A two-phase flow parameter measurement method based on spatiotemporal parallel feature fusion network

Through the spatial and temporal parallel feature fusion network and double helix high-frequency microwave sensor, the problems of poor real-time and low accuracy in the measurement of two-phase flow parameters are solved, and the rapid and accurate measurement of gas-solid and oil-water two-phase flow is achieved, which improves the real-time measurement Sex and objectivity.

CN115060744BActive Publication Date: 2024-11-26TIANJIN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210468325.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-29
Publication Date
2024-11-26
Estimated Expiration
2042-04-29

AI Technical Summary

Technical Problem

The existing two-phase flow parameter measurement methods have problems such as poor real-time performance and low measurement accuracy, making it difficult to achieve accurate and rapid measurement of gas-solid and oil-water two-phase flows.

Method used

Using a method based on spatiotemporal parallel feature fusion network, a double helix high-frequency microwave sensor is designed to collect and pre-process signals, and a data set is constructed, and a spatiotemporal parallel feature fusion network is used, combining long and short-term memory networks and convolutional neural networks to extract two The characteristics of the phase flow fluctuation signal are measured for flow and moisture content.

Benefits of technology

It realizes rapid and accurate measurement of gas-solid and oil-water two-phase flows, with high real-time and wide application range, reducing uncertainty in measurement and subjectivity of artificial operations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115060744B_ABST
    Figure CN115060744B_ABST
Patent Text Reader

Abstract

A two-phase flow parameter measurement method based on a time-space parallel feature fusion network includes: designing a double-helix high-frequency microwave sensor finite element simulation model through a finite element simulation model, optimizing parameters, and obtaining the optimal structure of the double-helix high-frequency microwave sensor; using a double-helix high-frequency microwave sensor to measure the fluctuation signal of gas-solid or oil-water two-phase flow; constructing a data set, specifically preprocessing the data, then using a sliding window to obtain samples from the data, adding corresponding labels to the samples, and then using 80% of the samples as a training set, 10% of the samples as a verification set, and the remaining 10% of the samples as a test set; constructing a time-space parallel feature fusion network to realize the feature fusion and flow measurement of the gas-solid two-phase flow flow fluctuation signal or the feature fusion and water content measurement of the oil-water two-phase flow fluctuation signal. The present invention can be used for solid flow measurement of gas-solid two-phase flow and water content measurement of oil-water two-phase flow, and has a wide range of applications.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a method for measuring gas-solid and oil-water two-phase flow parameters, and in particular to a method for measuring two-phase flow parameters based on a time-space parallel feature fusion network. Background Art

[0002] With the continuous development of modern science and technology, many industries have higher and higher requirements for the measurement accuracy and real-time performance of two-phase flow parameters, such as chemical, petroleum and many other industries. If the measurement accuracy and real-time performance of two-phase flow parameters can reach a certain level, it will have important guiding significance for the operation of these industries. At present, most of the measurement methods of two-phase flow parameters have the problems of poor real-time performance and low measurement accuracy, so accurate and fast measurement of two-phase flow parameters is a current technical problem. Existing two-phase flow measurement methods such as optical method, tomography method, thermodynamic method, nuclear magnetic resonance, etc. are difficult to achieve real-time and accurate measurement of two-phase flow parameters.

[0003] In recent years, soft measurement models have also developed rapidly in the field of multiphase flow. In particular, the application of deep learning methods and artificial intelligence technology has broadened the application scope of multiphase flow soft measurement models. Deep learning technology is a new theory that has emerged in recent years. It extracts the features of the measured object layer by layer in an unsupervised or supervised manner. Its features are highly objective and can accurately and comprehensively reflect the essence of the measured object. The data measured by the sensor can be fused through soft measurement methods, and the characteristic information of multiphase flow can be accurately and efficiently extracted by using intelligent and deep learning methods. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a two-phase flow parameter measurement method based on a time-space parallel feature fusion network that can be used for solid flow measurement of gas-solid two-phase flow and water content measurement of oil-water two-phase flow in order to overcome the deficiencies of the prior art.

[0005] The technical solution adopted by the present invention is: a two-phase flow parameter measurement method based on a time-space parallel feature fusion network, comprising the following steps:

[0006] 1) Design a finite element simulation model of a double-helix high-frequency microwave sensor through a finite element simulation model, optimize parameters, and obtain the optimal structure of the double-helix high-frequency microwave sensor;

[0007] 2) Use double-helix high-frequency microwave sensor to measure the fluctuation signal of gas-solid or oil-water two-phase flow;

[0008] 3) Construct a data set, specifically preprocess the data, then use a sliding window to obtain samples from the data, add corresponding labels to the samples, and then use 80% of the samples as a training set, 10% of the samples as a validation set, and the remaining 10% of the samples as a test set;

[0009] 4) Construct a spatiotemporal parallel feature fusion network to realize the feature fusion and flow measurement of gas-solid two-phase flow fluctuation signals or the feature fusion and water content measurement of oil-water two-phase flow fluctuation signals.

[0010] In the finite element simulation model of the double-helix high-frequency microwave sensor described in step 1), the interval angle between the excitation electrode and the protection electrode is 30°, the electrode pitch length is 90mm, the insulation layer thickness is 0.05mm, the inner diameter of the inner shielding layer is 40mm, and the thickness is 0.5mm; the input and output ports of the finite element simulation model of the double-helix high-frequency microwave sensor are set as wave ports, the characteristic impedance is 50 ohms, and the automatic subdivision finite element grid is set; the operating frequency is set to 0.8GHz; and the electrode angle is set to 70°.

[0011] The double-helix high-frequency microwave sensor described in step 1) is a hollow tube structure in which an inner shielding layer (6), an insulating layer (5), a measuring electrode layer and an outer shielding layer are arranged in sequence from the inside to the outside, wherein the inner shielding layer (6) is used to be sleeved on the outer wall of the measuring pipe, and the measuring electrode layer is formed by spirally winding an excitation electrode (1), a first protective electrode (3), a receiving electrode (2) and a second protective electrode (4) in parallel on the outer wall of the inner shielding layer (6), and a gap is formed between the excitation electrode (1), the first protective electrode (3), the receiving electrode (2) and the second protective electrode (4).

[0012] Step 2) comprises: when a gas-solid or oil-water two-phase flow enters a measuring pipeline section equipped with a double-helix high-frequency microwave sensor from a pipeline, the double-helix high-frequency microwave sensor collects a two-phase flow fluctuation signal, specifically, a sinusoidal excitation signal source generates an excitation signal, which is sent to an excitation electrode (1) of the double-helix high-frequency microwave sensor through a power divider to excite the gas-solid or oil-water two-phase flow, and is sent to an external detection unit; at the same time, a receiving electrode (2) sends the measured fluctuation signal under different flow rate conditions in the pipeline to another path of the external detection unit, the two paths of signals are mixed in the external detection unit, and the mixed signals are passed through an adder to obtain a microwave difference frequency signal.

[0013] Step 3) includes:

[0014] (3.1) First, the microwave difference frequency signal measured by the double-helix high-frequency microwave sensor is preprocessed, and the formula is as follows:

[0015]

[0016] Among them, I i Represents the i-th data, I mean and I std are the mean and standard deviation of the data, is the i-th data after preprocessing;

[0017] (3.2) The preprocessed signal is segmented using a non-overlapping windowing method, with a window length of H and a data acquisition length of L. samples, among which, represents rounding down; the microwave difference frequency signal is windowed and intercepted, and the actual value is used as the data label, and a total of N fluctuation samples with label values ​​are obtained;

[0018] (3.3) Randomly divide N fluctuation samples into a data set with training set, validation set and test set, and the specific ratio is [training set: validation set: test set] = [8:1:1].

[0019] The spatiotemporal parallel feature fusion network described in step 4) is divided into two parallel sub-networks, the first sub-network is a 2-layer long short-term memory network module, which uses the information in the storage unit to capture the long-term dependency in the sequence, extract the time series features, and obtain the associated information, and the second sub-network is a convolutional neural network module composed of a 3-layer convolutional network, which extracts the local features of the sequence and mines the potential information after convolution and pooling the input data; the different features of the microwave difference frequency signal are extracted by merging the long short-term memory network module and the convolutional neural network module; the extracted microwave difference frequency signal is trained by two fully connected layers to finally obtain the predicted value.

[0020] The specific operation of the spatiotemporal parallel feature fusion network includes:

[0021] (4.1) Input the microwave difference frequency signal into the first subnetwork in time sequence for feature extraction. Let t be the current time and the input be x t , through the weight W and bias b, combined with the hidden layer output h of the previous time node t-1 , calculate the three functions of input gate p(t), forget gate f(t) and output gate o(t), and the calculation formula of memory unit M(t) is as follows:

[0022] M(t)=f(t)M(t-1)+p(t)(W*[h t-1 ,x t ]+b)

[0023] The calculation formula of the state output unit is:

[0024] h(t)=o(t)tanh(M(t))

[0025] The obtained hidden layer state output h(t) is combined with the input at time t+1 as the input of the next hidden layer and calculated in sequence;

[0026] (4.2) Inputting the microwave difference frequency signal into the second sub-network in time sequence for feature extraction; wherein the convolutional neural network module includes three convolutional layers connected in series, wherein the outputs of the first convolutional layer and the second convolutional layer are respectively connected to a maximum pooling layer to prevent overfitting;

[0027] (4.3) The different features extracted by the long short-term memory network module and the convolutional neural network module are fused through the merging layer, and the fused feature vector is input into two fully connected layers with the actual value as the label. The last fully connected layer is used as the output layer and the ReLU function is used as the activation function to finally output the predicted value.

[0028] In the spatiotemporal parallel feature fusion network,

[0029] Set hyperparameters, where the hyperparameters are updated based on the training data using the Adam optimization algorithm through gradient back-up to the time-space parallel feature fusion network weights;

[0030] The mean square error, mean absolute error and mean absolute percentage error are used to construct the loss function between the predicted value and the label value of the spatiotemporal parallel feature fusion network, which is used to calculate the gap between the forward calculation result of each iteration of the spatiotemporal parallel feature fusion network and the true value. Minimizing the gap is used as the criterion to guide the training and optimization of the spatiotemporal parallel feature fusion network in the right direction.

[0031] The two-phase flow parameter measurement method based on the spatiotemporal parallel feature fusion network of the present invention has the following advantages:

[0032] 1. The double-helix high-frequency microwave sensor used has a small time constant and fast response speed. It can perform real-time monitoring and dynamic processing, and can quickly and accurately obtain sequence fluctuation signals. The double-helix structure can accurately reflect the frequency changes of materials during the flow process.

[0033] 2. The deep neural network used in the present invention is objective in parameters, eliminating the uncertainty in the test and the subjectivity of human operation. After being trained with a large amount of data and iterations, the deep convolutional neural network and long short-term memory network can have strong predictive performance and objectivity for such data.

[0034] 3. The present invention can be used to measure the solid flow rate of gas-solid two-phase flow and the water content of oil-water two-phase flow, and has a wide range of applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic diagram of the structure of the double-helix high-frequency microwave sensor in the present invention;

[0036] Figure 2 It is a schematic diagram of the spatiotemporal parallel feature fusion network structure established in the present invention. DETAILED DESCRIPTION

[0037] The following is a detailed description of a two-phase flow parameter measurement method based on a time-space parallel feature fusion network of the present invention in conjunction with the embodiments and drawings.

[0038] A two-phase flow parameter measurement method based on a time-space parallel feature fusion network of the present invention comprises the following steps:

[0039] 1) Design a finite element simulation model of a double-helix high-frequency microwave sensor through a finite element simulation model, optimize the parameters, and obtain the optimal structure of the double-helix high-frequency microwave sensor.

[0040] The finite element simulation model of the double-helix high-frequency microwave sensor has an interval angle of 30° between the excitation electrode and the protection electrode, an electrode pitch of 90 mm, an insulating layer thickness of 0.05 mm, an inner shielding layer inner diameter of 40 mm, and a thickness of 0.5 mm; the input and output ports of the finite element simulation model of the double-helix high-frequency microwave sensor are set as wave ports, the characteristic impedance is 50 ohms, and an automatic finite element mesh is set; first, the frequency characteristics of the sensor are investigated, the sweep step is 20 MHz, and the sweep range is 0.5 GHz to 2 GHz; the phase and amplitude responses at the resonance point frequency change regularly with the different two-phase flow fractions. After many experiments, we set the operating frequency to 0.8 GHz; different electrode angles are set, and simulation finds that for different electrode angles under an excitation signal of 0.8 GHz, the electrode angle at 70 degrees is smoother and the electric field distribution is more uniform, so the electrode angle is set to 70°.

[0041] like Figure 1 As shown, the double-helix high-frequency microwave sensor specifically comprises an inner shielding layer 6, an insulating layer 5, a measuring electrode layer and an outer shielding layer (not shown in the figure) arranged in sequence from the inside to the outside to form a hollow tube structure, wherein the inner shielding layer 6 is used to be sleeved on the outer wall of the measuring pipe, and the measuring electrode layer is formed by spirally winding an excitation electrode 1, a first protective electrode 3, a receiving electrode 2 and a second protective electrode 4 arranged side by side on the outer wall of the inner shielding layer 6 in sequence, and a gap is formed between the excitation electrode 1, the first protective electrode 3, the receiving electrode 2 and the second protective electrode 4.

[0042] 2) Use double-helix high-frequency microwave sensor to measure the fluctuation signal of gas-solid or oil-water two-phase flow. Including:

[0043] When a gas-solid or oil-water two-phase flow enters a measuring pipeline section equipped with a double-helix high-frequency microwave sensor from a pipeline, the double-helix high-frequency microwave sensor collects a two-phase flow fluctuation signal. Specifically, a sinusoidal excitation signal source generates an excitation signal, which is sent to an excitation electrode (1) of the double-helix high-frequency microwave sensor through a power divider to excite the gas-solid or oil-water two-phase flow, and is sent to an external detection unit. At the same time, a receiving electrode (2) sends the measured fluctuation signal under different flow rate conditions in the pipeline to another path of the external detection unit. The two paths of signals are mixed in the external detection unit, and the mixed signals are passed through an adder to obtain a microwave difference frequency signal.

[0044] 3) Construct a data set, specifically preprocess the data, then use a sliding window to obtain samples from the data, add corresponding labels to the samples, and then use 80% of the samples as a training set, 10% of the samples as a validation set, and the remaining 10% of the samples as a test set. Specifically include:

[0045] (3.1) First, the microwave difference frequency signal measured by the double-helix high-frequency microwave sensor is preprocessed, and the formula is as follows:

[0046]

[0047] Among them, I i Represents the i-th data, I mean and I std are the mean and standard deviation of the data, is the i-th data after preprocessing;

[0048] (3.2) The preprocessed signal is segmented using a non-overlapping windowing method, with a window length of H and a data acquisition length of L. samples, among which, represents rounding down; the microwave difference frequency signal is windowed and intercepted, and the actual value is used as the data label, and a total of N fluctuation samples with label values ​​are obtained;

[0049] (3.3) Randomly divide N fluctuation samples into a data set with training set, validation set and test set, and the specific ratio is [training set: validation set: test set] = [8:1:1].

[0050] 4) Construct a spatiotemporal parallel feature fusion network to realize the feature fusion and flow measurement of gas-solid two-phase flow fluctuation signals or the feature fusion and water content measurement of oil-water two-phase flow fluctuation signals.

[0051] like Figure 2As shown, the spatiotemporal parallel feature fusion network is divided into two parallel sub-networks. The first sub-network is a 2-layer long short-term memory network module, which uses the information in the storage unit to capture the long-term dependency in the sequence, extract the time series features, and obtain the associated information. The second sub-network is a convolutional neural network module composed of a 3-layer convolutional network. After convolution and pooling operations on the input data, the local features of the sequence are extracted and potential information is mined; the different features of the microwave difference frequency signal are extracted by merging the long short-term memory network module and the convolutional neural network module; the extracted microwave difference frequency signal is trained through two fully connected layers to finally obtain the predicted value.

[0052] Setting hyperparameters in the spatiotemporal parallel feature fusion network, where the hyperparameters are updated based on the training data by gradient reverse using the Adam (adaptive moment estimation) optimization algorithm;

[0053] The mean square error (MSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) are used to construct the loss function between the predicted value and the label value of the spatiotemporal parallel feature fusion network, which is used to calculate the gap between the forward calculation result of each iteration of the spatiotemporal parallel feature fusion network and the true value. Minimizing the gap is used as the criterion to guide the training and optimization of the spatiotemporal parallel feature fusion network in the right direction.

[0054] The specific operation of the spatiotemporal parallel feature fusion network includes:

[0055] (4.1) Input the microwave difference frequency signal into the first subnetwork in time sequence for feature extraction. Let t be the current time and the input be x t , through the weight W and bias b, combined with the hidden layer output h of the previous time node t-1 , calculate the three functions of input gate p(t), forget gate f(t) and output gate o(t), and the calculation formula of memory unit M(t) is as follows:

[0056] M(t)=f(t)M(t-1)+p(t)(W*[h t-1 ,x t ]+b)

[0057] The calculation formula of the state output unit is:

[0058] h(t)=o(t)tanh(M(t))

[0059] The obtained hidden layer state output h(t) is combined with the input at time t+1 as the input of the next hidden layer and calculated in sequence;

[0060] (4.2) The microwave difference frequency signal is input into the second sub-network in time sequence for feature extraction; wherein the convolutional neural network module includes three convolutional layers (Conv1D) connected in series, wherein the outputs of the first convolutional layer and the second convolutional layer are respectively connected to a maximum pooling layer (Maxpooling1D) to prevent overfitting;

[0061] (4.3) The different features extracted by the long short-term memory network module and the convolutional neural network module are fused through the merging layer, and the fused feature vector is input into two fully connected layers with the actual value as the label. The last fully connected layer is used as the output layer and the ReLU function is used as the activation function to finally output the predicted value.

Claims

1. A two-phase flow parameter measurement method based on a spatiotemporal parallel feature fusion network, characterized in that: The steps include: 1) Design a finite element simulation model of a double-helix high-frequency microwave sensor through a finite element simulation model, optimize parameters, and obtain the optimal structure of the double-helix high-frequency microwave sensor; The finite element simulation model of the double-helix high-frequency microwave sensor has an interval angle of 30° between the excitation electrode and the protection electrode, an electrode pitch length of 90 mm, an insulating layer thickness of 0.05 mm, an inner shielding layer inner diameter of 40 mm, and a thickness of 0.5 mm; the input and output ports of the finite element simulation model of the double-helix high-frequency microwave sensor are set as wave ports, the characteristic impedance is 50 ohms, and an automatic finite element mesh is set; the operating frequency is set to 0.8 GHz; and the electrode angle is set to 70°; The double-helix high-frequency microwave sensor is a hollow tube structure formed by sequentially arranging an inner shielding layer (6), an insulating layer (5), a measuring electrode layer and an outer shielding layer from the inside to the outside, wherein the inner shielding layer (6) is used to be sheathed on the outer wall of the measuring pipe, and the measuring electrode layer is formed by spirally winding an excitation electrode (1), a first protective electrode (3), a receiving electrode (2) and a second protective electrode (4) in parallel on the outer wall of the inner shielding layer (6), and a gap is formed between the excitation electrode (1), the first protective electrode (3), the receiving electrode (2) and the second protective electrode (4); 2) Using double-helix high-frequency microwave sensor to measure the fluctuation signal of gas-solid or oil-water two-phase flow; including: When a gas-solid or oil-water two-phase flow enters a measuring pipeline section equipped with a double-helix high-frequency microwave sensor from a pipeline, the double-helix high-frequency microwave sensor collects the two-phase flow fluctuation signal. Specifically, a sinusoidal excitation signal source generates an excitation signal, which is sent to an excitation electrode (1) of the double-helix high-frequency microwave sensor through a power divider to excite the gas-solid or oil-water two-phase flow, and is sent to an external detection unit; at the same time, a receiving electrode (2) sends the measured fluctuation signal under different flow rate conditions in the pipeline to another path of the external detection unit, and the two paths of signals are mixed in the external detection unit. The mixed signal is passed through an adder to obtain a microwave difference frequency signal. 3) Construct a data set, specifically preprocess the data, then use a sliding window to obtain samples from the data, add corresponding labels to the samples, and then use 80% of the samples as a training set, 10% of the samples as a validation set, and the remaining 10% of the samples as a test set; including: (3.1) First, the microwave difference frequency signal measured by the double-helix high-frequency microwave sensor is preprocessed, and the formula is as follows: Among them, I i Represents the i-th data, I mean and I std are the mean and standard deviation of the data, is the i-th data after preprocessing; (3.2) The preprocessed signal is segmented using a non-overlapping windowing method, with a window length of H and a data acquisition length of L. samples, among which, represents rounding down; the microwave difference frequency signal is windowed and intercepted, and the actual value is used as the data label, and a total of N fluctuation samples with label values ​​are obtained; (3.3) Randomly divide N fluctuation samples into a data set with training set, validation set and test set, with the specific ratio of [training set: validation set: test set] = [8:1:1]; 4) Constructing a spatiotemporal parallel feature fusion network to realize the feature fusion and flow measurement of gas-solid two-phase flow flow fluctuation signals or the feature fusion and water content measurement of oil-water two-phase flow fluctuation signals; The spatiotemporal parallel feature fusion network is divided into two parallel sub-networks. The first sub-network is a 2-layer long short-term memory network module, which uses the information in the storage unit to capture the long-term dependency in the sequence, extract the time series features, and obtain the associated information. The second sub-network is a convolutional neural network module composed of a 3-layer convolutional network. After convolution and pooling operations are performed on the input data, the local features of the sequence are extracted and potential information is mined; different features of the microwave difference frequency signal are extracted by fusing the long short-term memory network module and the convolutional neural network module through the merging layer; the extracted microwave difference frequency signal is trained through two fully connected layers to finally obtain the predicted value; The specific operation of the spatiotemporal parallel feature fusion network includes: (4.1) Input the microwave difference frequency signal into the first subnetwork in time sequence for feature extraction. Let t be the current time and the input be x t , through the weight W and bias b, combined with the hidden layer output h of the previous time node t-1 , calculate the three functions of input gate p(t), forget gate f(t) and output gate o(t), and the calculation formula of memory unit M(t) is as follows: M(t)=f(t)M(t-1)+p(t)(W*[h t-1 ,x t ]+b) The calculation formula of the state output unit is: h(t)=o(t)tanh(M(t)) The obtained hidden layer state output h(t) is combined with the input at time t+1 as the input of the next hidden layer and calculated in sequence; (4.2) Inputting the microwave difference frequency signal into the second sub-network in time sequence for feature extraction; wherein the convolutional neural network module includes three convolutional layers connected in series, wherein the outputs of the first convolutional layer and the second convolutional layer are respectively connected to a maximum pooling layer to prevent overfitting; (4.3) The different features extracted by the long short-term memory network module and the convolutional neural network module are fused through the merging layer, and the fused feature vector is input into two fully connected layers with the actual value as the label. The last fully connected layer is used as the output layer and the ReLU function is used as the activation function to finally output the predicted value.

2. A two-phase flow parameter measurement method based on a time-space parallel feature fusion network according to claim 1, characterized in that: In the spatiotemporal parallel feature fusion network described in step 4), Set hyperparameters, where the hyperparameters are updated based on the training data using the Adam optimization algorithm through gradient back-up to the time-space parallel feature fusion network weights; The mean square error, mean absolute error and mean absolute percentage error are used to construct the loss function between the predicted value and the label value of the spatiotemporal parallel feature fusion network, which is used to calculate the gap between the forward calculation result of each iteration of the spatiotemporal parallel feature fusion network and the true value. Minimizing the gap is used as the criterion to guide the training and optimization of the spatiotemporal parallel feature fusion network in the right direction.