Adaptive multi-factor coupling model-based monitoring system and method for electric submersible pumps

Abstract

A kind of electric submersible pump monitoring system and method based on adaptive multi-factor coupling model.The data acquisition unit of monitoring system of the application is used to collect the three-phase current, voltage and frequency data of motor;The local processing unit is used for data preprocessing and preliminary analysis;Central control unit is used for model calculation, state evaluation and fault diagnosis;Man-machine interface is used to display monitoring results and receive operation instructions;The model used by central control unit is adaptive multi-factor coupling model;The steps of monitoring method are as follows: collecting electric submersible pump electrical parameters and working condition data through data acquisition unit, inputting fluid characteristic value;The local processing unit carries out data preprocessing and abnormal monitoring;Central control unit carries out model calculation, state evaluation and fault diagnosis;Monitoring results and optimization suggestions are displayed through man-machine interface.The beneficial effect is that the working state of electric submersible pump is estimated by real-time data, and the calculation efficiency is high and the adaptive ability is strong.

Description

Technical Field

[0001] This invention relates to a submersible electric pump monitoring system, and more particularly to a submersible electric pump monitoring system and method based on an adaptive multi-factor coupling model, belonging to the field of oil and gas development technology. Background Technology

[0002] Submersible electric pumps (SAPs) are critical equipment in oil extraction, and their operating status directly affects oil production efficiency and equipment lifespan. Existing SSP monitoring technologies mainly include traditional parameter monitoring, complex model analysis, artificial intelligence methods, and spectrum analysis. Traditional parameter monitoring relies on simple judgments by measuring basic parameters such as current and voltage. While computationally efficient, it has low accuracy and cannot accurately reflect the actual operating status of the SSP. Complex model analysis uses detailed models of the motor and SSP, such as finite element analysis or dynamic simulation models. These methods offer high accuracy but are computationally complex, typically taking minutes to hours, making real-time monitoring difficult. Artificial intelligence methods use neural networks and other AI technologies for state prediction. While providing good prediction results, they require extensive historical data for training, and the models are opaque, difficult to interpret and adjust. Spectrum analysis diagnoses faults by analyzing the spectral characteristics of current or vibration signals. This requires complex signal processing algorithms; a typical FFT has a computational complexity of O(NlogN) (meaning the computation time increases with the amount of data N at a rate proportional to the logarithm of N), making it unsuitable for scenarios with high real-time requirements. It is evident that existing technologies either lack sufficient precision or have high computational complexity, making it difficult to achieve real-time, accurate, and economical monitoring of submersible electric pumps on the production site. Summary of the Invention

[0003] To overcome the aforementioned shortcomings of existing submersible electric pump monitoring technologies, this invention provides a submersible electric pump monitoring system and method based on an adaptive multi-factor coupling model.

[0004] The technical solution adopted by this invention to solve its technical problem is: a submersible electric pump monitoring system based on an adaptive multi-factor coupling model, comprising a submersible electric pump, a data acquisition unit located at the output terminal of the submersible electric pump motor frequency converter for acquiring three-phase current, voltage, and frequency data of the submersible electric pump motor; a local processing unit for data preprocessing and preliminary analysis; a central control unit for model calculation, status assessment, and fault diagnosis; and a human-machine interface for displaying monitoring results and receiving operation commands; wherein the model used by the central control unit is an adaptive multi-factor coupling model.

[0005] A monitoring method for a submersible electric pump monitoring system based on an adaptive multi-factor coupling model, comprising the following steps:

[0006] S1. Collect the motor electrical parameters and operating condition data of the submersible electric pump through the data acquisition unit, and input the fluid characteristic values;

[0007] S2, The local processing unit performs data preprocessing and anomaly detection;

[0008] S3, The central control unit performs model calculations, status assessments, and fault diagnosis;

[0009] S4. Display monitoring results and optimization suggestions through a human-computer interaction interface.

[0010] Furthermore, the model calculation in step S3 is based on an adaptive multi-factor coupling model to calculate the motor's input power P, load torque T, and system efficiency η, as follows:

[0011] Calculate the input power P.

[0012]

[0013] in,

[0014] P -- Motor input active power, W,

[0015] U -- Effective value of line voltage, V,

[0016] I -- Effective value of line current, in A.

[0017] --Power factor;

[0018] Estimate the load torque T.

[0019] T=k1(t)*P / f+k2(t)*f+k3(t)*μ(Q)*Q (2)where:

[0020] T -- Estimated motor load torque, N·m

[0021] f -- Inverter output frequency, Hz

[0022] η -- System efficiency

[0023] Q -- Flow rate of the submersible electric pump, m 3 / s,

[0024] μ(Q) -- fluid viscosity function,

[0025] k1(t), k2(t), k3(t) -- coefficients that adapt to time;

[0026] Calculate the system efficiency η.

[0027] η=[k4(t)*ρ*g*Q*H] / P*(GOR_ref / GOR)+k5(t) (3)

[0028] in:

[0029] ρ -- fluid density, kg / m³ 3 ;

[0030] g -- acceleration due to gravity, m / s² 2 ;

[0031] H—Head of the submersible electric pump, in meters;

[0032] GOR -- Gas-to-oil ratio, m 3 / m 3 ;

[0033] GOR_ref -- Reference gas-oil ratio, m 3 / m 3 ;

[0034] k4(t) and k5(t) are time-adaptive coefficients.

[0035] The steps of the adaptive parameter adjustment algorithm for the adaptive multi-factor coupling model are as follows:

[0036] A1. Start: Algorithm begins;

[0037] A2. Initialize parameters k1(0) to k5(0): Set the initial parameters using historical data or empirical values;

[0038] A3. Calculate the error function E;

[0039] A4. Update parameters using gradient descent;

[0040] A5. Apply parameter constraints to ensure that the updated parameters are within a reasonable range;

[0041] A6. Determine if convergence has occurred:

[0042] If convergence is not achieved, return to step A3 and continue iterating;

[0043] If convergence has been achieved, proceed to the termination step;

[0044] A7. End: This round of parameter adjustments is complete.

[0045] Furthermore, step A3 calculates the error function E:

[0046] E = w1(TT) measured ) 2 +w2(η-η measured ) 2 +w3(PP measured ) 2 (4)

[0047] in:

[0048] w1, w2, and w3 are weighting coefficients;

[0049] T measured η measured P measured These are the measured values ​​of T, η, and P;

[0050] The weighting coefficients w1, w2, and w3 are all greater than 0 and less than 1;

[0051] The weighting coefficients w1+w2+w3=1.

[0052] Furthermore, step A4 updates the parameters using gradient descent:

[0053]

[0054] in,

[0055] α -- Adaptive learning rate;

[0056] λ -- Regularization parameter, dynamically adjusted using Bayesian optimization methods;

[0057] i = 1, 2, 3, 4, 5;

[0058] Furthermore, step A5 employs a soft constraint method to limit the parameters within a reasonable range:

[0059] 0.5ki_initial≤ki≤1.5ki_initial (i=1, 2, 3, 4, 5).

[0060] Furthermore, the state evaluation in step S3 uses an adaptive threshold for state evaluation.

[0061] T_upper=T rated *(1.2+0.1*sin(2π*t / 24)+0.05*sin(2π*t / (24*7))) (6)

[0062] T_lower=T rated *(0.8-0.1*sin(2π*t / 24)-0.05*sin(2π*t / (24*7))) (7)

[0063] η_lower=η rated *(0.9-0.05*sin(2π*t / 24)-0.03*sin(2π*t / (24*7))) (8)

[0064] Where t -- current hour,

[0065] Trated --Rated torque, N·m;

[0066] H rated --Rated efficiency.

[0067] The rules for state assessment are as follows:

[0068] Normal state: T_lower≤T≤T_upper and η≥η_lower;

[0069] Mild anomaly: T_lower*0.9≤T <T_lower,

[0070] Or T_upper < T ≤ T_upper * 1.1,

[0071] Or η_lower*0.9≤η<η_lower;

[0072] Critical anomaly: T < T_lower * 0.9

[0073] Or T > T_upper * 1.1,

[0074] Or η < η_lower*0.9.

[0075] Furthermore, the fault diagnosis in step S3 employs a deep learning-based fault diagnosis function, and the steps are as follows:

[0076] B1. Data preprocessing: Normalize and slide window segment the time series data;

[0077] B2. Feature Extraction: Temporal features are extracted using a bidirectional LSTM network;

[0078] B3. Attention Mechanism: Introduce a self-attention layer to enhance the weights of key time steps and features;

[0079] B4. Classifier: A classifier is used for multi-class fault identification;

[0080] B5. Model Training: The model employs the cross-entropy loss function and the Adam optimizer, and uses early stopping to prevent overfitting.

[0081] The beneficial effects of this invention are:

[0082] (1) An adaptive multi-factor coupling model is adopted to evaluate the working status of the submersible electric pump in real time by collecting real-time submersible electric pump parameters and operating data, so as to provide a basis for condition monitoring and fault diagnosis.

[0083] (2) High computational efficiency, the model only contains basic algebraic operations, which is far lower than the computational analysis of traditional complex models;

[0084] (3) It has strong adaptability and can adapt to changes in different working conditions and equipment status by adjusting model parameters in real time;

[0085] (4) Low resource requirements: The system can run on a microcontroller with a main frequency of only 80MHz, making it suitable for deployment in field devices;

[0086] (5) To enable real-time, accurate and economical monitoring of submersible electric pumps at the production site. Attached Figure Description

[0087] Figure 1 This is a structural block diagram of the submersible electric pump monitoring system of the present invention.

[0088] Figure 2 This is a schematic diagram illustrating the working principle of the adaptive multi-factor coupling model of the present invention.

[0089] Figure 3 This is a flowchart of the parameter adaptive adjustment algorithm of the present invention. Detailed Implementation

[0090] The present invention will be further described below with reference to the accompanying drawings and embodiments. However, those skilled in the art should understand that the present invention is not limited to the specific embodiments listed, and any embodiment that conforms to the spirit of the present invention should be included within the scope of protection of the present invention.

[0091] In the description of this invention, it should be noted that the terms "vertical," "upper," "lower," "left," "right," and "horizontal," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing or simplifying the invention, 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. Therefore, they should not be construed as limitations on this invention. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0092] See appendix Figure 1 This invention discloses a submersible electric pump monitoring system based on an adaptive multi-factor coupling model, comprising: a submersible electric pump, a data acquisition unit, a central control unit, a local processing unit, and a human-machine interface.

[0093] The data acquisition unit is located at the output end of the submersible electric pump motor inverter and is used to acquire the three-phase current, voltage and frequency data of the submersible electric pump motor.

[0094] The local processing unit is used for data preprocessing and preliminary analysis.

[0095] The central control unit is used for model calculation, status assessment, and fault diagnosis.

[0096] The human-machine interface is used to display monitoring results and receive operation commands.

[0097] The central control unit uses an adaptive multi-factor coupling model.

[0098] The monitoring method based on an adaptive multi-factor coupling model for the submersible electric pump monitoring system comprises the following steps:

[0099] S1. Collect the motor electrical parameters and operating condition data of the submersible electric pump through the data acquisition unit, and input the fluid characteristic values;

[0100] S2, The local processing unit performs data preprocessing and anomaly detection;

[0101] S3, The central control unit performs model calculations, status assessments, and fault diagnosis;

[0102] S4. Display monitoring results and optimization suggestions through a human-computer interaction interface.

[0103] See appendix Figure 2 The model calculation in step S3 is based on an adaptive multi-factor coupling model to calculate the motor's input power P, load torque T, and system efficiency η, as follows:

[0104] 1. Calculate the input power P

[0105]

[0106] in,

[0107] P -- Motor input active power, W,

[0108] U -- Effective value of line voltage, V,

[0109] I -- Effective value of line current, in A.

[0110] --Power factor;

[0111] 2. Estimate the load torque T

[0112] T=k1(t)*P / f+k2(t)*f+k3(t)*μ(Q)*Q (2)

[0113] in:

[0114] T -- Estimated motor load torque, N·m

[0115] f -- Inverter output frequency, Hz

[0116] η -- System efficiency

[0117] Q -- Flow rate of the submersible electric pump, m 3 / s,

[0118] μ(Q) -- fluid viscosity function,

[0119] k1(t), k2(t), k3(t) -- coefficients that adapt to time;

[0120] 3. Calculate the system efficiency η

[0121] η=[k4(t)*ρ*g*Q*H] / P*(GOR_ref / GOR)+k5(t) (3)

[0122] in:

[0123] ρ -- fluid density, kg / m³ 3 ;

[0124] g -- acceleration due to gravity, m / s² 2 ;

[0125] H—Head of the submersible electric pump, in meters;

[0126] GOR -- Gas-to-oil ratio, m 3 / m 3 ;

[0127] GOR_ref -- Reference gas-oil ratio, m 3 / m 3 ;

[0128] k4(t) and k5(t) are time-adaptive coefficients.

[0129] The core innovation of the adaptive multi-factor coupling model lies in the introduction of adaptive coefficients k1(t) to k5(t), which can be dynamically adjusted over time, enabling the model to adapt to different operating conditions and equipment status changes. Furthermore, the model also considers the effects of fluid viscosity μ(Q) and the gas-oil ratio GOR, improving the accuracy of the calculations.

[0130] See appendix Figure 3 The steps of the adaptive parameter adjustment algorithm for the adaptive multi-factor coupling model are as follows:

[0131] A1. Start: Algorithm begins;

[0132] A2. Initialize parameters k1(0) to k5(0): Set the initial parameters using historical data or empirical values;

[0133] A3. Calculate the error function E;

[0134] A4. Update parameters using gradient descent;

[0135] A5. Apply parameter constraints to ensure that the updated parameters are within a reasonable range;

[0136] A6. Determine if convergence has occurred:

[0137] If convergence is not achieved, return to step A3 and continue iterating;

[0138] If convergence has been achieved, proceed to the termination step;

[0139] A7. End: This round of parameter adjustments is complete.

[0140] The complete process of the parameter adaptive adjustment algorithm can ensure the continuous optimization and adaptability of the adaptive multi-factor coupling model.

[0141] Furthermore, step A3 calculates the error function E:

[0142] E = w1(TT) measured ) 2 +w2(η-η measured ) 2 +w3(PP measured ) 2 (4)

[0143] in:

[0144] w1, w2, and w3 are weighting coefficients;

[0145] T measured η measured P measured These are the measured values ​​of T, η, and P.

[0146] The weighting coefficients w1, w2, and w3 are all greater than 0 and less than 1.

[0147] The weighting coefficients w1+w2+w3=1.

[0148] Furthermore, step A4 updates the parameters using gradient descent:

[0149]

[0150] in,

[0151] α -- Adaptive learning rate;

[0152] λ -- Regularization parameter, dynamically adjusted using Bayesian optimization methods;

[0153] i = 1, 2, 3, 4, 5;

[0154] Furthermore, step A5 employs a soft constraint method to limit the parameters within a reasonable range:

[0155] 0.5ki_initial≤ki≤1.5ki_initial(i=1, 2, 3, 4, 5)

[0156] Furthermore, the state assessment in step S3 uses an adaptive threshold for state assessment.

[0157] T_upper=T rated *(1.2+0.1*sin(2π*t / 24)+0.05*sin(2π*t / (24*7))) (6)

[0158] T_lower=T rated *(0.8-0.1*sin(2π*t / 24)-0.05*sin(2π*t / (24*7))) (7)

[0159] η_lower=η rated *(0.9-0.05*sin(2π*t / 24)-0.03*sin(2π*t / (24*7))) (8)

[0160] Where t -- current hour,

[0161] T rated --Rated torque, N·m;

[0162] H rated --Rated efficiency.

[0163] The rules for state assessment are as follows:

[0164] Normal state: T_lower≤T≤T_upper and η≥η_lower;

[0165] Mild anomaly: T_lower*0.9≤T <T_lower,

[0166] Or T_upper < T ≤ T_upper * 1.1,

[0167] Or η_lower*0.9≤η<η_lower;

[0168] Critical anomaly: T < T_lower * 0.9

[0169] Or T > T_upper * 1.1,

[0170] Or η < η_lower*0.9.

[0171] The fault diagnosis in step S3 employs a deep learning-based fault diagnosis function, and the steps are as follows:

[0172] B1. Data preprocessing: Normalize and slide window segment the time series data;

[0173] B2. Feature Extraction: Temporal features are extracted using a bidirectional LSTM network;

[0174] B3. Attention Mechanism: Introduce a self-attention layer to enhance the weights of key time steps and features;

[0175] B4. Classifier: A classifier is used for multi-class fault identification;

[0176] B5. Model Training: The model employs the cross-entropy loss function and the Adam optimizer, and uses early stopping to prevent overfitting.

[0177] Example 1: Submersible Electric Pump Monitoring System

[0178] A monitoring system for submersible electric pumps based on an adaptive multi-factor coupling model includes: a submersible electric pump, a data acquisition unit, a local processing unit, a central control unit, and a human-machine interface. Figure 1 As shown, the arrows indicate the data flow and show the information transmission path between units. The entire system forms a closed-loop workflow, from data acquisition to analysis and processing, and then to result display and manual intervention, realizing intelligent monitoring and management of submersible electric pumps.

[0179] The monitoring system implementation and adaptive multi-factor coupling model are as follows:

[0180] 1. Data Acquisition Unit

[0181] Located at the output of the submersible electric pump motor inverter, it is used to collect three-phase current, voltage, and frequency data of the submersible electric pump motor, including:

[0182] Current transformer: Hall effect current sensor, range 0~100A, accuracy ±0.5%;

[0183] Voltage transformer: High-precision resistor voltage divider network, range 0~1000V, accuracy ±0.1%;

[0184] Sampling frequency: 1kHz.

[0185] 2. Local processing unit

[0186] Used for data preprocessing and preliminary analysis, including:

[0187] Processor: ARM Cortex-M4 microcontroller, 80MHz clock speed;

[0188] Storage: 256KB SRAM + 1MB Flash.

[0189] 3. Central control unit

[0190] Used for model calculation, condition assessment, and fault diagnosis, including:

[0191] Processor: Industrial-grade ARM Cortex-A53 quad-core processor, clock speed 1.2GHz;

[0192] Storage: 2GB RAM + 16GB eMMC.

[0193] The model used is an adaptive multi-factor coupling model.

[0194] 4. Human-computer interaction interface

[0195] It is used to display monitoring results and receive operation commands, including: a 7-inch LCD touch screen with a resolution of 800×480.

[0196] 5. Environmental adaptability

[0197] Operating temperature range: -20℃~+60℃;

[0198] Protection rating: IP66, suitable for complex oilfield environments.

[0199] Example 2: Monitoring Methods and Model Calculation

[0200] The monitoring method based on an adaptive multi-factor coupling model for the submersible electric pump monitoring system comprises the following steps:

[0201] S1. Collect the motor electrical parameters and operating condition data of the submersible electric pump through the data acquisition unit, and input fluid characteristic values, including:

[0202] Electrical parameters: line voltage U, line current I, power factor ;

[0203] Operating parameters: frequency f, flow rate Q, head H;

[0204] Fluid properties: density ρ, gravitational acceleration g, gas-oil ratio GOR.

[0205] S2, the local processing unit performs data preprocessing and anomaly detection.

[0206] S3, the central control unit performs model calculations, status assessments, and fault diagnosis.

[0207] S4. Display monitoring results and optimization suggestions through a human-computer interaction interface.

[0208] The model calculation in step S3 is based on an adaptive multi-factor coupling model to calculate the input power P, load torque T, and system efficiency η of the submersible electric pump motor, as follows:

[0209] 1. Calculate the input power P

[0210]

[0211] 2. Estimate the load torque T

[0212] T=k1(t)*P / f+k2(t)*f+k3(t)*μ(Q)*Q (2)

[0213] 3. Calculate the system efficiency η

[0214] η=[k4(t)*ρ*g*Q*H] / P*(GOR_ref / GOR)+k5(t) (3)

[0215] The adaptive parameter adjustment algorithm for the adaptive multi-factor coupling model comprises the following steps:

[0216] A1. Start: Algorithm begins;

[0217] A2. Initialize parameters k1(0) to k5(0): Set the initial parameters using historical data or empirical values;

[0218] A3. Calculate the error function E;

[0219] A4. Update parameters using gradient descent;

[0220] A5. Apply parameter constraints to ensure that the updated parameters are within a reasonable range;

[0221] A6. Determine if convergence has occurred:

[0222] If convergence is not achieved, return to step A3 and continue iterating;

[0223] If convergence has been achieved, proceed to the termination step;

[0224] A7. End: This round of parameter adjustments is complete.

[0225] Parameter adaptive adjustment algorithm:

[0226] 1. Determining initial parameters

[0227] The initial values ​​of k1(0) to k5(0) are determined using historical data and the least squares method.

[0228] 2. Parameter Update

[0229] Define the multi-objective error function E:

[0230] E = w1(TT) measured ) 2 +w2(η-η measured ) 2 +w3(PP measured ) 2 (4)

[0231] 3. Parameter Constraints

[0232] Soft constraint methods are used to limit the parameters within a reasonable range:

[0233] 0.5ki_initial≤ki≤1.5ki_initial (i=1, 2, 3, 4, 5)

[0234] Update parameters using an improved stochastic gradient descent method:

[0235]

[0236] A Bayesian optimization method is used to dynamically adjust α and λ to balance convergence speed and generalization ability.

[0237] 4. Convergence Judgment

[0238] Early stopping is used to prevent overfitting. When the validation set error has not improved for N consecutive times (e.g., N=5), parameter updates are stopped.

[0239] Example 3: Condition Assessment and Intelligent Fault Diagnosis

[0240] 1. Adaptive threshold calculation

[0241] T_upper=T rated *(1.2+0.1*sin(2π*t / 24)+0.05*sin(2π*t / (24*7))) (6)

[0242] T_lower=T rated *(0.8-0.1*sin(2π*t / 24)-0.05*sin(2π*t / (24*7))) (7)

[0243] η_lower=η rated *(0.9)-0.05*sin(2π*t / 24)-0.03*sin(2π*t / (24*7))) (8)

[0244] 2. Status Assessment Rules

[0245] Normal state: T_lower≤T≤T_upper and η≥η_lower.

[0246] Mild anomaly: T_lower*0.9≤T<T_lower;

[0247] Or T_upper < T ≤ T_upper * 1.1;

[0248] Or η_lower*0.9≤η<η_lower.

[0249] Critical anomaly: T < T_lower * 0.9;

[0250] Or T > T_upper * 1.1;

[0251] Or η < η_lower*0.9.

[0252] 3. Fault diagnosis based on deep learning

[0253] The fault diagnosis employs a deep learning-based fault diagnosis function, and the steps are as follows:

[0254] B1. Data preprocessing: Normalize and slide window segment the time series data;

[0255] B2. Feature Extraction: Temporal features are extracted using a bidirectional LSTM network;

[0256] B3. Attention Mechanism: Introduce a self-attention layer to enhance the weights of key time steps and features;

[0257] B4. Classifier: A classifier is used for multi-class fault identification;

[0258] B5. Model Training: The model employs the cross-entropy loss function and the Adam optimizer, and uses early stopping to prevent overfitting.

[0259] 4. Predictive maintenance

[0260] Based on the output probability of the fault diagnosis model, combined with the exponential smoothing prediction method, the remaining useful life (RUL) of the equipment is estimated.

[0261] When the predicted RUL is below a threshold, the system provides maintenance recommendations. For example, when the predicted RUL is less than 500 hours, the system provides maintenance recommendations.

[0262] This invention discloses a monitoring system and method for submersible electric pumps based on an adaptive multi-factor coupling model. The core of this system lies in proposing an adaptive multi-factor coupling model that considers not only electrical parameters but also the influence of fluid characteristics and the gas-oil ratio, and has the following characteristics:

[0263] (1) Multi-factor coupling: The model takes into account the combined effects of electrical parameters, fluid characteristics and gas-oil ratio, which improves the estimation accuracy;

[0264] (2) Adaptive capability: Through an innovative parameter adaptive adjustment algorithm, the model can adapt to factors such as equipment aging and changes in oil well conditions;

[0265] (3) Computational efficiency: Despite the introduction of more factors, the model still maintains a computational complexity of O(1), which is much lower than that of traditional complex model analysis.

[0266] (4) Intelligent fault diagnosis: Deep learning method is used, combined with LSTM and attention mechanism to improve the accuracy of fault identification;

[0267] (5) Low resource requirements, the system can run on edge computing devices and is suitable for on-site deployment.

[0268] This invention significantly improves the accuracy and adaptability of submersible electric pump monitoring by introducing an adaptive multi-factor coupling model and an innovative parameter adjustment algorithm; the deep learning fault diagnosis method further enhances the intelligence level of the system, enabling this monitoring system to better cope with complex and ever-changing oil well conditions and provide reliable decision support for oilfield production management.

[0269] It should be noted that the above embodiments are examples and not limitations of the present invention, and those skilled in the art will be able to design many alternative embodiments without departing from the scope of the claims of this patent.

Claims

1. A monitoring system for a submersible electric pump based on an adaptive multi-factor coupling model, comprising a submersible electric pump, characterized in that: The data acquisition unit is located at the output end of the submersible electric pump motor inverter and is used to acquire the three-phase current, voltage and frequency data of the submersible electric pump motor. Local processing unit for data preprocessing and preliminary analysis; The central control unit is used for model calculation, condition assessment, and fault diagnosis. The human-computer interaction interface is used to display monitoring results and receive operation commands; The central control unit uses an adaptive multi-factor coupling model.

2. A monitoring method for a submersible electric pump monitoring system based on an adaptive multi-factor coupling model as described in claim 1, characterized in that: The steps of the submersible electric pump monitoring method based on the adaptive multi-factor coupling model are as follows: S1. Collect the motor electrical parameters and operating condition data of the submersible electric pump through the data acquisition unit, and input the fluid characteristic values; S2, The local processing unit performs data preprocessing and anomaly detection; S3, The central control unit performs model calculations, status assessments, and fault diagnosis; S4. Display monitoring results and optimization suggestions through a human-computer interaction interface.

3. The monitoring method according to claim 2, characterized in that: The model calculation in step S3 is based on an adaptive multi-factor coupling model to calculate the motor's input power P, load torque T, and system efficiency η, as follows: Calculate the input power P. in, P -- Motor input active power, W, U -- Effective value of line voltage, V, I -- Effective value of line current, in A. --Power factor; Estimate the load torque T. T=k1(t)*P / f+k2(t)*f+k3(t)*μ(Q)*Q (2) in: T -- Estimated motor load torque, N·m f -- Inverter output frequency, Hz η -- System efficiency Q -- Flow rate of the submersible electric pump, m 3 / s, μ(Q) -- fluid viscosity function, k1(t), k2(t), k3(t) -- coefficients that adapt to time; Calculate the system efficiency η. η=[k4(t)*ρ*g*Q*H] / P*(GOR_ref / GOR)+k5(t) (3) in: ρ -- fluid density, kg / m³ 3 ; g -- acceleration due to gravity, m / s² 2 ; H - Head of the submersible electric pump, in meters; GOR -- Gas-to-oil ratio, m 3 / m 3 ; GOR_ref -- Reference gas-oil ratio, m 3 / m 3 ; k4(t) and k5(t) are time-adaptive coefficients.

4. The monitoring method according to claim 3, characterized in that: The steps of the adaptive parameter adjustment algorithm for the adaptive multi-factor coupling model are as follows: A1. Start: Algorithm begins; A2. Initialize parameters k1(0) to k5(0): Set the initial parameters using historical data or empirical values; A3. Calculate the error function E; A4. Update parameters using gradient descent; A5. Apply parameter constraints to ensure that the updated parameters are within a reasonable range; A6. Determine if convergence has occurred: If convergence is not achieved, return to step A3 and continue iterating; If convergence has been achieved, proceed to the termination step; A7. End: This round of parameter adjustments is complete.

5. The monitoring method according to claim 4, characterized in that: Step A3 calculates the error function E: E=w1(TT measured ) 2 +w2(η-η measured ) 2 +w3(PP measured ) 2 (4) in: w1, w2, and w3 are weighting coefficients; T measured η measured P measured These are the measured values ​​of T, η, and P; The values ​​of the weighting coefficients w1, w2, and w3 are all greater than 0 and less than 1; The weighting coefficients w1+w2+w3=1.

6. The monitoring method according to claim 5, characterized in that: Step A4 uses gradient descent to update the parameters: in, α -- Adaptive learning rate; λ -- Regularization parameter, dynamically adjusted using Bayesian optimization methods; i＝1，2，3，4，5。 7. The monitoring method according to claim 6, characterized in that: Step A5 employs a soft constraint method to limit the parameters within a reasonable range: 0.5ki_initial≤ki≤1.5ki_initial (i=1, 2, 3, 4, 5).

8. The monitoring method according to claim 2, characterized in that: The state evaluation in step S3 uses an adaptive threshold. T_upper=T rated *(1.2+0.1*sin(2π*t / 24)+0.05*sin(2π*t / (24*7))) (6) T_lower=T rated *(0.8-0.1*sin(2π*t / 24)-0.05*sin(2π*t / (24*7))) (7) η_lower=η rated *(0.9-0.05*sin(2π*t / 24)-0.03*sin(2π*t / (24*7))) (8) Where t -- current hour, T rated --Rated torque, N·m; H rated --Rated efficiency.

9. The monitoring method according to claim 8, characterized in that: The rules for state assessment are as follows: Normal state: T_lower≤T≤T_upper and η≥η_lower; Mild anomaly: T_lower*0.9≤T<T_lower Or T_upper < T ≤ T_upper * 1.1, Or η_lower*0.9≤η<η_lower Critical anomaly: T < T_lower * 0.9 Or T > T_upper * 1.1, Or η < ηlower * 0.

9.

10. The monitoring method according to claim 2, characterized in that: The fault diagnosis in step S3 employs a deep learning-based fault diagnosis function, and the steps are as follows: B1. Data preprocessing: Normalize and slide window segment the time series data; B2. Feature Extraction: Temporal features are extracted using a bidirectional LSTM network; B3. Attention Mechanism: Introduce a self-attention layer to enhance the weights of key time steps and features; B4. Classifier: A classifier is used for multi-class fault identification; B5. Model Training: The model employs the cross-entropy loss function and the Adam optimizer, and uses early stopping to prevent overfitting.

Images