Oil well liquid production rate prediction method and device based on pumping unit working condition

Abstract

The present disclosure relates to an oil well liquid production rate prediction method and device based on the working condition of a pumping unit. The method comprises: determining a second parameter based on a first parameter of the pumping unit, obtaining a work graph representing the working condition of the pumping unit, wherein the first parameter comprises an electrical parameter, and the second parameter comprises a mechanical parameter of the working condition of the pumping unit; determining the working condition type of the pumping unit based on the work graph; determining a liquid production rate correction coefficient based on the working condition type; and predicting the liquid production rate of the pumping well by using the liquid production rate correction coefficient. The embodiments of the present disclosure can save the cost of liquid production rate prediction and improve the prediction accuracy.

Description

Technical Field

[0001] This disclosure relates to information processing technology, and in particular to a method and apparatus for predicting the production of pumping wells based on dynamometer charts. Background Technology

[0002] Timely and accurate production forecasting of pumping unit wells is crucial for ensuring efficient well operation. Currently, commonly used methods for measuring pumping unit well production require external sensors for mechanical parameters, resulting in high hardware investment and maintenance workload, which fails to meet the needs of continuous dynamic condition evaluation. Therefore, there is an urgent need for a low-cost and accurate method for measuring pumping unit well production to improve measurement precision. Summary of the Invention

[0003] This disclosure proposes a method, device, electronic equipment, and storage medium for predicting oil well production based on pumping unit operating conditions, which addresses the problems of low accuracy and high cost in the prediction of oil well production by pumping units in the prior art.

[0004] According to a first aspect of this disclosure, a method for predicting oil well production based on pumping unit operating conditions is provided, comprising: determining a second parameter based on a first parameter of the pumping unit; obtaining a dynamometer diagram representing the operating state of the pumping unit, wherein the first parameter includes electrical parameters and the second parameter includes mechanical parameters representing the operating state of the pumping unit; determining the operating condition type of the pumping unit based on the dynamometer diagram; determining a production correction coefficient based on the operating condition type; and predicting the production of the oil well using the production correction coefficient.

[0005] In some possible implementations, obtaining the dynamometer diagram representing the operating status of the pumping unit includes obtaining at least one of the surface dynamometer diagram and / or the downhole pump dynamometer diagram of the pumping unit.

[0006] In some possible implementations, the method for obtaining the surface dynamometer card of the pumping unit includes: determining the suspension point displacement and suspension point load of the pumping unit well using the second parameter; obtaining the surface dynamometer card based on the suspension point displacement and suspension point load; and / or A method for obtaining the downhole pump dynamometer diagram of the pumping unit well includes: determining the downhole pump dynamometer diagram based on the surface dynamometer diagram of the pumping unit.

[0007] In some possible implementations, determining the production volume correction coefficient based on the operating condition type includes: obtaining the operating condition probability and correction coefficient corresponding to the operating condition type; and determining the correction coefficient based on the operating condition probability and correction coefficient.

[0008] In some possible implementations, obtaining the working condition probability corresponding to the working condition type includes: obtaining a first image and a second image based on the dynamometer; performing a first feature extraction and a second feature extraction process on the first image to obtain a first feature and a second feature, and performing a first feature extraction process on the second image to obtain a third feature; wherein the first feature extraction process includes image encoding processing, and the second feature extraction process includes semantic feature encoding processing; performing a fusion encoding and decoding process on the first feature, the second feature, and the third feature to obtain a fusion encoding and decoding feature; and obtaining the working condition probability based on the fusion encoding and decoding feature.

[0009] In some possible implementations, predicting the production rate of the pumping well using the production rate correction coefficient includes: determining the effective stroke of the pumping unit based on the dynamometer card; determining the liquid loss and multiphase mixture volume factor within a single stroke; obtaining a first production rate using the liquid loss and multiphase mixture volume factor pair; and correcting the first production rate using the production rate correction coefficient to obtain the predicted production rate of the pumping well.

[0010] In some possible implementations, determining the operating condition type of the pumping unit based on the dynamometer map includes: obtaining a first image and a second image based on the dynamometer map; performing first feature extraction and second feature extraction processing on the first image to obtain a first feature and a second feature; performing first feature extraction processing on the second image to obtain a third feature; performing fusion encoding and decoding processing on the first feature, the second feature, and the third feature to obtain fusion encoding and decoding features; and obtaining the operating condition probability based on the fusion decoding features.

[0011] According to a second aspect of this disclosure, an oil well production prediction device based on pumping unit operating conditions is provided. The device includes: an acquisition module for determining a second parameter based on a first parameter of the pumping unit and acquiring a dynamometer diagram representing the operating state of the pumping unit, wherein the first parameter includes electrical parameters and the second parameter includes mechanical parameters representing the operating state of the pumping unit; a diagnosis module for determining the operating condition type of the pumping unit based on the dynamometer diagram; a determination module for determining a production correction coefficient based on the operating condition type; and a calculation module for predicting the production of the oil well using the production correction coefficient.

[0012] According to a third aspect of this disclosure, an electronic device is provided, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method described in any one of the first aspects.

[0013] According to a fourth aspect of this disclosure, a computer-readable storage medium is provided that stores computer program instructions thereon, which, when executed by a processor, implement the method described in any one of the first aspects.

[0014] In the oil well production prediction method based on pumping unit operating conditions in this embodiment, online monitoring of electrical parameters is first used as the front end, and the dynamometer diagram is inverted by combining the motion relationship of the pumping unit mechanism and the dynamic balance relationship of the system. The dynamometer diagram is then used to realize the automatic identification of different downhole operating conditions, and the operating condition categories are further mapped to the corresponding production correction coefficients. A corrected single-well production calculation model is then established, which improves the physical applicability and digital intelligence level of pumping unit well production calculation.

[0015] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure.

[0016] Other features and aspects of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the specification, serve to illustrate the technical solutions of this disclosure.

[0018] Figure 1 This is a flowchart illustrating the prediction of oil well production based on pumping unit operating conditions in an embodiment of this disclosure. Figure 2 This is a schematic diagram of the structure of the pumping unit operating condition diagnostic model in an embodiment of this disclosure; Figure 3 This is a schematic diagram of the feature points of the power diagram according to an embodiment of this disclosure. Detailed Implementation

[0019] Various exemplary embodiments, features, and aspects of this disclosure will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.

[0020] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments.

[0021] In this document, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0022] Furthermore, to better illustrate this disclosure, numerous specific details are set forth in the following detailed description. Those skilled in the art will understand that this disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art have not been described in detail in order to highlight the main points of this disclosure.

[0023] The method disclosed herein can be executed by an image processing device, such as a terminal device, a server, or other processing device. The terminal device can be a user equipment (UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal digital assistant (PDA), handheld device, computing device, in-vehicle device, wearable device, etc. In some possible implementations, the method can be implemented by a processor calling computer-readable instructions stored in memory.

[0024] It is understood that the various method embodiments mentioned above in this disclosure can be combined with each other to form combined embodiments without violating the principle and logic. Due to space limitations, this disclosure will not elaborate further.

[0025] Figure 1 This is a flowchart of the oil well production prediction method based on pumping unit operating conditions in an embodiment of this disclosure; as shown... Figure 1 As shown, the method for predicting oil well production based on pumping unit operating conditions includes: S10: Determine the second parameter based on the first parameter of the pumping unit, and obtain the dynamometer diagram representing the working state of the pumping unit. The first parameter includes the electrical parameters of the pumping unit drive motor, and the second parameter includes the mechanical parameters of the pumping unit's operating state. In some possible implementations, a first parameter, including electrical parameters, can be obtained using an electrical parameter acquisition device. This first parameter may include the voltage, current, and power of the drive motor during the operation of the pumping unit. The first parameter is directly acquired parameter information. The second parameter can be a parameter calculated using the first parameter, including the pumping unit crankshaft angle and crankshaft output torque. This disclosure does not specify the exact parameters of the first and second parameters; anything that can be used to measure the indicator diagram can be used as the first and second parameters of this disclosure.

[0026] S20: Determine the operating condition type of the pumping unit based on the dynamometer diagram; In some possible embodiments, a pre-trained deep learning model can be used to process the dynamometer card to obtain the corresponding operating condition type, thereby enabling the detection of the pumping unit's operating status and fault type. The operating condition types in this disclosure include, but are not limited to: normal operating condition, wax deposition, pump leakage, pipe leakage, insufficient liquid supply, and gas interference. This disclosure does not limit the specific type of operating condition and can adaptively set according to on-site requirements.

[0027] S30: Determine the liquid production correction coefficient based on the aforementioned operating condition type; In some possible implementations, a corresponding correction factor can be adaptively configured for each type of operating condition. This correction factor can be determined based on the product volume obtained without this correction process and the actual product volume.

[0028] S40: Predict the production rate of the pumping well using the production rate correction coefficient.

[0029] Once the correction factor is obtained, it can be used to correct the calculated production rate, resulting in a more accurate and optimized oil well production rate.

[0030] Based on the configuration of the embodiments of this disclosure, this disclosure proposes a method for calculating the production volume of pumping wells that integrates optimization procedures such as electrical parameter inversion and operating condition diagnosis, achieving a unified representation from surface electrical signals to downhole pump conditions and actual fluid discharge capacity. Furthermore, the embodiments of this disclosure also provide methods for selecting production correction coefficients corresponding to operating condition types and for correcting production volume. This not only provides adaptive production volume correction schemes for different operating conditions but also further improves the accuracy of production volume calculation. Since this disclosure determines a second parameter including mechanical parameters using a first parameter including electrical parameters, this process does not require additional force parameter sensing hardware, reducing hardware costs.

[0031] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings. First, a first parameter can be obtained in the embodiments of this disclosure. In the embodiments of this disclosure, the method of obtaining the first parameter may include at least one of the following methods: A) Directly acquire the first parameter, including the electrical parameters of the drive motor, using electrical parameter acquisition equipment; wherein, the electrical parameter acquisition equipment can be at least one of voltage and current sensors, current transformers, voltage transformers, voltage transmitters, three-phase multi-function power meters, and three-phase power acquisition modules. The above-mentioned equipment is not a specific limitation of this disclosure and can also be obtained through other equipment. The above-mentioned equipment can be the electrical acquisition equipment inherent in the pumping unit control cabinet, etc., eliminating the need for additional installation and reducing the cost of external equipment.

[0032] B) Transmitting and receiving the first parameter via an electronic device; in this embodiment of the disclosure, the first parameter transmitted by other electronic devices can be received via wired communication and / or wireless communication, and the communication method is not a specific limitation of this disclosure.

[0033] C) Read the first parameter from the storage medium or database; In this embodiment of the present disclosure, the first parameter can be read from the local storage device or server according to the received read instruction, and the present disclosure does not specifically limit this.

[0034] After obtaining the first parameter, the second parameter can be derived from it. The electrical parameters of the pumping unit well reflect the dynamic changes in the input energy at the motor end with the pumping cycle. Essentially, it is the external response of the downhole load after being transmitted step-by-step through the sucker rod string, connecting rod mechanism, gearbox, and motor system. Compared with directly acquiring the dynamometer card from the suspension point using load and displacement sensors, electrical parameter acquisition has advantages such as low installation and maintenance costs, strong online continuity, and suitability for large-scale deployment in platform well clusters. Therefore, this disclosure first establishes an inversion model from electrical parameters to the dynamometer card, providing surface boundary conditions for subsequent production calculations.

[0035] In this embodiment of the disclosure, the method for obtaining the second parameter using the first parameter includes: taking the voltage and current parameters of the drive motor and calculating the corresponding active power; optimizing the active power; using the optimized power for transmission conversion to obtain the effective mechanical power of the crankshaft; and obtaining the second parameter based on the effective mechanical power of the crankshaft.

[0036] First, this embodiment of the disclosure obtains the voltage and current parameters of the drive motor and calculates the corresponding active power. During the operation of the pumping unit, the line voltage of the drive motor is continuously collected. Line current Power factor If the data acquisition device can directly output active power, then the measured active power is recorded as follows: If the system only collects three-phase voltage, current, and power factor, for an approximately symmetrical three-phase AC system, the following can be calculated: in, : Active power input to the pumping unit drive motor, in watts (W); The phase difference angle between the fundamental voltage wave and the fundamental current wave at this moment is also called the power factor angle, which is dimensionless.

[0037] When collecting synchronous instantaneous three-phase voltage and current, the following methods can be used: in, This refers to the instantaneous voltage of phase A. This refers to the instantaneous current of phase A. This is the instantaneous voltage of phase B; This refers to the instantaneous current of phase B. This refers to the instantaneous voltage of phase C. This refers to the instantaneous current of phase C. In practical applications, if the power metering device directly provides the active power curve, the measured power parameters should be used first to avoid introducing additional errors from secondary calculations of the effective values ​​of voltage and current and the power factor.

[0038] Secondly, the active power is optimized. The original power curve is easily affected by grid harmonics, inverter switching noise, communication packet loss, and transient interference. Therefore, it is necessary to perform outlier identification, singular data interpolation, and smoothing filtering in sequence. Let the preprocessed active power be... . This represents singular data detection and interpolation. This indicates digital filtering. Hampel filtering can be used to suppress pulse anomalies, and then moving average, Savitzky-Golay or low-pass filtering can be used to reduce high-frequency fluctuations.

[0039] Then, the optimized power is used for transmission conversion to obtain the effective mechanical power of the crankshaft. Specifically, the effective output mechanical power of the motor can be obtained using the drive motor losses, by subtracting the overall motor losses from the optimized active power. Thus, the effective output mechanical power of the motor shaft is obtained. ,in This embodiment of the disclosure can utilize the no-load power P0 as the comprehensive loss of the motor. Then, the effective mechanical power of the crankshaft can be obtained from the effective output mechanical power of the motor. The effective output mechanical power of the motor is then multiplied sequentially by the belt drive efficiency and the gearbox efficiency to obtain the effective mechanical power of the crankshaft, which can be expressed as follows: .

[0040] The two transmission efficiencies are defined as the combined transmission efficiency. ,Right now ,at this time .

[0041] Having obtained the effective mechanical power of the crankshaft, a second parameter can be derived based on this effective mechanical power. This second parameter may include the crankshaft angular velocity and the crankshaft output torque. First, the crankshaft angular velocity can be obtained using the pumping unit strokes, and then the crankshaft output torque can be obtained using the effective mechanical power and angular velocity.

[0042] The method for calculating the crankshaft angular velocity is as follows: , where N represents the number of pumping strokes.

[0043] The crankshaft output torque M(t) can be obtained by using the ratio between the effective mechanical power of the crankshaft and its angular velocity, i.e. Substituting into the above formula, we can obtain Considering that the pumping unit may experience positive torque (the motor provides energy to the pumping unit crankshaft) or negative torque (energy feedback from the pumping unit mechanical system to the motor) within a single stroke cycle, and given the different efficiency mechanisms of forward and reverse transmission, the expression for the crankshaft output torque is modified accordingly. in, This indicates that the pumping unit is driven by an electric motor. This indicates that the oil pumping unit feeds mechanical energy back to the motor.

[0044] Finally, M(t) within one stroke cycle is determined according to the crank angle. By performing equal-angle resampling, the crankshaft output torque corresponding to the crank angle can be obtained. .

[0045] With the second parameter obtained, the dynamometer card of the pumping unit can be further obtained. In this embodiment, the dynamometer card can be at least one of a surface dynamometer card (suspension point dynamometer card) and / or a downhole pump dynamometer card. In a preferred embodiment, the pump dynamometer card can be used for subsequent production calculations. The methods for determining the surface dynamometer card and the downhole pump dynamometer card are described in detail below.

[0046] First, the method for obtaining the surface dynamometer diagram of the pumping unit includes: determining the suspension point displacement and suspension point load of the pumping unit well using the second parameter; and obtaining the surface dynamometer diagram based on the suspension point displacement and suspension point load.

[0047] First, the suspension point displacement is determined using the second parameter, including the crank angle. With the swing beam The correspondence between these two relationships yields the suspension point displacement. Since the suspension point of the pumping unit's head reciprocates with the swing of the traveling beam, the suspension point displacement can be determined by the change in the traveling beam's swing angle. Among them, S( () indicates crank angle The corresponding suspension point displacement, This represents the maximum swing angle corresponding to the extreme swing position of the walking beam. The current crank angle represents the swing angle of the walking beam, and A represents the length of the walking beam forearm.

[0048] Determining the suspension point load using the second parameter includes: using the torque factor and the crankshaft output torque to determine the suspension point load. Firstly, the following parameters can be used: the length A of the walking beam forearm, the crank radius R, the length C of the walking beam rear wall, and the instantaneous angle between the connecting rod axis and the crank radius direction. The instantaneous angle between the connecting rod axis and the walking beam rear arm axis The torque factor is obtained by the following expression, where TF( () represents the torque factor.

[0049] Then, the suspension point load is calculated using the torque factor and the crankshaft output torque, expressed as: .in, B is the suspension point load, and A is the unbalanced weight of the pumping unit structure. Crank maximum balance torque, The crank hysteresis angle is denoted as B. Here, B represents the combined effect of the structural components such as the downstream beam, connecting rod, and crossbeam, which is equivalent to a force at the suspension point when the self-weights of these components are not completely balanced in actual operation.

[0050] Based on the above, the suspension point displacement and suspension point load can be obtained. By obtaining continuous load and displacement points, and sorting and closing them according to the motion sequence from the lower dead point to the upper dead point and then back from the upper dead point to the lower dead point, an electrical parameter inversion suspension point dynamometer diagram can be formed.

[0051] Furthermore, given the surface dynamometer card, a downhole pump dynamometer card can also be obtained. In this embodiment, the method for obtaining the downhole pump dynamometer card of the pumping unit well includes: determining the damping attenuation coefficient using the one-dimensional damping wave equation of the sucker rod string; performing a Fourier transform on the suspension point displacement and suspension point load to obtain the corresponding Fourier series; obtaining the corrected displacement of the suspension point displacement and the corrected load of the suspension point load based on the damping attenuation coefficient and the Fourier series; and obtaining the downhole pump dynamometer card based on the corrected displacement and the corrected load.

[0052] The sucker rod string can be regarded as a longitudinally vibrating elastic body under damping, and its dynamic behavior can be described by a one-dimensional damped wave equation as shown in the following figure.

[0053] Where u(x,t) represents the axial displacement of the rod at depth x and time t; a is the equivalent damping coefficient; and c is the rod wave velocity. This equation characterizes the propagation law of the excitation at the suspension point in the rod, where the damping term describes the influence of well fluid viscous resistance, coupling resistance, and rod motion energy loss on the wave process. For the nth harmonic, when it propagates from the surface to the pump hanger depth L, it will simultaneously experience damping attenuation and phase delay. The damping attenuation coefficient can be expressed as follows.

[0054] Among them, R n The attenuation coefficient of the nth harmonic, L: pump hanger depth.

[0055] Correspondingly, the suspension point displacement and suspension point load obtained from the electrical parameter inversion are considered as boundary conditions at the upper end of the rod. Since both are periodic signals, they can be expanded using Fourier series. This disclosure does not provide specific details regarding the Fourier transform; those skilled in the art can obtain it using existing methods.

[0056] Furthermore, the corrected displacement and corrected load can be determined using the damping attenuation coefficient. The corrected displacement is calculated as follows: For a multi-stage sucker rod string composed of rod segments of different diameters and lengths, the surface suspension point displacement and load signals need to be propagated downhole segment by segment according to the structural parameters of each rod segment. The calculation method for the corrected load is as follows: Where A n Bn and Bn are the Fourier coefficients of the nth order cosine and sine terms of the ground suspension load signal, respectively, and A[0] is the DC component of the ground suspension load signal. This is for phase delay.

[0057] Once the pumping unit dynamometer diagram is obtained, operational condition diagnosis can be further performed. In this embodiment of the disclosure, the dynamometer diagram used for operational condition diagnosis can be a surface suspension point dynamometer diagram or a downhole pump dynamometer diagram, or both can be used simultaneously as input for operational condition diagnosis.

[0058] This disclosure embodiment uses a deep learning model to perform operating condition diagnosis. In this embodiment, the input to the deep learning model is a first image and a second image obtained based on the dynamometer map. The deep learning model then performs feature processing on the first and second images to detect the dynamometer map type. Specifically, obtaining the operating condition probability corresponding to the operating condition type includes: obtaining a first image and a second image based on the dynamometer map; performing first feature extraction and second feature extraction processing on the first image to obtain a first feature and a second feature; performing first feature extraction processing on the second image to obtain a third feature; performing fusion encoding and decoding processing on the first feature, the second feature, and the third feature to obtain fused encoding and decoding features; and obtaining the operating condition probability based on the fused decoding and decoding features.

[0059] The first image and the second image are obtained based on the work diagram in at least one of the following ways: A) Fill the area surrounded by displacement and load in the ground dynamometer diagram with color to obtain a first image; or, connect the two dynamometer diagrams with and without filling to obtain a first image; perform frequency domain feature extraction on the first image to obtain a second image.

[0060] B) Fill the area enclosed by displacement and load in the downhole pump dynamometer with color to obtain a first image; or, connect the two dynamometers with and without filling to obtain a first image; perform frequency domain feature extraction on the first image to obtain a second image.

[0061] C) Fill the area enclosed by displacement and load of the surface dynamometer chart with color to obtain the first image, or connect the two surface dynamometer charts with and without color to obtain the first image; fill the area enclosed by displacement and load of the downhole pump dynamometer chart with color to obtain the second image, or connect the two downhole pump dynamometer charts with and without color to obtain the first image. D) Fill the area enclosed by displacement and load in the surface dynamometer card with color, and connect the corresponding feature channels with the frequency domain of the filled image to obtain the first image; fill the area enclosed by displacement and load in the downhole pump dynamometer card with color, and connect the corresponding feature channels with the frequency domain of the filled image to obtain the second image. E) Use the standard dynamometer card of the pumping unit well under normal operating conditions as the first image, and the acquired surface dynamometer card and / or downhole pump dynamometer card as the second image, or use the filled-in dynamometer cards as the first and second images respectively. This standard dynamometer card can be the dynamometer card of the pumping unit stored when the pumping unit is restored to normal operation.

[0062] The fill color can be set by the user, as long as it can distinguish the area within and outside the enclosed region; those skilled in the art can set it themselves. The fill colors of the surface dynamometer card and the downhole pump dynamometer card can be different or the same. Furthermore, the frequency domain feature extraction method can include Fourier transform, wavelet transform, and other frequency domain feature extraction methods; this disclosure does not specifically limit this method. Based on frequency domain feature extraction, high-frequency and low-frequency features can be obtained. Weighting the high-frequency and low-frequency features yields the corresponding frequency domain image, which can be used as the second image in embodiments A and B, and the frequency domain feature extraction result in embodiment D. The weights can be set according to requirements, for example, both being 0.5, but this is not a specific limitation of this disclosure.

[0063] Figure 2 The diagram illustrates the structure of the working condition diagnosis model in this embodiment. Given a first image and a second image, a two-branch process is performed on the first image: a first feature extraction and a second feature extraction process are performed on the first image to obtain first and second features. The first branch uses a first model to perform the first feature extraction process on the first image, extracting coded features from the image. In some embodiments, the first model can be a SAM model, retaining its large-scale pre-trained general segmentation prior. The image encoder of this model effectively extracts edge information, shape information, and color texture features of the image, realizing the basic feature representation of the working condition map. The second branch uses a second model to perform the second feature extraction on the first image, specifically extracting semantic features from the first image. Specifically, in the second branch, the second model first performs target-layer semantic feature extraction on the first image to obtain primary semantic features. After obtaining the primary semantic features, these features are further converted into semantic cue features adapted to the first features, i.e., the second features. The second model includes a semantic feature extraction module and a feature adaptation module. The semantic feature extraction module can be a DINO model, such as the DINO-V2-B version, and the feature adaptation module can be a DINO SemanticAdapter. This disclosure selects to use DINO and SAM dual-branch features extraction respectively. Based on the differences in feature extraction methods of different models, image features and semantic features corresponding to the first image are obtained, and the first image is fully and completely quantified, which is beneficial to subsequent working condition diagnosis.

[0064] In this embodiment, a first feature extraction process is also performed on the second image to obtain a third feature. In this embodiment, the second image is also processed using a first model to obtain coded features. This first model can also identify information such as structure, texture, and color in the second image corresponding to the power map. In this embodiment, when the second image is a frequency domain feature, the second image can be obtained by connecting the channels of high-frequency and low-frequency features, thereby realizing the extraction of semantic information from the frequency domain features. When the second image is a downhole pump power map, coded features corresponding to the first feature can be obtained. In a preferred embodiment, the image encoding can be performed on the second image after standardization processing to reduce noise information in the second image. The standardization processing method may include mean-variance normalization, but this disclosure does not specifically limit it.

[0065] Given the first feature, second feature, and third feature, a further fusion encoding / decoding process can be performed, including: performing a first fusion process on the first feature, second feature, and third feature to obtain image semantic fusion features; performing channel matching processing on the second feature to obtain semantic encoding features that match the image semantic fusion features; performing position hint and type hint parsing on the semantic encoding features to obtain encoding hint features; and performing decoding processing using the encoding hint features and the image semantic fusion features to obtain fusion encoding / decoding features.

[0066] Specifically, a first fusion process is performed on the first feature, the second feature, and the third feature to obtain image semantic fusion features, including: performing mean processing on the first feature and the third feature to obtain a mean feature; applying a gated weighting to the second feature using a gate factor, and then summing it with the mean feature to obtain the image semantic fusion features. This embodiment of the present disclosure can achieve dual fusion of semantic and image features, enriching feature information and improving detection accuracy. Specifically, performing mean processing on the first feature and the third feature can achieve fusion enhancement of image-level encoded features. The gate factor can achieve selective feature enhancement, accurately extracting semantic information related to the working condition from the second feature, and then fusing it with the image encoded features to obtain more accurate image semantic information. In this embodiment of the present disclosure, the gate factor is a learnable parameter obtained through training the network model, and its value is in the range greater than or equal to 0 and less than or equal to 1. The expression for the fusion enhancement is: Ff = Fsd + gFq, g = σ ( α ) Among them, Fs d The mean feature is the first and third features. Fqα is the second feature, g is the gating factor, σ is the activation function, and α is the learnable parameter.

[0067] In other embodiments, the gating factor can be determined by the dynamometer feature point processing branch of the working condition detection model. The input of this branch is the measured feature information of the dynamometer, which includes graphic feature information based on the dynamometer, such as the feature point position, curvature at the feature point, dynamometer area, and slope between adjacent feature points.

[0068] The dynamometer feature point processing branch includes: dividing the original dynamometer sequence into four physical segments based on the displacement coordinates of four feature points: the upstroke unloaded segment AB, the loaded segment BC, the downstroke fully loaded segment CD, and the unloaded segment DA; inputting the measured feature information into multiple neurons to determine the segment weights w_1, w_2, w_3, w_4 for each physical segment; merging the segment weights into a weight sequence and inputting the weight sequence into the encoding module to obtain the gating factor.

[0069] Merging the segment weights into a weight sequence includes at least one of the following methods: connecting the segment weights to form a weight sequence; performing convolution processing on the connected weight sequences to obtain a new weight sequence.

[0070] The encoding module performs a two-branch process on the input weight sequence. The first branch process includes performing a linear transformation, sigmoid activation, and activation on the weight sequence. The second branch process includes performing a linear transformation, feature dimensionality reduction, activation, and state space transformation on the weight sequence. The two features obtained from the two-branch process are then multiplied and subjected to a linear transformation to obtain a gating factor corresponding to the weight sequence. In some possible implementations, the weight sequences processed by the first and second branches can be the same, formed by merging the segment weights. Alternatively, in some other implementations, the weight sequences input to the first and second branches can be obtained by merging weights in different directions.

[0071] For example, the input weight sequence of one of the first and second branches is the weight sequence obtained by performing weight merging on the segment weights in the length direction, and the input of the other branch is the weight sequence obtained by performing weight merging on the segment weights in the height direction. Multi-dimensional feature integration can be achieved by merging in different directions.

[0072] Furthermore, the encoding process for the weight sequence includes: using the first branch to perform a linear transformation and sigmoid activation on the weight sequence to obtain the first activation feature; using the second branch to perform a linear transformation, feature dimensionality reduction (such as convolution), sigmoid activation, and state space transformation on the weight sequence to obtain the state space feature; and performing a product operation and linear transformation on the first activation feature and the state space feature to obtain the encoded feature corresponding to the weight sequence, i.e., the gating factor. The linear transformations in the first and second branches use different weight parameters, which are learned during training. The convolutional processing in the second branch uses a 1*1 kernel, or it could be 3*3; this disclosure does not specifically limit this. Additionally, the state space transformation uses an SSM model, and the state equation and output equation are set by those skilled in the art according to their needs and are not specifically limited here. State space transformation enables accurate representation of features in the distance dimension, thereby improving model performance.

[0073] In some possible implementations, channel matching processing is performed on the semantic features obtained by the second feature extraction process to obtain semantic coding features that match the image semantic fusion features. The semantic features can be the primary semantic features in the above embodiments or the second features adapted to the first features. The following embodiments are referred to as semantic features.

[0074] The semantic features obtained from the second feature extraction process are subjected to channel matching processing to obtain semantic coding features that match the image semantic fusion features. This includes: performing channel integration processing on the semantic features to adjust the number of channels to be consistent with the number of channels in the image semantic fusion features; and performing width and height interpolation processing on the adjusted semantic features to obtain semantic coding features that match the dimensions of the image semantic fusion features. Channel integration can be performed using 1*1 convolution, and the interpolation method can use the resample function, but these are not considered specific limitations of this disclosure.

[0075] In this embodiment of the disclosure, after obtaining semantic encoded features, position and type hint parsing is performed on the semantic encoded features to obtain encoded hint features. Specifically, performing position and type hint parsing on the semantic encoded features to obtain encoded hint features includes: performing activation processing on the semantic encoded features using a first activation function to obtain a prediction mask; performing power map localization detection on the semantic encoded features to obtain detection hint features; and performing encoding processing on the prediction mask and detection hint features respectively to obtain encoded hint features. Both the prediction mask and the detection hint features can represent the basic fault shape and type of the power map. The encoding processing of the prediction mask and the detection hint features can be implemented using convolutional processing, such as using a CNN network to encode the mask features, but this is not a specific limitation of this disclosure. Furthermore, the first activation function can be a sigmoid function, and the module performing power map localization detection can be a YOLO detection head, used for coarse detection of the working condition type and coarse localization of the power map, respectively, to serve as hint information.

[0076] Given the encoded cue features, feature decoding can be performed using these features and image semantic fusion features. This decoding process yields fused encoded-decoded features, which represent the probabilities of each operating condition corresponding to the input power map. The type with the highest probability is identified as the operating condition type. Feature decoding can be implemented using the decoder connection activation function of the SAM model. Through this configuration, the dual encoding and fusion of semantic and image features can effectively improve the accuracy of operating condition identification by detecting and recognizing operating condition types.

[0077] Furthermore, to further improve the accuracy of the working condition detection model, this embodiment of the present disclosure optimizes the loss function during the model training process, thereby improving model accuracy. Specifically, the method for training the working condition detection model includes: using the working condition detection model to perform working condition type detection on the input training set, obtaining a prediction mask, detection prompt features, and fusion encoding / decoding features corresponding to the training set; performing a prediction loss on the prediction mask using at least two loss functions to obtain a first loss; performing a position loss prediction using the detection prompt features to obtain a second loss; and performing a working condition category loss prediction using the fusion encoding / decoding features to obtain a third loss. A weighted sum is performed on the first loss, second loss, and third loss to obtain the model loss; and the parameters of the fault detection model are adjusted using the model loss until the training termination condition is met. The loss functions for performing the prediction mask, the position loss, and the fault detection can include one or more of the following: cross-entropy loss, mean squared error (MSE), L2 loss, L1 loss, Smooth L1 loss, Huber loss, KL divergence (relative entropy), softmax loss, and Focal loss. This disclosure does not impose specific limitations on these functions. The training set includes paired first and second images. Extensive training under various conditions can improve the model's stability and robustness. The weights of the loss functions can be set according to requirements and are not specifically limited here.

[0078] In other embodiments of this disclosure, a multi-model optimization ensemble approach can be used for working condition type diagnosis, thereby reducing the instability caused by performance differences between different models. Embodiments of this disclosure can employ at least two working condition detection models. The fault detection based on the first and second images provided in the above embodiments can be one of these at least two models. Other models can include at least one or more of machine learning and deep learning models. Deep learning-based models can be one or more of CNN, RNN, GCN, etc., while machine learning-based models can include one or more of decision trees, Naive Bayes, support vector machines, multilayer perceptrons, random forests, gradient boosting trees, etc. For ease of description, subsequent embodiments will be uniformly described using working condition detection models.

[0079] When at least two working condition detection models are determined, the fusion weight coefficients of each working condition detection model can be determined during the training phase, and the working condition type of the input dynamometer can be determined based on the fusion weight coefficients.

[0080] Specifically, during the training phase, the fusion weight coefficients of each working condition detection model are determined, including: dividing the training dataset according to a first preset ratio to obtain a first training set and a first test set; dividing the first training set according to an n-fold cross-validation method to obtain n sub-training sets and sub-test sets, training the working condition detection model using each sub-training set, testing using the sub-test sets, and obtaining the test index of each working condition detection model on the n sub-test sets; and obtaining the fusion weight coefficients of the working condition detection model using the test indexes.

[0081] In some embodiments, the test metrics may include the area under the receiver operating characteristic curve (AUC) and the coordinate axes, recall, and precision. Other evaluation metrics applicable to classification tasks may also be included. Obtaining the fusion weight coefficients of the work condition detection model using the test metrics includes: using the initial weight list of the work condition detection model for each test metric; and for each work condition detection model, performing weight fusion on the initial weight list of each test metric to obtain optimized weights. The weight fusion includes linear fusion and nonlinear fusion. Nonlinear fusion includes weight fusion using exponential functions and power functions.

[0082] The initial weights for each test indicator are determined using the working condition detection model. This includes: for each working condition detection type and each test indicator, after training on each sub-training set and testing on the corresponding sub-test set, the test indicator is obtained. The ratio between the average of the test indicators of the model on the sub-test set and the average of the test indicators of all working condition detection models on the same sub-test set is determined as the initial weight of the working condition detection model for that test indicator in that set. The initial weights obtained from the n sub-test sets are combined to obtain an initial weight list.

[0083] Given the initial weight list for each test indicator in the working condition detection model, weight fusion can be performed on the initial weight list. Specifically, in the case of linear weight fusion, the weight coefficients for each test indicator are fitted using linear fitting, resulting in a weighted sum of the initial weight coefficients. The fusion weight coefficients of the working condition detection model are then determined using the average of the weighted sums of all test indicators.

[0084] Given the fusion weight coefficients of each working condition detection model, the working condition detection model trained in the subtest set corresponding to the best test index in the n-fold cross-validation can be used. The prediction probability of the working condition for the first test set is multiplied by the fusion weight coefficients to obtain the prediction result of the working condition detection model. The prediction results of each working condition detection model are then voted on to obtain the final working condition type.

[0085] Based on the above embodiments, accurate diagnosis of operating conditions based on dynamometer diagrams can be achieved, providing a basis for determining subsequent correction coefficients and ensuring the calculation of liquid production.

[0086] Based on the obtained operating condition type, a production volume correction coefficient can be further determined based on the operating condition type. This includes: calculating a first production volume based on the dynamometer cards of multiple pumping wells under the same operating condition; and using the average of the ratios between the first production volume and the actual production volume to obtain the correction coefficient corresponding to the operating condition type for that same operating condition. The method for calculating the first production volume is illustrated in the following embodiments, which are the production volumes obtained without operating condition correction according to the embodiments of this disclosure. In the embodiments of this disclosure, the correction coefficient for insufficient fluid supply is 0.84, the correction coefficient for gas influence is 0.78, the correction coefficient for fixed valve leakage is 0.72, the correction coefficient for traveling valve leakage is 0.66, the correction coefficient for severe mechanical abnormalities is 0, and the correction coefficient for normal operating conditions is 1. The above is only an illustrative example; correction coefficients for other operating conditions can be obtained using the above method.

[0087] Furthermore, in this embodiment of the present disclosure, predicting the production volume of the pumping well using the production volume correction coefficient includes: determining the effective stroke of the pumping unit based on the dynamometer card; determining the liquid loss and multiphase mixture volume factor within a single stroke; obtaining a first production volume using the liquid loss and multiphase mixture volume factor; and correcting the first production volume using the production volume correction coefficient to obtain the predicted production volume of the pumping well.

[0088] In this embodiment, a downhole pump dynamometer is preferably used to determine the effective stroke, but a surface dynamometer can also be used in other embodiments. Taking the downhole pump dynamometer as an example, determining the effective stroke of the pumping unit based on the dynamometer includes: performing feature extraction processing on the downhole pump dynamometer to obtain the effective stroke of the pumping unit. Specifically, firstly, noise reduction and optimization processing is performed on the displacement and load corresponding to the downhole pump dynamometer to obtain optimized displacement and load; the rate of curvature change of each point on the dynamometer formed by the optimized displacement and load is calculated, and four feature points are determined on the dynamometer during the upper and lower strokes based on the rate of curvature change; the effective stroke of the pumping unit is determined using the four feature points.

[0089] The process involves denoising and optimizing the displacement and load corresponding to the downhole pump dynamometer card. This includes: performing mean denoising on the displacement and load corresponding to the downhole pump dynamometer card to obtain denoised displacement and load; obtaining the maximum and minimum displacements based on the denoised displacements, and the minimum and maximum loads based on the denoised loads; normalizing the denoised displacements based on the maximum and minimum displacements to obtain optimized displacements; and normalizing the denoised loads based on the maximum and minimum loads to obtain optimized loads. The mean sliding window for the mean denoising process can be 5, meaning that the displacement and load values ​​of the current feature point are replaced by the mean values ​​of five feature points.

[0090] The expression for performing normalization is as follows.

[0091] in, This represents the i-th displacement. Let s represent the i-th load, min represent the minimum value, max represent the maximum value, and s pg and f pg These represent the optimized displacement and load, respectively.

[0092] Furthermore, the rate of curvature change at each point on the dynamometer card, composed of the optimized displacement and load, can be calculated, and four characteristic points on the dynamometer card during the upstroke and downstroke can be determined based on the rate of curvature change. The rate of curvature change can be obtained based on the ratio between the directed rotation angle and the arc length determined from n consecutive points on the dynamometer card. Determining the directed rotation angle from n consecutive points on the dynamometer card involves using the midpoint of the n points as a reference point, taking the angle between the first line segment between the first point and the midpoint and the second line segment between the midpoint and the nth point as the directed rotation angle, and the arc length as the sum of the arc lengths of each segment between the n consecutive points. The ratio between this directed rotation angle and the arc length is determined as the rate of curvature change at the reference point.

[0093] Reference Figure 3 The schematic diagram of the characteristic points of the stroke diagram shows that the point with the smallest displacement during the upstroke can be determined as the first characteristic point A by utilizing the switching characteristics of the upstroke and downstroke. The point with the largest change in curvature during the upstroke can be determined as the second characteristic point B and the third characteristic point C. The point with the largest change in curvature during the downstroke can be determined as the fourth characteristic point D and the first characteristic point A. In addition, the third characteristic point is also the point with the largest displacement.

[0094] Given four feature points, the first distance between the second and third feature points, and the second distance between the first and fourth feature points can be calculated. The smaller of the first and second distances is then determined as the effective stroke S. peThe distance between two points can be calculated using Euclidean distance or other methods, which can be set according to the needs of those skilled in the art.

[0095] If an effective stroke is obtained, the volume factor of crude oil and the volume factor of well water in the pumping well can be further obtained, and the first production volume can be calculated based on the effective stroke, the volume factor of crude oil and the volume factor of well water.

[0096] The crude oil volume factor is obtained by calculation using the AL-Marhoun principle, the expression of which is: Where b0 to b8 are set constants, respectively , , , , , , and p is the downhole pressure, t is the temperature of the oil well, and p b R is the bubble point pressure. s The volume of natural gas dissolved in a unit volume of crude oil at temperature t and pressure P is the dissolved gasoline ratio, B. ob p is the crude oil volume coefficient under bubble point pressure. s This is the suction pressure of the oil pump.

[0097] The bubble point pressure is calculated using the following formula.

[0098] The volume of natural gas dissolved in a unit volume of crude oil at temperature T and pressure P is the dissolved gasoline ratio R. s The calculation method is as follows.

[0099] in, The relative density of natural gas at 689.5 kPa is the oil volume factor. The relative density of crude oil under standard atmospheric pressure conditions. The relative density of natural gas, C1C2C2 values ​​are taken from Table 1 below.

[0100] Table 1. C1C2C2 Values The principle for obtaining the volume factor of oil well water is that, under standard pressure and temperature conditions, the volume factor of oil well water is Bw.

[0101] Where D0 to D1 are the correlation coefficients, calculated as follows.

[0102] t represents the downhole temperature. The correlation coefficient D and the natural gas saturated water coefficient a in this embodiment are shown in the table below.

[0103] Table 2. Correspondence between correlation coefficient and natural gas saturated water coefficient Based on the above, the effective process S can be obtained. pe The volume factor of crude oil, B0, and the volume factor of well water, B0. w Based on this, the first liquid production volume can be further calculated. The principle behind this is obtained using the following formula.

[0104] Among them, A p Let N be the cross-sectional area of ​​the pump plunger, and N be the number of strokes per unit time. This refers to the leakage rate of the oil pump during one stroke. This represents the average value of the gap between the pump barrel and the plunger. The diameter of the oil pump plunger; l z This refers to the length of the oil pump plunger. ρ represents the frictional force between the fluids inside the well, pd is the discharge pressure of the oil pump, and ps is the suction pressure of the oil pump. The average speed during the upstroke, where e is the eccentricity between the pump barrel bore axis and the plunger axis. n is the volume coefficient of a multiphase mixture. w This refers to the water content of the mixture under standard conditions. Regarding the above parameters, the embodiments of this disclosure can be directly measured and configured according to the specific conditions of the oil well.

[0105] Given the initial production rate, the corrected pumping well production rate can be obtained using the correction factor determined by the operating condition. Based on the above embodiment, under operating condition category c, there are a total of In the case of valid samples, the correction coefficient can be expressed as: in, This is the production correction factor for the c-th type of operating condition; This is the set of valid samples belonging to operating condition c; The first product volume of the i-th sample without operating condition correction; Let be the actual production volume of the i-th sample. After identifying the current operating conditions of the oil well, the final corrected production volume is: Based on the above, the production calculation in this embodiment is not simply based on traditional theoretical displacement, but rather on the actual pump end motion state obtained from pump dynamometer diagram inversion. It incorporates factors such as effective stroke, pump leakage, multiphase volume changes, gas content, and temperature and pressure property changes into a unified calculation framework. This model can dynamically determine correction terms based on the pump diameter, stroke rate, pump hanger depth, water cut, well fluid viscosity, downhole pressure and temperature, and pump condition characteristics of different oil wells, thereby improving the adaptability of production calculation under low permeability, fluctuating fluid supply, and complex well conditions. This method better reflects the actual fluid discharge process of the pump and the downhole hydrodynamic behavior, providing more reliable basic data for continuous evaluation of single-well production and optimization of operating procedures.

[0106] Those skilled in the art will understand that, in the above-described method of the specific implementation, the order in which each step is written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0107] In addition, this disclosure also provides a camouflage object detection device, electronic equipment, computer-readable storage medium, and program, all of which can be used to implement any of the oil well production prediction methods based on pumping unit operating conditions provided in this disclosure. The corresponding technical solutions and descriptions are described in the corresponding section of the method and will not be repeated here.

[0108] The oil well production prediction device based on pumping unit operating conditions includes: an acquisition module for determining a second parameter based on a first parameter of the pumping unit and acquiring a dynamometer diagram representing the operating state of the pumping unit, wherein the first parameter includes electrical parameters and the second parameter includes mechanical parameters representing the operating state of the pumping unit; a diagnosis module for determining the operating condition type of the pumping unit based on the dynamometer diagram; a determination module for determining a production correction coefficient based on the operating condition type; and a calculation module for predicting the production of the oil well using the production correction coefficient.

[0109] In some embodiments, the apparatus provided in this disclosure may have functions or include modules that can be used to perform the methods described in the above method embodiments. Specific implementations can be referred to the descriptions in the above method embodiments, and for brevity, will not be repeated here. This disclosure also proposes a computer-readable storage medium storing computer program instructions that, when executed by a processor, implement the above methods. The computer-readable storage medium may be a non-volatile computer-readable storage medium.

[0110] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.

Claims

1. A method for predicting oil well production based on pumping unit operating conditions, characterized in that, include: The second parameter is determined based on the first parameter of the pumping unit, and a dynamometer diagram representing the working state of the pumping unit is obtained. The first parameter includes electrical parameters, and the second parameter includes mechanical parameters representing the operating state of the pumping unit. The operating condition type of the pumping unit is determined based on the dynamometer diagram; Determine the liquid production correction coefficient based on the aforementioned operating condition type; The production rate of the pumping well is predicted using the production rate correction factor.

2. The method according to claim 1, characterized in that, The acquisition of the dynamometer diagram representing the operating status of the pumping unit includes: Obtain at least one of the surface dynamometer diagram and / or downhole pump dynamometer diagram of the pumping unit.

3. The method according to claim 2, characterized in that, The method for obtaining the ground dynamometer diagram of the pumping unit includes: The second parameter is used to determine the suspension point displacement and suspension point load of the pumping unit well; The ground dynamometer diagram is obtained based on the suspension point displacement and suspension point load. and / or The method for obtaining the downhole pump dynamometer diagram of the pumping unit well includes: The downhole pump dynamometer diagram is determined based on the surface dynamometer diagram of the pumping unit.

4. The method according to claim 1, characterized in that, Determining the production volume correction factor based on the aforementioned operating condition type includes: Obtain the working condition probability and correction coefficient corresponding to the working condition type; The correction coefficient is determined based on the operating condition probability and the correction coefficient.

5. The method according to claim 4, characterized in that, The step of obtaining the working condition probability corresponding to the working condition type includes: The first image and the second image are obtained based on the work diagram; The first image is subjected to a first feature extraction process and a second feature extraction process to obtain a first feature and a second feature, and the second image is subjected to a first feature extraction process to obtain a third feature; wherein the first feature extraction process includes image encoding process and the second feature extraction process includes semantic feature encoding process; Perform fusion encoding and decoding processing on the first feature, the second feature, and the third feature to obtain the fusion encoding and decoding feature; The operating condition probability is obtained based on the fused encoding and decoding features.

6. The method according to claim 1, characterized in that, Predicting the production rate of the pumping well using the production rate correction factor includes: The effective stroke of the pumping unit is determined based on the aforementioned dynamometer diagram. Determine the liquid loss during a single stroke and the volume coefficient of the multiphase mixture; The first product volume is obtained by using the liquid loss and the multiphase mixture volume coefficient. The first production volume is corrected using the production volume correction coefficient to obtain the predicted production volume of the oil well.

7. The method according to claim 1, characterized in that, The determination of the operating condition type of the pumping unit based on the dynamometer diagram includes: The first image and the second image are obtained based on the work diagram; The first image is processed by performing a first feature extraction and a second feature extraction to obtain a first feature and a second feature, and the second image is processed by performing a first feature extraction to obtain a third feature. Perform fusion encoding and decoding processing on the first feature, the second feature, and the third feature to obtain the fusion encoding and decoding feature; The operating condition probability is obtained based on the fused decoding features.

8. A device for predicting oil well production based on pumping unit operating conditions, characterized in that, The device includes: The acquisition module determines the second parameter based on the first parameter of the pumping unit and acquires the dynamometer diagram representing the working state of the pumping unit. The first parameter includes electrical parameters, and the second parameter includes mechanical parameters representing the operating state of the pumping unit. The diagnostic module is used to determine the operating condition type of the pumping unit based on the dynamometer diagram. The determination module is used to determine the liquid production correction coefficient based on the operating condition type; A calculation module is used to predict the production of the oil well using the production correction coefficient.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images