A closed-loop control method and system for submersible pumps based on multi-sensor fusion

Abstract

This invention relates to the field of submersible pump control technology, and discloses a closed-loop control method and system for submersible pumps based on multi-sensor fusion. The method acquires comprehensive operational data by deploying multiple types of sensors at the submersible pump inlet and motor end and synchronizing them in time. This method innovatively introduces a mechanism for storing and analyzing historical operating condition snapshots, and determines the stable operating state of the submersible pump based on the fluctuation degree or mean change rate and fluctuation range of various sensor data, ensuring the effectiveness of data acquisition.

Description

Technical Field

[0001] This invention relates to the field of submersible pump control technology, and more specifically, to a closed-loop control method and system for submersible pumps based on multi-sensor fusion. Background Technology

[0002] In submersible pump closed-loop control, multi-sensor fusion technology is typically used as a core approach. By collecting signals such as downhole pressure, temperature, motor current, power, vibration, and surface fluid production, the system fuses these signals to determine whether the submersible pump is currently in the high-efficiency zone, the critical fluid supply zone, or the insufficient fluid supply zone, and outputs frequency adjustment commands accordingly. When well conditions are relatively stable, the multi-sensor fusion system can accurately distinguish the fluid supply status based on these multi-source characteristics, keeping the submersible pump within its optimal operating range.

[0003] However, in long-term stable production and frequency regulation scenarios for wells with declining production, the formation fluid supply capacity typically decreases slowly over weeks or months, the fluid level recovery rate slows down, and the gas cut increases. At this point, the condition boundary determination stage in the multi-sensor fusion system is prone to problems. As production time increases, at the same pump frequency, the pressure recovery and production response relationships shift compared to the initial stage, and the true critical condition characteristics appear to have moved earlier. However, existing multi-sensor fusion systems typically still use the state boundaries established at the initial stage of operation to interpret the current multi-source characteristics, continuing to identify combinations of characteristics that are already close to insufficient fluid supply as normal fluid supply or a state where frequency can be increased.

[0004] This state judgment bias leads to an overestimation of frequency regulation, further deteriorating the operating conditions at the submersible pump inlet and causing new changes in pressure and electrical parameters. Because existing conventional methods lack an adaptive adjustment mechanism for the migration of operating state boundaries with historical feedback, these new changes are again interpreted within the old boundaries, resulting in continuous boundary misjudgments and accumulated control biases. This causes the closed-loop control to fail to promptly pull the submersible pump back from the critical supply zone, instead maintaining or increasing the frequency, leading to increased fluctuations in oil output, higher energy consumption, and the risk of equipment overload.

[0005] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0006] This invention provides a closed-loop control method and system for submersible pumps based on multi-sensor fusion, aiming to solve the problem that in the existing closed-loop control of submersible pumps, the multi-sensor fusion system is prone to deviation in the determination of the working condition boundary in the long-term stable production frequency adjustment scenario of the production decline well, resulting in the frequency adjustment decision being too high, which in turn causes aggravated fluctuations in oil production, increased energy consumption, and the risk of equipment overload.

[0007] The technical solution of this application is as follows:

[0008] Firstly, this application discloses a closed-loop control method for a submersible pump based on multi-sensor fusion, specifically including:

[0009] Fiber optic pressure and temperature sensors are deployed at the suction inlet of the submersible pump, and current and vibration sensors are deployed at the motor end of the submersible pump. The time synchronization of each sensor is marked based on a synchronous clock bus to obtain various time-synchronized sensor data.

[0010] The submersible pump is driven to run according to the frequency adjustment command sent by the ground frequency converter. When the submersible pump reaches a stable operating state, it captures various sensor data to form a historical operating condition snapshot and stores the historical operating condition snapshot in the historical data pool, which is deployed in non-volatile memory.

[0011] The stable operating status is determined in the following ways:

[0012] When the fluctuation of multiple sensor data is lower than a preset first threshold, it is determined to be in the first stable state; or, when the mean change rate of multiple sensor data is lower than a preset second threshold and the fluctuation range is lower than a preset third threshold, it is determined to be in the second stable state, and a historical working condition snapshot is reconstructed based on the envelope of multiple sensor data.

[0013] Obtain the preset initial judgment boundary, which is pre-calibrated based on the submersible pump factory test data and prior historical data of wells of the same model;

[0014] Historical operating condition snapshot sequences are extracted from the historical data pool. Pressure and power response characteristics reflecting the decline in formation fluid supply capacity and envelope width characteristics reflecting the degree of gas intrusion are extracted and trend analysis is performed to obtain long-term evolution trend information of formation fluid supply capacity.

[0015] Based on trend information, calculate the response offset of the current submersible pump operating condition relative to its initial operating state;

[0016] Based on the response offset, the preset initial decision boundary is corrected to obtain the adaptive decision boundary;

[0017] Based on the adaptive decision boundary, the current operating condition of the submersible pump is identified. The operating condition includes at least one of the following: normal state, insufficient fluid supply state, gas intrusion state, and wear state.

[0018] Based on the identification results, a frequency adjustment command is generated and sent to the ground frequency converter to adjust the speed of the submersible pump;

[0019] When any data from multiple sensors is continuously missing or exceeds the preset range, the system switches to fault protection mode. Fault protection mode is executed as follows: the current inverter output frequency is locked and an alarm signal is generated.

[0020] Furthermore, based on the response offset, the initial decision boundary is corrected, including:

[0021] The offset direction and magnitude of each sensing feature are obtained based on the response offset, and a boundary correction factor is established for each sensing feature.

[0022] The boundary correction factors of each sensing feature are constructed into a dimension-matched combined weight matrix according to the number of sensing features. The combined weight matrix is ​​used to fuse the correction factors to generate a boundary translation vector and a rotation matrix.

[0023] By applying the boundary translation vector and rotation matrix to the mathematical expression of the initial decision boundary, the corrected decision boundary is obtained.

[0024] Building upon the above, this application further proposes reconstructing historical operating condition snapshots based on the envelopes of multiple sensor data, including:

[0025] The system acquires the current short-term fluctuation range of various sensor data and dynamically adjusts the sliding window length based on the current short-term fluctuation range, where the larger the fluctuation range, the shorter the window length.

[0026] Based on the dynamically adjusted sliding window, the sliding mean and sliding variance of various sensor data are calculated.

[0027] When the rate of change of the moving mean is lower than the preset mean drift tolerance threshold and the range of the moving variance is lower than the preset variance fluctuation threshold, it is determined to be a dynamic steady state interval.

[0028] Within the dynamic steady-state range, search for local extreme points, connect the maximum points to form the upper envelope, and connect the minimum points to form the lower envelope;

[0029] The arithmetic midline of the upper and lower envelopes is calculated as the expected value of the feature, and the average distance between the upper and lower envelopes is calculated as the envelope width feature. The expected value of the feature and the envelope width feature are then encapsulated as a snapshot of historical operating conditions.

[0030] Furthermore, trend analysis includes:

[0031] Feature decomposition was performed on the historical operating condition snapshot sequence to obtain pressure and power response features, product fluid composition features, envelope width features, and frequency and current features.

[0032] Trend components were extracted from each feature, cross-validation and correlation analysis were performed to identify the superimposed effects of different factors, and the long-term evolution trend information of formation fluid supply capacity was obtained after weighted fusion.

[0033] Through this technical solution, this application can achieve a comprehensive and in-depth analysis of the long-term evolution trend of formation fluid supply capacity. Through multi-feature decomposition and cross-validation, it effectively identifies potential influencing factors under complex operating conditions and provides reliable trend information for adaptive boundary correction.

[0034] In some preferred embodiments, the response offset of the current submersible pump operating condition relative to its initial operating state is calculated, including:

[0035] A weighting coefficient is assigned to each sensing feature. The current operating condition feature value is compared with the initial state feature value to obtain the original offset. After weighting, the offset is mapped to the offset space through a nonlinear mapping function, and the result is combined to obtain the response offset.

[0036] Through this technical solution, this application can achieve accurate quantification of response offset. By introducing weighting coefficients and nonlinear mapping functions, the offset can more accurately reflect the actual changes in the submersible pump's operating conditions, thereby enhancing the sensitivity of adaptive control.

[0037] As a technical improvement, the weighted values ​​are mapped to the offset space through a nonlinear mapping function, and the resulting combinations yield the response offset, including:

[0038] Obtain the stage identifier of the current formation fluid supply capacity decline, select the corresponding mapping function from the preset nonlinear mapping function set according to the stage identifier, input the weighted offset into the mapping function, and obtain the mapped offset.

[0039] To improve the plan, it is necessary to obtain indicators of the current decline in formation fluid supply capacity, including:

[0040] Acquire multi-source downhole data, including formation pressure data, production volume data, and water cut data;

[0041] Perform quality checks on downhole multi-source data, and identify and process missing data points, abnormal data points, and delayed data points;

[0042] Based on the processed data, the formation pressure decrease rate, the cumulative decrease in fluid production, and the water cut change trend were calculated.

[0043] By comprehensively judging the rate of formation pressure decline, the magnitude of cumulative fluid production decline, and the trend of water cut changes, the stage indicator of the current decline in formation fluid supply capacity can be obtained.

[0044] Based on the above, a comprehensive judgment is made, including:

[0045] Obtain the well type and formation condition information of the current well, and match the stage segmentation parameter set from the well type and formation feature library;

[0046] The relative importance of dynamically assessing the rate of formation pressure decline, the magnitude of cumulative fluid production decline, and the trend of water cut changes;

[0047] The phase identifier is determined by weighting the components according to their relative importance, matching them with the phase division parameter set, and combining them with development phase information.

[0048] Through this technical solution, this application can achieve intelligent judgment of the decline stage of formation fluid supply capacity. By combining well type, formation conditions and dynamic assessment of the importance of each indicator, the determination of stage identification is more accurate and personalized, adapting to the complexity of different oil wells.

[0049] As a further improvement, the relative importance of dynamically assessing the formation pressure decline rate, the cumulative decrease in production, and the trend of water cut changes includes:

[0050] Preprocessing of formation pressure data, fluid production data, and water cut data;

[0051] Calculate the short-term volatility and long-term trend stability of each data point under different time windows;

[0052] The confidence levels of each indicator are adjusted based on the short-term volatility and the stability of the long-term trend.

[0053] By combining the correlation between various indicators and the stages of decline in formation fluid supply capacity in historical data, a preliminary importance level is calculated;

[0054] By monitoring the vibration intensity and motor temperature of the submersible pump in real time, the degree of disturbance to the downhole environment is assessed; the initial importance is then adjusted based on the degree of disturbance to obtain the relative importance.

[0055] Secondly, this application also discloses a closed-loop control system for a submersible pump based on multi-sensor fusion, used to execute the above method, specifically including:

[0056] The downhole sensor assembly, deployed at the suction inlet and motor end of the submersible pump, includes:

[0057] Fiber optic pressure and temperature sensors are deployed at the inlet of the submersible pump.

[0058] Current and vibration sensors are deployed at the motor end of the submersible pump;

[0059] A synchronous clock bus is connected to the fiber optic pressure sensor, fiber optic temperature sensor, current sensor, and vibration sensor respectively, and is used to mark the time synchronization of each sensor.

[0060] Surface control components, deployed at the wellhead surface, include:

[0061] The fiber optic sensor signal demodulator is connected to the fiber optic pressure sensor and fiber optic temperature sensor of the downhole sensor assembly via optical fiber.

[0062] The controller is connected to an optical fiber sensor signal demodulator, a current sensor, a vibration sensor, and a synchronous clock bus. It includes a non-volatile memory and a processor. The non-volatile memory stores a computer program and a historical data pool. When the processor executes the computer program, it implements the steps of the above method.

[0063] The frequency converter, connected to the power supply cable of the controller and the submersible pump, is used to receive frequency modulation commands generated by the controller and adjust the speed of the submersible pump.

[0064] An alarm, connected to the controller, is used to issue an alarm signal in fault protection mode.

[0065] Beneficial effects

[0066] This application provides a closed-loop control method and system for submersible pumps based on multi-sensor fusion. By deploying multiple types of sensors at the pump's suction inlet and motor end and synchronizing them in time, comprehensive operational data is acquired. This method innovatively introduces a mechanism for storing and analyzing historical operating condition snapshots, and determines the stable operating state of the submersible pump based on the fluctuation level or mean change rate and fluctuation range of various sensor data, ensuring the effectiveness of data acquisition. Attached Figure Description

[0067] Figure 1 This is a schematic flowchart of a closed-loop control method for a submersible pump based on multi-sensor fusion provided in an embodiment of the present invention;

[0068] Figure 2 This is a schematic diagram of a closed-loop control system for a submersible pump based on multi-sensor fusion, provided in an embodiment of the present invention. Detailed Implementation

[0069] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0070] Reference Figure 1 , Figure 1 This is a flowchart illustrating a closed-loop control method for a submersible pump based on multi-sensor fusion, provided in an embodiment of the present invention, including:

[0071] S11, fiber optic pressure and temperature sensors are deployed at the suction port of the submersible pump, and current and vibration sensors are deployed at the motor end of the submersible pump. The time synchronization of each sensor is marked based on the synchronous clock bus to obtain various time-synchronized sensor data.

[0072] S12 drives the submersible pump to run according to the frequency adjustment command sent by the ground frequency converter. When the submersible pump reaches a stable operating state, it captures various sensor data to form a historical operating condition snapshot and stores the historical operating condition snapshot in the historical data pool. The historical data pool is deployed in non-volatile memory.

[0073] The stable operating status is determined in the following ways:

[0074] When the fluctuation levels of multiple sensor data are below a preset first threshold, it is determined to be in a first stable state; or...

[0075] When the mean change rate of multiple sensor data is lower than the preset second threshold and the fluctuation range is lower than the preset third threshold, it is determined to be in the second stable state, and the historical working condition snapshot is reconstructed based on the envelope of multiple sensor data.

[0076] Obtain the preset initial judgment boundary, which is pre-calibrated based on the submersible pump factory test data and prior historical data of wells of the same model;

[0077] Historical operating condition snapshot sequences are extracted from the historical data pool. Pressure and power response characteristics reflecting the decline in formation fluid supply capacity and envelope width characteristics reflecting the degree of gas intrusion are extracted and trend analysis is performed to obtain long-term evolution trend information of formation fluid supply capacity.

[0078] Based on trend information, the response offset of the current submersible pump operating condition relative to its initial operating state is calculated; based on the response offset, the preset initial judgment boundary is corrected to obtain the adaptive judgment boundary.

[0079] Based on the adaptive decision boundary, the current operating condition of the submersible pump is identified. The operating condition includes at least one of the following: normal state, insufficient fluid supply state, gas intrusion state, and wear state.

[0080] Based on the identification results, a frequency adjustment command is generated and sent to the ground frequency converter to adjust the speed of the submersible pump. When any data from multiple sensors is continuously missing or exceeds the preset range, the system switches to fault protection mode. Fault protection mode is executed in the following manner: the current frequency converter output frequency is locked and an alarm signal is generated.

[0081] This application introduces an adaptive decision boundary correction mechanism, which can effectively solve the problem of misjudgment of the working condition boundary in the long-term stable production frequency adjustment scenario of the traditional multi-sensor fusion system in the production decline well. This avoids the risks of high frequency adjustment decision, deterioration of working condition, increase in energy consumption and equipment overload, and significantly improves the accuracy and reliability of the submersible pump closed-loop control.

[0082] The method proposed in this application is mainly applied in the field of oil and gas extraction, particularly for oil wells using submersible pumps for artificial lift. A submersible pump is a device that converts electrical energy into mechanical energy, generating centrifugal force through impeller rotation to lift downhole fluids to the surface. During the operation of a submersible pump, its working condition is affected by various factors such as formation fluid supply capacity, gas content, and equipment wear. To achieve efficient and stable operation of the submersible pump, closed-loop control is required.

[0083] Fiber optic pressure and temperature sensors utilize optical fibers as sensing elements to detect pressure and temperature by measuring changes in light signals. They offer advantages such as high temperature resistance, high pressure resistance, and resistance to electromagnetic interference, making them particularly suitable for harsh downhole environments. Current and vibration sensors monitor the current and vibration of the submersible pump motor, respectively; these parameters are crucial indicators of the submersible pump's operating status and health. A synchronous clock bus is a communication mechanism to ensure consistent data acquisition times from different sensors. Time synchronization markers ensure alignment of multi-source sensor data in the time dimension, laying the foundation for subsequent data fusion and analysis. A historical operating condition snapshot is a set of multi-source sensor data captured during stable submersible pump operation, representing the pump's characteristics under that specific condition. A historical data pool is a database storing historical operating condition snapshots, typically deployed in non-volatile memory to ensure long-term data preservation and reliability. Initial judgment boundaries are pre-set thresholds or zones based on the submersible pump's factory test data and prior historical data from wells of the same model, used to initially determine the submersible pump's operating condition. The "adaptive decision boundary" is obtained by correcting the initial decision boundary based on the long-term operating trend information and response offset of the submersible pump, and can better adapt to changes in downhole operating conditions.

[0084] Fiber optic pressure and temperature sensors are deployed at the inlet of the submersible pump (SMP), while current and vibration sensors are deployed at the motor end. Each sensor is time-synchronized using a synchronous clock bus to obtain various time-synchronized sensor data. The placement of the sensors is crucial for obtaining accurate downhole information. For example, fiber optic pressure and temperature sensors can be installed near the SMP inlet to directly measure the pressure and temperature of the fluid entering the pump. Current sensors can be integrated into the SMP motor control unit to monitor motor current consumption in real time. Vibration sensors can be mounted on the SMP body or motor housing to detect abnormal vibrations. The synchronous clock bus can use standard protocols such as NTP (Network Time Protocol) or PTP (Precision Time Protocol) to ensure that all sensor data is accurately timestamped.

[0085] The submersible pump is driven by frequency adjustment commands sent by the ground-based frequency converter. When the submersible pump reaches a stable operating state, it captures various sensor data to form a historical operating condition snapshot, which is stored in a historical data pool deployed in non-volatile memory. The ground-based frequency converter controls the submersible pump's speed by adjusting the output frequency, thereby changing its head and flow rate. The stable operating state of the submersible pump can be determined in several ways. For example, a time window can be set; when the fluctuation of all sensor data (such as standard deviation or variance) is lower than a preset first threshold within this time window, the submersible pump is considered to be in a first stable state. Alternatively, the mean change rate and fluctuation range of the sensor data can be calculated; when the mean change rate is lower than a preset second threshold and the fluctuation range is lower than a preset third threshold, it is determined to be in a second stable state. In the second stable state, to more accurately reflect the operating condition characteristics, the historical operating condition snapshot can be reconstructed based on the envelope of various sensor data. Historical data pools can be implemented using non-volatile storage such as flash memory, solid-state drives (SSDs), or EEPROMs to ensure that data is not lost in the event of a power outage.

[0086] A preset initial decision boundary is obtained, which is pre-calibrated based on the submersible pump's factory test data and prior historical data from wells of the same model. Obtaining the initial decision boundary is fundamental to operating condition identification. For example, during the submersible pump's factory test, parameters such as pressure, temperature, current, and vibration can be measured at different speeds and flow rates, and the corresponding operating conditions (e.g., normal, gas intrusion, insufficient fluid supply, etc.) can be recorded. This data can be used to construct an initial operating condition classification model or decision rules. Simultaneously, historical operating data of the same model of submersible pump in other wells can be referenced to further refine and calibrate the initial decision boundary. These boundaries can be hyperplanes, hyperspheres, or more complex decision boundaries in multidimensional space.

[0087] Historical operating condition snapshot sequences are extracted from the historical data pool. Pressure and power response characteristics reflecting the decline in formation fluid supply capacity, and envelope width characteristics reflecting the degree of gas intrusion, are extracted and trend analysis is performed to obtain long-term evolution trend information of formation fluid supply capacity. The historical operating condition snapshot sequences contain the operating characteristics of submersible pumps at different time points. By analyzing these sequences, the long-term evolution trend of formation fluid supply capacity can be revealed. For example, average pressure and pressure recovery rate from pressure sensor data can be extracted as pressure response characteristics, and average current and power consumption from current sensor data can be extracted as power response characteristics. The envelope width characteristic of the degree of gas intrusion can be obtained through envelope analysis of pressure or vibration signals. Trend analysis can employ time series analysis methods, such as moving average, exponential smoothing, and regression analysis, to identify the declining trend and fluctuation cycle of formation fluid supply capacity.

[0088] Based on trend information, the response offset of the current submersible pump operating condition relative to its initial operating state is calculated. The response offset reflects the difference between the current operating condition and the initial operating condition. For example, if the formation fluid supply capacity decreases, the submersible pump may exhibit phenomena such as a decrease in suction inlet pressure and an increase in motor current fluctuations at the same speed. These changes can be quantified by calculating the difference or ratio between the characteristic values ​​of the current operating condition and the characteristic values ​​of the initial state.

[0089] Based on the response offset, the preset initial judgment boundary is corrected to obtain an adaptive judgment boundary. This adaptive judgment boundary correction mechanism is one of the core innovations of this application. For example, if the response offset indicates a continuous decline in formation fluid supply capacity, the pressure threshold originally used to determine insufficient fluid supply may need to be adjusted downwards to identify the risk of insufficient fluid supply earlier. This correction can be linear or non-linear, depending on the relationship between the response offset and boundary changes.

[0090] Based on adaptive decision boundaries, the current operating state of the submersible pump is identified. This operating state includes at least one of the following: normal state, insufficient fluid supply state, air intrusion state, and wear state. Operating state identification is a crucial step in closed-loop control. For example, when the submersible pump's inlet pressure remains consistently below the adaptively adjusted insufficient fluid supply threshold, the system can determine it to be in an insufficient fluid supply state. Abnormal high-frequency components or an increased envelope width in the vibration sensor data may indicate air intrusion or wear.

[0091] Based on the identification results, a frequency adjustment command is generated and sent to the ground-based frequency converter to regulate the submersible pump's speed. The generation of the frequency adjustment command is based on the operating condition identification results. For example, if insufficient fluid supply is identified, the system can generate a command to reduce the submersible pump's speed to decrease the pump's suction volume, restore the fluid level, and prevent cavitation. If gas intrusion is identified, the system may need to adjust the speed to optimize gas-liquid separation.

[0092] When any data from multiple sensors is continuously missing or exceeds a preset range, the system switches to fault protection mode. Fault protection mode is executed as follows: the current inverter output frequency is locked, and an alarm signal is generated. Fault protection mode is designed to handle sensor failures or abnormal data. For example, if fiber optic pressure sensor data is continuously missing for a certain period, or if current sensor data suddenly exceeds its normal operating range, the system should immediately enter fault protection mode to prevent the submersible pump from operating under incorrect commands, causing equipment damage or production accidents. Locking the current inverter output frequency avoids incorrect frequency adjustment operations when data is abnormal, while generating an alarm signal promptly notifies operators to intervene. The multi-sensor fusion-based closed-loop control method for submersible pumps proposed in this application effectively solves the problem of misjudgment of operating condition boundaries in traditional multi-sensor fusion systems during long-term stable production frequency adjustment scenarios in wells with declining production by introducing an adaptive judgment boundary correction mechanism.

[0093] Specifically, the core innovation of this application lies in:

[0094] First, by deploying multiple types of sensors at the submersible pump inlet and motor end, and using time synchronization marking based on a synchronous clock bus, the accuracy and consistency of multi-source sensor data were ensured, laying a solid foundation for subsequent operating condition analysis.

[0095] Secondly, this application introduces the concepts of historical operating condition snapshots and historical data pools. By capturing and storing various sensor data under stable operating conditions of submersible pumps, it provides data support for the long-term evolution trend analysis of formation fluid supply capacity.

[0096] Furthermore, this application extracts pressure and power response characteristics as well as envelope width characteristics by performing trend analysis on historical operating condition snapshot sequences, thereby obtaining long-term evolution trend information of formation fluid supply capacity. This trend information is crucial for revising the operating condition determination boundary.

[0097] Most importantly, this application calculates the response offset of the current submersible pump operating condition relative to its initial operating state based on the long-term evolution trend information of formation fluid supply capacity, and corrects the preset initial judgment boundary according to the response offset to obtain an adaptive judgment boundary. This adaptive correction mechanism enables the operating condition judgment boundary to be dynamically adjusted with the actual changes in downhole operating conditions, thereby avoiding misjudgments caused by boundary solidification in traditional methods.

[0098] Finally, based on adaptive boundary judgment, the current operating state of the submersible pump is identified, and frequency modulation commands are generated according to the identification results, thus realizing closed-loop control of the submersible pump. Simultaneously, this application also designs a fault protection mode to cope with abnormal sensor data, improving the robustness and safety of the system.

[0099] This application, by introducing a response offset and an adaptive decision boundary correction mechanism, can dynamically adjust the operating condition decision boundary according to the long-term evolution trend of formation fluid supply capacity. For example, when the system detects a decreasing trend in formation fluid supply capacity, it automatically moves the decision boundary for insufficient fluid supply forward, thereby identifying the critical fluid supply state of the submersible pump earlier and more accurately. This adaptive adjustment capability enables the closed-loop control system to promptly pull the submersible pump back from the critical fluid supply zone, avoiding the risks of deteriorating operating conditions, increased energy consumption, and equipment overload caused by excessively high frequency tuning decisions. Therefore, this application significantly improves the accuracy, reliability, and adaptability of the submersible pump closed-loop control, which has important practical significance for achieving long-term, efficient, and stable oil well production.

[0100] In some embodiments described above, a scheme is proposed to correct a preset initial judgment boundary based on the response offset to obtain an adaptive judgment boundary. However, in practical applications, the operating conditions of submersible pumps are affected by various factors such as formation conditions and equipment aging, and their characteristic responses do not change simply linearly, but exhibit complex, multi-dimensional offsets. If only simple translation or scaling is used to correct the judgment boundary, it may not be able to fully capture the complex changes in the interaction between various sensing features, resulting in the corrected judgment boundary not accurately reflecting the true state of the current operating condition, thereby affecting the accuracy of operating condition identification and the effectiveness of the control strategy.

[0101] In this regard, this application further proposes a specific method for correcting the initial determination boundary based on the response offset, including:

[0102] The offset direction and magnitude of each sensing feature are obtained based on the response offset, and a boundary correction factor is established for each sensing feature.

[0103] The boundary correction factors of each of the sensing features are constructed according to the number of sensing features to form a dimension-matched combined weight matrix. The combined weight matrix is ​​used to fuse the correction factors to generate a boundary translation vector and a rotation matrix.

[0104] The boundary translation vector and the rotation matrix are applied to the mathematical expression of the initial decision boundary to obtain the corrected decision boundary.

[0105] Specifically, response offset refers to the difference in response between the current operating condition of the submersible pump and its initial operating state. It reflects the performance drift of the submersible pump during long-term operation due to factors such as changes in formation fluid supply capacity, equipment wear, or gas intrusion. Sensing characteristics can be understood as key indicators extracted from various sensor data obtained from fiber optic pressure sensors, fiber optic temperature sensors, current sensors, and vibration sensors, such as average pressure, temperature fluctuation, effective current value, and vibration frequency. Offset direction and amplitude refer to the degree and trend of deviation of each sensing characteristic from its initial state; for example, pressure may continuously decrease, while current may gradually increase. The boundary correction factor quantifies the contribution of the offset direction and amplitude of each sensing characteristic to the correction of the decision boundary. It can be a scalar or a vector, used to indicate how the decision boundary should be adjusted for that characteristic.

[0106] The dimension-matched combined weight matrix is ​​a matrix constructed based on the number of sensing features, with its dimensions matching the number of sensing features. The elements in the matrix represent the relative importance or mutual influence of different sensing features during the boundary correction process. This combined weight matrix allows for the weighting and fusion of individual boundary correction factors, generating a comprehensive boundary translation vector and rotation matrix. The boundary translation vector is used to move the decision boundary as a whole in the multidimensional feature space to adapt to the overall drift of the working condition; the rotation matrix is ​​used to adjust the orientation or direction of the decision boundary to adapt to changes in the correlation or sensitivity between different features. The initial decision boundary is typically expressed as a hyperplane, hypersurface, or a set of inequalities in a multidimensional space, used to divide different working condition regions.

[0107] Applying the boundary translation vector and rotation matrix to this mathematical expression means performing a geometric transformation on the original decision boundary, causing it to be translated and rotated in the feature space, thereby obtaining a corrected decision boundary that can more accurately adapt to the current operating state of the submersible pump.

[0108] This application's solution establishes an independent boundary correction factor for each sensing feature and fuses them using a dimension-matched combined weight matrix to generate a boundary translation vector and a rotation matrix, achieving refined, multi-dimensional correction of the initial judgment boundary. This correction method not only considers the overall drift of the operating condition—that is, adjusting the center position of the boundary through the translation vector—but more importantly, it adjusts the geometry and orientation of the judgment boundary through the rotation matrix. This allows the judgment boundary to better adapt to the complex and dynamically changing relationships between different sensing features, thereby more accurately dividing different operating condition regions such as normal state, insufficient liquid supply state, gas intrusion state, and wear state in the multi-dimensional feature space. Therefore, this solution effectively solves the problem of insufficient adaptability of judgment boundaries in traditional methods under complex operating conditions, significantly improving the accuracy and robustness of operating condition identification.

[0109] In some preferred embodiments, it is assumed that the initial decision boundary is defined by a hyperplane in a three-dimensional feature space (e.g., composed of three sensing features: pressure, current, and vibration). After the submersible pump has been running for a period of time, trend analysis and response offset calculation reveal that the overall operating state of the submersible pump is shifting towards higher current, higher vibration, and lower pressure, and that air intrusion causes changes in the correlation between pressure and current features. At this point, the system establishes boundary correction factors for pressure, current, and vibration based on the offset direction and magnitude of each sensing feature. For example, the pressure correction factor might indicate that the boundary should move towards lower pressure, while the current and vibration correction factors might indicate that the boundary should move towards higher current and higher vibration. Subsequently, these correction factors are input into a pre-trained combined weight matrix, which considers the relative importance and mutual influence of each feature under different operating conditions. Through the fusion calculation of this matrix, a boundary translation vector is generated to move the hyperplane as a whole towards higher current, higher vibration, and lower pressure; simultaneously, a rotation matrix is ​​generated to fine-tune the tilt angle of the hyperplane, enabling it to better separate normal operating conditions from air intrusion conditions. Finally, the translation vector and rotation matrix are applied to the mathematical expression of the initial hyperplane to obtain a corrected hyperplane in the feature space after translation and rotation, which serves as the new adaptive decision boundary.

[0110] Specifically, the above-mentioned reconstruction of historical operating condition snapshots based on the envelopes of multiple sensor data includes:

[0111] The current short-term fluctuation range of the various sensor data is obtained, and the length of the sliding window is dynamically adjusted according to the current short-term fluctuation range, wherein the larger the fluctuation range, the shorter the window length.

[0112] Based on the dynamically adjusted sliding window, the sliding mean and sliding variance of the various sensor data are calculated.

[0113] When the rate of change of the moving mean is lower than a preset mean drift tolerance threshold and the range of the moving variance is lower than a preset variance fluctuation threshold, it is determined to be a dynamic steady state interval.

[0114] Within the dynamic steady-state range, local extreme points are searched, and the maximum points are connected to form the upper envelope, while the minimum points are connected to form the lower envelope.

[0115] The arithmetic midline of the upper envelope and the lower envelope is calculated as the expected value of the feature, and the average distance between the upper envelope and the lower envelope is calculated as the envelope width feature. The expected value of the feature and the envelope width feature are encapsulated into the historical operating condition snapshot.

[0116] The current short-term fluctuation amplitude refers to the range or intensity of changes in the values ​​of various sensor data within a short time window, reflecting the activity level of the data in an instant or recently. The dynamic adjustment of the sliding window length is adaptively adjusted based on the magnitude of the current short-term fluctuation amplitude. Specifically, when the fluctuation amplitude is large, the sliding window length is shortened to respond more quickly to data changes and capture transient features; conversely, when the fluctuation amplitude is small, the sliding window length is appropriately extended to smooth the data and reduce noise.

[0117] Furthermore, based on the dynamically adjusted sliding window, the moving mean and moving variance are calculated for various sensor data. The moving mean refers to the average value of the data within the sliding window, used to reflect the central trend of the data; the moving variance refers to the dispersion of the data within the sliding window, used to reflect the volatility of the data.

[0118] Specifically, when the rate of change of the moving average is lower than a preset mean drift tolerance threshold, and the range of the moving variance is lower than a preset variance fluctuation threshold, the current data can be determined to be in a dynamic steady-state range. The mean drift tolerance threshold and variance fluctuation threshold are preset parameters used to define the acceptable fluctuation range of the data in terms of mean and variance, ensuring that the determined range has sufficient stability. The dynamic steady-state range refers to a state where the mean and variance of multiple sensor data remain relatively stable within a certain period, indicating that the submersible pump's operating condition is in a relatively stable phase.

[0119] Within the dynamic steady-state range, the data envelope is constructed by searching for local extrema. Specifically, all local maxima in the data sequence are connected to form the upper envelope; all local minima are connected to form the lower envelope. The upper and lower envelopes together delineate the boundaries of data fluctuations, intuitively reflecting the range and trend of data fluctuations.

[0120] Therefore, the arithmetic median of the upper and lower envelopes is calculated as the expected value of the feature, which represents the average level or trend center of data fluctuations. Simultaneously, the average distance between the upper and lower envelopes is calculated as the envelope width feature, which quantifies the amplitude or intensity of data fluctuations. Finally, the expected value and the envelope width feature are encapsulated to form the historical operating condition snapshot, which is used for subsequent operating condition analysis and trend judgment.

[0121] This application's solution introduces steps such as dynamically adjusting the sliding window length, calculating the sliding mean and sliding variance, determining the dynamic steady-state interval, and reconstructing historical operating condition snapshots based on the envelope. This aims to more accurately and robustly identify the stable operating state of a submersible pump. Through these technical solutions, the problem of traditional methods failing to accurately identify the stable operating state of a submersible pump under complex and variable operating conditions can be effectively solved. The mechanism of dynamically adjusting the sliding window length allows the system to adaptively process sensor data with different levels of fluctuation, improving the sensitivity and accuracy of steady-state determination. The dynamic steady-state interval determination based on the sliding mean and sliding variance effectively eliminates the influence of instantaneous disturbances and noise, ensuring that the captured historical operating condition snapshots are more representative. Furthermore, reconstructing historical operating condition snapshots using the envelope allows for a more comprehensive capture of data fluctuation characteristics, including the central trend and fluctuation amplitude, thus providing a more refined and reliable data foundation for subsequent operating condition identification and trend analysis, thereby improving the accuracy and response speed of the submersible pump's closed-loop control.

[0122] In some embodiments of this application, in order to obtain long-term evolution trend information on formation fluid supply capacity, it is necessary to perform trend analysis on historical operating condition snapshot sequences. Specifically, the trend analysis includes:

[0123] The historical operating condition snapshot sequence is subjected to feature decomposition to obtain the pressure and power response features, product liquid composition features, envelope width features, and frequency and current features.

[0124] Trend components are extracted from each feature, cross-validation and correlation analysis are performed to identify the superimposed effects of different factors, and the long-term evolution trend information of the formation's fluid supply capacity is obtained after weighted fusion.

[0125] The historical operating condition snapshot sequence refers to a series of snapshot data reflecting the operating status of the submersible pump (SPMP) extracted from the historical data pool. Each historical operating condition snapshot is formed by capturing data from multiple sensors when the SPMP reaches a stable operating state. Feature decomposition can be understood as the process of data dimensionality reduction and information extraction from these historical operating condition snapshot sequences, aiming to separate key features with physical significance and trend indication from the original, complex sensor data. Specifically, through feature decomposition, pressure and power response characteristics can be obtained, reflecting the impact of decreased formation fluid supply capacity on SPMP operating parameters; production fluid composition characteristics, which may be obtained through other sensors or indirect methods, reflecting changes in the composition of downhole fluids; envelope width characteristics, reflecting the degree of gas intrusion or fluid fluctuations; and frequency and current characteristics, which are directly related to the SPMP's operating status and energy consumption.

[0126] Further, trend component extraction of each feature involves using time series analysis methods, such as moving average, exponential smoothing, and wavelet analysis, to separate the long-term trends of the features obtained from the above decomposition. This step aims to filter out short-term fluctuations and noise, focusing on the essential changes in formation fluid supply capacity over time. Subsequently, cross-validation and correlation analysis are performed to identify the cumulative effects of different factors (such as formation pressure, production rate, water cut, and gas intrusion) on submersible pump operation and formation fluid supply capacity. For example, by analyzing the relationship between pressure and power response characteristics and envelope width characteristics, it is possible to distinguish between simple insufficient fluid supply and complex conditions accompanied by gas intrusion. Finally, through weighted fusion, the feature information after trend component extraction, cross-validation, and correlation analysis is integrated to obtain the long-term evolution trend information of the formation fluid supply capacity. The weighting coefficients can be set according to the sensitivity and reliability of each feature to the decline in formation fluid supply capacity to ensure the accuracy and representativeness of the fusion results.

[0127] This application's solution comprehensively captures key indicators of submersible pump (SPP) operating status by performing multi-dimensional feature decomposition on historical operating condition snapshot sequences. By extracting trend components from these features, instantaneous noise and short-term fluctuations can be effectively filtered out, revealing the long-term, stable evolution of formation fluid supply capacity. Furthermore, through cross-validation and correlation analysis, this solution can gain a deeper understanding of the combined effects of different formation factors (such as formation pressure, produced fluid composition, and gas intrusion) on SPP operating conditions, avoiding the limitations of relying on single indicators. Finally, through weighted fusion, these refined feature information are integrated, ensuring the accuracy and reliability of the obtained long-term evolution trend information of formation fluid supply capacity, providing a solid data foundation for subsequent operating condition identification and frequency modulation command generation.

[0128] This application further proposes a method for calculating the response offset of the current submersible pump operating condition relative to its initial operating state, including:

[0129] A weighting coefficient is assigned to each sensing feature. The current operating condition feature value is compared with the initial state feature value to obtain the original offset. After weighting, the offset is mapped to the offset space through a nonlinear mapping function, and the response offset is obtained by combining the weights.

[0130] Specifically, assigning weight coefficients to each sensing feature means allocating a weight value to each sensing feature (such as pressure, temperature, current, vibration, etc.) based on its sensitivity to changes in the submersible pump's operating condition, its importance, and its indicative role in specific failure modes. These weight coefficients can be preset or dynamically adjusted based on expert experience, historical data analysis, or machine learning methods. The purpose is to highlight the contribution of key features to operating condition deviations and weaken the interference of secondary features.

[0131] The process of comparing the current operating condition feature value with the initial state feature value to obtain the original offset involves calculating the difference or ratio between the measured value of each sensor feature under the current operating condition and the corresponding feature value of the submersible pump under the initial operating state (usually referring to the baseline state of healthy and stable operation). This quantifies the degree of deviation between the current operating condition and the initial baseline. For example, the difference between the current pressure value and the initial pressure value, or the ratio between the current current value and the initial current value, can be calculated to obtain the original offset reflecting the degree of deviation of a single feature. In practical applications, weighting and mapping to the offset space through a nonlinear mapping function means first multiplying the original offset of each sensor feature obtained above with its corresponding weight coefficient to obtain the weighted offset.

[0132] Subsequently, these weighted offsets are used as input and transformed through a preset nonlinear mapping function. The nonlinear mapping function can be a polynomial function, exponential function, logarithmic function, sigmoid function, or other mathematical model capable of capturing nonlinear relationships. Its purpose is to transform the weighted offsets from the original linear or approximately linear space to an offset space that better reflects the complex changes in the actual operating conditions of the submersible pump. This mapping enhances the sensitivity to small but important changes in operating conditions and reasonably compresses or amplifies large offsets to adapt to the response characteristics of different operating stages. Therefore, the combined response offset refers to the integration of the offsets of various sensing features after nonlinear mapping. This combination can be achieved through vector superposition, feature fusion algorithms, or multidimensional coordinate representation, ultimately forming a comprehensive response offset that fully characterizes the overall deviation of the current submersible pump operating condition from its initial operating state. This response offset is a multidimensional vector or a comprehensive index, and its dimension and form depend on the specific combination strategy.

[0133] This application's solution effectively addresses the shortcomings in accuracy and sensitivity that traditional methods may suffer from when calculating response offsets by introducing weighting coefficients and nonlinear mapping functions. Through this technical solution, the application significantly improves the accuracy and sensitivity of submersible pump (SPP) response offset calculations. Specifically, by introducing weighting coefficients, the system can differentiate the processing based on the actual contribution of each sensing feature to changes in operating conditions, preventing key feature information from being overwhelmed and thus improving the early identification capability of potential operating condition anomalies. Furthermore, the application of the nonlinear mapping function allows the response offset to more realistically reflect the complex nonlinear changes in SPP operating conditions, especially in the critical stage of transitioning from normal to abnormal conditions, providing a more refined and accurate offset representation. This precise and sensitive response offset calculation provides a more solid foundation for the subsequent dynamic correction of adaptive judgment boundaries, thereby improving the accuracy and robustness of SPP operating condition identification, effectively reducing the risk of misjudgment and missed judgment, and ensuring the safe and efficient operation of the SPP.

[0134] This application further proposes that the weighted average is mapped to the offset space through a nonlinear mapping function, and the resulting combination yields the response offset, including:

[0135] Obtain the stage identifier of the current formation fluid supply capacity decline, select the corresponding mapping function from the preset nonlinear mapping function set according to the stage identifier, input the weighted offset into the mapping function, and obtain the mapped offset.

[0136] Specifically, identifying the stage of decline in current formation fluid supply capacity involves analyzing multi-source downhole data, such as formation pressure, production, and water cut, to comprehensively determine the current production stage of the well, such as initial decline, mid-term stable decline, or late-stage exhaustion. This stage indicator reflects the overall evolution trend and current state of formation fluid supply capacity.

[0137] The preset set of nonlinear mapping functions is a library containing multiple different nonlinear mapping functions, each pre-designed and calibrated to adapt to different stages of formation fluid supply capacity decline. For example, for the initial decline stage, a mapping function more sensitive to small offsets might be selected; for the later exhaustion stage, a mapping function more robust to large offsets might be selected. These functions can be polynomial functions, exponential functions, logarithmic functions, sigmoid functions, or other nonlinear models. Selecting the corresponding mapping function from the preset set of nonlinear mapping functions based on the stage identifier means that the system dynamically selects the nonlinear mapping function that best matches the characteristics of the current stage based on the currently identified formation fluid supply capacity decline stage. Therefore, inputting the weighted offset into the mapping function to obtain the mapped offset means taking the original offsets of each sensing feature after weighting coefficient processing as input, and transforming them through the selected nonlinear mapping function to obtain a more accurate mapped offset that better reflects the actual current operating conditions.

[0138] This application's solution introduces a stage indicator for the decline in formation fluid supply capacity, enabling dynamic selection of the nonlinear mapping function. During the long-term operation of a submersible pump (SPP), the formation fluid supply capacity gradually decreases, with the rate and manifestation of this decline differing significantly at different stages. Through the aforementioned technical solution, this application can dynamically adjust the nonlinear mapping function according to the different stages of formation fluid supply capacity decline, thereby significantly improving the accuracy and adaptability of the response offset calculation. Compared to a solution using a fixed mapping function, this application can more precisely capture the changing characteristics of the SPP's operating conditions at different production stages, avoiding error accumulation caused by mapping function mismatch. Consequently, the obtained response offset can more accurately reflect the deviation between the actual operating state and the initial state of the SPP, providing a more reliable basis for subsequent adaptive boundary correction, thus improving the overall accuracy and robustness of the SPP closed-loop control system, effectively extending the SPP's service life and optimizing oil production efficiency.

[0139] In some embodiments described above in this application, in order to calculate the response offset of the current submersible pump operating condition relative to its initial operating state, it is necessary to obtain the stage identifier of the current formation fluid supply capacity decline. To this end, this application further proposes a specific method for obtaining the stage identifier of the current formation fluid supply capacity decline.

[0140] The step of identifying the stage where the current formation fluid supply capacity is declining includes:

[0141] Acquiring multi-source downhole data includes formation pressure data, production data, and water cut data. Specifically, multi-source downhole data refers to various types of data collected from the downhole environment of the oil well, which can directly or indirectly reflect the supply status of formation fluids. Formation pressure data reflects changes in formation energy, production data indicates the well's production capacity, and water cut data reflects the extent of formation water intrusion. This data is typically obtained through downhole sensors or periodic testing.

[0142] The downhole multi-source data undergoes quality checks to identify and process missing, outlier, and delayed data points. Quality checks aim to ensure data accuracy and reliability, as raw data acquisition may contain various noises and errors. Missing data points refer to data that was not acquired at a specific time point and can be filled using interpolation, prediction, or other methods. Outlier data points are those that significantly deviate from the normal range, possibly caused by sensor malfunction or transient disturbances; they can be identified and smoothed or removed using statistical analysis, machine learning, or other methods. Delayed data points indicate data acquisition or transmission with time lags, requiring time calibration to ensure data synchronization.

[0143] Based on the processed data, the formation pressure decline rate, cumulative production decline, and water cut trend were calculated. Specifically, the formation pressure decline rate refers to how quickly the formation pressure changes over time, reflecting the rate of formation energy consumption. The cumulative production decline refers to the reduction in the total production of the oil well relative to a baseline value within a certain period, directly reflecting the decline in formation fluid supply capacity. The water cut trend refers to the direction and magnitude of water cut changes over time, indicating the advancement of the water drive front or the severity of formation water invasion. The calculation of these indicators helps to quantify the changing characteristics of formation fluid supply capacity.

[0144] By comprehensively assessing the formation pressure decline rate, the cumulative fluid production decline rate, and the water cut change trend, a stage identifier for the current formation fluid supply capacity decline is obtained. This comprehensive assessment involves integrating and analyzing multiple calculated indicators to fully evaluate the actual state of the formation fluid supply capacity. The stage identifier is a classification or grading of the degree of formation fluid supply capacity decline, such as initial decline, intermediate decline, and late decline. Through this comprehensive assessment, the current stage of formation fluid supply capacity decline in the oil well can be more accurately identified.

[0145] This application's solution, by acquiring and processing multi-source downhole data, can comprehensively and accurately reflect the true state of formation fluid supply capacity. Through the aforementioned technical solution, this application can utilize multi-source, high-quality downhole data to comprehensively and accurately assess the actual situation of formation fluid supply capacity decline. Compared to relying solely on limited or unprocessed data for judgment, this solution, through refined data processing and comprehensive analysis of multiple indicators, significantly improves the accuracy and reliability of identifying the stages of formation fluid supply capacity decline. This precise stage identification allows for the selection of the most suitable mapping function from a pre-defined set of nonlinear mapping functions based on the different stages of formation fluid supply capacity decline when calculating the response offset. This more accurately maps the weighted offset to the offset space, ultimately improving the adaptability and control accuracy of the submersible pump closed-loop control system to formation changes.

[0146] In some embodiments described above, this application proposes to obtain the stage indicator of the current decline in formation fluid supply capacity by comprehensively judging the rate of formation pressure decline, the magnitude of cumulative fluid production decline, and the trend of water cut changes. However, in actual oil and gas field development, the well types, formation conditions, and development stages of different wells vary greatly. If a single or fixed judgment logic is used, the identification of the stage of decline in formation fluid supply capacity may not be accurate enough, thereby affecting the accuracy of subsequent submersible pump frequency adjustment commands.

[0147] In this regard, this application further proposes the following comprehensive judgments:

[0148] Obtain the well type and formation condition information of the current well, and match the stage segmentation parameter set from the well type and formation feature library;

[0149] The relative importance of dynamically assessing the formation pressure decline rate, the cumulative decrease in fluid production, and the trend of water cut change;

[0150] The relative importance is weighted and combined, matched with the stage division parameter set, and combined with development stage information to determine the stage identifier.

[0151] Specifically, obtaining the current well type and formation condition information refers to the system automatically or manually inputting the specific type of the currently operating submersible pump (SMP) well (e.g., vertical well, horizontal well, branch well, etc.) and the geological characteristics of its formation (e.g., reservoir lithology, permeability, porosity, fluid properties, etc.). This information is used to retrieve and match the most suitable set of stage division parameters for the current well condition from a pre-established well type and formation feature database. This parameter set may include critical thresholds, judgment rules, or weight configurations for different stages, with the aim of providing a customized benchmark for subsequent comprehensive judgment.

[0152] The dynamic assessment of the relative importance of the formation pressure decline rate, the cumulative production decline, and the water cut change trend can be understood as intelligently adjusting the weight of these three key indicators in the comprehensive judgment based on real-time monitoring data and historical experience. For example, in the early stages of well development, the formation pressure decline rate may be a more sensitive indicator; while in the later stages, the water cut change trend may better reflect the depletion of formation fluid supply capacity. The purpose of this dynamic assessment is to enable the judgment model to adapt to different stages of the well's life cycle and the complexity of formation response.

[0153] In practical applications, a weighted combination based on the relative importance is performed and matched with the stage division parameter set, combined with development stage information, to determine the stage identifier. This means that after dynamically assessing the relative importance of each indicator, the values ​​of these indicators are assigned corresponding weights and then weighted and summed or combined using other multi-indicator fusion algorithms. The combined result is compared with the stage division parameter set matched from the well type and formation feature database, for example, to determine whether it falls within a predefined stage interval. Simultaneously, the judgment result is corrected or confirmed by combining the current development stage information of the oil well (e.g., early, middle, late stage), ultimately accurately determining the stage identifier of the current formation fluid supply capacity decline. The purpose is to ensure the accuracy and reliability of the stage identifier, providing a solid foundation for subsequent submersible pump control.

[0154] This application's solution incorporates well type and formation condition information, and matches a customized set of stage division parameters from a feature library. This transforms the comprehensive judgment from a general approach to a personalized judgment tailored to specific oil wells, thus avoiding misjudgments caused by differences in geological conditions. Through this technical solution, this application overcomes the judgment biases caused by traditional methods in identifying the decline stage of formation fluid supply capacity, which fails to fully consider the dynamic importance of well type, formation conditions, and indicators. Specifically, by introducing a customized set of stage division parameters and a mechanism for dynamically evaluating the relative importance of indicators, the identification of the decline stage of formation fluid supply capacity becomes more accurate and adaptive. This provides more reliable operating condition judgments for the closed-loop control of submersible pumps, effectively avoiding low submersible pump operating efficiency, equipment damage, or oil well production losses due to misjudgments, significantly improving the intelligent management level and economic benefits of oil well production.

[0155] This application further proposes the relative importance of the formation pressure decline rate, the cumulative production decline rate, and the water cut change trend in the above-mentioned dynamic assessment, including:

[0156] Preprocessing is performed on the formation pressure data, the fluid production data, and the water cut data;

[0157] Calculate the short-term volatility and long-term trend stability of each data point under different time windows;

[0158] The confidence levels of each indicator are adjusted based on the short-term volatility and the long-term trend stability.

[0159] By combining the correlation between various indicators and the stages of decline in formation fluid supply capacity in historical data, a preliminary importance level is calculated;

[0160] The degree of disturbance to the downhole environment is assessed by real-time monitoring of the vibration intensity and motor temperature of the submersible pump.

[0161] The initial importance is corrected based on the degree of disturbance to obtain the relative importance.

[0162] Specifically, preprocessing of formation pressure, production, and water cut data aims to eliminate noise, outliers, and missing values ​​in the raw data, ensuring the accuracy of subsequent analysis. Preprocessing methods can include, but are not limited to, filtering, interpolation, and smoothing techniques. Calculating the short-term fluctuation amplitude and long-term trend stability of each data point at different time windows involves analyzing the preprocessed formation pressure, production, and water cut data at different time scales. Short-term fluctuation amplitude reflects instantaneous changes in the data over a short period, such as daily or weekly fluctuations; long-term trend stability reflects the overall trend of the data over a longer time scale, such as monthly or quarterly trends. These calculations can be achieved using statistical methods (such as standard deviation, moving average, and regression analysis).

[0163] In practical applications, adjusting the confidence levels of various indicators based on short-term fluctuations and long-term trend stability means that when data fluctuations are large or trends are unstable, the confidence level will be lowered accordingly, and vice versa. Confidence level adjustments can be performed using fuzzy logic, expert systems, or machine learning-based models. Furthermore, by combining the correlation between each indicator and the stage of formation fluid production decline in historical data, a preliminary importance is calculated. This involves using historical downhole data to analyze the performance characteristics of the three indicators—formation pressure decline rate, cumulative fluid production decline, and water cut change trend—at different stages of formation fluid production decline (e.g., early, middle, and late stages) and their correlation strength with stage determination. This correlation analysis can employ methods such as statistical correlation coefficients, decision trees, or neural networks.

[0164] Furthermore, by monitoring the vibration intensity and motor temperature of the submersible pump in real time, the degree of disturbance in the downhole environment is assessed, aiming to obtain real-time information on the downhole operating environment. The vibration intensity and motor temperature of the submersible pump are key indicators reflecting the stability of downhole operating conditions and the presence of abnormal disturbances (such as gas lock, pump wear, scaling, etc.). When the vibration intensity or motor temperature rises abnormally, it indicates that there may be significant disturbances in the downhole environment, which may affect the accuracy or representativeness of formation pressure, production rate, and water cut data. Therefore, the relative importance is obtained by correcting the initial importance based on the degree of disturbance. This means using the real-time monitored degree of downhole environmental disturbance as a correction factor to adjust the initial importance calculated based on historical data. For example, when the degree of downhole disturbance is high, the importance weight of certain disturbance-sensitive indicators (such as the formation pressure drop rate) can be appropriately reduced to avoid misjudgments caused by instantaneous disturbances.

[0165] The proposed solution employs refined preprocessing and multi-scale analysis of formation pressure, production, and water cut data to more accurately capture the true changes in the data and assess its reliability. Through this technical solution, the proposed solution enables more precise and robust dynamic assessment of the decline in formation fluid supply capacity.

[0166] refer to Figure 2 , Figure 2 This is a schematic diagram of a closed-loop control system for a submersible pump based on multi-sensor fusion, provided by an embodiment of the present invention, used to execute the above method, characterized in that it includes:

[0167] The downhole sensor assembly, deployed at the suction inlet and motor end of the submersible pump, includes:

[0168] Fiber optic pressure sensors and fiber optic temperature sensors are deployed at the inlet of the submersible pump.

[0169] A current sensor and a vibration sensor are deployed at the motor end of the submersible pump;

[0170] A synchronous clock bus is connected to the fiber optic pressure sensor, the fiber optic temperature sensor, the current sensor, and the vibration sensor respectively, and is used to perform time synchronization marking for each sensor;

[0171] Surface control components, deployed at the wellhead surface, include:

[0172] The fiber optic sensor signal demodulator is connected to the fiber optic pressure sensor and the fiber optic temperature sensor of the downhole sensor assembly via optical fiber.

[0173] The controller is connected to the fiber optic sensor demodulator, the current sensor, the vibration sensor, and the synchronous clock bus. It includes a non-volatile memory and a processor. The non-volatile memory stores a computer program and a historical data pool. When the processor executes the computer program, it implements the steps of the above method.

[0174] A frequency converter, connected to the power supply cable of the controller and the submersible pump, is used to receive frequency adjustment commands generated by the controller and adjust the speed of the submersible pump;

[0175] An alarm, connected to the controller, is used to issue an alarm signal in fault protection mode.

[0176] The multi-sensor fusion-based closed-loop control system for submersible pumps proposed in this application constructs a complete architecture capable of real-time monitoring of the submersible pump's operating status and adaptively adjusting control strategies by deploying collaborative components both downhole and on the surface. The system acquires multi-source sensor data through downhole sensor components and utilizes a synchronous clock bus to ensure data time consistency. The fiber optic sensor signal demodulator in the surface control component processes the fiber optic sensor data, while the controller, as the core processing unit, executes pre-stored computer programs to achieve intelligent identification and adaptive boundary correction of the submersible pump's operating conditions. The frequency converter precisely adjusts the submersible pump speed according to the frequency modulation commands generated by the controller, and the alarm provides timely warnings when the system detects an anomaly. Therefore, this system effectively solves the misjudgment problem caused by the solidification of operating condition boundaries in traditional control methods during long-term operation, ensuring the continuous, efficient, and stable operation of the submersible pump in complex and ever-changing downhole environments.

[0177] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A closed-loop control method for a submersible pump based on multi-sensor fusion, characterized in that, include: A fiber optic pressure sensor and a fiber optic temperature sensor are deployed at the suction inlet of the submersible pump, and a current sensor and a vibration sensor are deployed at the motor end of the submersible pump. The time synchronization of each sensor is marked based on a synchronous clock bus to obtain multiple sensor data with time synchronization. The submersible pump is driven to operate according to the frequency adjustment command sent by the ground frequency converter. When the submersible pump reaches a stable operating state, various sensor data are captured to form a historical operating condition snapshot, and the historical operating condition snapshot is stored in the historical data pool, which is deployed in a non-volatile memory. The stable operating state is determined in the following way: When the fluctuation level of the various sensor data is lower than a preset first threshold, it is determined to be in a first stable state; or... When the mean change rate of the multiple sensor data is lower than a preset second threshold and the fluctuation range is lower than a preset third threshold, it is determined to be a second stable state, and the historical working condition snapshot is reconstructed based on the envelope of the multiple sensor data. Obtain a preset initial judgment boundary, which is pre-calibrated based on the submersible pump factory test data and prior historical data of wells of the same model; Historical operating condition snapshot sequences are extracted from the historical data pool. Pressure and power response characteristics reflecting the decline in formation fluid supply capacity and envelope width characteristics reflecting the degree of gas intrusion are extracted and trend analysis is performed to obtain long-term evolution trend information of formation fluid supply capacity. Based on the trend information, calculate the response offset of the current submersible pump operating condition relative to its initial operating state; Based on the response offset, the preset initial decision boundary is corrected to obtain an adaptive decision boundary; Based on the adaptive judgment boundary, the current operating condition of the submersible pump is identified, and the operating condition includes at least one of the following: normal state, insufficient fluid supply state, gas intrusion state, and wear state. Based on the identification results, a frequency adjustment command is generated and sent to the ground frequency converter to adjust the speed of the submersible pump; When any of the multiple sensor data is continuously missing or exceeds the preset range, the system switches to fault protection mode. The fault protection mode is executed in the following manner: the current inverter output frequency is locked and an alarm signal is generated.

2. The method according to claim 1, characterized in that, The step of correcting the initial determination boundary based on the response offset includes: The offset direction and magnitude of each sensing feature are obtained based on the response offset, and a boundary correction factor is established for each sensing feature. The boundary correction factors of each of the sensing features are constructed according to the number of sensing features to form a dimension-matched combined weight matrix. The combined weight matrix is ​​used to fuse the correction factors to generate a boundary translation vector and a rotation matrix. The boundary translation vector and the rotation matrix are applied to the mathematical expression of the initial decision boundary to obtain the corrected decision boundary.

3. The method according to claim 1, characterized in that, The reconstructing of the historical operating condition snapshot based on the envelope of the multiple sensor data includes: The current short-term fluctuation range of the various sensor data is obtained, and the length of the sliding window is dynamically adjusted according to the current short-term fluctuation range, wherein the larger the fluctuation range, the shorter the window length. Based on the dynamically adjusted sliding window, the sliding mean and sliding variance of the various sensor data are calculated. When the rate of change of the moving mean is lower than a preset mean drift tolerance threshold and the range of the moving variance is lower than a preset variance fluctuation threshold, it is determined to be a dynamic steady state interval. Within the dynamic steady-state range, local extreme points are searched, and the maximum points are connected to form the upper envelope, while the minimum points are connected to form the lower envelope. The arithmetic midline of the upper envelope and the lower envelope is calculated as the expected value of the feature, and the average distance between the upper envelope and the lower envelope is calculated as the envelope width feature. The expected value of the feature and the envelope width feature are encapsulated into the historical operating condition snapshot.

4. The method according to claim 1, characterized in that, The trend analysis includes: The historical operating condition snapshot sequence is subjected to feature decomposition to obtain the pressure and power response features, product liquid composition features, envelope width features, and frequency and current features. Trend components are extracted from each feature, cross-validation and correlation analysis are performed to identify the superimposed effects of different factors, and the long-term evolution trend information of the formation's fluid supply capacity is obtained after weighted fusion.

5. The method according to claim 1, characterized in that, The calculation of the response offset of the current submersible pump operating condition relative to its initial operating state includes: A weighting coefficient is assigned to each sensing feature. The current operating condition feature value is compared with the initial state feature value to obtain the original offset. After weighting, the offset is mapped to the offset space through a nonlinear mapping function, and the response offset is obtained by combining the weights.

6. The method according to claim 5, characterized in that, The weighted values ​​are mapped to the offset space through a nonlinear mapping function, and the resulting combination yields the response offset, including: Obtain the stage identifier of the current formation fluid supply capacity decline, select the corresponding mapping function from the preset nonlinear mapping function set according to the stage identifier, input the weighted offset into the mapping function, and obtain the mapped offset.

7. The method according to claim 6, characterized in that, The step of identifying the stage where the current formation fluid supply capacity is declining includes: Acquire downhole multi-source data, including formation pressure data, fluid production data, and water cut data; The downhole multi-source data is subjected to quality checks to identify and process missing data points, abnormal data points, and delayed data points. Based on the processed data, the formation pressure decrease rate, the cumulative decrease in fluid production, and the water cut change trend were calculated. By comprehensively judging the formation pressure decline rate, the cumulative fluid production decline rate, and the water cut change trend, the stage indicator of the current formation fluid supply capacity decline can be obtained.

8. The method according to claim 7, characterized in that, The comprehensive judgment includes: Obtain the well type and formation condition information of the current well, and match the stage segmentation parameter set from the well type and formation feature library; The relative importance of dynamically assessing the formation pressure decline rate, the cumulative decrease in fluid production, and the trend of water cut change; The relative importance is weighted and combined, matched with the stage division parameter set, and combined with development stage information to determine the stage identifier.

9. The method according to claim 8, characterized in that, The dynamic assessment of the relative importance of the formation pressure decline rate, the cumulative production decline, and the water cut change trend includes: Preprocessing is performed on the formation pressure data, the fluid production data, and the water cut data; Calculate the short-term volatility and long-term trend stability of each data point under different time windows; The confidence levels of each indicator are adjusted based on the short-term volatility and the long-term trend stability. By combining the correlation between various indicators and the stages of formation fluid supply decline in historical data, a preliminary importance level is calculated; The degree of disturbance to the downhole environment is assessed by real-time monitoring of the vibration intensity and motor temperature of the submersible pump. The initial importance is corrected based on the degree of disturbance to obtain the relative importance.

10. A closed-loop control system for a submersible pump based on multi-sensor fusion, used to execute the method according to any one of claims 1-9, characterized in that, include: The downhole sensor assembly, deployed at the suction inlet and motor end of the submersible pump, includes: Fiber optic pressure sensors and fiber optic temperature sensors are deployed at the inlet of the submersible pump. A current sensor and a vibration sensor are deployed at the motor end of the submersible pump; A synchronous clock bus is connected to the fiber optic pressure sensor, the fiber optic temperature sensor, the current sensor, and the vibration sensor respectively, and is used to perform time synchronization marking for each sensor; Surface control components, deployed at the wellhead surface, include: The fiber optic sensor signal demodulator is connected to the fiber optic pressure sensor and fiber optic temperature sensor of the downhole sensor assembly via optical fiber. The controller is connected to the fiber optic sensor demodulator, the current sensor, the vibration sensor, and the synchronous clock bus. It includes a non-volatile memory and a processor. The non-volatile memory stores a computer program and a historical data pool. When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-9. A frequency converter, connected to the power supply cable of the controller and the submersible pump, is used to receive frequency adjustment commands generated by the controller and adjust the speed of the submersible pump; An alarm, connected to the controller, is used to issue an alarm signal in fault protection mode.

Images