4 m × 0.4 m. If the porosity deviation exceeds 2% during model comparison, a model update is triggered.

[0125] For example, if the local formation porosity is detected to have increased to 15%, it indicates that a high-permeability sandstone layer may have been entered, and the formation type needs to be re-evaluated.

[0126] In one possible implementation, a genetic algorithm optimizes drilling angle and speed. The algorithm iteratively calculates the optimal parameters, targeting drilling efficiency and path stability.

[0127] For example, the initial drilling angle was adjusted to 14.5°, and the drilling speed was optimized from 5 m / h to 6.2 m / h to generate a preliminary path. The path deviation was fitted using the least squares method, calculating the horizontal deviation of the actual trajectory from the target trajectory to be 0.8 m and the vertical deviation to be 0.35 m. This determined the adjustment parameters, such as a tool face angle of 220°.

[0128] For example, simulation algorithms verify path stability. The digital twin model simulates the adjusted path and assesses formation stress and drill bit forces. If stability parameters, such as the vibration index, are below 0.2, the path is deemed feasible. After verifying the path, control parameters, such as a build rate of 3.5° / 30 meters, are extracted to generate the final path.

[0129] It should be noted that the control instructions are transmitted to the drilling equipment via the OPC UA protocol.

[0130] For example, the command might include adjusting the drill bit speed to 120 rpm and monitoring the execution status in real time. If feedback data shows that the deviation has dropped to 0.4 meters horizontally and 0.2 meters vertically, the path optimization is effective.

[0131] In one embodiment, the system enhances reliability through multi-faceted data fusion. Sensor data, model updates, and optimization algorithms mutually validate each other to ensure precise path adjustments. A real-time transport protocol ensures rapid command execution, reducing downhole operational delays.

[0132] Understandably, the implementation of the above method depends on sensor accuracy and algorithm efficiency. Reasonable parameter selection and model update strategies can significantly improve the adaptability of drilling paths and meet the needs of complex formations.

[0133] Furthermore, the reservoir geological model is updated based on the optimal drilling path, and pressure, temperature, and flow data are obtained from the seabed sensor network. The data is then transmitted to the edge computing node through the data communication system. The process includes:

[0134] Using a preset sampling frequency, pressure data, temperature data, and flow data are collected from the submarine sensor network in real time to obtain the original sensor data set;

[0135] Through the data communication system, the original sensor data set is encoded and compressed using differential coding technology to obtain a compressed data packet;

[0136] If the integrity check of the compressed data packet passes, the compressed data packet is transmitted to the edge computing node through the low-latency communication protocol to obtain the transmission data stream;

[0137] The transmission data stream is decoded and preprocessed at the edge computing node, and the sliding window averaging method is used to smooth the data fluctuations to obtain a smoothed data sequence;

[0138] According to the smoothed data sequence and the preset reservoir geological parameters, the particle filter algorithm is used to update the reservoir geological model to obtain an updated geological model;

[0139] To update the geological model, a genetic algorithm is used to optimize the drilling path and determine the optimal drilling trajectory;

[0140] If the deviation of the optimal drilling trajectory is less than a preset threshold, the optimal drilling trajectory is output to the drilling control system to obtain a path control instruction.

[0141] For example, in the process of updating the reservoir geological model using the optimal drilling path, the Kriging interpolation method based on the inverse distance weighted algorithm is used to merge the porosity data obtained by drilling (such as 12.5% ​​for Well A and 14.2% for Well B) with the original geological model. Through variogram analysis, the main range is determined to be 350 meters and the secondary range is 180 meters, and finally an updated model with a resolution of 10 meters × 10 meters × 2 meters is generated.

[0142] The submarine sensor network acquires pressure (e.g., 8.7 MPa ± 0.2 at P12 node), temperature (42.3°C at T09 node), and flow data (235 m3 at Q05 node) at a sampling interval of 5 minutes. 3 / d), using the improved LoRaWAN protocol with 16QAM modulation in the 470MHz frequency band, with a packet loss rate of less than 0.5%. The edge computing node deploys a time series prediction model based on LSTM, with 128 neurons in the input layer. The Adam optimizer (learning rate 0.001) is used to process noisy data and perform real-time prediction of pressure fluctuations (mean square error 0.2MPa). 2 ) and uses the DBSCAN clustering algorithm (ε=0.8, MinPts=15) to identify abnormal flow data. When a threshold is exceeded (e.g., a 30% surge in flow for two consecutive cycles), an early warning mechanism is automatically triggered. The entire system uses the OPCUA protocol for standardized data exchange, with latency controlled within 200ms, ensuring a closed-loop feedback loop between dynamic geological model updates and real-time monitoring data.

[0143] For example, when acquiring pressure, temperature, and flow data from a subsea sensor network, sensors are typically deployed at the wellhead of a subsea oil and gas well and at key pipeline nodes. Pressure sensors measure the fluid pressure within the well, temperature sensors monitor the fluid and ambient temperature, and flow sensors record the rate at which oil and gas flow through the pipeline. The default sampling frequency might be once per second to ensure that the data captures dynamic changes.

[0144] For example, in a deepwater oil field, the pressure data range is 20-50MPa, the temperature data is 50-120℃, and the flow data is 100-500m 3Through high-frequency acquisition, the original data set can reflect the real-time status of the reservoir and provide reliable input for subsequent modeling.

[0145] In one possible implementation, when encoding and compressing the original sensor data set, differential encoding technology reduces redundancy by recording data differences between adjacent sampling points.

[0146] Preferably, assuming a pressure data sequence of 50.0 MPa, 50.2 MPa, and 50.1 MPa over a certain period of time, differential encoding is performed and stored as 50.0, +0.2, and -0.1, significantly reducing the data volume. The compressed data packets undergo integrity checks, such as using a CRC checksum, to ensure that the data has not been tampered with before transmission. This approach effectively reduces bandwidth usage and improves data transmission efficiency.

[0147] Specifically, a dedicated fiber-based communication network can be used to transmit compressed data packets to edge computing nodes via a low-latency communication protocol.

[0148] For example, on an offshore platform, data packets are transmitted via 5G or fiber optic links with latency under 10 milliseconds. Once the data stream reaches the edge computing node, it needs to be decoded and restored. A sliding window averaging method is used to smooth data fluctuations.

[0149] For example, taking the average of five sampling points to process flow data eliminates abnormal spikes caused by turbulence and obtains a smooth data series. This preprocessing can improve the accuracy of subsequent modeling.

[0150] It is understood that the particle filter algorithm is used to update the reservoir geological model based on the smoothed data sequence and preset reservoir geological parameters. The particle filter simulates multiple sets of possible geological states and combines them with sensor data to weight and select the optimal state.

[0151] For example, the initial model assumes a reservoir permeability of 100 mD. If measured data indicates the permeability is too low, the algorithm updates the model to 80 mD. This dynamic update allows the model to more closely reflect actual formation characteristics.

[0152] In one embodiment, for the updated geological model, when the genetic algorithm optimizes the drilling path, the population size can be set to 100 and the number of iterations can be set to 150 generations, with the goal of minimizing the path length and drilling time.

[0153] For example, the algorithm outputs an optimal trajectory with an inclination of 80°, an azimuth of 15°, and a path length reduction of 5%. If the trajectory deviation falls below a preset threshold, such as a horizontal deviation of less than 3 meters, the trajectory is output to the drilling control system, which generates path control instructions to guide the drill rig in adjusting the drilling direction. This optimization improves drilling efficiency.

[0154] For example, after the path control instructions are transmitted to the drilling control system, the system adjusts the drill bit angle and speed according to the instructions.

[0155] For example, upon receiving a command for an 80° well inclination, the drilling rig automatically calibrates drilling parameters to ensure the trajectory meets design requirements. This closed-loop control improves drilling accuracy and reduces manual intervention.

[0156] Furthermore, the process of calculating the production status of each control unit and extracting abnormal fluctuation characteristics includes:

[0157] Obtain pressure, temperature, and flow data collected by edge computing nodes, clean the data using preset standardization rules, and obtain a standardized data set;

[0158] Using time series analysis methods, the normalized data set is segmented, the production state parameters of each control unit are calculated, and the state characteristic vector is determined;

[0159] If the fluctuation amplitude of the state feature vector exceeds the preset threshold, the abnormal fluctuation frequency characteristics are extracted through Fourier transform to obtain the abnormal fluctuation feature set;

[0160] According to the abnormal fluctuation feature set, the cluster analysis algorithm is used to group the features and determine the abnormal state category of each control unit;

[0161] Extract high-risk categories from abnormal state categories, generate abnormal fluctuation feature descriptions of control units through preset mapping rules, and determine the distribution pattern of abnormal fluctuation features;

[0162] According to the distribution pattern of abnormal fluctuation characteristics, the decision tree algorithm is used to generate the optimal control strategy and obtain the adjustment parameters for each control unit;

[0163] The adjustment parameters are sent to each control unit through the edge computing node, the production status is updated in real time, and closed-loop processing of abnormal fluctuations is completed.

[0164] For example, pressure sensors deployed at edge computing nodes collect pipeline pressure data at a sampling frequency of 100 Hz. After noise interference is eliminated using a Kalman filter, the standard deviation within a 10-second sliding window is calculated in real time. Anomaly detection is triggered when pressure fluctuations exceed a set threshold of 0.15 MPa. Temperature data is collected at a density of 5 points per second using a distributed fiber-optic temperature measurement system. A three-dimensional temperature distribution is constructed using a thermodynamic model. Anomalies are detected using an isolation forest algorithm that detects deviations from the normal range (20-80°C). Anomalies are identified when a temperature gradient exceeds 5°C / meter over three consecutive sampling periods. Flow data is collected at 0.5-second intervals using an electromagnetic flowmeter. Wavelet transforms are used to extract the frequency domain features of the flow signal. Anomalies are flagged when a sudden energy increase in the 0.1-1 Hz band exceeds 30% of the baseline value. All sensor data is timestamp-aligned and fed into an LSTM neural network. The number of hidden units is set to 64. A multivariate joint analysis is performed using the last 60 seconds of data as a time window. A control unit state is identified as abnormal when the sum of squared prediction residuals exceeds 0.05. For the detected abnormal features, the DBSCAN clustering algorithm (setting the neighborhood radius to 0.2 and the minimum number of samples to 5) was used for pattern classification. The feature combination of a sudden drop in pressure accompanied by a slow rise in temperature (pressure drop of 0.12 MPa and temperature rise of 2 degrees Celsius / minute) was identified as a pipeline leakage feature, triggering the automatic valve closing instruction.

[0165] Furthermore, if the abnormal fluctuation characteristics exceed the preset range, a neural network is used to analyze the time series regularity of the abnormal fluctuation characteristics and generate a differentiated water injection and oil production parameter adjustment strategy. The process includes:

[0166] If the abnormal fluctuation characteristics exceed the preset range, time series data are extracted from the collected data, and the time series data are denoised using data processing methods to obtain a smooth time series;

[0167] The smooth time series is analyzed by neural network, and the periodic characteristics in the smooth time series are extracted by long short-term memory network to obtain the fluctuation regularity characteristics.

[0168] If the fluctuation pattern matches the historical fluctuation pattern, the preset water injection and oil production parameter adjustment rules are queried based on the matching results to determine the preliminary adjustment parameters;

[0169] Based on the preliminary adjustment parameters, a genetic algorithm is used to optimize the water injection and oil production parameters to generate differentiated adjustment strategies;

[0170] Extract water injection parameters and oil production parameters based on differentiated adjustment strategies, and determine whether parameter adjustments meet production constraints through parameter comparison and analysis, thus obtaining a parameter set that meets the constraints.

[0171] If the parameter set meets the constraints, the stability of the adjusted parameters is verified through real-time data, and the fluctuation trend of the parameters after adjustment is determined by time series prediction method to obtain the final adjustment strategy;

[0172] Based on the final adjustment strategy, control instructions for water injection and oil production equipment are generated, and the control instructions are transmitted through the equipment interface to complete parameter adjustment.

[0173] For example, when the monitored oil well pressure fluctuation amplitude exceeds a preset threshold of ±15%, the system automatically triggers the LSTM neural network analysis module. This module uses a three-hidden layer structure (with 128 / 64 / 32 neurons, respectively) and uses the bottomhole pressure and flow rate data collected every minute for the past 30 days as input. The prediction model is trained using the Adam optimizer (learning rate 0.001). For example, if a well experiences a periodic pressure drop (an 18% drop at 2:00 PM daily), and analysis reveals a 7-hour phase difference with the water injection cycle of an adjacent well, the algorithm will generate a dynamic adjustment strategy: adjust the water injection pump frequency from 50Hz to 45Hz (a 10% decrease) and increase the oil pump stroke rate from 6 times / minute to 6.5 times / minute (an 8.3% increase). For abnormal fluctuations in water content (e.g., a sudden increase from 30% to 42% within 2 hours), the CNN network (convolution kernel size 3×3) will analyze nearly 200 sets of sensor data, identify that the anomaly is related to the permeability change of the injection layer, and then output a stratified water injection plan: increase the water injection volume of layer B from 80m 3 / d reduced to 65m 3 / d, and at the same time increase the water injection volume of the C layer from 60m 3 / d increased to 75m 3 All parameter adjustment instructions are sent to the on-site PLC controller in real time via the OPC-UA protocol. The rate of change of key indicators is continuously monitored during the adjustment process. If the indicators return to the normal fluctuation range (±5%) within 10 minutes, the strategy is deemed effective and stored in the case library for subsequent use in similar scenarios.

[0174] Furthermore, the process of generating the global optimization control parameters includes:

[0175] Using numerical simulation methods, the geological and production data of each control unit are obtained from the spatial layout data of the oil field, and the production efficiency of the control unit is calculated;

[0176] If the production efficiency of the control unit is lower than a preset threshold, the water injection parameters and oil production parameters are adjusted based on the differentiation strategy to generate preliminary adjustment parameters;

[0177] Through the particle swarm optimization algorithm, the inter-unit production data is obtained from the preliminary adjustment parameters, and the inter-unit production balance index is calculated;

[0178] If the inter-unit production balance index does not reach the preset threshold, the water injection parameters and oil production parameters are iteratively adjusted to generate optimized adjustment parameters;

[0179] According to the optimized adjustment parameters, dynamic production data is extracted from the oil field spatial layout data, and the updated control unit production efficiency is recalculated;

[0180] By comparing the updated control unit production efficiency with the initial efficiency, the convergence of the global optimization parameters is judged and the final global optimization control parameters are generated;

[0181] Based on the final global optimization control parameters and combined with the oilfield spatial layout data, the water injection and oil production execution plans of each control unit are generated.

[0182] For example, during the implementation of the differentiated water injection strategy, a permeability field model was established using geological modeling software, and the reservoir was divided into five control units (unit A with a permeability of 85mD and unit B with a permeability of 210mD). The Eclipse numerical simulator was used to calculate the water injection efficiency of each unit, and it was found that the injection-production ratio of unit C needed to be adjusted to 1.2 to maintain pressure balance. In the optimization of oil production parameters, a neural network prediction model was established based on historical production data, and eight parameters such as bottom hole pressure (1200psi to 1500psi) and water cut (45% to 78%) were input to output the optimal liquid production rate (unit D needs to be adjusted from 200m 3 / d increased to 230m 3 / d). The spatial layout analysis uses the Voronoi algorithm to divide the well network into production units. The connectivity index between units is calculated in combination with the GIS system (0.35 to 0.82). It is identified that there is a pressure drop gradient of 0.18 between units E and F that needs to be coordinated. The particle swarm optimization sets 30 particles. The objective function includes the production variance (weight 0.6) and the water content difference (weight 0.4). After 50 iterations, the global optimal solution reduces the standard deviation of the daily oil production of each unit from 15.7t to 6.3t. The final output is the water injection intensity of each unit (1.5m 3 / (d·m) to 3.2m 3 / (d·m)) and pump strokes (4 times / min to 7 times / min) are optimized parameter matrices.

[0183] Furthermore, the process of generating dynamic scheduling instructions for maintenance operations and logistics supplies includes:

[0184] Obtain weather forecast data, equipment status information, and production plan data, and generate a comprehensive data set through data fusion processing;

[0185] If the comprehensive data set is complete, the particle swarm optimization algorithm is used to calculate the control parameters and obtain the optimized parameter set;

[0186] Based on the optimized parameter set, a constraint satisfaction algorithm is used to generate a resource allocation plan and determine the resource allocation result;

[0187] If the resource allocation result satisfies the production plan data constraints, a maintenance operation instruction is generated to obtain a maintenance scheduling sequence;

[0188] Generate logistics supply instructions and determine the supply scheduling sequence based on maintenance scheduling sequence and logistics supply requirements;

[0189] Integrate maintenance scheduling sequence and supply scheduling sequence to generate dynamic scheduling instructions;

[0190] Get real-time information to update comprehensive data sets, and loop execution to achieve continuously optimized dynamic scheduling instructions.

[0191] For example, based on weather forecast data predicting heavy rainfall of 50 mm for eight consecutive hours over the next 24 hours, the system automatically triggers an equipment protection plan. Using genetic algorithm optimization, it relocates three excavators in the open-air work area to a rain shelter and adjusts the lifting operation window of two tower cranes to within the four-hour interval between rain events. Integrating real-time data from the equipment status monitoring system, the bearing temperature of one of the rollers exceeded the threshold of 85°C. Using a fuzzy logic control algorithm, the system calculates the optimal maintenance plan, scheduling a two-hour cooling maintenance period for the equipment during the rainy period and automatically generates a request list for five spare parts. The production planning database indicates that two priority A bridge pouring tasks must be completed within 36 hours. Using a multi-objective particle swarm algorithm, the system dynamically schedules concrete pump trucks, transport vehicles, and worker teams. The system calculates that the three pump trucks should prioritize continuous pouring of Pier 1, while simultaneously allocating two transport vehicles to provide intermittent supply to Pier 2. During the resource allocation phase, the system calculated the optimal fuel refueling plan based on a linear programming model. It dispatched refueling trucks to replenish 12 pieces of equipment with a total of 3,800 liters of diesel between operations. Using a path optimization algorithm, it mapped the shortest refueling route covering five operational locations, saving an estimated 45 minutes of transportation time. All dispatch instructions were pushed to mobile devices in real time, and the central control system's 3D visualization interface was simultaneously updated, displaying a heat map of the adjusted equipment distribution and a Gantt chart of the operational progress.

[0192] Furthermore, the process of generating a production parameter adjustment or production stoppage instruction and determining safety protection measures includes:

[0193] Obtain typhoon or equipment failure risk signals through submarine sensor networks and determine the risk signal type;

[0194] If the risk signal type is typhoon, the wind speed and wave height are calculated using the preset risk model to obtain the risk level;

[0195] According to the risk level, the corresponding preset risk model is matched from the emergency plan management system to generate production parameter adjustment instructions;

[0196] If the risk level exceeds the preset threshold, a production suspension instruction is generated and the suspension time range is determined;

[0197] The management system matches the safety protection measures database to obtain the protection measures plan corresponding to the production suspension order;

[0198] Use machine learning classification algorithms to optimize the protection measures and obtain the final security protection measures;

[0199] Extract execution parameters from the final security protection measures and generate device control instructions.

[0200] For example, a submarine sensor network uses sensors deployed on the seabed to monitor ocean environmental parameters such as water temperature, flow rate, and pressure in real time. If the water temperature rises abnormally above 28°C and the flow rate exceeds 2 meters per second, the system triggers a typhoon warning signal. Simultaneously, equipment failure risk signals are generated by sensors monitoring equipment operating status. For example, if the current fluctuation exceeds 10% of the rated value or the temperature exceeds 60°C, the system will determine that the equipment is at risk of failure. Based on the emergency plan management system, the system matches pre-set risk models. For example, the typhoon risk model analyzes data such as wind speed and wave height. If the wind speed exceeds level 12 and the wave height exceeds 6 meters, the system will generate a production parameter adjustment instruction, reducing the drilling speed of the drilling platform from 60 revolutions per minute to 30 revolutions per minute to reduce equipment load. If the equipment failure risk model analysis indicates that the equipment temperature continues to rise and exceeds 70°C, the system will generate a production suspension instruction and initiate safety measures, such as automatically shutting down the equipment power and activating the cooling system to ensure equipment safety. The system will also determine whether it is necessary to initiate an emergency evacuation procedure based on real-time data analysis. For example, when a typhoon path forecast shows that it will pass through the work area within 24 hours, the system will notify all personnel in advance to evacuate to a safe area to ensure their safety.

[0201] Furthermore, the process of generating a reservoir management strategy includes:

[0202] Collect real-time production data from edge computing nodes and transmit it to the land data center using a preset encryption protocol to obtain an encrypted real-time data set;

[0203] If the received encrypted real-time data set is complete, the data is decrypted in the land data center, combined with the stored historical production data, and data cleaning methods are used to remove noise to obtain the integrated production data set;

[0204] For the integrated production data set, long short-term memory network is used to perform time series analysis, extract long-term trend features, and obtain trend feature sets;

[0205] Based on the trend feature set, the parameters of the digital twin model are updated, and the model accuracy is judged using the preset model verification method to obtain the updated digital twin model;

[0206] The updated digital twin model is used to simulate the reservoir operation status, and a rule engine is used to generate preliminary reservoir management strategies to obtain a candidate set of strategies.

[0207] If the strategies in the strategy candidate set meet the preset constraints, the optimal strategy is selected from them to generate the final reservoir management strategy and obtain the management strategy output;

[0208] According to the management policy output, the collection parameters of the edge computing nodes are adjusted to optimize the collection process of real-time production data and obtain the optimized collection configuration.

[0209] For example, when acquiring real-time production data from edge computing nodes, sensors deployed at oil well sites can collect parameters such as pressure, temperature, and flow rate, with a sampling frequency set to once per second. This data is then transmitted to an edge gateway via the MQTT protocol. The edge gateway preprocesses the data, for example, using a Kalman filter to remove noise and compress the data into minute-by-minute averages to reduce bandwidth usage. The processed data is then transmitted to a land-based data center via a 5G network, with latency kept under 100 milliseconds. At the data center, real-time data is combined with historical production data, including daily production and water injection rates over the past five years. Trend mining is performed using a long short-term memory (LSTM) network. The LSTM model has an input layer with 128 neurons and a hidden layer with 64 neurons. The output layer provides a production forecast for the next 30 days. The model is trained using the Adam optimizer with a learning rate of 0.001, 100 training cycles, and a mean squared error (MSE) loss function. The final prediction error is kept within 5%. Based on the LSTM predictions, the digital twin model is dynamically updated daily, adjusting parameters such as reservoir permeability and porosity. The updated model simulates the impact of different water injection schemes on production and generates optimal reservoir management strategies. For example, it recommends adjusting the daily water injection rate from 500 cubic meters to 550 cubic meters to increase cumulative production over the next month. Throughout the entire process, data collection, transmission, analysis, and decision-making are all automated, requiring no human intervention.

[0210] Furthermore, the process of determining the final production deployment plan includes:

[0211] Reservoir management strategies, geological data, production performance data, and oilfield spatial layout data are obtained to construct a proxy model that incorporates fluid migration patterns, resulting in a proxy model simulation framework. Using this proxy model simulation framework and combined with a set of constraints, well location and pipeline network adjustment plans are simulated to obtain a preliminary set of adjustment plans. If the preliminary set of adjustment plans satisfies the preset set of constraints, the oilfield spatial layout data is used to calculate the spatial adaptability of each plan and obtain a spatial adaptability score. Based on the spatial adaptability score and the optimization objective function, adjustment plans that meet the optimization objective are screened and candidate layout optimization recommendations are determined. For these candidate layout optimization recommendations, the geological data and fluid migration patterns are integrated to verify their feasibility and obtain validated optimization recommendations. These validated optimization recommendations are then combined with production performance data to generate a final production deployment plan, which is then determined. Based on the final production deployment plan, the oilfield spatial layout data is updated to generate adjusted spatial layout data, resulting in the optimized spatial layout of oilfield A.

[0212] For example, in the reservoir management strategy, the well location and pipeline network adjustment plan is first simulated through the agent model, and the historical production data is trained using the random forest algorithm based on machine learning to predict the production changes under different well location layouts.

[0213] For example, by inputting parameters such as well depth, permeability, and porosity, the model predicts that adding a new well between well locations A and B could increase daily production by 15%. Then, combined with oilfield spatial layout data, a geographic information system (GIS) was used for spatial analysis to generate layout optimization recommendations.

[0214] For example, by analyzing the distance between well locations and the pipeline network, they optimized the pipeline route, reducing pipeline length by approximately 500 meters and thus lowering construction costs. Finally, based on this optimized layout, a linear programming algorithm was used to determine the final production deployment plan, ensuring that production capacity requirements were met while maximizing economic benefits.

[0215] For example, by optimizing the switching strategy for production wells, annual returns are expected to increase by 8%. Throughout the entire process, data analysis and algorithm application are closely integrated to ensure that every decision is scientific and actionable.

[0216] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A multi-system combined self-regulation method for offshore oil fields, characterized in that: include: Acquire multi-sensor data from drilling equipment and calculate drilling trajectory deviations based on pre-established geological models; If the drilling trajectory deviation exceeds a preset threshold, the geological parameters in the digital twin model are updated based on the multi-sensor data, and a genetic algorithm is used to optimize the drilling angle and speed to obtain the optimal drilling path; Update the reservoir geological model according to the optimal drilling path, obtain pressure, temperature and flow data from the seabed sensor network, and transmit it to the edge computing node through the data communication system; Analyze the pressure, temperature, and flow data at the edge computing node, calculate the production status of each control unit, and extract abnormal fluctuation characteristics; If the abnormal fluctuation characteristics exceed the preset range, a neural network is used to analyze the time series regularity of the abnormal fluctuation characteristics to generate a differentiated water injection and oil production parameter adjustment strategy; Based on the differentiated water injection and oil production parameter adjustment strategy, combined with the oilfield spatial layout data, a numerical simulation method is used to evaluate the production efficiency of each control unit, and a particle swarm optimization algorithm is used to coordinate the production balance between units and generate global optimized control parameters; Based on the global optimization control parameters, combined with weather forecasts, equipment status and production plan data, an intelligent scheduling algorithm is used to calculate resource allocation plans and generate dynamic scheduling instructions for maintenance operations and logistics supplies; If the submarine sensor network detects a typhoon or equipment failure risk signal, the emergency plan management system matches the preset risk model, generates production parameter adjustment or shutdown instructions, and determines safety protection measures; According to the security protection measures, real-time production data is obtained from the edge computing node and transmitted to the land data center, and combined with historical production data, long-term trends are mined using a long-short-term memory network to update the digital twin model and generate a reservoir management strategy; Based on the reservoir management strategy, an agent model is used to simulate well location and pipeline network adjustment plans. Combined with the oilfield spatial layout data, layout optimization suggestions are generated and the final production deployment plan is determined.

2. The method according to claim 1, characterized in that The process of acquiring multi-sensor data from drilling equipment and calculating drilling trajectory deviations based on pre-established geological models includes: Acquiring multi-sensor data based on drilling equipment; wherein the multi-sensor data includes drill bit position and orientation data and geological parameters, and storing the data as a raw data set; The data preprocessing method is used to denoise and standardize the original data set to obtain the processed sensor data set; The processed sensor data set is fused through the Kalman filter algorithm to generate drill bit position and direction data and determine the fused positioning data; Generating a real-time drilling trajectory using a geometric calculation method based on the fused positioning data and a preset geological model to obtain drilling trajectory data; If there is a deviation between the drilling trajectory data and the expected trajectory of the preset geological model, the trajectory deviation is calculated by the least square method to obtain a deviation value; Based on the deviation value, the PID control algorithm is used to adjust the drilling equipment parameters and generate a control instruction data set; The drilling equipment is driven by the control instruction data set, the drill bit position and direction are updated, and new sensor data are obtained.

3. The method according to claim 1, characterized in that The process of obtaining the optimal drilling path includes: If the drilling trajectory deviation exceeds a preset threshold, real-time drilling parameters are obtained from the multi-sensor data, and the sensor data is processed using a filtering algorithm to obtain denoised drilling parameters; Based on the denoised drilling parameters, the geological parameters in the digital twin model are updated, and the spatial distribution of the geological parameters is calculated using an interpolation algorithm to determine the updated geological model. If the deviation between the updated geological model and the preset geological model exceeds a threshold, a genetic algorithm is used to optimize the drilling angle and speed based on the drilling parameters and the geological model to obtain a preliminary optimized drilling path; Calculating the deviation between the drilling trajectory and the target path using the preliminary optimized drilling path, fitting the trajectory deviation using the least squares method, and determining trajectory adjustment parameters; Adjusting parameters according to the trajectory, updating the drilling path in the digital twin model, and verifying the stability of the adjusted path using a simulation algorithm to obtain a verified drilling path; If the stability of the verified drilling path meets a preset threshold, extracting drilling control parameters from the verified path to determine a final drilling path; According to the final drilling path, a drilling equipment control instruction is generated and sent to the drilling equipment using a real-time transmission protocol to obtain the drilling status after execution.

4. The method according to claim 1, wherein The process of updating the reservoir geological model based on the optimal drilling path, acquiring pressure, temperature, and flow data from the seabed sensor network, and transmitting the data to the edge computing node through the data communication system includes: Using a preset sampling frequency, pressure data, temperature data, and flow data are collected from the submarine sensor network in real time to obtain the original sensor data set; The original sensor data set is encoded and compressed using a differential encoding technique through a data communication system to obtain a compressed data packet; If the integrity check of the compressed data packet passes, the compressed data packet is transmitted to the edge computing node through a low-latency communication protocol to obtain a transmission data stream; Decoding and preprocessing the transmission data stream at the edge computing node, smoothing data fluctuations using a sliding window averaging method to obtain a smoothed data sequence; updating the reservoir geological model using a particle filter algorithm based on the smoothed data sequence and preset reservoir geological parameters to obtain an updated geological model; Based on the updated geological model, a genetic algorithm is used to optimize the drilling path and determine the optimal drilling trajectory; If the deviation of the optimal drilling trajectory is less than a preset threshold, the optimal drilling trajectory is output to the drilling control system to obtain a path control instruction.

5. The method according to claim 1, wherein The process of calculating the production status of each control unit and extracting abnormal fluctuation characteristics includes: Obtain pressure, temperature, and flow data collected by edge computing nodes, clean the data using preset standardization rules, and obtain a standardized data set; Using a time series analysis method, the normalized data set is segmented, the production state parameters of each control unit are calculated, and the state characteristic vector is determined; If the fluctuation amplitude of the state feature vector exceeds a preset threshold, the abnormal fluctuation frequency feature is extracted by Fourier transform to obtain an abnormal fluctuation feature set; According to the abnormal fluctuation feature set, a cluster analysis algorithm is used to group the features and determine the abnormal state category of each control unit; Extracting high-risk categories from the abnormal state categories, generating abnormal fluctuation feature descriptions of the control unit through preset mapping rules, and determining a distribution pattern of the abnormal fluctuation features; According to the distribution pattern of the abnormal fluctuation characteristics, a decision tree algorithm is used to generate an optimized control strategy to obtain adjustment parameters for each control unit; The adjustment parameters are sent to each control unit through the edge computing node, the production status is updated in real time, and closed-loop processing of abnormal fluctuations is completed.

6. The method according to claim 1, characterized in that If the abnormal fluctuation characteristics exceed the preset range, a neural network is used to analyze the time series regularity of the abnormal fluctuation characteristics and generate a differentiated water injection and oil production parameter adjustment strategy, including the following process: If the abnormal fluctuation characteristics exceed the preset range, time series data is extracted from the collected data, and the time series data is denoised using a data processing method to obtain a smoothed time series; Performing regularity analysis on the smoothed time series through a neural network, extracting periodic features in the smoothed time series using a long short-term memory network, and obtaining fluctuation regularity features; If the fluctuation regularity matches the historical fluctuation pattern, the preset water injection and oil production parameter adjustment rules are queried based on the matching results to determine the preliminary adjustment parameters; Based on the preliminary adjustment parameters, a genetic algorithm is used to optimize the water injection and oil production parameters to generate a differentiated adjustment strategy; Extracting water injection parameters and oil production parameters based on the differentiated adjustment strategy, determining whether the parameter adjustment meets the production constraints through parameter comparison and analysis, and obtaining a parameter set that meets the constraints; If the parameter set meets the constraints, the stability of the adjusted parameters is verified through real-time data, and the fluctuation trend of the parameters after adjustment is determined by time series prediction method to obtain the final adjustment strategy; According to the final adjustment strategy, control instructions for water injection and oil production equipment are generated, and the control instructions are transmitted through the equipment interface to complete the parameter adjustment.

7. The method according to claim 1, characterized in that The process of generating global optimization control parameters includes: Using numerical simulation methods, the geological and production data of each control unit are obtained from the spatial layout data of the oil field, and the production efficiency of the control unit is calculated; If the production efficiency of the control unit is lower than a preset threshold, the water injection parameters and the oil production parameters are adjusted based on the differentiation strategy to generate preliminary adjustment parameters; Through the particle swarm optimization algorithm, the inter-unit production data is obtained from the preliminary adjustment parameters, and the inter-unit production balance index is calculated; If the inter-unit production balance index does not reach the preset threshold, the water injection parameters and oil production parameters are iteratively adjusted to generate optimized adjustment parameters; According to the optimized adjustment parameters, dynamic production data is extracted from the oil field spatial layout data, and the updated control unit production efficiency is recalculated; By comparing the updated control unit production efficiency with the initial efficiency, the convergence of the global optimization parameters is judged and the final global optimization control parameters are generated; Based on the final global optimization control parameters and combined with the oilfield spatial layout data, the water injection and oil production execution plans of each control unit are generated.

8. The method according to claim 1, characterized in that The process of generating dynamic dispatch instructions for maintenance operations and logistics supplies includes: Obtain weather forecast data, equipment status information, and production plan data, and generate a comprehensive data set through data fusion processing; If the comprehensive data set is complete, the control parameters are calculated using a particle swarm optimization algorithm to obtain an optimized parameter set; Based on the optimized parameter set, a constraint satisfaction algorithm is used to generate a resource allocation plan and determine the resource allocation result; If the resource allocation result satisfies the production plan data constraints, a maintenance operation instruction is generated to obtain a maintenance scheduling sequence; Generate logistics supply instructions and determine the supply scheduling sequence according to the maintenance scheduling sequence and logistics supply requirements; Fusion of the maintenance scheduling sequence and the supply scheduling sequence to generate dynamic scheduling instructions; Get real-time information to update comprehensive data sets, and loop execution to achieve continuously optimized dynamic scheduling instructions.

9. The method according to claim 1, characterized in that The process of generating production parameter adjustment or shutdown instructions and determining safety protection measures includes: Obtain typhoon or equipment failure risk signals through submarine sensor networks and determine the risk signal type; If the risk signal type is typhoon, the wind speed and wave height are calculated using the preset risk model to obtain the risk level; According to the risk level, a corresponding preset risk model is matched from the emergency plan management system to generate a production parameter adjustment instruction; If the risk level exceeds the preset threshold, a production suspension instruction is generated and the suspension time range is determined; The management system matches the safety protection measures database to obtain the protection measures plan corresponding to the production suspension order; Use machine learning classification algorithms to optimize the protection measures and obtain the final security protection measures; Execution parameters are extracted from the final security protection measures to generate device control instructions.

10. The method according to claim 1, characterized in that The process of generating a reservoir management strategy includes: Collect real-time production data from edge computing nodes and transmit it to the land data center using a preset encryption protocol to obtain an encrypted real-time data set; If the received encrypted real-time data set is complete, the data is decrypted in the land data center, combined with the stored historical production data, and data cleaning methods are used to remove noise to obtain the integrated production data set; For the integrated production data set, long short-term memory network is used to perform time series analysis, extract long-term trend features, and obtain trend feature sets; According to the trend feature set, the parameters of the digital twin model are updated, and the model accuracy is judged using a preset model verification method to obtain an updated digital twin model; The updated digital twin model is used to simulate the reservoir operation status, and a rule engine is used to generate preliminary reservoir management strategies to obtain a candidate set of strategies. If the strategies in the strategy candidate set meet the preset constraints, the optimal strategy is selected from them to generate the final reservoir management strategy and obtain the management strategy output; According to the management policy output, the collection parameters of the edge computing nodes are adjusted to optimize the collection process of real-time production data and obtain the optimized collection configuration.

Images