Oil pumping unit frequency converter dynamic regulation and control device and method based on deep learning

Abstract

The invention relates to the technical field of oil pumping unit control, and provides an oil pumping unit frequency converter dynamic regulation and control device and method based on deep learning, and the device comprises a motor operation data detection device which is used for the motor operation state of an oil pumping unit, and comprises a phase voltage, a phase current and a power factor; the main control unit is configured to predict a voltage value of a frequency converter of the oil pumping unit based on an aggregation model of score and weight parameters and a deep learning model according to the detected load and the motor operation state of the oil pumping unit, and send the voltage value to the full-power adaptive regulator; and the full-power self-adaptive regulator is used for connecting the main control unit and the frequency converter, is arranged on the output side of the frequency converter, and ensures that the motor of the oil pumping unit runs in the optimal interval by self-adaptively regulating and controlling the regulation and control instruction of the frequency converter according to the received predicted voltage value. The device disclosed by the invention can realize intelligent tracking of the motor along with load change, and can realize harmonic suppression and energy feedback.

Description

Technical Field

[0001] The present disclosure relates to the technical field related to oil pumping unit control, and more specifically, to a dynamic control device and method for an oil pumping unit inverter based on deep learning. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] With the continuous development of oilfield enterprises and the improvement of energy conservation and emission reduction requirements, the motors in the existing pumping systems in oil production do not match the actual working conditions. The main technical problems are as follows:

[0004] 1) The motor speed does not match the working conditions: Due to the complexity and diversity of reservoir types, the fixed speed of the motor limits the reasonable stroke frequency of the pump, resulting in a low filling coefficient of the pump, and the motor does a lot of useless work, resulting in energy waste. For example, in the working conditions of heavy oil wells, due to the high viscosity of the oil, the actual required pumping frequency is 1-2 times / minute, but the minimum motor speed is 4 times / minute, resulting in a "half-pumping" state and serious power loss.

[0005] 2) Low motor efficiency: The fluid production of oil wells fluctuates greatly, and the existing motors cannot be adjusted according to the dynamic changes in fluid production, resulting in low motor efficiency and increased system energy consumption.

[0006] Due to the limitation of motor speed, the motor speed of oil wells is too high, the pumping frequency cannot be adjusted down, and it cannot work at a reasonable frequency, resulting in a low filling coefficient of the pump, an increase in useless work done by the motor, and serious energy waste. Taking the analysis of the working conditions of heavy oil wells as an example, first, due to the high viscosity of heavy oil wells (i.e., high resistance), the motor speed must be reduced, and due to the limitation of motor speed, its minimum speed can only be set to 4 times / minute, but the actual working condition only requires 1-2 times / minute, resulting in a "half-pumping" state of 2-3 times / minute, which causes power loss. Second, the liquid production of the oil well changes from high to low in each cycle, and the liquid production in the continuous cycle shows a large fluctuation. The matching motor cannot be dynamically adjusted with the fluctuation of liquid production, resulting in low motor efficiency.

[0007] 3) Current energy-saving research focuses on improving mechanical structures and / or selecting new motors to improve motor efficiency, but these methods are costly and difficult to apply on a large scale. Summary of the invention

[0008] In order to solve the above problems, the present invention proposes a deep learning-based dynamic control device and method for a pumping unit inverter, which can realize intelligent tracking of the motor as the load changes, and can achieve harmonic control and energy feedback.

[0009] In order to achieve the above objectives, the present disclosure adopts the following technical solutions:

[0010] One or more embodiments provide a deep learning-based dynamic control device for a pumping unit frequency converter, including:

[0011] The motor operation data detection device is used for the motor operation status of the oil pumping unit, including phase voltage, phase current and power factor;

[0012] A main control unit is configured to predict a voltage value of the frequency converter of the oil pumping unit based on the detected load and the motor operation state of the oil pumping unit, based on an aggregation model of score and weight parameters and a deep learning model, and send the prediction to a full power adaptive regulator;

[0013] The full-power adaptive regulator is used to connect the main control unit and the inverter. It is set on the output side of the inverter. According to the received predicted voltage value, it adaptively adjusts the control instructions of the inverter to ensure that the motor of the oil pump runs in the optimal range.

[0014] One or more embodiments provide a method for dynamic control of a pumping unit frequency converter based on deep learning, comprising the following steps:

[0015] Model typical electric traction equipment in oil fields, simulate the operation status of the pumping unit and analyze the energy consumption, obtain the operation law of the pumping unit and the correlation data between the motor load, current, voltage and efficiency;

[0016] The associated data is input into the trained recurrent neural network prediction model, as well as the motor operation data detected by the motor operation data detection device, to predict the optimal voltage value of the inverter.

[0017] Compared with the prior art, the present invention has the following beneficial effects:

[0018] In the present disclosure, the operation of the oil pump is dynamically controlled by combining a deep learning model and an aggregation model. First, the motor operation data detection device collects key parameters of the motor, such as phase voltage, phase current, and power factor. Then, the main control unit uses a deep learning model to analyze these data and predict the optimal voltage output of the inverter based on historical data and real-time load conditions. The full-power adaptive regulator adjusts the working instructions of the inverter according to the predicted voltage value to ensure that the motor always operates within the optimal operating range, so as to realize intelligent adjustment of the input voltage according to different loads. In this way, the device can effectively solve the problem of mismatch between motor speed and working conditions, reduce useless power consumption of the motor, and improve the energy utilization efficiency of the system.

[0019] The control device avoids energy waste caused by unreasonable motor speed by dynamically adjusting the running state of the motor; it adjusts the working state of the motor in real time to ensure that the motor can run in an efficient manner in each cycle, thereby significantly improving the efficiency of the motor. The device of this embodiment has strong adaptability and low cost; compared with improving efficiency by improving mechanical structure or using new motors, this control method based on deep learning has lower cost and is easier to apply on a large scale in existing systems, and has strong market promotion value.

[0020] The present invention completes the control of the inverter through hardware communication, realizes the intelligent tracking technology of the motor voltage that changes with the load, customizes the V / F control curve to explore the control law of the inverter, and can realize full-power voltage regulation at a fixed frequency.

[0021] The advantages of the present disclosure and the advantages of additional aspects will be described in detail in the following specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings constituting a part of the present disclosure are used to provide a further understanding of the present disclosure. The exemplary embodiments of the present disclosure and the description thereof are used to explain the present disclosure but do not constitute a limitation of the present disclosure.

[0023] Figure 1 is a structural block diagram of a dynamic control device for a pumping unit frequency converter according to Embodiment 1 of the present disclosure;

[0024] Figure 2 It is a structural diagram of the host computer system of Embodiment 1 of the present disclosure;

[0025] Figure 3 is a flow chart of a modeling method in a process of generating associated data in Embodiment 2 of the present disclosure;

[0026] Figure 4 It is a structural diagram of an LSTM neural network model based on a self-attention mechanism (Self-Attention) of Embodiment 2 of the present disclosure;

[0027] FIG5( a ) is a schematic diagram of a load power cycle variation of 0-5.5 kW under the first test condition of the test verification of Example 2 of the present disclosure;

[0028] FIG5( b ) is a schematic diagram of a second test condition of the test verification of Example 2 of the present disclosure, with the load power cycle changing from 0 to 9 kW;

[0029] Figure 6 It is the voltage change value optimized based on the dynamic control method under the first test condition of Example 2 of the present disclosure;

[0030] Figure 7is the current value optimized by the dynamic control method of Example 2 of the present disclosure under the first test conditions;

[0031] Figure 8 is the optimized input power value obtained by optimizing the dynamic control method of Example 2 under the first test conditions of Example 2 of the present disclosure;

[0032] Fig. 9 is the voltage change value after optimization based on the dynamic control method of Example 2 of Example 2 of the present disclosure under the second test conditions;

[0033] Fig.10 is the current value optimized based on the dynamic control method of Example 2 under the second test conditions of Example 2 of the present disclosure;

[0034] Fig.11 is the optimized input power value obtained by optimizing the dynamic control method of Example 2 under the second test conditions of Example 2 of the present disclosure;

[0035] Fig.12 This is a diagram of a power and electric energy numerical display interface before the energy feedback unit is added to the dynamic control device of the oil pumping unit frequency converter of Example 1 of the present disclosure;

[0036] Fig.13 This is a diagram of the power and electric energy numerical display interface after the energy feedback unit is added to the dynamic control device of the oil pump inverter of Example 1 of the present disclosure. DETAILED DESCRIPTION

[0037] The present disclosure is further described below in conjunction with the accompanying drawings and embodiments.

[0038] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present disclosure belongs.

[0039] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof. It should be noted that, in the absence of conflict, the various embodiments in the present disclosure and the features in the embodiments can be combined with each other. The embodiments will be described in detail below in conjunction with the accompanying drawings.

[0040] Example 1

[0041] In the technical solutions disclosed in one or more embodiments, Figures 1 to 13 As shown, a dynamic control device for a pumping unit frequency converter based on deep learning includes: a full-power adaptive regulator, a power factor sensor, a main control unit, a voltage sensor, a current sensor, a power quality management unit, an energy feedback unit and a monitoring system;

[0042] The motor operation data detection device includes a voltage sensor, a current sensor and a power factor sensor, which are used to detect the phase voltage, phase current and power factor to estimate the current load and the motor operation state of the oil pump;

[0043] A main control unit is configured to predict a voltage value of the frequency converter of the oil pumping unit based on the detected load and the motor operation state of the oil pumping unit, based on an aggregation model of score and weight parameters and a deep learning model, and send the prediction to a full power adaptive regulator;

[0044] The full-power adaptive regulator is used to connect the main control unit and the inverter. It is set at the output side of the inverter. According to the received predicted voltage value, it adaptively adjusts the control command of the inverter to ensure that the motor of the pumping unit runs in the optimal range.

[0045] In this embodiment, the operation of the oil pump is dynamically controlled by combining a deep learning model with an aggregation model. First, the motor operation data detection device collects key parameters of the motor, such as phase voltage, phase current, and power factor. Then, the main control unit uses a deep learning model to analyze these data and predict the optimal voltage output of the inverter based on historical data and real-time load conditions. The adaptively controlled full-power adaptive regulator adjusts the working instructions of the inverter according to the predicted voltage value to ensure that the motor always operates within the optimal operating range, so as to realize intelligent adjustment of the input voltage according to different loads. In this way, the device can effectively solve the problem of mismatch between motor speed and working conditions, reduce useless power consumption of the motor, and improve the energy utilization efficiency of the system.

[0046] By dynamically adjusting the running state of the motor, energy waste caused by unreasonable motor speed is avoided; the working state of the motor is adjusted in real time to ensure that the motor can operate in an efficient manner in each cycle, thereby significantly improving the efficiency of the motor. The device of this embodiment has strong adaptability and low cost; compared with improving efficiency by improving mechanical structure or using new motors, this deep learning-based control method has lower cost and is easier to apply on a large scale in existing systems, and has strong market promotion value.

[0047] In some embodiments, the voltage transformer and the current transformer are used to obtain voltage and current signals of the same phase;

[0048] The main control unit measures the power factor, phase voltage and phase current through voltage transformer, current transformer and power factor sensor, estimates the current load and motor operation status, predicts the voltage value that the inverter should have, and sets the inverter output predicted voltage value through 485 communication. The motor runs in the best efficiency range, and realizes automatic speed control according to the pumping unit parameter adjustment (stroke adjustment) needs.

[0049] The main control unit is configured to perform the following processes:

[0050] Step 11: Model typical electric traction equipment in oil fields, simulate the operation status of the pumping unit and analyze the energy consumption, and obtain the correlation data between the operation law of the pumping unit and the motor operation data; the motor operation data includes load, current, voltage, efficiency, etc.;

[0051] Step 12: input the associated data and the motor operation data detected by the motor operation data detection device into the trained recurrent neural network prediction model to predict the optimal voltage value of the inverter;

[0052] Optionally, the full-power adaptive regulator is a same-frequency full-power adaptive V / F voltage regulation balanced power mode; based on the received predicted voltage value, adaptive regulation obtains the control instruction of the inverter, including the following steps:

[0053] Step 21, system modeling and parameter estimation: for typical electric traction equipment in oil fields, a mathematical model for predicting the correlation between voltage value and control command of the inverter is established, and the current state and dynamic characteristics of the typical electric traction equipment in oil fields are estimated;

[0054] Optionally, the mathematical model may adopt a state space model or a nonlinear time-varying model;

[0055] Step 22: design an adaptive rule, introduce a multivariable control strategy into the gain adjustment algorithm, optimize and adjust multiple key parameters, and use a genetic algorithm or a particle swarm optimization algorithm to solve the mathematical model;

[0056] Step 23: Update the control instructions of the inverter in real time according to the solved parameters.

[0057] Furthermore, the control device also includes a power quality management unit and an energy feedback unit, which are arranged at the DC bus on the intermediate side of the inverter to improve the power factor through energy feedback;

[0058] The power quality management unit includes a filter, which is arranged at the output end of the current transformer and the voltage transformer to filter out harmonic interference.

[0059] The energy feedback unit includes a DC capacitor, a power electronic device IGBT and its drive circuit, a protection circuit and an absorption circuit; wherein the DC bus is supported by the DC capacitor, and the energy feedback unit retains the brake circuit of the original pumping unit inverter device used for the brake system, and serves as a backup protection system for the pumping unit's dedicated filter energy feedback device, and feeds back the electric energy of the inverter's braking regeneration energy to the grid side, thereby achieving energy saving and consumption reduction;

[0060] Specifically, the method for implementing energy feedback by the energy feedback unit includes the following process:

[0061] 1) When the energy of the single-well operating equipment causes the DC bus voltage to rise to the feedback threshold, the energy feedback unit is in the energy feedback state, feeding the excess energy back to the AC grid;

[0062] Specifically, the starting voltage of the feedback can be set or adjusted according to actual needs to meet the requirements of different working conditions.

[0063] 2) When the inverter is in the braking state, the DC side voltage of the inverter equipment increases, and when the DC bus voltage of the inverter rectifier exceeds the set energy feedback threshold, it automatically switches to the grid-connected power generation state, feeds back energy to the grid to keep the bus voltage constant to the set value, and feeds back the electric energy to the grid side through the energy feedback unit to achieve the purpose of electric energy feedback. The feedback action priority is higher than the action value of the inverter braking circuit;

[0064] The energy feedback unit of this embodiment is expected to improve the power factor PF from 0.83 to 0.97 through the implementation of energy feedback technology. Fig.12 and Fig.13 , Fig.12 , Fig.13 They are the power and energy before and after adding the energy feedback unit. By comparing the two figures, the power factor PF is increased from 0.83 to 0.97. When the actual pump is working normally, after the filter energy feedback unit is started, the maximum feedback power is 2kW, and the feedback time is about 2 / 25 cycles. 2×2 / 25×24×365=1401.6kWh. When the energy feedback unit is running, harmonics are eliminated to reduce device losses. It can be seen from the feedback power that the feedback power is about 400w. 400×24 / 25×24×365=3363.8kWh.

[0065] A further technical solution also includes a monitoring system, the main control unit is in communication connection with the monitoring system, and the monitoring system includes an upper computer system and a lower computer system;

[0066] The host computer system is used to obtain and display the real-time status of energy consumption of each device in the well area; the host computer system includes a single device operation database, an oilfield production management database, and a single device energy consumption management database, and inputs real-time data (i.e., real-time status data of energy consumption) into the single device operation database;

[0067] The lower computer system is used to obtain the current, voltage and power parameters of a single inverter;

[0068] 1) Single device operation database: This database is used to store the real-time operation data of each device, including but not limited to key parameters such as current, voltage, power, and speed. These data are collected in real time by the lower computer system and transmitted to the upper computer system to ensure that the status of each device during operation can be reflected;

[0069] The data content of key parameters may include: equipment identification, timestamp, operating parameters (such as current, voltage, power, etc.), alarm information, equipment health status, etc. The database can be used to query and analyze historical data, monitor equipment performance trends, identify potential problems, and perform preventive maintenance.

[0070] 2) Oilfield production management database: used to store the operating data and production management information of all relevant equipment in the oilfield production process; it integrates the operating status, production parameters, process flow and other information of each equipment in the well area, forming the core data source for the overall production management of the oilfield;

[0071] Specifically, the data content may include the production status of each equipment in the well area, equipment operating time, production efficiency, output, equipment failure records, maintenance history, personnel operation records, etc.; through data aggregation and analysis, managers can optimize production processes, improve production efficiency, and make global decisions.

[0072] 2) Single device energy consumption management database: used to record and manage the energy consumption data of each device, including total energy consumption, energy efficiency, and energy cost. By collecting parameters such as current, voltage, and power of a single device, the database can calculate the energy consumption of the device, help managers understand the energy usage of the device, and identify high-energy-consuming devices.

[0073] Specifically, the data content may include equipment identification, energy consumption data (such as total daily, weekly, and monthly energy consumption), energy efficiency (such as energy consumption per unit of output), cost data, energy consumption trend analysis, etc.; energy consumption reports can be generated based on the data in the database to support energy-saving transformation and energy management decisions.

[0074] In this embodiment, a host computer and a slave computer system should be designed and developed. The host computer can collect and display the real-time status of energy consumption of each traction equipment in the well area. The slave computer system is connected to a single inverter to collect the current, voltage, and power of the single inverter, and set a signal sampling point on the traction equipment to collect the motion conditions of the traction equipment through a motion sensor, enter the real-time data into the oil well production data management system, and then upload it to the oil well equipment energy consumption monitoring system database.

[0075] Example 2

[0076] Based on Example 1, this embodiment provides a dynamic control method for a pumping unit frequency converter based on deep learning, which can be executed in a main control unit, such as Figures 3 to 11 As shown, the following steps are included:

[0077] Step 1: Model typical electric traction equipment in oil fields, simulate the operation status of the pumping unit and analyze the energy consumption, and obtain the operation law of the pumping unit and the correlation data between the motor load, current, voltage and efficiency;

[0078] Step 2: input the associated data into the trained recurrent neural network prediction model, as well as the motor operation data detected by the motor operation data detection device, to predict the optimal voltage value of the inverter;

[0079] Among them, the motor operation data includes power factor, phase voltage, and phase current measured by sensors and the main control unit.

[0080] Furthermore, the output voltage of the inverter is set through 485 communication to ensure that the motor operates in the optimal efficiency range.

[0081] In step 1, typical electric traction equipment includes motor systems for driving various mechanical equipment, including but not limited to pumping units, oil pumps, compressors and drilling equipment; these equipment play a key role in oil field production, mainly used to transmit mechanical energy to different working units in the oil field, thereby realizing the extraction, processing and transportation of oil;

[0082] In step 1, CAE software such as ADAMS and MAXWELL can be used to model typical electric traction equipment in oil fields, simulate the operating status of the pumping unit and analyze the energy consumption, and obtain the movement law of the pumping unit and the correlation data between the motor load, current, voltage, and efficiency; these correlation data are of great guiding significance for optimizing the operating performance of the equipment, improving energy utilization efficiency, and reducing energy consumption.

[0083] ADAMS software, whose full name is Automatic Dynamic Analysis of Mechanical Systems, can be used to simulate the mechanical movement of the oil pump and analyze its movement laws and dynamic characteristics. MAXWELL software focuses on the simulation of electromagnetic fields and can accurately calculate the electromagnetic characteristics of the motor, thereby obtaining the dynamic changes of its load, current and voltage.

[0084] MAXWELL is an electromagnetic field simulation software launched by ANSYS, which is mainly used for the design and analysis of motors, power transformers, inductors and other electromagnetic equipment;

[0085] Specifically, Figure 3 As shown in the figure, the modeling method in the generation process of associated data is:

[0086] Step 1.1, construct an initial aggregation model, which includes local nodes and aggregation nodes, and takes typical electric traction equipment in each oil field as local nodes;

[0087] Step 1.2, detect the local node and transmit the detected motor operation data to the aggregation node;

[0088] Step 1.3, the aggregation node calculates the parameters and weights of each local node; the aggregation parameters are aggregated to obtain an aggregation model based on the score and weight parameters, and the aggregation model is sent to each detection node to determine whether the data is optimized data. If so, the voltage, current, power and other related data are obtained; if not, step 1.2 is executed to continue the cycle until the optimized data is satisfied;

[0089] Among them, the detection data includes its armature reaction reactance parameters, etc. By changing the number of rotors and comparing the parameter sizes of the rotors under different conditions, the optimal situation of the rotor under different conditions can be obtained; the armature reaction reactance parameters are calculated based on the phase voltage, phase current and power factor;

[0090] Optionally, in the process of generating the associated data, for each detection node, a method of determining whether the data is optimized is used, and a simulation method using two simulation softwares, ADAMS and MAXWELL, is used. The specific steps are as follows:

[0091] Step 11: Use ADAMS to simulate the input detection data to obtain mechanical load data, including torque, speed, position, etc. The mechanical load data is imported into MAXWELL as input conditions to form a linkage simulation;

[0092] This means that the load on the motor in MAXWELL is consistent with the actual mechanical operating state of the pump. This load changes over time and can reflect the actual operating conditions of the pump under different working conditions.

[0093] Step 12: Motor performance feedback: Electromagnetic data generated by MAXWELL simulation, and whether the current operating state is the optimal state according to the electromagnetic data;

[0094] Electromagnetic data includes motor torque, speed, current and voltage, etc. Feeding electromagnetic data back to ADAMS simulation can further adjust the operating status of the mechanical model. Through this two-way data interaction, researchers can coordinately optimize the motor-mechanical system of the pumping unit to ensure that the motor output can always meet the needs of the mechanical system.

[0095] In this embodiment, by combining the simulation results of ADAMS and MAXWELL, researchers can deeply analyze the overall performance of the pumping unit under different working conditions, including the motion stability of the mechanical system, the energy efficiency of the motor, electromagnetic loss, etc. This comprehensive analysis not only helps to identify the weak links of the system, but also provides a theoretical basis for design improvement and energy-saving optimization.

[0096] In the above scheme, through the organic combination of two computer-aided software, researchers can fully understand the working characteristics of the pumping unit and optimize its design and operating parameters to improve the overall efficiency of the equipment and reduce energy consumption.

[0097] Furthermore, the recurrent neural network prediction model adopts an LSTM neural network model based on a self-attention mechanism. The LSTM neural network model based on a self-attention mechanism includes an input layer, an LSTM network layer, a self-attention layer, and an output layer connected in sequence.

[0098] Input layer, used to obtain input data x1, x2, x3…xn;

[0099] The LSTM network layer is used to predict the voltage value of the inverter of the motor according to the acquired motor operation status;

[0100] The self-attention layer is used to adjust the attention weight and the importance of the output data of the LSTM network layer;

[0101] Output layer, used to output result data y1, y2, y3…yn;

[0102] The process of building an LSTM neural network model based on the self-attention mechanism includes: determining the input and output and the number of neurons, determining the optimization algorithm, and determining the activation function.

[0103] Optionally, the optimization algorithm can use the adaptive parameter Adam algorithm, which introduces first-order momentum and second-order momentum based on the gradient descent algorithm. The learning rate can be adjusted adaptively, which also effectively avoids the turbulence problem in the optimization process. The Adam algorithm is applied in the recurrent neural network to avoid artificially setting an inappropriate learning rate, which affects the prediction results of the model.

[0104] The activation function of the output layer is set to the tanh function.

[0105] This embodiment uses hardware communication to complete the system's control of the inverter, implements intelligent tracking technology for motor voltage that changes with load, customizes V / F control curves to explore inverter control rules, and implements full-power voltage regulation at a fixed frequency; manual parameter adjustment can be achieved by key input model;

[0106] In addition, a monitoring system including a host computer system and a slave computer system was designed and developed. The host computer can collect and display the real-time status of energy consumption of each traction equipment in the well area. The slave computer system is connected to a single frequency converter to collect the current, voltage, and power of the single frequency converter, and sets signal sampling points on the traction equipment to collect the motion conditions of the traction equipment through motion sensors. The real-time data is entered into the oil well production data management system and then uploaded to the oil well equipment energy consumption monitoring system database.

[0107] Through the dynamic control method of this embodiment, by dynamically outputting the control instructions of the frequency converter in real time according to the operating status of the motor and the corresponding load, the comprehensive power saving rate of the oil pump motor can be improved by more than 5%; the system tracking dynamic adjustment time is less than 10 minutes; the harmonic filtering rate is ≥95%; the feedback current distortion rate: ≤3% (rated output), the feedback power is 2kW; and the full power voltage regulation of the oil pump can be achieved at a fixed frequency.

[0108] In order to verify the dynamic control device and method of the oil pumping unit inverter based on deep learning, experimental verification was carried out.

[0109] 1. System control motor test

[0110] The test conditions are 60s load cycle, 15s loading, 15s holding, 15s unloading, 15s holding. As shown in Figure 5(a), the first test is a 0-5.5kW load power cycle. As shown in Figure 5(b), the second test is a 0-9kW load power cycle. The horizontal axis of Figure 5 is time s, and the vertical axis is load power Kw.

[0111] The test condition is that the inverter frequency is 50Hz, and the actual measured load range is 900W-5.5kW. The current change value, voltage value and input power value after optimization by the motor intelligent control system are as follows: Figure 6 , Figure 7 , Figure 8 As shown. Among them, Figure 6 The horizontal axis is power W, and the vertical axis is current A; Figure 7 The horizontal axis is power W, and the vertical axis is voltage A; Figure 8 The horizontal axis is load power W, and the vertical axis is input power W.

[0112] The test conditions are as follows: the inverter frequency is 50Hz, the load range is 200W-2kW, and the current change value, voltage value and input power value after optimization by the motor intelligent control system are as follows: Fig. 9 , Fig.10 , Fig.11 shown.

[0113] Two 200W-2kW cycle load tests obtained the corresponding optimal voltage values, which were 218V and 246V respectively, as shown in Table 1.

[0114] Table 1 Optimal voltage values ​​corresponding to periodic loads

[0115] Load power 200w-2kw Optimum voltage (V) 218

[0116] Optimizing performance under test cycle load conditions:

[0117] The test and test conditions are as follows: the load cycle is 60s, loading is 15s, holding is 15s, unloading is 15s, holding is 15s. The load power is 200W-2kW in sequence, and the optimized power change value is obtained through the optimization of the motor intelligent control system.

[0118] Table 2 Frequency conversion energy saving efficiency table

[0119]

[0120]

[0121] The above test data can verify that the device can realize the intelligent tracking technology of motor voltage that changes with load, and at the same time verify the correctness of the neural network algorithm. The algorithm data is basically consistent with the test data.

[0122] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. For those skilled in the art, the present disclosure may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

[0123] Although the above describes the specific implementation methods of the present disclosure in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present disclosure. Technical personnel in the relevant field should understand that on the basis of the technical solution of the present disclosure, various modifications or variations that can be made by those skilled in the art without creative work are still within the scope of protection of the present disclosure.

Claims

1. A dynamic control device for a pumping unit frequency converter based on deep learning, characterized in that: include: The motor operation data detection device is used for the motor operation status of the oil pumping unit, including phase voltage, phase current and power factor; A main control unit is configured to predict a voltage value of the frequency converter of the oil pumping unit based on the detected load and the motor operation state of the oil pumping unit, based on an aggregation model of score and weight parameters and a deep learning model, and send the prediction to a full power adaptive regulator; The full-power adaptive regulator is used to connect the main control unit and the inverter. It is set on the output side of the inverter. According to the received predicted voltage value, it adaptively adjusts the control instructions of the inverter to ensure that the motor of the oil pump operates in the optimal range.

2. The deep learning-based dynamic control device for pumping unit frequency converter according to claim 1, characterized in that: The main control unit is configured to perform the following processes: Model typical electric traction equipment in oil fields, simulate the operation status of the pumping unit and analyze the energy consumption, and obtain the correlation data between the operation law of the pumping unit and the motor operation data; The associated data, as well as the motor operation data detected by the motor operation data detection device, are input into the trained recurrent neural network prediction model to predict the optimal voltage value of the inverter.

3. The deep learning-based dynamic control device for pumping unit frequency converter according to claim 1, characterized in that: The full-power adaptive regulator is a full-power adaptive V / F voltage regulation balanced power mode with the same frequency. Based on the received predicted voltage value, the adaptive regulation obtains the control instruction of the inverter, including the following steps: For typical electric traction equipment in oil fields, a mathematical model is established to predict the correlation between voltage value and inverter control command, and the current state and dynamic characteristics of typical electric traction equipment in oil fields are estimated; Design adaptive rules, introduce multivariable control strategies into the gain adjustment algorithm, optimize and adjust multiple key parameters, and use genetic algorithms or particle swarm optimization algorithms to solve mathematical models; The control instructions of the inverter are updated in real time according to the solved parameters.

4. The deep learning-based dynamic control device for pumping unit frequency converter according to claim 1, characterized in that: The control device also includes a power quality management unit and an energy feedback unit, which are arranged at the DC bus on the intermediate side of the inverter to improve the power factor through energy feedback; The power quality management unit includes a filter, which is arranged at the output end of the current transformer and the voltage transformer of the motor operation data detection device to filter out harmonic interference.

5. The deep learning-based dynamic control device for pumping unit frequency converter according to claim 4, characterized in that: The energy feedback unit is configured to perform the following process: When the energy of the single-well operating equipment causes the DC bus voltage to rise to the feedback threshold, the energy feedback unit is in the energy feedback state, feeding the excess energy back to the AC grid; When the inverter is in the braking state, the DC side voltage of the inverter equipment increases, and when the DC bus voltage of the inverter rectifier part exceeds the set energy feedback threshold, it automatically switches to the grid-connected power generation state, feeds energy back to the grid to keep the bus voltage constant to the set value, and feeds electric energy back to the grid side through the energy feedback unit.

6. The deep learning-based dynamic control device for pumping unit frequency converter according to claim 1, characterized in that: It also includes a monitoring system, the main control unit is connected to the monitoring system in communication, and the monitoring system includes an upper computer system and a lower computer system; The host computer system is used to obtain and display the real-time status of energy consumption of each device in the well area; the host computer system includes a single device operation database, an oilfield production management database, and a single device energy consumption management database, and inputs real-time data into the single device operation database.

7. A dynamic control method for a pumping unit frequency converter based on deep learning, characterized in that: The steps include: Model typical electric traction equipment in oil fields, simulate the operation status of the pumping unit and analyze the energy consumption, obtain the operation law of the pumping unit and the correlation data between the motor load, current, voltage and efficiency; The associated data, as well as the motor operation data detected by the motor operation data detection device, are input into the trained recurrent neural network prediction model to predict the optimal voltage value of the inverter.

8. The deep learning-based dynamic control method for pumping unit frequency converter according to claim 7, characterized in that: The modeling method in the process of generating associated data is: Step 1.1, construct an initial aggregation model, which includes local nodes and aggregation nodes, and takes typical electric traction equipment in each oil field as local nodes; Step 1.2, detect the local node and transmit the detected motor operation data to the aggregation node; Step 1.3, the aggregation node calculates the parameters and weights of each local node; The aggregation parameters are aggregated to obtain an aggregation model based on the score and weight parameters, and the aggregation model is sent to each detection node to determine whether the data is optimized data. If so, the associated data of voltage, current, and power are obtained; if not, step 1.2 is executed to continue the cycle until the optimized data is met.

9. The method for dynamic control of a pumping unit frequency converter based on deep learning according to claim 7, characterized in that: In the process of generating associated data, for each detection node, a method of judging whether it is an optimized data is used. The simulation method of two simulation softwares, ADAMS and MAXWELL, is used. The specific steps are as follows: The input test data is simulated through ADAMS to obtain mechanical load data, and the mechanical load data is imported into MAXWELL as input conditions; The electromagnetic data generated by MAXWELL simulation is used to determine whether the current operating state is the optimal state.

10. The method for dynamic control of a pumping unit frequency converter based on deep learning according to claim 7, characterized in that: The recurrent neural network prediction model adopts an LSTM neural network model based on the self-attention mechanism. The LSTM neural network model based on the self-attention mechanism includes an input layer, an LSTM network layer, multiple parallel self-attention layers and an output layer connected in sequence.

Images