Control valve opening prediction model training prediction method, device, equipment and system
By performing feature extraction and neural network model training on the sound wave signal of fluid flowing through the control valve, real-time detection of the opening degree of the control valve is achieved, solving the problem that the opening status of the control valve cannot be accurately detected in the existing technology, and improving the underground production Management level and control accuracy.
Patent Information
- Application Number
- CN202210571532.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-05-24
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2042-05-24
AI Technical Summary
The prior art cannot detect the opening state during the control valve adjustment process in real time and accurately, resulting in the opening degree of the control valve during the downhole production process not meeting expectations, affecting production efficiency and accuracy.
By obtaining the sound wave signal of fluid flowing through the control valve, extracting time and frequency domain characteristics, building a neural network model, and real-time prediction of the control valve opening degree, achieving accurate detection of the control valve opening degree.
It improves the management level and control accuracy of the intelligent well production system, ensures real-time and accurate detection of the opening degree of the control valve, and enhances the stability of underground production.
Smart Images

Figure CN115018138B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of oil and gas exploitation, and in particular to a control valve opening degree prediction model training prediction method, device, equipment and system. Background Art
[0002] The control valve is one of the core equipment of the intelligent well production system for oil and gas extraction. Its main function is to change the area of fluid passing between the downhole oil pipe and the annulus, thereby controlling the flow state of the downhole fluid.
[0003] At present, the opening degree of the control valve at the end of the adjustment operation is usually determined by the ground pressure time, pressure size and position sensor reading parameters during the control valve adjustment process. Due to the complex downhole working conditions during the actual operation, the control valve may be blocked, scaled on the surface, etc., which may cause the actual opening degree of the control valve to fall short of expectations. However, the existing methods cannot detect the opening state of the control valve in real time and accurately during the adjustment process. Summary of the invention
[0004] In view of the above problems, the present invention provides a control valve opening degree prediction model training prediction method, device, equipment and system, which can detect the opening state of the control valve during the adjustment process in real time.
[0005] According to one aspect of the present invention, a control valve opening degree prediction model training method is provided, comprising: obtaining a sample acoustic wave signal marked with a control valve opening degree label, the sample acoustic wave signal being an acoustic wave signal of a fluid flowing through a control valve; extracting sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling of the sample acoustic wave signal; inputting a feature matrix composed of sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling into a neural network for training, thereby obtaining a control valve opening degree prediction model.
[0006] In the control valve opening prediction model training method provided by the embodiment of the present invention, the collected acoustic wave signal of the fluid flowing through the control valve is used as the original data for analysis and processing. Specifically, by obtaining a sample acoustic wave signal marked with a control valve opening label, and extracting the time domain features and frequency domain features of the sample acoustic wave signal, the time domain features are specifically sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient and sample zero-crossing rate, and the frequency domain features are specifically sample spectrum flow and sample spectrum rolling, so as to facilitate subsequent neural network analysis and identification, by inputting the feature matrix composed of the extracted sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling with the control valve opening label into the neural network for training, a control valve opening prediction model is obtained, and then when the control valve opening prediction model is applied to the actual intelligent well production system, by real-time collection of the acoustic wave signal of the fluid flowing through the control valve, the opening of the control valve can be detected in real time and accurately, which is conducive to improving the management level of the intelligent well production system.
[0007] In an optional manner, after obtaining a sample acoustic wave signal marked with a control valve opening degree label, the method further includes: performing energy normalization processing on the sample acoustic wave signal to obtain a normalized result of the sample acoustic wave signal; extracting sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling of the sample acoustic wave signal, including: extracting sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling of the normalized result of the sample acoustic wave signal. Since the amplitudes of the sample acoustic wave signals obtained under different conditions vary greatly, the sample acoustic wave signals are energy normalized and the sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling of the normalized results of the sample acoustic wave signals are extracted. This can reduce the dispersion problem caused by the amplitude difference in subsequent data processing, ensure the reliability of the training data during the subsequent training of the control valve opening prediction model, and thereby improve the accuracy of the prediction results when the control valve opening prediction model is applied.
[0008] In an optional manner, the sample acoustic wave signal is obtained through a sample acquisition device, the sample acquisition device includes a control valve, a control device and an acoustic wave detection device, the control valve is used to adjust the flow of the fluid, the control device is used to control the opening degree of the control valve, and the acoustic wave detection device is used to detect the acoustic wave signal of the fluid flowing through the control valve; obtaining a sample acoustic wave signal marked with a control valve opening degree label, including: setting the opening degree: setting the control valve opening degree to a first opening degree through the control device; obtaining a label: obtaining the first opening degree through the control device; obtaining training data: obtaining the first sample acoustic wave signal when the control valve opening degree is at the first opening degree through the acoustic wave detection device; training data labeling: labeling the obtained first opening degree as the label of the first sample acoustic wave signal; traversing other opening degrees: repeating setting the opening degree, obtaining a label, obtaining training data and training data labeling until traversing all opening degree nodes. Through the above method, sample acoustic wave signals marked with different opening degree labels can be obtained, simulating different opening degree conditions in the well, increasing the diversity of training data, and thereby improving the training effect of the control valve opening degree prediction model.
[0009] In an optional manner, a circulation pipeline is connected between the two ends of the control valve, and a constant flow pump device is arranged on the circulation pipeline; before setting the opening degree, the method further includes: setting the displacement: setting the displacement of the fluid flowing into the control valve from the circulation pipeline to the first displacement by the constant flow pump device; training data labeling, including: obtaining the first displacement, labeling the obtained first displacement and the first opening degree as the label of the first sample acoustic wave signal; after traversing other opening degrees, the method further includes: traversing other displacements: repeating setting the displacement, setting the opening degree, obtaining the label, obtaining the training data, labeling the training data and traversing other opening degrees until traversing all the displacement nodes. Through the above method, sample acoustic wave signals labeled with labels corresponding to different opening degrees for each displacement node can be obtained, simulating multiple displacements and multiple opening degree conditions in the well, further increasing the diversity of training data, and improving the training effect of the control valve opening degree prediction model.
[0010] In an optional manner, a back pressure control device is provided at a position upstream of the constant flow pump device on the circulation pipeline; before setting the displacement, the method further includes: setting the back pressure: setting the back pressure of the fluid flowing into the circulation pipeline from the control valve to the first back pressure through the back pressure control device; training data labeling, including: obtaining the first back pressure, labeling the obtained first back pressure, first displacement and first opening degree as the label of the first sample acoustic wave signal; after traversing other displacements, the method further includes: traversing other back pressures: repeatedly setting the back pressure, setting the displacement, setting the opening degree, obtaining the label, obtaining the training data, labeling the training data, traversing other opening degrees and traversing other displacements until all the back pressure nodes are traversed. Through the above method, a sample acoustic wave signal labeled with the label of each displacement node and each opening degree node corresponding to each back pressure node can be obtained, simulating multiple back pressures, multiple displacements and multiple opening degree conditions in the well, further increasing the diversity of the training data, and improving the training effect of the control valve opening degree prediction model.
[0011] In an optional manner, a heat storage device is provided between the constant flow pump device and the back pressure control device on the circulation pipeline, and the heat storage device is used to store fluid; before setting the back pressure, the method further includes: setting the temperature: setting the temperature of the fluid input into the control valve from the circulation pipeline to a first temperature through the heat storage device; training data labeling, including: obtaining the first temperature, and labeling the obtained first temperature, first back pressure, first displacement and first opening degree as the label of the first sample acoustic wave signal; after traversing other back pressures, the method further includes: traversing other temperatures: repeatedly setting back pressure, setting displacement, setting opening degree, obtaining labels, obtaining training data, training data labeling, traversing other opening degrees, traversing other displacements and traversing other back pressures until all temperature nodes are traversed. Through the above method, sample acoustic wave signals labeled with labels corresponding to each back pressure node, each displacement node and each opening degree node of each temperature node can be obtained, simulating multiple temperatures, multiple back pressures, multiple displacements and multiple opening degree conditions in the well, further increasing the diversity of training data, and improving the training effect of the control valve opening degree prediction model.
[0012] According to another aspect of the present invention, a control valve opening degree prediction method is provided, comprising: acquiring an acoustic wave signal; extracting energy entropy, signal average amplitude, short-time energy, short-time energy autocorrelation coefficient, zero-crossing rate, spectrum flow and spectrum rolling of the acoustic wave signal; inputting the energy entropy, signal average amplitude, short-time energy, short-time energy autocorrelation coefficient, zero-crossing rate, spectrum flow and spectrum rolling into a control valve opening degree prediction model to obtain a predicted control valve opening degree.
[0013] In the control valve opening degree prediction method provided in an embodiment of the present invention, the acoustic wave signal at the control valve channel collected during the actual production operation of the smart well is analyzed and processed through the control valve opening degree prediction model, so that the opening degree of the control valve can be detected in real time and accurately, thereby improving the control accuracy of downhole production and enhancing the management level of the smart well production system.
[0014] According to another aspect of the present invention, a control valve opening degree prediction model training device is provided, comprising: an acquisition unit, used to acquire a sample acoustic wave signal marked with a control valve opening degree label, the sample acoustic wave signal being an acoustic wave signal of a fluid flowing through a control valve; an extraction unit, used to extract sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling of the sample acoustic wave signal; a training unit, used to input a feature matrix consisting of sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling into a neural network for training, so as to obtain a control valve opening degree prediction model.
[0015] According to another aspect of the present invention, there is provided a control valve opening degree prediction model training device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; the memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute the operation of the control valve opening degree prediction model training method as described in any one of the above items.
[0016] According to another aspect of the present invention, a control valve opening degree prediction model training system is provided, including a control valve opening degree prediction model training device and a sample collection device, the sample collection device including: a casing; an oil pipe, which is arranged in the casing, and the top of the oil pipe is connected to the bottom of the casing through a circulation pipeline; a control valve, which is arranged on the oil pipe, and the control valve has a valve hole connecting the inside of the oil pipe and the inside of the casing, and a sliding sleeve for covering the valve hole is slidably arranged on the control valve, and the fluid enters the casing, the control valve, and the oil pipe in turn, and then enters the casing again through the circulation pipeline to form a circulation flow; a control device, which is connected to the control valve, and the control valve is connected to the oil pipe. The control valve is electrically connected to the outside of the sleeve, and the control device is used to control the opening degree of the valve hole by sliding the control sleeve; the optical cable is wound around the outer wall of the valve hole on the control valve; the acoustic wave detection device is connected to the optical cable on the outside of the sleeve, and is used to detect the acoustic wave signal of the fluid flowing through the valve hole through the optical cable; the control valve opening degree prediction model training equipment is electrically connected to the control device and the acoustic wave detection device respectively, and the control valve opening degree prediction model training equipment is used to obtain the opening degree of the control valve through the control device, and is also used to obtain the sample acoustic wave signal of the fluid flowing through the valve hole when the control valve is at the opening degree through the acoustic wave detection device.
[0017] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:
[0019] Figure 1 A flow chart of a control valve opening degree prediction model training method provided by an embodiment of the present invention;
[0020] Figure 2 A flowchart of a control valve opening degree prediction model training method provided by another embodiment of the present invention;
[0021] Figure 3 for Figure 1 or Figure 2 Flow chart of sub-steps of step S110;
[0022] Figure 4 A flow chart of a control valve opening degree prediction method provided by an embodiment of the present invention;
[0023] Figure 5 A schematic diagram of the structure of a control valve opening degree prediction model training device provided by an embodiment of the present invention;
[0024] Figure 6 A schematic diagram of the structure of a control valve opening degree prediction model training device provided by an embodiment of the present invention;
[0025] Figure 7 A schematic diagram of the structure of a control valve opening degree prediction model training system provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0026] The following embodiments of the technical solution of the present invention are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are therefore only used as examples, and cannot be used to limit the protection scope of the present invention.
[0027] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which the present invention belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention; the terms "including" and "having" in the specification and claims of the present invention and the above-mentioned figure descriptions and any variations thereof are intended to cover non-exclusive inclusions.
[0028] In the description of the embodiments of the present invention, technical terms such as "first" and "second" are only used to distinguish different objects and cannot be understood as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features.
[0029] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0030] In the description of the embodiments of the present invention, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).
[0031] In recent years, in the field of oil and gas extraction, intelligent well production systems have received more and more attention. Compared with traditional downhole production systems, intelligent well production systems are equipped with monitoring and completion equipment such as packers, sensors, and control valves downhole, which can realize remote production control, thereby optimizing production and improving recovery rates. They are often used in offshore horizontal wells and multi-branch well production.
[0032] Control valves, such as downhole flow control valves, are one of the core equipment of the intelligent well production system. The opening degree of the control valve is adjusted downhole by ground hydraulic or electrical means, thereby achieving flow control of the downhole production layer. The main function of the control valve is to change the fluid passage area between the oil pipe and the annulus, thereby controlling the flow state of the downhole fluid. During the downhole flow control valve adjustment process, the opening degree of the downhole flow control valve at the end of the adjustment is determined by the ground pressure time, pressure size and position sensor reading parameters.
[0033] The inventor of this application noticed that due to the complex actual working conditions downhole, the downhole flow control valve has channel blockage, sand in the channel, scaling on the valve body surface, etc., which causes the valve core movement resistance to increase, resulting in problems such as the control valve opening not meeting expectations and control valve control signal failure. In addition, after the control valve is adjusted, the downhole pressure fluctuation is likely to cause the control valve to drift in the later production process and affect the opening. The current method can only determine the opening after the control valve is adjusted, and cannot detect the opening in real time and accurately during the control valve adjustment process.
[0034] Based on this, an embodiment of the present invention provides a control valve opening prediction model training method on the one hand, in which the collected acoustic wave signal of the fluid flowing through the control valve is used as the original data for analysis and processing. Specifically, by obtaining a sample acoustic wave signal marked with a control valve opening label, and extracting the time domain features and frequency domain features of the sample acoustic wave signal, the time domain features are specifically sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient and sample zero-crossing rate, and the frequency domain features are specifically sample spectrum flow and sample spectrum rolling, so as to facilitate subsequent neural network analysis and identification, by inputting the feature matrix composed of the extracted sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling with the control valve opening label into the neural network for training, a control valve opening prediction model is obtained, and then when the control valve opening prediction model is applied to the actual intelligent well production system, by real-time collection of the acoustic wave signal of the fluid flowing through the control valve, the opening of the control valve can be detected in real time and accurately, which is conducive to improving the management level of the intelligent well production system.
[0035] See also Figure 1 , the figure shows the process of a control valve opening degree prediction model training method provided by an embodiment of the present invention, the method is executed by a computing device that needs to train the control valve opening degree prediction model, such as a mobile phone, a computer, a server, etc. As shown in the figure, the method includes:
[0036] S110: Acquire a sample sound wave signal marked with a control valve opening degree label, where the sample sound wave signal is a sound wave signal of fluid flowing through the control valve.
[0037] In this step, the acoustic wave signal of the fluid flowing through the control valve can be detected and acquired by an acoustic wave sensor.
[0038] S130: extracting sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling edge of the sample sound wave signal.
[0039] In this step, sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient and sample zero-crossing rate are the time domain characteristics of the sample acoustic wave signal, and sample spectrum flow and sample spectrum rolling edge are the frequency domain characteristics of the sample acoustic wave signal. By extracting these characteristics of the sample acoustic wave signal, it helps the subsequent neural network to accurately and reliably analyze and identify the sample acoustic wave signal in the time domain and frequency domain, thereby ensuring the accuracy of the control valve opening prediction model predicting data.
[0040] It should be noted that in this step, the sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling edge extracted from the sample sound wave signal all have the control valve opening degree marked on the sample sound wave signal.
[0041] S150: Inputting a feature matrix consisting of sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling edge into a neural network for training to obtain a control valve opening degree prediction model.
[0042] In this step, a BP artificial neural network algorithm tool module is used to build a learning network model. The network structure includes an input layer, a hidden layer, and an output layer, wherein the number of neurons in the input layer is equal to the number of features extracted from the sample acoustic wave signal, that is, the input layer contains 7 neurons, and the output layer contains 1 neuron, which is used to output the predicted control valve opening degree result. The number of hidden layers and the number of neurons in the hidden layer can be optimized and adjusted according to the training effect during the training process. The model effect is evaluated by the loss function, and the loss function can use the mean square error loss function, the mean absolute error loss function, the quantile loss function, the cross entropy loss function, the Hinge loss function, etc.
[0043] In the control valve opening prediction model training method provided by the embodiment of the present invention, the collected acoustic wave signal of the fluid flowing through the control valve is used as the original data for analysis and processing. Specifically, by obtaining a sample acoustic wave signal marked with a control valve opening label, and extracting the time domain features and frequency domain features of the sample acoustic wave signal, the time domain features are specifically sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient and sample zero-crossing rate, and the frequency domain features are specifically sample spectrum flow and sample spectrum rolling, so as to facilitate subsequent neural network analysis and identification, by inputting the feature matrix composed of the extracted sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling with the control valve opening label into the neural network for training, a control valve opening prediction model is obtained, and then when the control valve opening prediction model is applied to the actual intelligent well production system, by real-time collection of the acoustic wave signal of the fluid flowing through the control valve, the opening of the control valve can be detected in real time and accurately, which is conducive to improving the management level of the intelligent well production system.
[0044] In order to improve the accuracy of the prediction results of the control valve opening degree prediction model, the present invention further proposes an implementation method, please refer to Figure 2 , the figure shows the steps of a control valve opening degree prediction model training method provided by another embodiment. As shown in the figure, after the above step S110, the method further includes:
[0045] S120: performing energy normalization processing on the sample sound wave signal to obtain a normalization result of the sample sound wave signal.
[0046] In this step, the formula for energy normalization of the sample sound wave signal is:
[0047]
[0048] Among them, X is the sample sound wave signal, is the average value of the acoustic wave signal, and X' is the normalized result of the sample acoustic wave signal.
[0049] In a preferred embodiment, the normalized result of the sample acoustic wave signal is divided into a training set and a verification set in a ratio of 7:3. The training set is used to input into the neural network for training. After the training is completed, the verification set is substituted into the optimized neural network, and the effect of the model prediction is verified by a loss function, such as a mean square error loss function. The verification accuracy represents the error between the model training result and the actual result. When the verification accuracy meets the requirements, that is, the error between the training result and the actual result is less than or equal to the preset threshold, the control valve opening degree prediction model training is completed and can be used for prediction. When the verification accuracy does not meet the requirements, the number of hidden layers in the neural network and the number of neurons contained in the hidden layer are adjusted accordingly according to the training results, and retraining and verification are performed until the verification accuracy meets the requirements.
[0050] The above step 130 includes:
[0051] S131: extracting sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling edge of the normalized result of the sample sound wave signal.
[0052] Since the amplitudes of the sample acoustic wave signals obtained under different conditions vary greatly, the sample acoustic wave signals are energy normalized and the sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling of the normalized results of the sample acoustic wave signals are extracted. This can reduce the dispersion problem caused by the amplitude difference in subsequent data processing, ensure the reliability of the training data during the subsequent training of the control valve opening prediction model, and thereby improve the accuracy of the prediction results when the control valve opening prediction model is applied.
[0053] Regarding the above step 110, the present invention further proposes an implementation method, wherein the sample acoustic wave signal is obtained by a sample acquisition device, and the sample acquisition device includes a control valve, a control device, and an acoustic wave detection device. The control valve is used to adjust the flow rate of the fluid, the control device is used to control the opening degree of the control valve, and the acoustic wave detection device is used to detect the acoustic wave signal of the fluid flowing through the control valve. Figure 3 The process of step 110 sub-steps is shown in FIG. As shown in the figure, step 110 includes:
[0054] S1104: Setting the opening degree: setting the opening degree of the control valve to a first opening degree through the control device.
[0055] S1106: Obtain a tag: obtain a first opening degree through a control device.
[0056] In the above steps S1104 and S1106, the computing device that executes the control valve opening degree prediction model training method is connected to the control device signal, so that the computing device realizes the setting of the control valve opening degree through the control device, and the computing device can obtain the control signal of the control device for setting the control valve opening degree, that is, obtain the opening degree of the control valve.
[0057] S1108: Acquire training data: acquire a first sample sound wave signal when the opening degree of the control valve is at a first opening degree through a sound wave detection device.
[0058] In this step, when the control valve is at the first opening degree, the fluid will flow and pass through the control valve. The acoustic wave detection device detects the acoustic wave signal generated when the fluid passes through the control valve and sends it to the computing device that executes the control valve opening degree prediction model training method to obtain the first sample acoustic wave signal.
[0059] S1110: Training data labeling: labeling the obtained first opening degree as a label of the first sample sound wave signal.
[0060] In this step, the first opening degree is marked as a label of the first sample sound wave signal to form training data for inputting into the prediction model.
[0061] S1112: traverse other opening degrees: repeat the above steps S1104, S1106, S1108 and S1110 until all opening degree nodes are traversed.
[0062] In this step, the opening degree node can be any of several values in the range of 0-100%, for example, the opening degree nodes include 20%, 40%, 60%, 80% and 100%, which represents the ratio of the opening area of the control valve channel to the total area of the control valve channel. It should be noted that the numerical value of the above-mentioned opening degree node is an implementation method listed for the convenience of explanation, and does not limit the specific setting method of the opening degree node. It can be understood that in this step, the above-mentioned step S1104 is repeated, that is, the control valve opening degree is set to the second opening degree through the control device, and the first opening degree and the second opening degree are two different opening degree nodes. By setting the opening degree, obtaining a label, obtaining training data and annotating the training data at each opening degree node, the collection of sample acoustic wave signals with different opening degree labels is realized.
[0063] Through the above method, sample acoustic wave signals marked with different opening degree labels can be obtained to simulate different opening degree conditions in the well, increase the diversity of training data, and thus improve the training effect of the control valve opening degree prediction model.
[0064] Furthermore, in some embodiments, a circulation pipeline is connected between the two ends of the control valve, and a constant flow pump device is arranged on the circulation pipeline. Figure 3 As shown in the figure, before the above step S1104, the method further includes:
[0065] S1103: Setting the displacement: The displacement of the fluid flowing from the circulation pipeline into the control valve is set to a first displacement by a constant flow pump device.
[0066] In the above step S1110 , a first displacement is also obtained, and the obtained first displacement and first opening degree are marked as labels of the first sample sound wave signal.
[0067] After the above step S1112, the method further includes:
[0068] S1113: traverse other displacements: repeat the above steps S1103, S11104, S1106, S1108, S1110 and S1112 until all displacement nodes are traversed.
[0069] Specifically, the displacement node can be any number of values between 0 and 100%, for example, the displacement nodes include 20%, 40%, 60%, 80% and 100%, wherein 100% represents the maximum displacement, and 20%, 40%, 60% and 80% represent the percentage of the current displacement to the maximum displacement. It should also be noted that the values of the above displacement nodes are an implementation method listed for the convenience of explanation, and do not limit the specific setting method of the displacement nodes.
[0070] After the displacement is set in step S1103, steps S1104, S1106, S1108, S1110 and S1112 are performed in sequence. When performing step S1112, the opening degree of the control valve is reset to another opening degree node, while the displacement of the fluid flowing into the control valve from the circulation pipeline remains at the first displacement set in the previous process, thereby ensuring that sample acoustic wave signals corresponding to different opening degrees of the first displacement can be collected. After step S1112 traverses all the opening nodes, that is, after collecting the sample sound wave signals when the first displacement corresponds to all the opening nodes, step S1113 is performed, that is, the displacement of the fluid flowing into the control valve from the circulation pipeline is set to the second displacement through the constant current pump device. The first displacement and the second displacement are two different displacement nodes. Similarly, after setting the second displacement, it is necessary to traverse all the opening nodes when performing step S1112 to ensure that the sample sound wave signals when the second displacement corresponds to all the opening nodes can be collected. The subsequent steps are similar until the sample sound wave signals of each displacement node corresponding to all the opening nodes are collected.
[0071] Through the above method, sample acoustic wave signals marked with labels corresponding to different opening degrees for each displacement node can be obtained, and various displacement and opening degree working conditions in the well can be simulated to further increase the diversity of training data and improve the training effect of the control valve opening degree prediction model.
[0072] Further, in some embodiments, a back pressure control device is provided on the circulation pipeline upstream of the constant flow pump device. Figure 3 As shown in the figure, before the above step S1103, the method further includes:
[0073] S1102: Setting back pressure: The back pressure of the fluid flowing into the circulation pipeline from the control valve is set to a first back pressure through a back pressure control device.
[0074] In the above step S1110, a first back pressure is also obtained, and the obtained first back pressure, first displacement and first opening degree are marked as labels of the first sample sound wave signal.
[0075] After the above step S1113, the method further includes:
[0076] S1114: Traverse other back pressures: repeat the above steps S1102, S1103, S1104, S1106, S1108, S1110, S1112 and S1113.
[0077] As mentioned above, the back pressure node is to set several corresponding back pressure values in advance in the computing device.
[0078] Through the above method, sample acoustic wave signals marked with labels of each back pressure node corresponding to each displacement node and each opening degree node can be obtained, simulating various back pressure, displacement and opening degree conditions in the well, further increasing the diversity of training data and improving the training effect of the control valve opening degree prediction model.
[0079] Furthermore, in some embodiments, a heat tracing storage device is provided between the constant flow pump device and the back pressure control device on the circulation pipeline, and the heat tracing storage device is used to store fluid. Figure 3 As shown in the figure, before the above step S1102, the method further includes:
[0080] S1101: Setting the temperature: setting the temperature of the fluid input into the control valve from the circulation pipeline to a first temperature through a heat storage device.
[0081] In the above step S1110, the first temperature is also acquired, and the acquired first temperature, first back pressure, first displacement and first opening degree are marked as labels of the first sample sound wave signal.
[0082] After the above step S1114, the method further includes:
[0083] S1115: traverse other temperatures: repeat the above steps S1101, S1102, S1103, S1104, S1106, S1108, S1110, S1112, S1113 and S1114 until all temperature nodes are traversed.
[0084] As mentioned above, the back pressure node is to pre-set several corresponding temperature values in the computing device.
[0085] Through the above method, sample acoustic wave signals marked with labels of each temperature node corresponding to each back pressure node, each displacement node and each opening degree node can be obtained, simulating various underground temperatures, various back pressures, various displacements and various opening degree conditions, further increasing the diversity of training data and improving the training effect of the control valve opening degree prediction model.
[0086] Further, in an optional embodiment, the fluid includes single-phase water, single-phase oil, single-phase gas, oil-water two-phase mixture, water-gas two-phase mixture, oil-gas two-phase mixture or oil-water-gas three-phase mixture. The fluid can be set to single-phase water first, and then Figure 3 All the steps shown in FIG. 1 are repeated, and in step S1110, the fluid is also labeled as single-phase water as the label of the first sample acoustic wave signal. Then the fluid is changed to single-phase oil, and the steps are repeated. Figure 3 In all the steps shown in FIG. 1 , in step S1110, the fluid is marked as single-phase oil as the label of the first sample acoustic wave signal. Then the fluid is changed to single-phase gas and the process is repeated. Figure 3 In all the steps shown in FIG. 1 , also in step S1110 , the fluid is marked as single-phase gas as a label of the first sample acoustic wave signal.
[0087] Then the fluid is replaced with an oil-water two-phase mixture of the first oil-water ratio, and Figure 3 All the steps shown in, in step S1110, the fluid is an oil-water two-phase mixture of the first oil-water ratio as the label of the first sample acoustic wave signal, and after step S1115 is completed, the fluid is also replaced with an oil-water two-phase mixture of the second oil-water ratio, and all the steps are performed again until all the ratio nodes are traversed. Among them, the first oil-water ratio and the second oil-water ratio are two different mass ratios between oil and water, the first oil-water ratio and the second oil-water ratio are two nodes in the ratio node, and the specific number of ratio nodes and the specific value of each ratio node can be set accordingly according to the specific situation, and will not be repeated here. Then the fluid is replaced with a water-gas two-phase mixture of the first water-gas ratio, and the steps performed thereafter are the same as the above steps. Then the fluid is replaced with a water-gas two-phase mixture of the first oil-gas ratio, and the same steps as above are performed.
[0088] Finally, the fluid is replaced with a three-phase mixture of the first oil-water-gas ratio and Figure 3 All the steps shown in the figure are performed, in step S1110, the fluid is marked as an oil-water-gas three-phase mixture with a first oil-water-gas ratio as the label of the first sample acoustic wave signal, and after step S1115 is completed, the fluid is also replaced with an oil-water-gas three-phase mixture with a second oil-water-gas ratio, and all the steps are performed again until all ratio nodes are traversed.
[0089] By changing the fluid type, sample acoustic wave signals marked with labels of each temperature node corresponding to each back pressure node, each displacement node and each opening degree node can be collected when the fluid is single-phase water, single-phase oil, single-phase gas, two-phase mixture of oil and water in different proportions, two-phase mixture of water and gas, two-phase mixture of oil and gas and three-phase mixture of oil, water and gas in different proportions. This enriches the training sample content, simulates various downhole working conditions, and helps to fully improve the training effect of the control valve opening degree prediction model.
[0090] According to another aspect of the embodiment of the present invention, a control valve opening degree prediction method is also provided. Figure 4 , the figure shows the process of the control valve opening degree prediction method provided by the embodiment of the present invention. As shown in the figure, the method includes:
[0091] S210: Acquire sound wave signals;
[0092] S220: extracting energy entropy, signal average amplitude, short-time energy, short-time energy autocorrelation coefficient, zero-crossing rate, spectrum flow and spectrum rolling edge of the acoustic wave signal;
[0093] S230: Inputting energy entropy, signal average amplitude, short-time energy, short-time energy autocorrelation coefficient, zero-crossing rate, spectrum flow and spectrum rolling edge into a control valve opening degree prediction model to obtain a predicted control valve opening degree.
[0094] In step S230, the control valve opening degree prediction model is obtained by the control valve opening degree prediction model training method in any of the above-mentioned embodiments.
[0095] In the control valve opening degree prediction method provided in an embodiment of the present invention, the acoustic wave signal at the control valve channel collected during the actual production operation of the smart well is analyzed and processed through the control valve opening degree prediction model, so that the opening degree of the control valve can be detected in real time and accurately, thereby improving the control accuracy of downhole production and enhancing the management level of the smart well production system.
[0096] According to another aspect of the present invention, a control valve opening degree prediction model training device is also provided. Figure 5, the figure shows the structure of the control valve opening prediction model training device provided by an embodiment of the present invention. As shown in the figure, the control valve opening prediction model training device 300 includes: an acquisition unit 310, an extraction unit 330, and a training unit 350. The acquisition unit 310 is used to acquire a sample acoustic wave signal marked with a control valve opening label, and the sample acoustic wave signal is an acoustic wave signal of a fluid flowing through a control valve. The extraction unit 330 is used to extract the sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling of the sample acoustic wave signal. The training unit 350 is used to input the feature matrix composed of sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling into a neural network for training to obtain a control valve opening prediction model.
[0097] Please continue reading Figure 5 In some embodiments, the control valve opening prediction model training device 300 further includes a processing unit 320, which is used to perform energy normalization processing on the sample acoustic wave signal to obtain a normalized result of the sample acoustic wave signal. The extraction unit 330 is used to extract sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow, and sample spectrum rolling edge of the normalized result of the sample acoustic wave signal.
[0098] In some embodiments, the sample acoustic wave signal is acquired through a sample acquisition device, and the sample acquisition device includes a control valve, a control device, and an acoustic wave detection device. The control valve is used to adjust the flow rate of the fluid, the control device is used to control the opening degree of the control valve, and the acoustic wave detection device is used to detect the acoustic wave signal of the fluid flowing through the control valve. The acquisition unit 110 is used to set the opening degree: set the control valve opening degree to a first opening degree through the control device, for acquiring a label: acquire the first opening degree through the control device, for acquiring training data: acquire the first sample acoustic wave signal when the control valve opening degree is at the first opening degree through the acoustic wave detection device, for training data labeling: label the acquired first opening degree as the label of the first sample acoustic wave signal, for traversing other opening degrees: repeat setting the opening degree, acquiring the label, acquiring the training data, and training data labeling until all opening degree nodes are traversed.
[0099] Further, in some embodiments, a circulation pipeline is connected between the two ends of the control valve, and a constant flow pump device is arranged on the circulation pipeline. The acquisition unit 310 is used to set the displacement before setting the opening degree: the displacement of the fluid flowing into the control valve from the circulation pipeline is set to the first displacement by the constant flow pump device, for training data labeling: obtaining the first displacement, labeling the obtained first displacement and the first opening degree as the label of the first sample sound wave signal, for traversing other displacements after traversing other opening degrees: repeating setting the displacement, setting the opening degree, obtaining the label, obtaining the training data, training data labeling and traversing other opening degrees until all displacement nodes are traversed.
[0100] Further, in some embodiments, a back pressure control device is provided at a position upstream of the constant flow pump device on the circulation pipeline. The acquisition unit 310 is used to set the back pressure before setting the displacement: the back pressure of the fluid flowing into the circulation pipeline from the control valve is set to the first back pressure by the back pressure control device, for training data labeling: obtaining the first back pressure, labeling the obtained first back pressure, first displacement and first opening degree as the label of the first sample acoustic wave signal, for traversing other back pressures after traversing other displacements: repeating setting back pressure, setting displacement, setting opening degree, obtaining label, obtaining training data, training data labeling, traversing other opening degrees and traversing other displacements until all back pressure nodes are traversed.
[0101] Further, in some embodiments, a heat tracing storage device is provided between the constant flow pump device and the back pressure control device on the circulation pipeline, and the heat tracing storage device is used to store fluid. The acquisition unit is used to set the temperature before setting the back pressure: the temperature of the fluid input into the control valve from the circulation pipeline is set to the first temperature through the heat tracing storage device, for training data labeling: acquiring the first temperature, labeling the acquired first temperature, first back pressure, first displacement and first opening degree as the label of the first sample acoustic wave signal, for traversing other temperatures after traversing other back pressures: repeating setting back pressure, setting displacement, setting opening degree, acquiring labels, acquiring training data, training data labeling, traversing other opening degrees, traversing other displacements and traversing other back pressures until all temperature nodes are traversed.
[0102] According to another aspect of the embodiment of the present invention, a control valve opening degree prediction model training device is also provided. Figure 6 , the figure shows the structure of the control valve opening degree prediction model training device provided by an embodiment of the present invention. The specific embodiment of the present invention does not limit the specific implementation method of the control valve opening degree prediction model training device.
[0103] like Figure 6As shown in , the control valve opening degree prediction model training device may include: a processor (processor) 402, a communication interface (Communications Interface) 404, a memory (memory) 406, and a communication bus 408.
[0104] The processor 402, the communication interface 404, and the memory 406 communicate with each other via the communication bus 408. The communication interface 404 is used to communicate with other devices such as a client or other server network elements. The processor 402 is used to execute the program 410, which can specifically execute the relevant steps in the above-mentioned embodiment of the method for training the control valve opening prediction model.
[0105] Specifically, the program 410 may include program code including computer executable instructions.
[0106] The processor 402 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiment of the present invention. The one or more processors included in the valve opening degree prediction model training device may be processors of the same type, such as one or more CPUs; or may be processors of different types, such as one or more CPUs and one or more ASICs.
[0107] The memory 406 is used to store the program 410. The memory 406 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.
[0108] The program 410 can be specifically called by the processor 402 to enable the valve opening degree prediction model training device to perform the following operations:
[0109] Acquire a sample sound wave signal marked with a control valve opening degree label, the sample sound wave signal being a sound wave signal of fluid flowing through the control valve;
[0110] Extracting sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling edge of the sample sound wave signal;
[0111] The characteristic matrix composed of sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling edge is input into the neural network for training to obtain the control valve opening degree prediction model.
[0112] According to another aspect of the present invention, a control valve opening degree prediction model training system is also provided. Figure 7 , the figure shows the structure of the control valve opening degree prediction model training system provided by an embodiment of the present invention. As shown in the figure, the control valve opening degree prediction model training system 500 includes a sample collection device 510 and the control valve opening degree prediction model training device 400. The sample collection device 510 includes: a casing 511, an oil pipe 512, a control valve 513, a control device 514, an optical cable 515 and an acoustic wave detection device 516. The oil pipe 512 is arranged in the casing 511, and the top of the oil pipe 512 is connected to the bottom of the casing 511 through a circulation pipeline 517. The control valve 513 is arranged on the oil pipe 512. The control valve 513 has a valve hole 5131 connecting the inside of the oil pipe 512 and the inside of the casing 511. A sleeve 5132 for covering the valve hole 5131 is slidably arranged on the control valve 513. The fluid enters the casing 511, the control valve 513, and the oil pipe 512 in sequence and then enters the casing 511 again through the circulation pipeline 517 to form a circulation flow. The control device 514 is electrically connected to the control valve 513 outside the casing 511. The control device 514 is used to control the opening degree of the valve hole 5131 by controlling the sliding sleeve 5132 to slide. The optical cable 515 is wound around the outer wall of the valve hole 5131 on the control valve 513. The acoustic wave detection device 516 is connected to the optical cable 515 outside the casing 511, and is used to detect the acoustic wave signal of the fluid flowing through the valve hole 5131 through the optical cable 151. The control valve opening degree prediction model training device 400 is electrically connected to the control device 514 and the acoustic wave detection device 516 respectively. The control valve opening degree prediction model training device 400 is used to obtain the opening degree of the control valve 513 through the control device 514, and is also used to obtain the sample acoustic wave signal of the fluid flowing through the valve hole 5131 when the control valve 513 is at this opening degree through the acoustic wave detection device 516.
[0113] Due to the complex underground working conditions, it is difficult to collect data and obtain annotation information. Therefore, in order to facilitate the acquisition of training data, in this embodiment, the sample collection device 510 is set as an indoor test device for collecting data.
[0114] The casing 511 and the oil pipe 512 are both hollow pipes sealed at both ends. Specifically, the oil pipe 512 and the casing 511 can be connected as a whole by flange sealing at both ends. The internal space of the valve hole 5131 on the control valve 513 forms a channel connecting the inside of the casing 511 and the inside of the oil pipe 512. The optical cable 515 is wound around the outer wall of the valve hole 5131 of the control valve 513, and it is necessary to ensure that the optical cable 515 does not block the valve hole 5131 and does not affect the sliding of the sliding sleeve 5132.
[0115] In a data acquisition process, a circulating pipeline 517 is used to allow the fluid to circulate in the casing 511, the control valve 513 and the oil pipe 512. The control device 514 sends a control signal and controls the sleeve 5132 to move from bottom to top to open the first valve hole 5131 by hydraulic drive or electric drive, and sends the control signal to the control valve opening degree prediction model training device 400. At the same time, the acoustic wave detection device 516 detects the acoustic wave signal of the fluid flowing through the opened valve hole 5131 through the optical cable 515, and sends the acoustic wave signal as a sample acoustic wave signal to the control valve opening degree prediction model training device 400. The control valve opening degree prediction model training device 400 determines the opening degree of the control valve 513 through the control signal, specifically, by calculating the ratio of the area of the opened valve hole 5131 to the total area of all valve holes 5131, to obtain the opening degree of the control valve 513, and marks the opening degree as a label of the sample acoustic wave signal, which is used as training data for subsequent prediction models. Thereafter, the valve hole 5131 is gradually opened according to the set opening degree nodes, and the acoustic wave signals of the fluid flowing through the valve hole 5131 are collected one by one to obtain the acoustic wave signals when the valve hole 5131 is at different opening degrees.
[0116] Please continue reading Figure 7 In some embodiments, a constant flow pump device 5171 is provided on the circulation pipeline 517, and the constant flow pump device 5171 is used to control the flow rate of the fluid flowing from the circulation pipeline 517 into the casing 511. The constant flow pump device 5171 is electrically connected to the control valve opening degree prediction model training device 400, and the control valve opening degree prediction model training device 400 is used to obtain the current flow value through the constant flow pump device 5171. During the data collection process, multiple flow nodes can be pre-set, and sample acoustic wave signals corresponding to each opening degree node are collected at each flow node.
[0117] Please continue reading Figure 7 In some embodiments, a back pressure control device 5172 is further provided on the circulation pipeline 517 upstream of the constant flow pump device 5171, and the back pressure control device 5172 is used to control the back pressure of the fluid flowing from the oil pipe 512 into the circulation pipeline 517. The back pressure control device 5172 is electrically connected to the control valve opening degree prediction model training device 400, and the control valve opening degree prediction model training device 400 is used to obtain the current back pressure value through the back pressure control device 5172. During the data collection process, multiple back pressure nodes can be pre-set, and sample acoustic wave signals are collected when each back pressure node corresponds to each flow node and each opening degree node.
[0118] Please continue reading Figure 7In some embodiments, a heat storage device 5173 is further provided on the circulation pipeline 517 between the constant flow pump device 5171 and the back pressure control device 5172, and the heat storage device 5173 is used to store and heat the fluid. The heat storage device 5173 is electrically connected to the control valve opening degree prediction model training device 400, and the control valve opening degree prediction model training device 400 is used to obtain the current temperature value of the fluid through the heat storage device 5173. During the data collection process, multiple temperature nodes can be pre-set, and sample acoustic wave signals corresponding to each temperature node, each back pressure node, each flow node, and each opening degree node are collected.
[0119] Furthermore, in some embodiments, the fluid type can be replaced by the heat tracing storage device 5173, and the fluid can include single-phase water, single-phase oil, single-phase gas, oil-water two-phase mixture, water-gas two-phase mixture, oil-gas two-phase mixture, and oil-water-gas three-phase mixture. During the data collection process, for each type of fluid, sample acoustic wave signals corresponding to each temperature node, each back pressure node, each flow node, and each opening node are collected.
[0120] For the sake of convenience, a specific collection method is provided as follows:
[0121] The fluids include single-phase water A1, single-phase oil A2, single-phase gas A3, oil-water two-phase mixture A4, water-gas two-phase mixture A5, oil-gas two-phase mixture A6 and oil-water-gas three-phase mixture A7.
[0122] The opening degree nodes include B1 and B2. The flow nodes include C1 and C2. The back pressure nodes include D1 and D2. The temperature nodes include E1 and E2.
[0123] The collected sample sound wave signals are marked with X, and X1, X2, etc. represent samples collected under different conditions (i.e., sample sound wave signals with different labels). The collection results are shown in Table 1 below:
[0124] Table 1
[0125]
[0126]
[0127] It should be noted that the above acquisition results are for the sake of convenience of explanation, taking the number of opening nodes, flow nodes, back pressure nodes and temperature nodes as two for example. In other embodiments, the number of opening nodes, flow nodes, back pressure nodes and temperature nodes is generally more than two, such as more than ten or even dozens, to ensure that more working conditions in the well can be simulated and improve the training effect of the prediction model.
[0128] In the control valve opening prediction model training system 500 provided in the embodiment of the present invention, the sample collection device 510 can simulate the acoustic signal monitoring of the complex fluid flow in the well, and the control valve opening prediction model training system 500 has a simple overall structure and high simulation accuracy, which meets the conditions for laboratory research on the state of the downhole flow control valve in actual application. By simulating the state of the downhole flow control valve in actual application in the laboratory in advance to collect training data, it is more convenient to obtain data and train the prediction model. By applying the trained prediction model to the actual mining process of the smart well to perform real-time monitoring of the control valve opening, the risks in the production operation of the smart well can be reduced, and the control accuracy of the downhole production can be improved.
[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention is described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention. In particular, as long as there is no structural conflict, the various technical features mentioned in each embodiment can be combined in any way. The present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions that fall within the scope of the claims.
Claims
1. A control valve opening degree prediction model training method, characterized in that: include: A sample acoustic wave signal marked with a control valve opening degree label is obtained, wherein the sample acoustic wave signal is an acoustic wave signal of a fluid flowing through the control valve; the sample acoustic wave signal is obtained through a sample acquisition device, wherein the sample acquisition device includes a control valve, a control device and an acoustic wave detection device, wherein the control valve is used to adjust the flow rate of the fluid, the control device is used to control the opening degree of the control valve, and the acoustic wave detection device is used to detect the acoustic wave signal of the fluid flowing through the control valve; the obtaining of the sample acoustic wave signal marked with the control valve opening degree label comprises: setting the opening degree: setting the control valve opening degree to a first opening degree through the control device; obtaining a label: obtaining the first opening degree through the control device; obtaining training data: obtaining a first sample acoustic wave signal when the control valve opening degree is at the first opening degree through the acoustic wave detection device; training data labeling: labeling the obtained first opening degree as a label of the first sample acoustic wave signal; traversing other opening degrees: repeating the setting of the opening degree, the obtaining of the label, the obtaining of the training data and the labeling of the training data until all opening degree nodes are traversed; Performing energy normalization processing on the sample sound wave signal to obtain a normalized result of the sample sound wave signal; Extracting sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling edge of the normalized result of the sample sound wave signal; The characteristic matrix composed of the sample energy entropy, the sample signal average amplitude, the sample short-time energy, the sample short-time energy autocorrelation coefficient, the sample zero-crossing rate, the sample spectrum flow and the sample spectrum rolling edge is input into the neural network for training to obtain a control valve opening degree prediction model.
2. The control valve opening degree prediction model training method according to claim 1 is characterized in that: A circulation pipeline is connected between the two ends of the control valve, and a constant flow pump device is arranged on the circulation pipeline; Before setting the opening degree, the method further includes: Setting the displacement: setting the displacement of the fluid flowing from the circulation pipeline into the control valve to a first displacement by the constant flow pump device; The training data annotation includes: Acquire the first displacement, and mark the acquired first displacement and the first opening degree as labels of the first sample sound wave signal; After traversing other opening degrees, the method further includes: Traversing other displacements: repeating the steps of setting the displacement, setting the opening degree, obtaining the label, obtaining the training data, labeling the training data, and traversing other opening degrees until all displacement nodes are traversed.
3. The control valve opening degree prediction model training method according to claim 2 is characterized in that: A back pressure control device is provided on the circulation pipeline at a position upstream of the constant flow pump device; Before setting the displacement, the method further includes: Setting the back pressure: setting the back pressure of the fluid flowing into the circulation pipeline from the control valve to a first back pressure by the back pressure control device; The training data annotation includes: acquiring the first back pressure, and marking the acquired first back pressure, the first displacement, and the first opening degree as labels of the first sample sound wave signal; After traversing other displacements, the method further includes: Traversing other back pressures: repeating the steps of setting the back pressure, setting the displacement, setting the opening degree, obtaining the label, obtaining the training data, labeling the training data, traversing other opening degrees, and traversing other displacements until all back pressure nodes are traversed.
4. The control valve opening degree prediction model training method according to claim 3 is characterized in that: A heat tracing storage device is provided on the circulation pipeline between the constant flow pump device and the back pressure control device, and the heat tracing storage device is used to store the fluid; Before setting the back pressure, the method further includes: Setting temperature: setting the temperature of the fluid input into the control valve from the circulation pipeline to a first temperature through the heat storage device; The training data annotation includes: acquiring the first temperature, and marking the acquired first temperature, the first back pressure, the first displacement, and the first opening degree as labels of the first sample sound wave signal; After traversing other back pressures, the method further includes: Traversing other temperatures: repeating the steps of setting the back pressure, setting the displacement, setting the opening degree, obtaining the label, obtaining the training data, labeling the training data, traversing other opening degrees, traversing other displacements, and traversing other back pressures until all temperature nodes are traversed.
5. A control valve opening degree prediction method, characterized in that: include: Acquiring sound wave signals; Extracting energy entropy, signal average amplitude, short-time energy, short-time energy autocorrelation coefficient, zero-crossing rate, spectrum flow and spectrum rolling edge of the acoustic wave signal; The energy entropy, the signal average amplitude, the short-time energy, the short-time energy autocorrelation coefficient, the zero-crossing rate, the spectrum flow and the spectrum rolling edge are input into a control valve opening degree prediction model to obtain a predicted control valve opening degree; wherein the control valve opening degree prediction model is obtained using the control valve opening degree prediction model training method described in any one of claims 1 to 4.
6. A control valve opening degree prediction model training device, characterized in that: include: An acquisition unit is used to acquire a sample acoustic wave signal labeled with a control valve opening degree label, wherein the sample acoustic wave signal is an acoustic wave signal of a fluid flowing through the control valve; the sample acoustic wave signal is acquired through a sample acquisition device, wherein the sample acquisition device includes a control valve, a control device and an acoustic wave detection device, wherein the control valve is used to adjust the flow rate of the fluid, the control device is used to control the opening degree of the control valve, and the acoustic wave detection device is used to detect the acoustic wave signal of the fluid flowing through the control valve; the acquisition of the sample acoustic wave signal labeled with the control valve opening degree label comprises: setting the opening degree: setting the control valve opening degree to a first opening degree through the control device; acquiring a label: acquiring the first opening degree through the control device; acquiring training data: acquiring a first sample acoustic wave signal when the control valve opening degree is at the first opening degree through the acoustic wave detection device; training data labeling: labeling the acquired first opening degree as a label of the first sample acoustic wave signal; traversing other opening degrees: repeating the setting of the opening degree, the acquiring of the label, the acquiring of the training data and the labeling of the training data until all opening degree nodes are traversed; A processing unit, used for performing energy normalization processing on the sample sound wave signal to obtain a normalized result of the sample sound wave signal; An extraction unit, used to extract sample energy entropy, sample signal average amplitude, sample short-time energy, sample short-time energy autocorrelation coefficient, sample zero-crossing rate, sample spectrum flow and sample spectrum rolling edge of the normalized result of the sample sound wave signal; The training unit is used to input the characteristic matrix composed of the sample energy entropy, the sample signal average amplitude, the sample short-time energy, the sample short-time energy autocorrelation coefficient, the sample zero-crossing rate, the sample spectrum flow and the sample spectrum rolling edge into the neural network for training to obtain a control valve opening degree prediction model.
7. A control valve opening prediction model training device, characterized in that: include: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform the operation of the control valve opening degree prediction model training method as described in any one of claims 1-4.
8. A control valve opening degree prediction model training system, characterized in that: The control valve opening degree prediction model training device and the sample acquisition device according to claim 7 are included, and the sample acquisition device comprises: casing; An oil pipe is arranged in the casing, and the top of the oil pipe is connected with the bottom of the casing through a circulation pipeline; A control valve is arranged on the oil pipe, the control valve has a valve hole communicating the inside of the oil pipe with the inside of the casing, a sleeve is slidably arranged on the control valve for covering the valve hole, and the fluid sequentially enters the casing, the control valve, and the oil pipe, and then enters the casing again through the circulation pipeline to form a circulation flow; A control device, electrically connected to the control valve outside the sleeve, the control device being used to control the opening degree of the valve hole by controlling the sliding of the sleeve; An optical cable is wound around the outer wall of the valve hole on the control valve; an acoustic wave detection device, connected to the optical cable outside the casing, and used for detecting an acoustic wave signal of the fluid flowing through the valve hole through the optical cable; The control valve opening degree prediction model training device is electrically connected to the control device and the acoustic wave detection device respectively. The control valve opening degree prediction model training device is used to obtain the opening degree of the control valve through the control device, and is also used to obtain the sample acoustic wave signal of the fluid flowing through the valve hole when the control valve is at the opening degree through the acoustic wave detection device.
Citation Information
Patent Citations
Prediction method for transient opening height of one-way valve for multiphase mixing transmission pump
CN107918709A
Dense medium intelligent control method and system
CN112138861A
Bearing monitoring method, system and equipment based on operation state sound radiation signals
CN113158366A
Valve inner leakage detection method based on sound waves
CN113295346A
Oil and gas transportation pipeline harmful flow pattern forecasting and control method based on Labview and Matlab
CN113962144A
