An electric submersible screw pump downhole anti-evacuation method and control system

Abstract

This application discloses a downhole anti-cavitation method and control system for an electric submersible screw pump, comprising the following steps: real-time acquisition of pressure parameters at the pump inlet; calculation of the current pressure value p and pressure change rate v based on the pressure parameters; construction of a two-dimensional state space based on preset pressure and rate thresholds, dividing the two-dimensional state space into multiple control regions, and determining the current control region based on the current pressure value p and pressure change rate v; execution of corresponding control actions based on the determined control region, including maintaining operation, frequency reduction adjustment, pulse frequency increase, and protective shutdown; after performing frequency reduction adjustment or pulse frequency increase, an observation period is entered, pressure parameters are re-acquired, and the control effect is evaluated; based on the evaluation results, a decision is made to maintain the current state or execute the next control action. The use of flexible frequency conversion adjustment and stepped frequency reduction enables the unit to operate continuously during low-production periods, avoiding the impact of frequent start-stop cycles on the motor and pump body.

Description

Technical Field

[0001] This application belongs to the field of petroleum extraction technology, specifically relating to a method and control system for preventing air cavitation in downhole electric submersible screw pumps. Background Technology

[0002] Electric submersible screw pumps (ESPs) are widely used in the production of heavy oil, sandy, and gas-bearing wells as a highly efficient artificial lift method. The downhole motor drives the rotor to perform planetary motion within the stator (rubber bushing), lifting the fluid to the surface. However, when the wellbore's fluid supply capacity is insufficient, fluid interruption at the pump inlet causes dry friction between the rotor and stator. This can lead to serious malfunctions such as stator rubber burnout and rotor wear within a short period, necessitating pump maintenance, significantly increasing operating costs, and impacting production uptime.

[0003] Currently, existing anti-cavitation technologies mainly employ intermittent pumping based on submersion control. This method involves installing a pressure sensor at the pump inlet to convert the measured pressure into submersion (i.e., the height of the liquid column above the pump inlet), and pre-setting upper and lower limits for submersion. During production, when the submersion drops to the lower limit, the control system triggers an emergency shutdown to prevent dry running caused by cavitation; after the submersion naturally recovers to the upper limit, the system automatically restarts and resumes production. This approach, through intermittent pumping, periodically matches the pump's operation with the formation's fluid supply capacity.

[0004] However, the intermittent pumping method based on the upper and lower limits of submersion intensity has its protection action occurring when or after the submersion intensity has dropped to the lower limit. Before shutdown, the equipment is already in a critical state of evacuation, which can easily cause some damage to the stator rubber. It relies solely on the upper and lower limits of submersion intensity for two direct states: running and stopped. Frequent starts and stops generate large current surges and mechanical stresses, affecting the service life of the motor, frequency converter, and screw pump unit.

[0005] Manually set submersion limits are not easily adapted to dynamic operating conditions such as formation pressure decay and changes in fluid properties. As formation energy decreases, the original submersion limits may become inapplicable, leading to frequent false shutdowns or insufficient protection, requiring frequent manual adjustment of thresholds and increasing the workload of on-site operation and maintenance.

[0006] Intermittent pumping leads to production interruptions. After each shutdown, it is necessary to wait for the fluid level to recover before restarting, which reduces production efficiency and oil well recovery rate. For low-permeability reservoirs or wells with poor fluid supply capacity, shutdowns and waiting times can last for hours or even days, affecting production output.

[0007] The system is unable to distinguish the specific types of pre-evacuation precursors (such as liquid supply failure, gas lock, sand sticking), and can only adopt a unified shutdown protection strategy. For the precursor of gas lock, frequency reduction or shutdown may instead exacerbate the gas lock; for the precursor of sand sticking, if the shutdown is not timely, it may cause serious equipment damage.

[0008] Therefore, there is an urgent need for a solution to prevent evacuation that can be flexibly regulated. Summary of the Invention

[0009] This application provides a method and control system for preventing evacuation of downhole electric submersible screw pumps, which solves the problems of shortened unit life and low production efficiency caused by passive shutdown and fixed thresholds in the prior art.

[0010] The technical solution adopted in this application is as follows: A method for preventing evacuation of downhole electric submersible screw pumps includes the following steps: Step S1: Real-time collect the pressure parameters at the suction port of the electric submersible screw pump; Step S2: Calculate the current pressure value p and the pressure change rate v according to the pressure parameters, where the pressure change rate v is obtained by calculating the pressure change slope within a predetermined time window through a linear regression algorithm and performing filtering processing; Step S3: Construct a two-dimensional state space according to the preset pressure threshold and rate threshold, divide the two-dimensional state space into multiple control regions, and determine the current control region according to the current pressure value p and pressure change rate v; Step S4: Execute the corresponding control actions according to the determined control region, and the control actions include maintaining operation, frequency reduction adjustment, pulse frequency increase, and protective shutdown; Step S5: After executing the frequency reduction adjustment or pulse frequency increase, enter an observation period, re-collect the pressure parameters and evaluate the control effect, and decide whether to maintain the current state or execute the next control action according to the evaluation result.

[0011] Further, in the step S3, the pressure threshold includes the lower limit p1 of the safety zone, the median line p2 of the regulation zone, and the shutdown protection threshold p3, where p1>p2>p3; the rate threshold includes the rapid decline threshold v1, the medium-speed decline threshold v2, and the slow decline threshold v3, where v1<v2<v3<0; the division of the control region includes: Safety zone: p>p1 or v>0; Shallow regulation zone: p2<p≤p1 and v2<v≤v3; Deep regulation zone: p3<p≤2 or v1<v≤v2; Danger zone: p≤3 or v≤v1.

[0012] Furthermore, the corresponding control actions performed in step S4 specifically include: maintaining the current frequency operation when in the safe zone; performing a slight or medium frequency reduction when in the shallow control zone; performing a stepped frequency reduction when in the deep control zone; and performing an emergency shutdown when in the danger zone.

[0013] Furthermore, the stepped frequency reduction includes: limiting the single frequency reduction amplitude to a preset percentage of the rated frequency, waiting for a first preset observation period after frequency reduction, monitoring the pressure response during the observation period, stopping frequency reduction if the pressure rises, and continuing to perform the next step of stepped frequency reduction if the pressure continues to fall but does not enter the danger zone, until the preset minimum operating frequency is reached.

[0014] Furthermore, the pressure threshold and rate threshold are dynamically updated using a self-learning method, including: in the offline training phase, collecting historical normal production data of oil wells to establish a pressure distribution model under normal operating conditions; and in the online update phase, updating each threshold based on the latest data using an exponentially weighted moving average algorithm at regular intervals or after each control event.

[0015] Furthermore, in the offline training phase, a variational autoencoder is used to establish a pressure distribution model under normal operating conditions, with a preset quantile of the pressure distribution as the initial pressure threshold; the rate threshold is determined based on the pressure change rate distribution within a predetermined time period before historical evacuation events.

[0016] Furthermore, it also includes a pump condition identification step: collecting pressure, temperature, vibration, and current parameters to extract multi-dimensional features, inputting the multi-dimensional features into a pre-trained classifier to identify the current operating condition type, which includes fluid supply failure, airlock precursors, and sand jam precursors; the pump condition identification result obtained from the pump condition identification step includes: when fluid supply failure is identified, a frequency reduction control strategy is executed; when airlock precursors are identified, a pulse frequency increase control strategy is executed; when sand jam precursors are identified, an emergency shutdown is executed; when the pump condition identification result is inconsistent with the control action executed according to the determined control area, the control action is executed according to the pump condition identification result.

[0017] Furthermore, the pulse frequency boosting control strategy includes boosting the operating frequency to a preset multiple of the rated frequency, and then restoring the original frequency after a second preset time.

[0018] This application also includes, in another aspect, an intelligent control system for preventing cavitation in electric submersible screw pumps, comprising: The downhole sensing layer consists of a sensor array installed at the inlet of the electric submersible screw pump, including pressure sensors, temperature sensors, and vibration sensors, used to collect downhole operating parameters in real time. The wellhead control layer includes an edge controller and a frequency converter. The edge controller has a built-in two-dimensional state-space decision model and a self-learning module for executing the method described in any of the above-mentioned cases. The frequency converter receives control commands to flexibly adjust the speed of the electric submersible screw pump. The cloud-based intelligent layer connects to the wellhead control layer via a communication network. It is used to store historical data, train and update self-learning models, and send the updated model parameters to the edge controller.

[0019] Furthermore, the edge controller also has a built-in pump condition recognition model, which is used to identify liquid supply failure, airlock precursors, and sand jam precursors based on multi-dimensional features of pressure, temperature, vibration, and current.

[0020] Due to the adoption of the above technical solution, the beneficial effects achieved by this application are as follows: By introducing the pressure change rate, abnormal trends can be identified before cavitation occurs, allowing for early intervention and control to prevent dry running. Flexible variable frequency control replaces traditional direct shutdown, and a stepped frequency reduction strategy enables continuous operation of the unit even during low-production periods, avoiding the impact of frequent start-stop cycles on the motor and pump. Intelligent pump condition identification allows for pulsed frequency boosting for airlock exhaust, avoiding the airlock problems associated with traditional frequency reduction solutions and effectively improving production efficiency under complex operating conditions. Attached Figure Description

[0021] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of an anti-cavitation method in one embodiment of the present invention. Detailed Implementation

[0022] To more clearly illustrate the overall concept of this application, a detailed explanation is provided below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description to provide a thorough understanding of this application. However, this application may also be implemented in other ways different from those described herein. Therefore, the scope of protection of this application is not limited to the specific embodiments disclosed below. It should be noted that, unless otherwise specified, the embodiments of this application and the features thereof can be combined with each other.

[0024] Furthermore, it should be understood in the description of this application that the terms "top", "bottom", "inner", "outer", "axial", "radial", "circumferential", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0025] In this application, unless otherwise expressly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection, an electrical connection, or a communication connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0026] In this application, unless otherwise expressly specified and limited, the "above" or "below" of the second feature can mean that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. In the description of this specification, references to terms such as "an embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described can be combined in any suitable manner in one or more embodiments or examples.

[0027] This application provides a downhole anti-cavitation method and control system for electric submersible screw pumps. Compared with existing intermittent pumping methods based on submersion limits, which require waiting for the fluid level to recover to the upper limit after each shutdown before restarting, the production interruption time depends on the recovery speed of the formation fluid supply capacity. For wells with poor fluid supply capacity, this may result in prolonged production losses. This invention uses variable frequency continuous control and frequency reduction adjustment to match the pump's displacement with the formation fluid supply capacity in real time, eliminating the need for production interruption and improving production efficiency while protecting equipment. The following embodiments illustrate the invention in detail: Example 1: This application proposes a downhole sensing layer, including a sensor array installed at the inlet of the electric submersible screw pump, specifically comprising a pressure sensor, a temperature sensor, and a vibration sensor. The pressure sensor is used to acquire the pump inlet pressure in real time, with an accuracy requirement of 0.1% FS and a sampling frequency of no less than 1Hz. The temperature sensor monitors the pump body temperature to assist in identifying abnormal overheating. The vibration sensor monitors pump body vibration to assist in identifying mechanical faults such as sand jamming. The sensors transmit signals to the surface control cabinet via armored cables.

[0028] The wellhead control layer comprises an edge controller and a frequency converter. The edge controller employs a programmable logic controller (PLC) with floating-point arithmetic capabilities or an embedded edge computing gateway. It incorporates a two-dimensional state-space decision model, a self-learning module, and a pump condition identification model, with a cycle time of no more than 100 milliseconds, ensuring real-time control response. The frequency converter connects to the edge controller via an industrial communication protocol, receiving frequency commands and flexibly adjusting the speed of the electric submersible screw pump to achieve stepless speed regulation.

[0029] Understandably, the cycle time of the edge controller can be set according to well conditions and control requirements. For conventional oil wells with slow fluid supply changes, the cycle time can be set to 100-500 milliseconds; for oil wells with rapid fluid supply changes or a risk of gas lock-in, it can be set to 50-100 milliseconds to achieve a faster control response.

[0030] The cloud-based intelligent layer, deployed on the industrial internet platform, communicates with the wellhead control layer via wireless or wired networks. It is used to store historical production data, train and update various self-learning models, and send the optimized model parameters to the edge controller to achieve continuous iterative optimization of the model.

[0031] The control flow of this embodiment includes the following steps: Step S1: Real-time acquisition of pressure parameters at the inlet of the electric submersible screw pump. The edge controller reads the values ​​from the pressure sensor at a sampling frequency of 1Hz to obtain real-time pressure data. To ensure data quality, the controller filters the raw data and removes outliers.

[0032] Step S2: Calculate the pressure value p and pressure change rate v at the current moment based on the pressure parameters. The pressure change rate v is obtained by calculating the slope of pressure change within a predetermined time window using a linear regression algorithm and then filtering it.

[0033] The edge controller calculates the current pressure value *p* and the pressure change rate *v* based on the acquired pressure sequence. The pressure change rate *v* is calculated using a linear regression algorithm to determine the slope of pressure change within a predetermined time window, and then filtered to obtain a stable rate value. Both the linear regression algorithm and the filtering process described above can be implemented using existing technologies and will not be elaborated upon here.

[0034] Step S3: Construct a two-dimensional state space based on the preset pressure threshold and rate threshold, divide the two-dimensional state space into multiple control regions, and determine the current control region based on the current pressure value p and pressure change rate v.

[0035] The edge controller maps the current state (p, v) to a two-dimensional state space based on preset pressure and rate thresholds to determine its control region. The thresholds include pressure thresholds (lower limit of the safety zone p1, median line of the control zone p2, shutdown protection threshold p3) and rate thresholds (rapid descent threshold v1, medium-speed descent threshold v2, slow descent threshold v3). For example, the initial thresholds are set as follows: p1 = 1.2 MPa, p2 = 0.8 MPa, p3 = 0.4 MPa, v1 = -0.30 MPa / min, v2 = -0.15 MPa / min, v3 = -0.05 MPa / min. See Example 2 for the region division rules.

[0036] Step S4: Execute the corresponding control actions according to the determined control area. The control actions include maintaining operation, frequency reduction adjustment, pulse frequency increase, and protective shutdown.

[0037] Based on the region determined in step S3, the edge controller sends corresponding control commands to the frequency converter: When in a safe zone, maintain the current frequency and do not intervene; When in the shallow control range, perform a slight or medium frequency reduction adjustment; When in a deep control zone, implement step-wise frequency reduction; When in a danger zone, immediately perform an emergency shutdown and send an alarm message to the cloud.

[0038] Step S5: After performing frequency reduction adjustment or pulse frequency increase, enter the observation period, re-collect pressure parameters and evaluate the control effect, and decide whether to maintain the current state or perform the next control action based on the evaluation results.

[0039] After performing frequency reduction or pulse frequency increase, the controller enters an observation period to re-acquire pressure parameters and evaluate the control effect. If the pressure recovers and stabilizes, the current frequency is maintained; if the pressure continues to deteriorate but has not yet entered the danger zone, the next control action is executed; if the pressure enters the danger zone, an emergency shutdown is executed. This closed-loop feedback mechanism ensures the effectiveness of each control action.

[0040] Example 2: In step S3, the pressure thresholds include the lower limit of the safe zone p1, the median line of the control zone p2, and the shutdown protection threshold p3, where p1>p2>p3; the rate thresholds include the rapid descent threshold v1, the medium-speed descent threshold v2, and the slow descent threshold v3, where v1 <v2<v3<0。

[0041] The two-dimensional state space is divided into four mutually exclusive regions with clear boundaries. Any state (p, v) belongs to one and only one region. The division of the control regions includes: Safe region: p > p1 or v > 0. When in the safe region, maintain the current frequency operation.

[0042] For example, when the measured pressure p = 1.3 MPa, regardless of the rate, it is in the safe region; when the pressure p = 0.5 MPa but the pressure change rate v is positive (the pressure is rising), it is also in the safe region because the pressure recovery trend indicates that the liquid supply is recovering.

[0043] Shallow regulation region: p2 < p ≤ p1 and v2 < v ≤ v3. When in the shallow regulation region, perform a slight frequency reduction or a medium frequency reduction.

[0044] This region corresponds to the working condition where the liquid supply capacity starts to be insufficient but decreases slowly. The system only needs minor intervention to maintain production.

[0045] Deep regulation region: p3 < p ≤ p2 or v1 < v ≤ v2. When in the deep regulation region, perform a stepped frequency reduction.

[0046] This region corresponds to the working condition where the liquid supply capacity is significantly insufficient or the decrease rate accelerates, and active stepped frequency reduction intervention is required.

[0047] Dangerous region: p ≤ p3 or v ≤ v1. When in the dangerous region, perform an emergency shutdown.

[0048] This region corresponds to the working condition of severe liquid supply shortage or rapid pressure drop, and immediate shutdown protection is necessary.

[0049] In actual operation, the controller calculates the current state at preset time intervals and maps it to the above regions. For example, at a certain moment, the measured p = 0.9 MPa and v = -0.08 MPa / min, then the state falls in the shallow regulation region, and the controller performs a slight frequency reduction operation. At another moment, the measured p = 0.6 MPa and v = -0.20 MPa / min, then the state falls in the deep regulation region, and the controller performs a stepped frequency reduction operation. If the measured p = 0.3 MPa or v = -0.35 MPa / min, then the state falls in the dangerous region, and the controller performs an emergency shutdown.

[0050] Example 3: The stepped frequency reduction includes: limiting the single frequency reduction amplitude to a preset percentage of the rated frequency, waiting for an observation period of the first preset time after performing the frequency reduction, monitoring the pressure response during the observation period, stopping the frequency reduction if the pressure rises, and continuing to perform the next step of the stepped frequency reduction if the pressure continues to drop but does not enter the dangerous region until the preset minimum operating frequency is reached.

[0051] Stepwise frequency reduction gradually finds the liquid supply balance point by reducing the frequency in small increments multiple times, avoiding excessive frequency reduction at one time that could lead to over-adjustment or production interruption.

[0052] Assuming the current operating frequency is 50Hz, the rated frequency is 50Hz, and the minimum operating frequency is limited to 15Hz (i.e., 30% of the rated frequency), when the controller determines that stepped frequency reduction is necessary, it performs the following steps: Step 1: Calculate the single frequency reduction magnitude, limiting it to a preset percentage of the rated frequency (5%, or 2.5Hz, in this embodiment). The controller sends a command to the inverter to reduce the frequency by this magnitude.

[0053] Step 2: Enter the first preset observation period (60 seconds in this embodiment). During this period, the controller continuously monitors the pressure value p and the pressure change rate v.

[0054] Step 3: After the observation period, assess the stress response: If the pressure rises, it indicates that the frequency reduction is effective and the liquid supply capacity is matched. Stop the frequency reduction and maintain the current frequency operation. If the pressure continues to drop but does not enter the danger zone, it indicates that the fluid supply is still insufficient, and the frequency needs to be reduced further, returning to step one; If the pressure enters the danger zone, an emergency shutdown should be performed immediately.

[0055] For example, during a stepped frequency reduction process, if the pressure rises from 0.7 MPa to 0.75 MPa after the first frequency reduction, and the rate of increase improves, then the frequency reduction is stopped, and the current frequency is maintained. In another instance, if the pressure continues to drop after the first frequency reduction, then the second step of frequency reduction is executed, and the situation is observed again. If the pressure still does not rise after multiple consecutive frequency reductions and the minimum frequency has been reached, then a protective shutdown is performed. This achieves adaptive matching of the liquid supply capacity, avoiding both frequent shutdowns and production interruptions caused by excessive frequency reduction.

[0056] Example 4: This example details the dynamic threshold self-learning method, which enables the system to automatically adapt to the formation energy decay throughout the entire life cycle of the oil well without the need for frequent manual parameter adjustments.

[0057] The pressure and rate thresholds are dynamically updated using a self-learning method, including: in the offline training phase, historical normal production data of oil wells are collected to establish a pressure distribution model under normal operating conditions; in the online update phase, the thresholds are updated daily or after each control event based on the latest data using an exponentially weighted moving average algorithm.

[0058] During the offline training phase, a variational autoencoder is used to establish a pressure distribution model under normal operating conditions, with the preset quantile of the pressure distribution as the initial pressure threshold; the rate threshold is determined based on the pressure change rate distribution within a predetermined time period before historical evacuation events.

[0059] The offline training phase includes: During the initial deployment, the system collected pressure data from the well during a period of normal production (without pumping-out events) over a past period to form a training dataset. A variational autoencoder (VAE) was used to model the pressure distribution under normal operating conditions. A VAE is a generative model that can learn the probability distribution of data; its specific algorithm and training methods can be implemented using existing technologies and will not be elaborated here.

[0060] After training, the preset quantiles of the pressure distribution are calculated using the VAE model: The lower limit of the safety zone p1 is taken as the P85 quantile of the pressure distribution; the median line of the control zone p2 is taken as the P50 quantile of the pressure distribution; and the shutdown protection threshold p3 is taken as the P15 quantile of the pressure distribution.

[0061] For the rate thresholds, the system collects all historical evacuation events that led to shutdowns, extracts the pressure change rate data within a predetermined time period before each event, and constructs a rate sample set. Then, the distribution of this sample set is calculated: the rapid descent threshold v1 is taken from the P10 quantile of the rate distribution; the medium-speed descent threshold v2 is taken from the P50 quantile of the rate distribution; and the slow descent threshold v3 is taken from the P90 quantile of the rate distribution.

[0062] The online update phase includes: During system operation, a threshold update is performed once daily. The controller reads the pressure data from the previous 24 hours and calculates the corresponding quantiles of the pressure distribution for the current day. Then, the thresholds are updated using the Exponentially Weighted Moving Average (EWMA) algorithm, which can be implemented using existing technology and will not be elaborated here.

[0063] The rate threshold is also updated using the EWMA algorithm, but only immediately after each evacuation event, because the evacuation event provides the latest danger boundary information.

[0064] It should be noted that the P85, P50, and P15 quantiles are merely exemplary choices in this embodiment and are not intended to limit the invention. In practical applications, other quantile combinations can be selected based on the specific operating conditions of the oil well, such as P90, P50, P10, or P80, P50, P20, etc. The selection principle for quantiles is: the lower limit of the safety zone should cover the upper end of the normal pressure distribution, the shutdown protection threshold should cover the lower end of the normal pressure distribution, and the median line of the control zone should be taken as the center position of the pressure distribution. The above quantiles can be automatically calculated based on historical normal production data of the oil well using a dynamic self-learning method.

[0065] Example 5: This example illustrates the pump condition identification steps and differentiated control strategies. Through multi-sensor fusion technology, the specific causes of impending pump cavitation are identified, and targeted control actions are executed.

[0066] The process includes a pump condition identification step: collecting pressure, temperature, vibration, and current parameters to extract multi-dimensional features; inputting these features into a pre-trained classifier to identify the current operating condition type, which includes fluid supply failure, airlock precursors, and sand jam precursors. The pump condition identification results obtained from this step include: when fluid supply failure is identified, a frequency reduction control strategy is implemented; when airlock precursors are identified, a pulse frequency increase control strategy is implemented; when sand jam precursors are identified, an emergency shutdown is implemented; when the pump condition identification result is inconsistent with the control actions executed according to the determined control area, the control action is executed based on the pump condition identification result.

[0067] The system simultaneously collects data from four sensors: pressure, temperature, vibration, and current. The controller extracts features from the raw data to construct a multi-dimensional feature vector, including pressure features (mean, standard deviation, rate of change, fluctuation frequency), temperature features (mean, gradient), vibration features (peak value, root mean square value, frequency band energy), and current features (mean, harmonic content). All of these feature extraction methods can be implemented using existing technologies.

[0068] Specifically, random forest or XGBoost can be used as the classifier, with historical labeled data as training samples. The historical data is manually labeled and categorized into four types: normal operating conditions, fluid supply failure, signs of airlock, and signs of sand and cardiomyopathy. Random forest and XGBoost are well-known classification algorithms in the field, and their training methods will not be elaborated here. After training, the classifier is deployed in the edge controller to achieve real-time inference.

[0069] When the classifier identifies a specific operating condition, it executes differentiated control actions: Scenario 1: Fluid Supply Failure. When fluid supply failure is detected, a frequency reduction control strategy is implemented. For example, if at a certain moment the pressure slowly decreases, the temperature slowly increases, and the current steadily decreases, the classifier identifies this as fluid supply failure. The controller performs a stepped frequency reduction, lowering the operating frequency to match the fluid supply with the discharge rate, thus maintaining production.

[0070] Scenario 2: Airlock Precursor. When an airlock precursor is detected, a pulse frequency boosting control strategy is executed. For example, if there are drastic fluctuations in pressure, temperature, and current at a certain moment, the classifier identifies this as an airlock precursor. If the frequency is reduced using traditional methods, the displacement will be further reduced, exacerbating the airlock. This system executes pulse frequency boosting; the specific method is described in Example 6.

[0071] Scenario 3: Precursor Signs of Sand Blockage. When a precursor sign of sand blockage is detected, an emergency shutdown is initiated. For example, if at a certain moment the pressure drops, vibration increases sharply, and current fluctuates, the classifier identifies this as a precursor to sand blockage. The controller immediately executes an emergency shutdown to prevent sand particles from abrading the stator rubber and protect the screw pump unit.

[0072] When the pump condition identification result differs from the area division result, the control action shall be executed according to the pump condition identification result. This priority rule ensures that the system can correctly identify and handle special operating conditions, avoiding control failure due to misjudgment.

[0073] Example 6: The pulse frequency boosting control strategy includes boosting the operating frequency to a preset multiple of the rated frequency, and then restoring the original frequency after a second preset time.

[0074] Pulse frequency boosting is used to deal with the precursors of gas lock. By briefly increasing the pump speed and displacement, an instantaneous pressure shock is generated, which pushes the gas accumulated at the pump inlet into the wellbore or discharges it, thereby releasing the gas lock.

[0075] The specific implementation steps are as follows: Step 1: When the pump condition identification module determines that there are signs of impending airlock, the controller records the current operating frequency f0.

[0076] Step 2: The controller sends a command to the frequency converter to increase the frequency to a preset multiple of the rated frequency (1.1 to 1.2 times in this embodiment). The increase multiple should not be too high to avoid overloading the motor.

[0077] Step 3: Maintain high-frequency operation for a second preset time (5 to 10 seconds in this embodiment), which is sufficient to generate an effective pressure shock.

[0078] Step 4: After the time is up, the controller will restore the frequency to the original frequency f0.

[0079] Step 5: Enter the observation period, re-collect pressure data, and assess whether the airlock has been released. If the airlock is released, resume normal production; if the airlock is not released, repeat the pulse frequency increase 1 to 2 times; if multiple pulse frequency increases are ineffective, it is determined to be a severe airlock or fluid supply failure, and switch to frequency reduction control or shutdown.

[0080] For example, a well operating at 45Hz with a rated frequency of 50Hz was identified as a precursor to gas lock. The controller executed a pulse frequency boost: the frequency was increased to 55Hz, held for 8 seconds, and then returned to 45Hz. After observing for 30 seconds, the pressure rebounded, the fluctuations disappeared, the gas lock was released, and production returned to normal.

[0081] Pulse frequency boosting releases airlock through short-duration impacts, avoiding the predicament of airlock recurring with frequency reduction in traditional solutions, and improving production efficiency under complex operating conditions.

[0082] The minimum operating frequency for screw pumps is typically 30%-40% of the rated frequency. Below this value, pump efficiency drops sharply, and insufficient motor cooling may occur due to excessively low speed. Adjustments can be made based on the pump type in practical applications.

[0083] The variable frequency drive (VFD) anti-cavitation mechanism of this invention is based on the positive displacement characteristics of screw pumps. The displacement of a screw pump is directly proportional to its rotational speed; reducing the operating frequency linearly reduces the pump's displacement. When insufficient fluid supply is detected, the pump's suction capacity is reduced by lowering the frequency, thus rematching it with the wellbore's fluid supply capacity and preventing cavitation. Compared to traditional emergency shutdown protection, VFD enables flexible control during continuous production, protecting the equipment and preventing production interruptions.

[0084] Example 7: An intelligent control system for preventing runout of an electric submersible screw pump in the well, comprising: a downhole sensing layer, a sensor group installed at the inlet of the electric submersible screw pump, including a pressure sensor, a temperature sensor, and a vibration sensor, for real-time acquisition of downhole operating parameters; a wellhead control layer, including an edge controller and a frequency converter, the edge controller having a built-in two-dimensional state-space decision model and a self-learning module, for executing the method as described in the previous embodiment, and the frequency converter receiving control commands to flexibly adjust the speed of the electric submersible screw pump; and a cloud-based intelligent layer, connected to the wellhead control layer via a communication network, for storing historical data, training and updating the self-learning model, and sending the updated model parameters to the edge controller.

[0085] The edge controller also has a built-in pump condition recognition model, which is a random forest classifier or an XGBoost classifier, used to identify pump condition failure, airlock precursors and sand jam precursors based on multi-dimensional features of pressure, temperature, vibration and current.

[0086] The downhole sensing layer includes a sensor array integrated at the inlet of the electric submersible screw pump. The sensors employ an integrated design, combining pressure, temperature, and vibration sensors within a single protective housing, transmitting signals to the surface via a single cable. This application provides the following sensor technical parameters for reference: Pressure sensor: range 0-5 MPa, accuracy 0.1% FS, output 4-20 mA analog signal; Temperature sensor: range 0-150°C, accuracy ±0.5°C, output Pt100 resistance signal; Vibration sensor: range ±5g, frequency response 10-1000 Hz, output 4-20 mA analog signal. The sensors are installed as close as possible to the pump inlet to ensure measurement accuracy.

[0087] The wellhead control layer includes an edge controller and a frequency converter, installed within the wellhead control cabinet. The edge controller is a programmable logic controller (PLC) with floating-point capabilities or an embedded industrial computer, equipped with analog input modules, digital output modules, and supporting industrial communication protocols. The edge controller incorporates the following software modules: a two-dimensional state-space decision model for implementing control logic; a self-learning module for dynamic threshold updates; and a pump condition identification model for identifying operating conditions and implementing differentiated control, which can employ a random forest or XGBoost classifier.

[0088] The cloud-based intelligent layer is deployed on the industrial internet platform, and the platform's functions include: Data storage: Stores historical production data for all oil wells; Model training: Regularly retrain the VAE model and classifier model using the latest data; Model distribution: Distribute the updated model parameters to each wellhead controller; Visual monitoring: The operating status of the oil well is displayed in real time on the terminal interface, and alarm information is pushed.

[0089] Taking 10 ESP wells in an oilfield as an example, each well is equipped with a downhole sensing layer and a wellhead control layer. All wellhead controllers are connected to a cloud platform via a wireless network. The cloud platform centrally stores data, performs model training daily, and periodically distributes updated model parameters.

[0090] It should be noted that the specific values ​​of pressure thresholds, rate thresholds, and control parameters in the above embodiments are illustrative and not intended to limit the invention. In practical applications, these parameters should be obtained through dynamic self-learning methods or software simulation, based on factors such as the specific well depth, formation pressure, fluid properties, and pump specifications, or adjusted by on-site engineering technicians according to actual conditions.

[0091] The method and system for preventing air cavitation in electric submersible screw pumps provided by this invention are based on existing mature sensor, controller, and frequency converter hardware, and have good industrial feasibility. Through two-dimensional state-space decision-making, dynamic threshold self-learning, and intelligent pump condition identification, it can solve the technical problems of shortened unit life and low production efficiency caused by passive shutdown and threshold solidification in existing technologies.

[0092] For any parts not mentioned in this application, existing technologies may be used or referenced.

[0093] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0094] The above description is merely an embodiment of this application and is not intended to limit the scope 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 the claims of this application.

Claims

1. A method for preventing air cavitation in an electric submersible screw pump well, characterized in that, It includes the following steps: Step S1: Real-time collect the pressure parameters at the suction inlet of the electric submersible screw pump. Step S2: Calculate the current pressure value p and the pressure change rate v according to the pressure parameters, where the pressure change rate v is obtained by calculating the pressure change slope within a predetermined time window through a linear regression algorithm and then performing filtering processing. Step S3: Construct a two-dimensional state space according to the preset pressure threshold and rate threshold, divide the two-dimensional state space into multiple control regions, and determine the current control region according to the current pressure value p and the pressure change rate v. Step S4: Execute corresponding control actions according to the determined control region, and the control actions include maintaining operation, reducing frequency adjustment, pulse frequency increase, and protective shutdown. Step S5: After executing the reducing frequency adjustment or pulse frequency increase, enter the observation period, re-collect the pressure parameters and evaluate the control effect, and decide whether to maintain the current state or execute the next control action according to the evaluation result.

2. The method according to claim 1, characterized in that, In step S3, the pressure threshold includes the lower limit p1 of the safety zone, the median line p2 of the regulation zone, and the shutdown protection threshold p3, where p1>p2>p3; the rate threshold includes the rapid decline threshold v1, the medium-speed decline threshold v2, and the slow decline threshold v3, where v1<v2<v3<0; the division of the control regions includes: Safety zone: p>p1 or v>0; Shallow regulation zone: p2<p≤p1 and v2<v≤v3; Deep regulation zone: p3<p≤2 or v1<v≤v2; Dangerous zone: p≤3 or v≤v1.

3. The method according to claim 2, characterized in that, In step S4, the execution of the corresponding control actions specifically includes: maintaining the current frequency operation when in the safety zone; performing a slight frequency reduction or medium frequency reduction when in the shallow regulation zone; performing a stepwise frequency reduction when in the deep regulation zone; and performing an emergency shutdown when in the dangerous zone.

4. The method according to claim 3, characterized in that, The stepwise frequency reduction includes: limiting the single frequency reduction amplitude to a preset percentage of the rated frequency, waiting for an observation period of the first preset time after performing the frequency reduction, monitoring the pressure response during the observation period, stopping the frequency reduction if the pressure rebounds, and continuing to perform the next step of the stepwise frequency reduction if the pressure continues to drop but does not enter the dangerous zone until the preset minimum operating frequency is reached.

5. The method according to claim 2, characterized in that, The pressure threshold and the rate threshold are dynamically updated by a self-learning method, including: collecting historical normal production data of the oil well in the offline training stage to establish a pressure distribution model of the normal working condition; in the online update stage, updating each threshold daily at a fixed time or after each control event based on the latest data using the exponential weighted moving average algorithm.

6. The method according to claim 5, characterized in that, In the offline training stage, a variational autoencoder is used to establish a pressure distribution model of the normal working condition, and the preset quantiles of the pressure distribution are used as the initial pressure threshold; the rate threshold is determined according to the pressure change rate distribution within a predetermined time before the historical抽空 event.

7. The method according to claim 1, characterized in that, The system also includes a pump condition identification step: collecting pressure, temperature, vibration, and current parameters to extract multi-dimensional features, inputting these features into a pre-trained classifier to identify the current operating condition type, which includes fluid supply failure, airlock precursors, and sand jam precursors; the pump condition identification results obtained from the pump condition identification step include: when fluid supply failure is identified, a frequency reduction control strategy is executed; when airlock precursors are identified, a pulse frequency increase control strategy is executed; when sand jam precursors are identified, an emergency shutdown is executed; when the pump condition identification result is inconsistent with the control action executed according to the determined control area, the control action is executed based on the pump condition identification result.

8. The method according to claim 7, characterized in that, The pulse frequency boosting control strategy includes boosting the operating frequency to a preset multiple of the rated frequency, and then restoring the original frequency after a second preset time.

9. An intelligent control system for preventing cavitation in electric submersible screw pumps, characterized in that, include: The downhole sensing layer consists of a sensor array installed at the inlet of the electric submersible screw pump, including pressure sensors, temperature sensors, and vibration sensors, used to collect downhole operating parameters in real time. The wellhead control layer includes an edge controller and a frequency converter. The edge controller has a built-in two-dimensional state-space decision model and a self-learning module for executing the method described in any one of claims 1 to 8. The frequency converter receives control commands to flexibly adjust the speed of the electric submersible screw pump. The cloud-based intelligent layer connects to the wellhead control layer via a communication network. It is used to store historical data, train and update self-learning models, and send the updated model parameters to the edge controller.

10. The system according to claim 9, characterized in that, The edge controller also has a built-in pump condition recognition model, which is used to identify conditions such as fluid supply failure, airlock precursors, and sand jam precursors based on multi-dimensional features of pressure, temperature, vibration, and current.

Images