Model Predictive Control Method for Strengthening Health Perception of Subsea Production Systems in Offshore Oil

Abstract

The present invention relates to the technical field of petroleum engineering, and proposes a model predictive control method for enhancing the health perception of an offshore oil underwater production system. This method includes three core steps: multi-source sensor information fusion, health assessment and remaining useful life prediction, and model predictive control with enhanced health perception. By fusing multi-source sensor data, a health index and a dynamic Bayesian network model are constructed to predict the remaining useful life of the system, and the health status and production objectives are incorporated into the optimization framework to dynamically adjust the rotational speed of the electric submersible pump and the opening of the choke valve, achieving a balance between production efficiency and equipment life. The present invention improves the state monitoring accuracy through multi-source sensor data fusion, uses a dynamic Bayesian network to achieve flexible and real-time health assessment, and provides a scientific basis for maintenance decisions. The model predictive control with enhanced health perception optimizes the control instructions, ensures the safety and efficiency of the production process, effectively extends the service life of the system, and has important practical value.

Description

Technical Field

[0001] The present invention relates to the technical field of oil engineering, and proposes a model predictive control method for enhancing the health perception of an offshore oil underwater production system. Background Art

[0002] With the continuous growth of global energy demand, the development of offshore oil resources has become an important way to meet energy supply. As the core component of offshore oil exploitation, the stability and reliability of the underwater production system directly affect the production efficiency and economic benefits of oil fields. Electrical submersible pumps and chokes, as key equipment in the underwater production system, are responsible for providing downhole fluid lifting power and regulating oil well production respectively. However, the complexity of the underwater environment such as high pressure, low temperature, corrosive media, etc., and the performance degradation caused by long-term operation of equipment have significantly increased the failure risk of key equipment in the underwater production system, seriously threatening the safety and continuity of the underwater production system.

[0003] Traditional model predictive control methods have been widely used in underwater production systems, which mainly achieve production goals by optimizing control instructions. However, the existing model predictive control methods do not consider the health status and remaining service life of equipment, resulting in control instructions that may exceed the bearing range of equipment, accelerating equipment degradation and even causing failures. Therefore, it is particularly necessary to propose a model predictive control method for enhancing the health perception of an offshore oil underwater production system. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies of the prior art and propose a model predictive control method for enhancing the health perception of an offshore oil underwater production system.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions:

[0006] A model predictive control method for enhancing the health perception of an offshore oil underwater production system, including:

[0007] Multi-source sensor information fusion: Preprocess and align the data from pressure sensors, temperature sensors, and vibration sensors of each component in the underwater production system, use the sliding window method to solve the problem of inconsistent sampling frequencies, linearly interpolate the low-frequency data to align its timestamp with the high-frequency data, and perform mean calculation to generate output data with a unified frequency; Normalize the sensor data with different dimensions to the [0,1] interval for subsequent fusion calculation; Use the reliability allocation weights of each component, and based on the sensor data of each component, fuse the multi-source sensor data by the weighted average method;

[0008] Health assessment and remaining service life prediction: Based on the fused multi-source sensor data, construct a health index reflecting the current health status of the underwater production system HI; A dynamic Bayesian network model for the health status of the underwater production system is constructed using the sensor data of each component and the reliability allocation weights of each component. As time changes, when the sensor data of each component is updated, the new data is input into the dynamic Bayesian network of the health status of the underwater production system through the evidence nodes, and the dynamic update of the health index of the model is completed; According to the dynamic health index obtained from the dynamic Bayesian network of the health status of the underwater production system, the remaining service life of the underwater production system is calculated;

[0009] Model predictive control with enhanced health perception: Incorporate the equipment health status and production goals into the optimization framework simultaneously, dynamically adjust the speed of the electric submersible pump and the opening of the choke valve, and achieve a balance between maximizing production efficiency and extending equipment life; The inputs in the controller architecture are real-time sensor data, the prediction results of the remaining service life, and the expected oil and gas production targets, and the outputs are the control instructions for the speed of the electric submersible pump and the opening of the choke valve. The core modules include a state prediction model, a remaining service life constraint module, and an optimization solver; The objective function comprehensively considers the tracking error of oil and gas production and the health loss of the equipment, converts the predicted value of the remaining service life into the upper and lower limit constraints of the control input, and the state variables such as pressure and flow need to be within the safe range; Based on the system model, predict the state variables in the future time domain, construct the objective function and constraint conditions, call the quadratic programming solver to generate the optimal control sequence, send the optimal control instructions to the actuator, and update the system state to enter the next control cycle.

[0010] Furthermore, the multi-source sensor information fusion includes:

[0011] In the multi-source sensor information fusion step, the time length of the sliding window is T. For each window, collect the data of all sensors within this time period; For high-frequency sensors, there are Nh = fh × T data points within the window T; For low-frequency sensors, there are Nl = fl × T data points within the window T; Perform linear interpolation on the low-frequency data to align its timestamp with the high-frequency data; For the i-th data point yl(ti) of the low-frequency sensor, the interpolation at time t (ti ≤ t ≤ ti+1) is:

[0012] ;

[0013] Calculate the mean value of the data within each window to generate output data with a unified frequency; Let the data of the high-frequency sensor within the window T be { y h ( t 1 ), y h ( t 2 ), y h ( t Nh)}, the interpolation data of the low-frequency sensor is { y l ( t 1 ), y l ( t 2 ), y l ( t Nh ). The mean value within the window is:

[0014] ;

[0015] ;

[0016] The sliding window moves forward with a step size of ∆t to update the data within the window and repeat the above calculations.

[0017] Furthermore, in the multi-source sensor information fusion step, the sensor data with different dimensions is normalized to the [0, 1] interval to facilitate subsequent fusion calculations. For the sensor data y, the normalized data yn is:

[0018] .

[0019] Furthermore, in the multi-source sensor information fusion step, the reliability of each component is used to assign weights, and based on the sensor data of each component, the weighted average method is used to fuse the multi-source sensor data:

[0020] ;

[0021] where y f is the fused data, w i is the weight calibrated according to historical data.

[0022] Furthermore, in the health assessment and remaining useful life prediction step, based on the fused multi-source sensor data, a health index HI reflecting the current health state of the underwater production system is constructed:

[0023] ;

[0024] where y m is the reference value in the normal state of the underwater production system.

[0025] Further, in the health assessment and remaining service life prediction step, a dynamic Bayesian network model of the underwater production system health state is constructed using the sensor data of each component and the reliability allocation weights of each component; in this model, E 1 — E n The nodes are the evidence nodes of the sensor data of each component of the underwater production system n for updating each sensor data; Y 1 — Y n The nodes are the sensor data of each component of the underwater production system n ; W The nodes are the reliability allocation weight nodes; Y f The nodes are the multi-source sensor data fusion nodes; HI is the underwater production system health index node; Y m The nodes are the benchmark value nodes under the normal state of the underwater production system; as time changes, when each sensor data is updated, the new data is input into the dynamic Bayesian network of the underwater production system health state through the evidence nodes to complete the dynamic update of the health index of this model.

[0026] Further, in the health assessment and remaining service life prediction step, the remaining service life of the underwater production system is the time period from the current monitoring time to when the health index first reaches the failure threshold. According to the dynamic health index obtained from the dynamic Bayesian network of the underwater production system health state, the remaining service life of the underwater production system can be calculated as:

[0027] ;

[0028] where, RUL sys is the remaining service life of the underwater production system, G is the failure threshold.

[0029] Further, in the model predictive control step with enhanced health perception, the objective function comprehensively considers the oil and gas production tracking error and equipment health loss and can be expressed as:

[0030] ;

[0031] where, q ( k ) is the system output, r ( k ) is the expected production target, u ( k ) is the control input, Q andR are weight matrices, representing the costs of production tracking error and control input respectively, HI ( k ) is the health index, reflecting the current degradation state of the system, λ is the health weight factor, which is dynamically adjusted as the remaining useful life shortens:

[0032] ;

[0033] wherein, λ 0 is the initial weight, β is the decay coefficient of the health weight factor;

[0034] Furthermore, in the model predictive control step with enhanced health perception, the predicted value of the remaining useful life is converted into the upper and lower limit constraints of the control input:

[0035] ;

[0036] wherein, u max ( k ) gradually decreases as the remaining useful life shortens, avoiding equipment overload:

[0037] ;

[0038] wherein, u design is the designed maximum input, α is the decay coefficient of the control input upper limit;

[0039] State variables such as pressure and flow rate need to be within a safe range:

[0040] .

[0041] Furthermore, in the model predictive control step with enhanced health perception, the state variables in the future time domain are predicted based on the system model:

[0042] ;

[0043] wherein, f (∙) is the system state equation;

[0044] The objective function and constraint conditions are constructed, and the quadratic programming solver is called to generate the optimal control sequence:

[0045] ;

[0046] The optimal control instructions, that is, the speed of the electric submersible pump and the opening of the choke valve, are sent to the actuator, and the system state is updated to enter the next control cycle.

[0047] The beneficial effects of the present invention are as follows:

[0048] Through the effective fusion of multi-source sensor information, the state of the underwater production system can be monitored more precisely, thereby enhancing the accuracy of health assessment. In addition, the application of the dynamic Bayesian network makes the assessment of the health state more flexible and real-time, while the remaining useful life prediction provides a scientific basis for maintenance decisions. The model predictive control with enhanced health perception not only optimizes the control instructions but also ensures the safety and efficiency of the production process, which has important practical value in dealing with potential failures and extending the system life. Description of the Drawings

[0049] Figure 1 is the dynamic Bayesian network model of the health state of the underwater production system;

[0050] Figure 2 is a schematic diagram of the underwater production system;

[0051] Figure 3 is a schematic diagram of the model predictive control system with enhanced health perception for the offshore oil underwater production system.

[0052] In the figure, 101, underwater distribution unit, 102, hydraulic distribution module, 103, electronic distribution module, 104, underwater control module, 105, control pilot valve, 106, first solenoid valve, 107, second solenoid valve, 108, third solenoid valve, 109, underwater electronic module, 110, fourth solenoid valve, 111, fifth solenoid valve, 112, sixth solenoid valve, 113, underwater valve group, 114, chemical agent injection valve, 115, conversion valve, 116, annulus wing valve, 117, production wing valve, 118, production main valve, 119, annulus main valve, 120, wellhead mechanical module, 121, choke module, 122, downhole electric submersible pump module, 123, main control station 123, 201, multi-source sensor information fusion subsystem, 202, underwater production system monitoring data preprocessing and alignment module, 203, underwater production system sensor data normalization module, 204, underwater production system sensor data fusion module, 205, health assessment and remaining useful life prediction subsystem, 206, underwater production system health index calculation module, 207, underwater production system health state dynamic Bayesian network establishment module, 208, underwater production system remaining useful life prediction module, 209, model predictive control subsystem with enhanced health perception, 210, controller architecture establishment module, 211, objective function and constraint condition design module, 212, control instruction optimization module. Detailed Embodiments

[0053] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

[0054] Embodiment 1:

[0055] As Figure 1 shown, a model predictive control method for enhancing the health perception of an offshore oil underwater production system mainly includes three steps: multi-source sensor information fusion, health assessment and remaining useful life prediction, and health perception-enhanced model predictive control.

[0056] Among them, the specific steps of multi-source sensor information fusion are:

[0057] S101: Preprocessing and alignment of underwater production system monitoring data. Align the timestamps of the data from the pressure sensors, temperature sensors, and vibration sensors of each component of the underwater production system, and use the sliding window method to solve the problem of inconsistent sampling frequencies. Let the time length of the sliding window be T , for each window, collect the data of all sensors within this time period. For high-frequency sensors (sampling frequency f h ), within the window T there are N h = f h × T data points; for low-frequency sensors (sampling frequency f l ), within the window T there are N l = f l × T data points. Perform linear interpolation on the low-frequency data to align its timestamp with the high-frequency data. For the i th data point y l ([[]] t i ) of the low-frequency sensor, the interpolation at time t ([[]] t i ≤ t ≤ t i+1 ) is:

[0058] ;

[0059] Calculate the mean value of the data within each window to generate output data with a unified frequency. Let the data of the high-frequency sensor within the window T be {y h ( t 1 ), y h ( t 2 ), y h ( t Nh )}, the interpolation data of the low-frequency sensor is { y l ( t 1 ), y l ( t 2 ), y l ( t Nh )}, and the mean value within the window is:

[0060] ;

[0061] ;

[0062] The sliding window moves forward with a step size of ∆t to update the data within the window and repeat the above calculations.

[0063] S102: Normalize the sensor data of the subsea production system. Normalize the sensor data with different dimensions to the interval [0, 1] for subsequent fusion calculations. For the sensor data y , the normalized data y n is:

[0064] ;

[0065] S103: Fuse the sensor data of the subsea production system. Utilize the reliability of each component to allocate weights and fuse the multi-source sensor data based on the sensor data of each component using the weighted average method:

[0066] ;

[0067] Among them, y f is the fused data, w i is the weight calibrated according to historical data.

[0068] The specific steps for health assessment and remaining useful life prediction are:

[0069] S201: Calculate the health index of the underwater production system. Based on the fused multi-source sensor data, construct a health index that reflects the current health status of the underwater production system HI :

[0070] ;

[0071] Among them, y m is the reference value under the normal state of the underwater production system.

[0072] S202: Establish a dynamic Bayesian network for the health status of the underwater production system. Use the sensor data of each component and the reliability allocation weights of each component to construct a dynamic Bayesian network model for the health status of the underwater production system, as Figure 1 shown. In this model, E 1 — E n The nodes are the evidence nodes of the sensor data of each component of the underwater production system n for updating each sensor data; Y 1 — Y n The nodes are the sensor data of each component of the underwater production system n ; W The node is the reliability allocation weight node; Y f The node is the multi-source sensor data fusion node; HI is the health index node of the underwater production system; Y m The node is the reference value node under the normal state of the underwater production system. As time changes, when each sensor data is updated, the new data is input into the dynamic Bayesian network of the health status of the underwater production system through the evidence node, and the health index of the model is dynamically updated.

[0073] S203: Predict the remaining useful life of the underwater production system. The remaining useful life of the underwater production system is the time period from the current monitoring time to when the health index first reaches the failure threshold. According to the dynamic health index obtained from the dynamic Bayesian network of the health status of the underwater production system, the remaining useful life of the underwater production system can be calculated as:

[0074] ;

[0075] Among them, RUL sys is the remaining useful life of the underwater production system, G is the failure threshold.

[0076] The specific steps of model predictive control with enhanced health perception are as follows:

[0077] S301: Controller architecture establishment. The core idea of model predictive control for enhancing the health perception of underwater production systems is to incorporate both the equipment health status and production goals into the optimization framework, dynamically adjust the speed of the electrical submersible pump and the opening of the choke valve, and achieve a balance between maximizing production efficiency and extending equipment life. The inputs in the controller architecture are real-time sensor data, the predicted remaining useful life, and the expected oil and gas production goals; the outputs are the control commands for the speed of the electrical submersible pump and the opening of the choke valve; the core modules include a state prediction model, a remaining useful life constraint module, and an optimization solver.

[0078] S302: Design of the objective function and constraint conditions. The objective function comprehensively considers the tracking error of oil and gas production and equipment health degradation, and can be expressed as:

[0079] ;

[0080] where, q ( k ) is the system output, r ( k ) is the expected production goal, u ( k ) is the control input, Q and R are the weight matrices respectively, representing the costs of production tracking error and control input, HI ( k ) is the health index, reflecting the current degradation state of the system, λ is the health weight factor, which is dynamically adjusted as the remaining useful life shortens:

[0081] ;

[0082] where, λ 0 is the initial weight, β is the decay coefficient of the health weight factor.

[0083] Convert the predicted value of the remaining useful life into the upper and lower limit constraints of the control input:

[0084] ;

[0085] where, u max ( k ) gradually decreases as the remaining useful life shortens to avoid equipment overload:

[0086] ;

[0087] where, u design is the designed maximum input,α To control the input upper limit attenuation coefficient.

[0088] State variables such as pressure and flow rate need to be within a safe range:

[0089] ;

[0090] S303: Control instruction optimization. Predict state variables in the future time domain based on the system model:

[0091] ;

[0092] Among them, f (∙) is the system state equation.

[0093] Construct the objective function and constraint conditions, and call the quadratic programming solver to generate the optimal control sequence:

[0094] ;

[0095] Send the optimal control instructions, i.e., the speed of the electric submersible pump and the opening of the choke valve, to the actuator, and update the system state to enter the next control cycle.

[0096] Embodiment 2:

[0097] Such as Figure 2As shown, the subsea production system includes a subsea distribution unit 101, a subsea control module 104, a subsea valve group 113, and a wellhead mechanical module 120. Among them, the subsea distribution unit 101 is located outside the Christmas tree and installed on the subsea support, including a hydraulic distribution module 102 and an electronic distribution module 103. The hydraulic distribution module 102 is connected to the control pilot valve 105 through a hydraulic pipeline to provide hydraulic power for the control pilot valve 105. The electronic distribution module 103 is connected to the subsea electronic module 109 through a cable to provide power for the subsea electronic module 109. The subsea control module 104 is located between the subsea distribution unit 101 and the subsea Christmas tree and installed on the subsea support, including a control pilot valve 105, a subsea electronic module 109, a first solenoid valve 106, a second solenoid valve 107, a third solenoid valve 108, a fourth solenoid valve 110, a fifth solenoid valve 111, and a sixth solenoid valve 112. The control pilot valve 105 is connected to the first solenoid valve 106, the second solenoid valve 107, the third solenoid valve 108, the fourth solenoid valve 110, the fifth solenoid valve 111, and the sixth solenoid valve 112 through a hydraulic pipeline to provide hydraulic power for the six solenoid valves. The subsea electronic module 109 is connected to the first solenoid valve 106, the second solenoid valve 107, the third solenoid valve 108, the fourth solenoid valve 110, the fifth solenoid valve 111, and the sixth solenoid valve 112 through a cable to control the opening and closing of the six solenoid valves. The first solenoid valve 106 is connected to the annulus wing valve 116 through a hydraulic pipeline to control the opening and closing of the annulus wing valve 116. The second solenoid valve 107 is connected to the chemical injection valve 114 through a hydraulic pipeline to control the opening and closing of the chemical injection valve 114. The third solenoid valve 108 is connected to the annulus master valve 119 through a hydraulic pipeline to control the opening and closing of the annulus master valve 119. The fourth solenoid valve 110 is connected to the switching valve 115 through a hydraulic pipeline to control the opening and closing of the switching valve 115. The fifth solenoid valve 111 is connected to the production master valve 118 through a hydraulic pipeline to control the opening and closing of the production master valve 118. The sixth solenoid valve 112 is connected to the production wing valve 117 through a hydraulic pipeline to control the opening and closing of the production wing valve 117. The subsea valve group 113 is located on the Christmas tree and installed on the body of the subsea Christmas tree, including a chemical injection valve 114, a switching valve 115, an annulus wing valve 116, a production wing valve 117, a production master valve 118, and an annulus master valve 119. The chemical injection valve 114, the switching valve 115, the annulus wing valve 116, the production wing valve 117, the production master valve 118, and the annulus master valve 119 are connected to the oil pipeline to control the transportation of oil. The wellhead mechanical module 120 is located at the subsea wellhead and installed at the bottom of the subsea Christmas tree. The wellhead mechanical module 120 is connected to the oil pipeline to provide power for subsea oil and gas transportation. The choke module 121 is located downhole and installed at the oil outlet. The choke module 121 is connected to the oil pipeline to control the oil outlet pressure. The downhole ESP module 122 is located downhole and installed in the oil reservoir.The downhole electric submersible pump module 122 is connected to the choke module 121 and is used for extracting downhole petroleum; the main control station 123 is located on the water surface and installed in the control station; the main control station 123 is connected to the underwater distribution unit 101 through a cable and is used for controlling underwater components and collecting underwater sensor information.

[0098] Embodiment 3:

[0099] As Figure 3 shown, the model predictive control system for enhancing the health perception of an underwater production system for offshore oil includes a multi-source sensor information fusion subsystem 201 installed at an underwater wellhead, a health assessment and remaining service life prediction subsystem 205 installed at the underwater wellhead, and a model predictive control subsystem 209 for enhancing health perception installed at the underwater wellhead.

[0100] The multi-source sensor information fusion subsystem 201 includes a preprocessing and alignment module 202 for monitoring data of the underwater production system, a normalization module 203 for sensor data of the underwater production system, and a fusion module 204 for sensor data of the underwater production system; the preprocessing and alignment module 202 for monitoring data of the underwater production system is connected to the main control station 123 through a cable and is used for collecting and preprocessing sensor data of the underwater production system; the normalization module 203 for sensor data of the underwater production system is connected to the preprocessing and alignment module 202 for monitoring data of the underwater production system through a cable and is used for normalizing sensor data of the underwater production system; the fusion module 204 for sensor data of the underwater production system is connected to the normalization module 203 for sensor data of the underwater production system through a cable and is used for data fusion of sensors of the underwater production system.

[0101] The health assessment and remaining service life prediction subsystem 205 includes a health index calculation module 206 for the underwater production system, a dynamic Bayesian network establishment module 207 for the health state of the underwater production system, and a remaining service life prediction module 208 for the underwater production system; the health index calculation module 206 for the underwater production system is connected to the sensor data fusion module 204 of the underwater production system through a cable and is used for calculating the health index of the underwater production system; the dynamic Bayesian network establishment module 207 for the health state of the underwater production system is connected to the health index calculation module 206 for the underwater production system through a cable and is used for establishing a dynamic assessment model for the health state of the underwater production system; the remaining service life prediction module 208 for the underwater production system is connected to the dynamic Bayesian network establishment module 207 for the health state of the underwater production system through a cable and is used for calculating the health service life of the underwater production system.

[0102] The model predictive control subsystem 209 with enhanced health perception includes a controller architecture establishment module 210, an objective function and constraint design module 211, and a control instruction optimization module 212; the controller architecture establishment module 210 is connected to the remaining service life prediction module 208 of the underwater production system through a cable and is used to construct a controller architecture integrating life information; the objective function and constraint design module 211 is connected to the controller architecture establishment module 210 through a cable and is used to construct an optimization mathematical model for control; the control instruction optimization module 212 is connected to the objective function and constraint design module 211, the choke valve module 121, and the downhole electric submersible pump module 122 through a cable and is used to optimize the control instructions of the choke valve module 121 and the downhole electric submersible pump module 122.

[0103] As described above, it is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes should be covered within the protection scope of the present invention.

Claims

1. A model predictive control method for enhancing health perception of an offshore oil underwater production system, characterized in that: include: First, through multi-source sensor information fusion, the pressure, temperature, and vibration sensor data are preprocessed and aligned, and the data are normalized and fused; Secondly, a health index HI is constructed based on the fused data, and a dynamic Bayesian network model is used to predict the health status and remaining service life of the underwater production system. Finally, through model predictive control, the equipment health status and production goals are incorporated into the optimization framework, and the speed and nozzle opening of the electric submersible pump are dynamically adjusted to achieve a balance between maximizing production efficiency and extending equipment life. The controller receives real-time sensor data, remaining service life prediction and oil and gas production targets, outputs control instructions, and generates the optimal control sequence through a quadratic programming solver. The specific steps include: Multi-source sensor information fusion: pre-process and align the data from the pressure sensors, temperature sensors, and vibration sensors of various components of the underwater production system, use the sliding window method to solve the problem of inconsistent sampling frequencies, perform linear interpolation on low-frequency data to align it with the timestamp of high-frequency data, and perform mean calculation to generate output data with uniform frequency; Normalize sensor data of different dimensions to the interval [0,1] to facilitate subsequent fusion calculations; use the reliability of each component to assign weights, and use the weighted average method to fuse multi-source sensor data based on the sensor data of each component; Health assessment and remaining service life prediction: Based on the fused multi-source sensor data, a health index HI that reflects the current health status of the underwater production system is constructed; a dynamic Bayesian network model of the health status of the underwater production system is constructed using the sensor data of each component and the reliability allocation weight of each component. As time changes, when the data of each sensor is updated, the new data is input into the dynamic Bayesian network of the health status of the underwater production system through the evidence node to complete the dynamic update of the health index of the model; based on the dynamic health index obtained by the dynamic Bayesian network of the health status of the underwater production system, the remaining service life of the underwater production system is calculated; Model predictive control with enhanced health perception: The equipment health status and production targets are simultaneously incorporated into the optimization framework, and the speed and nozzle opening of the submersible pump are dynamically adjusted to achieve a balance between maximizing production efficiency and extending equipment life. The inputs in the controller architecture are real-time sensor data, remaining service life prediction results, and expected oil and gas production targets, and the outputs are submersible pump speed control instructions and nozzle opening control instructions. The core module includes a state prediction model, a remaining service life constraint module, and an optimization solver. The objective function comprehensively considers oil and gas production tracking errors and equipment health losses, and converts the remaining service life prediction value into upper and lower limit constraints of the control input. State variables such as pressure and flow must be within a safe range. Based on the system model, the state variables in the future time domain are predicted, the objective function and constraints are constructed, the quadratic programming solver is called to generate the optimal control sequence, the optimal control instructions are sent to the actuator, and the system state is updated to enter the next control cycle.

2. The model predictive control method for enhancing health perception of an offshore oil underwater production system according to claim 1 is characterized in that: The multi-source sensor information fusion includes: In the multi-source sensor information fusion step, the time length of the sliding window is T. For each window, the data of all sensors in the time period are collected; for high-frequency sensors, there are Nh=fh×T data points in window T; for low-frequency sensors, there are Nl=fl×T data points in window T; linear interpolation is performed on the low-frequency data to align it with the timestamp of the high-frequency data; for the i-th data point yl(ti) of the low-frequency sensor, the interpolation at time t (ti≤t≤ti+1) is: ； The data in each window is averaged to generate output data with uniform frequency. Let the data bits of the high-frequency sensor in window T be { y h ( t 1), y h ( t 2), y h ( t Nh )}, the interpolation data of the low-frequency sensor is { y l ( t 1), y l ( t 2), y l ( t Nh )}, the mean value in the window is: ； ； Sliding window with step size ∆t Move forward, update the data in the window and repeat the above calculation.

3. The model predictive control method for enhancing health perception of an offshore oil underwater production system according to claim 2 is characterized in that: In the multi-source sensor information fusion step, the sensor data of different dimensions are normalized to the interval [0,1] to facilitate subsequent fusion calculations. For sensor data y, the normalized data yn is: 。 4. The model predictive control method for enhancing health perception of an offshore oil underwater production system according to claim 3 is characterized in that: In the multi-source sensor information fusion step, the reliability of each component is used to assign weights, and the multi-source sensor data is fused using the weighted average method based on the sensor data of each component: ； in, y f is the fused data, w i The weights are calibrated based on historical data.

5. The model predictive control method for enhancing health perception of an offshore oil underwater production system according to claim 4 is characterized in that: In the health assessment and remaining service life prediction steps, a health index HI reflecting the current health status of the underwater production system is constructed based on the fused multi-source sensor data: ； in, y m It is the benchmark value under normal conditions of the underwater production system.

6. The model predictive control method for enhancing health perception of an offshore oil underwater production system according to claim 5 is characterized in that: In the health assessment and remaining service life prediction steps, the sensor data of each component and the reliability allocation weight of each component are used to construct a dynamic Bayesian network model of the health status of the underwater production system; in this model, E 1— E n Nodes are components of the underwater production system n The evidence node of the sensor data is used to update the data of each sensor; Y 1— Y n Nodes are components of the underwater production system n Sensor data; W Nodes are reliability-assigned weight nodes; Y f The node is a multi-source sensor data fusion node; HI It is the health index node of the underwater production system; Y m The node is the baseline value node under the normal state of the underwater production system; as time changes, when the data of each sensor is updated, the new data is input into the dynamic Bayesian network of the health status of the underwater production system through the evidence node to complete the dynamic update of the health index of the model.

7. The model predictive control method for enhancing health perception of an offshore oil underwater production system according to claim 6 is characterized in that: In the health assessment and remaining service life prediction steps, the remaining service life of the underwater production system is the time period from the current monitoring time to the time when the health index first reaches the fault threshold. According to the dynamic health index obtained by the dynamic Bayesian network of the health status of the underwater production system, the remaining service life of the underwater production system can be calculated as: ； in, RUL sys The remaining useful life of the subsea production system, G is the failure threshold.

8. The model predictive control method for enhancing health perception of an offshore oil underwater production system according to claim 7 is characterized in that: In the health-aware enhanced model predictive control step, the objective function comprehensively considers the oil and gas production tracking error and equipment health loss, which can be expressed as: ； in, q ( k ) is the system output, r ( k ) is the expected production target, u ( k ) is the control input, Q and R are weight matrices, representing the cost of production tracking error and control input respectively, HI ( k ) is the health index, reflecting the current degradation state of the system. λ is the health weight factor, which is adjusted dynamically as the remaining service life decreases: ； in, λ 0 is the initial weight, β is the attenuation coefficient of the health weight factor.

9. The model predictive control method for enhancing health perception of an offshore oil underwater production system according to claim 8 is characterized in that: In the health-aware enhanced model predictive control step, the remaining useful life prediction value is converted into the upper and lower limit constraints of the control input: ； in, u max ( k ) gradually decreases as the remaining service life shortens to avoid equipment overload: ； in, u design For the design maximum input, α is the upper limit attenuation coefficient of the control input; State variables such as pressure and flow must be within a safe range: 。 10. The model predictive control method for enhancing health perception of an offshore oil underwater production system according to claim 9, characterized in that: In the health-aware enhanced model predictive control step, the state variables in the future time domain are predicted based on the system model: ； in, f (∙) is the system state equation; Construct the objective function and constraints, and call the quadratic programming solver to generate the optimal control sequence: ； The optimal control instruction, namely the speed of the electric submersible pump and the nozzle opening, is sent to the actuator, and the system status is updated to enter the next control cycle.

Images