Multiphase flow metering system and method

Abstract

The invention relates to the technical field of flow metering, in particular to a multiphase flow metering system and method.The multiphase flow metering method comprises the steps that acoustic emission signals, microwave attenuation signals and differential pressure pulsation signals at the section of a pipeline are synchronously obtained through a distributed sensor array, and original sensing data are formed; performing time-frequency conjoint analysis on the original sensing data, and extracting a flow state feature vector; flow state evolution trend prediction: inputting the flow state feature vector into a long-short term memory network model optimized through transfer learning, and outputting a predicted flow state parameter set in the future 3 seconds; on the basis of the predicted flow state parameter set, a compensation coefficient matrix of the flow of each phase is generated in combination with a preset interface coupling factor database; according to the compensation coefficient matrix, the fusion weight of the multi-sensor data is adjusted in a self-adaptive mode, and an optimal metering model under the current flow state is constructed; according to the method, the flow state parameter set in the next three seconds can be accurately predicted on the basis of the real-time flow state characteristics by using the long-short-term memory network model optimized by transfer learning.

Description

Technical Field

[0001] The present invention relates to the technical field of flow measurement, and particularly to a multiphase flow measurement system and method. Background Art

[0002] With the continuous development of multiphase flow measurement technology in industrial processes and pipeline transportation, accurate and efficient multiphase flow measurement has become a key technical requirement in many fields. Multiphase flow measurement requires real-time collection and analysis of data from different sensors to accurately reflect the state and changes of the fluid in the pipeline. In multiphase flow measurement systems, common sensors include acoustic emission sensors, microwave attenuation sensors, and differential pressure sensors, which can provide information about fluid density, flow velocity, and flow pattern, etc.

[0003] Currently, flow pattern prediction and parameter modeling have become one of the core technologies of multiphase flow measurement. In traditional flow measurement methods, flow calculation usually relies on direct measurement of sensor data and simple statistical methods, but these methods often ignore the dynamic changes of the flow pattern and its impact on flow measurement. Therefore, using historical flow pattern data and real-time sensor data, predicting the evolution trend of the flow pattern through advanced algorithms and models, and compensating for the flow rate, is an important means to improve measurement accuracy and reliability. Summary of the Invention

[0004] Based on the above purpose, the present invention provides a multiphase flow measurement method.

[0005] A multiphase flow measurement method includes the following steps: S1, synchronous acquisition of multi-source heterogeneous data: synchronously obtain acoustic emission signals, microwave attenuation signals, and differential pressure pulsation signals at the pipeline cross-section through a distributed sensor array to form raw sensing data in the three-dimensional space-time domain; S2, extraction of dynamic flow pattern features: perform time-frequency joint analysis on the raw sensing data to extract flow pattern feature vectors, where the flow pattern feature vectors include flow pattern identification codes, interphase slip rates, and interface fluctuation intensity coefficients; S3, prediction of flow pattern evolution trend: input the flow pattern feature vectors into a long short-term memory network model optimized by transfer learning, and output a set of predicted flow pattern parameters within the next 3 seconds; S4, multiphase coupling compensation calculation: based on the set of predicted flow pattern parameters, combined with a preset interface coupling factor database, generate a compensation coefficient matrix for the flow rate of each phase; S5, construction of a dynamic weight measurement model: adaptively adjust the fusion weights of multi-sensor data according to the compensation coefficient matrix to construct an optimal measurement model under the current flow pattern.

[0006] Further, the S1 includes: S11, Multi-source sensor signal acquisition: Through a distributed sensor array, various sensing signals at the pipeline cross-section are acquired, including acoustic emission signals , microwave attenuation signals and differential pressure pulsation signals ; S12, Data synchronization: Through GPS signals, the sensing signals acquired by different sensors at time are synchronously stored, including acoustic emission signals, microwave attenuation signals, and differential pressure pulsation signals; S13, Spatial calibration and data arrangement: Each sensor position is calibrated through spatial coordinates to represent the position of the sensor on the pipeline cross-section. By pairing the sensing signals with their corresponding spatial coordinates, a three-dimensional data matrix at each moment is obtained, forming the original sensing data in the space-time domain .

[0007] Furthermore, the S2 includes: S21, Preprocessing of original sensing data: Filtering and denoising processing are performed on the original sensing data ; S22, Time-frequency joint analysis: The short-time Fourier transform (STFT) is used to perform time-frequency domain transformation on the preprocessed original sensing data, expressed as: , where is the time-frequency diagram obtained by the short-time Fourier transform, is time, is frequency; S23, Feature extraction: Flow regime feature vectors are extracted from the time-frequency diagram after time-frequency domain transformation. The flow regime feature vectors include flow pattern identification codes , interphase slip rates and interfacial fluctuation intensity coefficients .

[0008] Furthermore, the S3 includes: S31, Constructing a training dataset: According to the flow regime feature vectors at historical moments and the corresponding historical flow regime parameter sets , a training dataset is constructed. The input of the training dataset is a sequence of flow regime feature vectors, and the output is the flow regime parameter set at the corresponding moment , expressed as: ; ; where is the number of selected historical moments, The input sequence in the training dataset contains the flow pattern feature vectors of the past time instants, and the target output sequence in the training dataset contains the flow pattern parameter sets of the past time instants; S32, Long Short-Term Memory Network Model Training and Transfer Learning Optimization: Use the historical flow pattern feature vectors and the corresponding flow pattern parameter sets to train the Long Short-Term Memory Network model. The Long Short-Term Memory Network learns the temporal relationship between the input sequence and the target flow pattern parameter set to optimize the network weights and predict the future flow pattern parameter set; S33, Flow Pattern Parameter Set Prediction: Input the flow pattern feature vector at the current moment into the trained Long Short-Term Memory Network model to predict the flow pattern parameter set within the next 3 seconds, denoted as: ; Through the forward propagation of the Long Short-Term Memory Network, the flow pattern parameter set at future time instants is denoted as: ; where correspond to the prediction results at 1 second, 2 seconds, and 3 seconds in the future respectively, is the prediction function generated by the Long Short-Term Memory Network; S34, Output Prediction Results: Output the obtained flow pattern parameter set within the next 3 seconds , including flow rate and pressure.

[0009] Furthermore, the S4 includes: S41, Matching Predicted Flow Pattern Parameters with Interface Coupling Factors: Based on the flow regime identification code in the predicted flow pattern parameter set , look up the corresponding coupling factor in the interface coupling factor database, denoted as: ; where represents the interface coupling factor matching the flow pattern feature , is the operation of querying from the coupling factor database; S42, Calculating the Compensation Coefficient Matrix for Each Phase Flow Rate: Combine the predicted flow pattern parameter set and the interface coupling factor , calculate according to the flow pattern features to generate the compensation coefficient matrix for each phase flow rate.

[0010] Further, S42 includes: S421, calculating the flow compensation factor: performing a weighted calculation on the interface coupling factor and the flow regime parameter to obtain the compensation factor for each phase , expressed as: ; wherein, represents the flow rate or other physical parameter of the th phase, and is the compensation factor for this phase; S422, generating a compensation coefficient matrix: combining the compensation factors of all phases to form a compensation coefficient matrix , expressed as: ; wherein, is the compensation factor for each phase, and is the final compensation coefficient matrix.

[0011] Further, S5 includes: S51, calculating the weighted fusion coefficient for each sensor according to the compensation coefficient matrix ; S52, adjusting the multi-sensor data fusion weights to construct an optimal measurement model.

[0012] Further, the weighted fusion coefficient for each sensor in S51 is expressed as: ; wherein, is the compensation factor of the th phase, is the original measurement data of the th sensor, and is the weighted fusion coefficient.

[0013] Further, the construction of the optimal measurement model in S52 is expressed as: ; wherein, is the weighted coefficient of the th sensor, is the measurement data of the th sensor, and is the optimal measurement model generated based on the weighted fusion data.

[0014] A multiphase flow metering system for implementing the above multiphase flow metering method includes the following modules: Multi-source heterogeneous data acquisition module: It is used to synchronously obtain acoustic emission signals, microwave attenuation signals, and differential pressure pulsation signals at the pipe cross-section through a distributed sensor array, and form raw sensing data in the three-dimensional space-time domain; Time-frequency joint analysis module: It is used to perform time-frequency joint analysis on the raw sensing data and extract flow state feature vectors, including flow pattern identification codes, inter-phase slip rates, and interface fluctuation intensity coefficients; Flow state evolution prediction module: It is used to input the flow state feature vectors into a long short-term memory network model optimized by transfer learning and output a set of predicted flow state parameters within the next 3 seconds; Multi-phase coupling compensation calculation module: It is used to generate a compensation coefficient matrix for the flow rates of each phase based on the set of predicted flow state parameters and in combination with a pre-set interface coupling factor database; Dynamic weight measurement model construction module: It is used to adaptively adjust the fusion weights of multi-sensor data according to the compensation coefficient matrix and construct an optimal measurement model under the current flow state.

[0015] Advantages of the present invention: In the present invention, by using a long short-term memory network model optimized by transfer learning, it is possible to accurately predict a set of flow state parameters within the next 3 seconds based on real-time flow state features. By combining historical flow state features and real-time sensor data, the LSTM model can effectively capture the temporal characteristics of flow state evolution, overcome the problem that traditional methods cannot predict flow state changes in real time, and achieve accurate prediction of future flow states.

[0016] In the present invention, by combining the set of predicted flow state parameters with a pre-set interface coupling factor database, it is possible to generate a compensation coefficient matrix for the flow rates of each phase, which is adaptively adjusted according to the results of real-time flow state prediction, improving the accuracy of flow compensation calculation. This innovative compensation calculation method can dynamically adjust the compensation factor according to the actual changes in the flow state, thereby overcoming the errors that may be caused by traditional fixed compensation factor methods under different flow state conditions and ensuring the accuracy and reliability of flow measurement under complex and variable flow states. Description of the drawings

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only those of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0018] Figure 1 It is a schematic diagram of the method flow of the embodiment of the present invention; Figure 2 It is a schematic diagram of the system module of the embodiment of the present invention. Detailed implementation manners

[0019] To make the objectives, technical solutions, and advantages of the present invention more clear and understandable, the present invention will be further described in detail below in conjunction with specific embodiments.

[0020] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should have the ordinary meanings understood by those with ordinary skills in the field to which the present invention belongs. The "first", "second", and similar terms used in the present invention do not indicate any order, quantity, or importance, but are only used to distinguish different components. Words such as "including" or "comprising" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. Words such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right", etc. are only used to represent relative position relationships, and when the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0021] As Figure 1 shown, a multiphase flow metering method includes the following steps: S1, multi-source heterogeneous data synchronous acquisition: Synchronously obtain acoustic emission signals, microwave attenuation signals, and differential pressure pulsation signals at the pipeline cross-section through a distributed sensor array to form raw sensing data in the three-dimensional space-time domain; S2, dynamic flow regime feature extraction: Perform time-frequency joint analysis on the raw sensing data to extract flow regime feature vectors, where the flow regime feature vectors include flow pattern identification codes, inter-phase slip rates, and interfacial fluctuation intensity coefficients; S3, flow regime evolution trend prediction: Input the flow regime feature vectors into a long short-term memory network model optimized by transfer learning, and output a set of predicted flow regime parameters within the next 3 seconds; S4, multiphase coupling compensation calculation: Based on the set of predicted flow regime parameters, combined with a pre-set interfacial coupling factor database, generate a compensation coefficient matrix for the flow rates of each phase; S5, dynamic weight metering model construction: Adaptively adjust the fusion weights of multi-sensor data according to the compensation coefficient matrix to construct an optimal metering model under the current flow regime.

[0022] S1 includes: S11, multi-source sensor signal acquisition: Through a distributed sensor array, collect various sensing signals at the pipeline cross-section, including acoustic emission signals (representing high-frequency vibration signals generated in the pipeline), microwave attenuation signals (representing the attenuation effect of the fluid medium in the pipeline on microwaves) and differential pressure pulsation signals (indicating the pressure fluctuation generated by the fluid flow in the pipeline), where is time, is the th sensor, (ultrasonic sensors are used to collect acoustic emission signals, microwave reflectometers are used to collect microwave attenuation signals, and piezoelectric differential pressure sensors are used to collect differential pressure pulsation signals); S12, data synchronization: Through GPS signals, synchronously store the sensing signals collected by different sensors at time including acoustic emission signals, microwave attenuation signals, and differential pressure pulsation signals; S13, spatial calibration and data arrangement: The position of each sensor is calibrated through spatial coordinates indicating the position of the sensor on the pipeline cross-section. By pairing the sensing signals with their corresponding spatial coordinates, a three-dimensional data matrix at each moment is obtained, forming the original sensing data in the space-time domain , expressed as: ; where each row represents the measurement result of a sensor at time containing three signal types.

[0023] S2 includes: S21, preprocessing of original sensing data: Filter and denoise the original sensing data to remove noise interference; S22, time-frequency joint analysis: Use the short-time Fourier transform (STFT) to perform time-frequency domain transformation on the preprocessed original sensing data, expressed as: , where is the time-frequency diagram obtained by the short-time Fourier transform, is time, is frequency, represents the short-time Fourier transform; S23, feature extraction: Extract the flow regime feature vector from the time-frequency diagram after time-frequency domain transformation. The flow regime feature vector includes the flow pattern identification code , the interfacial slip rate and the interfacial fluctuation intensity coefficient ; The flow regime feature vector specifically includes: Flow pattern identification code: By performing K-means analysis on the time-frequency diagram, the flow pattern identification code is obtained, indicating the current flow regime type (e.g., bubbly flow, stratified flow, mixed flow, etc.); Interfacial slip rate: By calculating the relative velocity between different phases, the interfacial slip rate is obtained , that is, the velocity difference between the gas-liquid phase or the liquid-solid phase, is expressed as: ; Among them, and are the velocities of the two phases respectively, is the cross-sectional length of the pipeline for fluid flow; Interface fluctuation intensity coefficient: The interface fluctuation intensity coefficient is extracted through the amplitude information of the frequency components in the time-frequency diagram represents the fluctuation intensity of the interfacial phase between phases, and is expressed as: ; Among them, is the standard deviation, is the mean value is the result of the short-time Fourier transform.

[0024] S3 includes: S31, constructing a training data set: According to the flow pattern feature vectors at historical moments and the corresponding historical flow pattern parameter sets , a training data set is constructed. The input of the training data set is a sequence of flow pattern feature vectors, and the output is the flow pattern parameter set at the corresponding moment , which is expressed as: ; ; Among them, is the number of historical moments selected, The input sequence in the training data set contains the flow pattern feature vectors of the past moments, is the target output sequence in the training data set, which contains the flow pattern parameter sets of the past moments; S32, training of the long short-term memory network model and optimization of transfer learning: Use the historical flow pattern feature vectors and the corresponding flow pattern parameter sets to train the long short-term memory network model; The transfer learning technique is used in the training process. By transferring the model parameters from the pre-trained source model (based on different flow pattern data sets), it can quickly adapt to the target flow pattern data set; The long short-term memory network optimizes the network weights by learning the temporal relationship between the input sequence and the target flow pattern parameter set , and predicts the future flow pattern parameter set; The output (hidden state) of the long short-term memory network is recursively updated through the following formula: ; Among them, is the weight matrix of the long short-term memory network, is the bias term, is the sigmoid activation function, is the hyperbolic tangent activation function; S33, prediction of the flow regime parameter set: Input the flow regime feature vector at the current moment into the trained long short-term memory network model to predict the flow regime parameter set within the next 3 seconds, expressed as: ; Through the forward propagation of the long short-term memory network, the flow regime parameter set at the future moment is expressed as: ; Among them, correspond to the prediction results at 1 second, 2 seconds, and 3 seconds in the future respectively, is the prediction function generated by the long short-term memory network; S34, output the prediction result: Output the obtained flow regime parameter set within the next 3 seconds, including flow rate and pressure.

[0025] S4 includes: S41, match the predicted flow regime parameters with the interface coupling factor: Based on the flow regime identification code in the predicted flow regime parameter set , search for the corresponding coupling factor from the interface coupling factor database, expressed as: ; Among them, represents the interface coupling factor matching the flow regime feature , is the operation of querying from the coupling factor database; S42, calculate the compensation coefficient matrix of each phase flow rate: Combine the predicted flow regime parameter set and the interface coupling factor , calculate according to the flow regime characteristics to generate the compensation coefficient matrix of each phase flow rate.

[0026] S42 includes: S421, calculate the flow rate compensation factor: Perform weighted calculation on the interface coupling factor and the flow regime parameter to obtain the compensation factor of each phase, expressed as: ; Among them, represents the The flow rate or other physical parameters of the phase, is the compensation factor for the phase; S422, generating a compensation coefficient matrix: Combining the compensation factors of all phases into a compensation coefficient matrix , expressed as: ; where, is the compensation factor for each phase, is the final compensation coefficient matrix.

[0027] S5 includes: S51, according to the compensation coefficient matrix , calculating the weighted fusion coefficient of each sensor; S52, adjusting the multi-sensor data fusion weights to construct an optimal measurement model.

[0028] The weighted fusion coefficient of each sensor in S51 is expressed as: ; where, is the compensation factor of the th phase, is the original measurement data of the th sensor, is the weighted fusion coefficient.

[0029] The construction of the optimal measurement model in S52 is expressed as: ; where, is the weighted coefficient of the th sensor, is the measurement data of the th sensor, is the optimal measurement model generated based on the weighted fusion data.

[0030] As Figure 2 shown, a multiphase flow metering system for implementing the above-mentioned multiphase flow metering method includes the following modules: Multi-source heterogeneous data acquisition module: Used to synchronously obtain acoustic emission signals, microwave attenuation signals, and differential pressure pulsation signals at the pipeline cross-section through a distributed sensor array, forming raw sensing data in the three-dimensional space-time domain; Time-frequency joint analysis module: Used to perform time-frequency joint analysis on the raw sensing data to extract flow state feature vectors, including flow pattern identification codes, inter-phase slip rates, and interface fluctuation intensity coefficients; Flow state evolution prediction module: It is used to input the flow state feature vector into the long short-term memory network model optimized by transfer learning and output the predicted flow state parameter set within the next 3 seconds; Multiphase coupling compensation calculation module: It is used to generate the compensation coefficient matrix of each phase flow rate based on the predicted flow state parameter set and in combination with the preset interface coupling factor database; Dynamic weight measurement model construction module: It is used to adaptively adjust the fusion weights of multi-sensor data according to the compensation coefficient matrix and construct the optimal measurement model under the current flow state.

[0031] Those of ordinary skill in the art should understand that the discussion of any of the above embodiments is only exemplary and is not intended to imply that the scope of the present invention is limited to these examples; under the idea of the present invention, the technical features in the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations in different aspects of the present invention as described above, which are not provided in detail for the sake of brevity.

Claims

1. A multiphase flow measurement method, characterized in that: The following steps are involved: S1, synchronous acquisition of multi-source heterogeneous data: The acoustic emission signal, microwave attenuation signal and differential pressure pulsation signal at the pipeline cross section are synchronously acquired through a distributed sensor array to form the original sensor data in the three-dimensional space-time domain; S2, dynamic flow pattern feature extraction: perform time-frequency joint analysis on the original sensor data to extract the flow pattern feature vector, which includes the flow pattern identification code, interphase slip rate and interface fluctuation intensity coefficient; S3, flow state evolution trend prediction: the flow state feature vector is input into the long short-term memory network model optimized by transfer learning, and the predicted flow state parameter set within the next 3 seconds is output; S4, multiphase coupling compensation calculation: based on the predicted flow state parameter set and combined with the preset interface coupling factor database, the compensation coefficient matrix of each phase flow is generated; S5, dynamic weight metering model construction: adaptively adjust the fusion weight of multi-sensor data according to the compensation coefficient matrix to build the optimal metering model under the current flow state.

2. A multiphase flow measurement method according to claim 1, characterized in that: The S1 includes: S11, multi-source sensor signal acquisition: through the distributed sensor array, collect multiple sensor signals at the pipeline cross section, including acoustic emission signals , microwave attenuation signal and differential pressure pulsation signal ; S12, data synchronization: through GPS signals, synchronize and store different sensors at all times The collected sensor signals include acoustic emission signals, microwave attenuation signals, and differential pressure pulsation signals; S13, spatial calibration and data collation: Each sensor position is determined by spatial coordinates Calibrate to indicate sensor By pairing the sensor signal with its corresponding spatial coordinates, the three-dimensional data matrix at each moment is obtained to form the original sensor data in the space-time domain. .

3. A multiphase flow measurement method according to claim 2, characterized in that: The S2 includes: S21, raw sensor data preprocessing: Perform filtering and denoising; S22, time-frequency joint analysis: Short-time Fourier transform is used to transform the preprocessed raw sensor data into time-frequency domain, which is expressed as: ,in, is the time-frequency diagram obtained by short-time Fourier transform, For time, is the frequency; S23, feature extraction: extract the flow state feature vector from the time-frequency diagram after time-frequency domain transformation , the flow pattern feature vector includes the flow pattern identification code , interphase slip rate and the interface wave intensity coefficient .

4. A multiphase flow measurement method according to claim 3, characterized in that: The S3 includes: S31, build training data set: according to the flow feature vector at historical moments and the corresponding historical flow state parameter set , construct a training data set, the training process includes: input sequence in the training data set , the output is the target output sequence in the training data set , Including the flow parameter set at the corresponding time ; S32, Long Short-Term Memory Network Model Training and Transfer Learning Optimization: Using Historical Flow Feature Vectors and the corresponding flow parameter set The long short-term memory network model is trained. The long short-term memory network learns the input sequence and target flow parameter set The temporal relationship of , predict the future flow state parameter set; S33, flow state parameter set prediction: the flow state feature vector at the current moment Input into the trained long short-term memory network model to predict the flow parameter set within the next 3 seconds, expressed as: ； Through the forward propagation of the long short-term memory network, the flow parameter set at the future moment It is expressed as: ； in, The prediction results for the next 1 second, 2 seconds, and 3 seconds respectively. is the prediction function generated by the long short-term memory network; S34, output prediction results: output the flow state parameter set within the next 3 seconds , including flow and pressure.

5. A multiphase flow measurement method according to claim 4, characterized in that: The S4 includes: S41, Matching predicted flow parameters with interface coupling factors: Based on the predicted flow parameter set Flow type identification code , find the corresponding coupling factor from the interface coupling factor database and express it as: ； in, Representation and flow characteristics Matching interface coupling factor, It is an operation to query from the coupling factor database; S42, calculate the compensation coefficient matrix of each phase flow: combined with the predicted flow parameter set and interface coupling factor , calculate according to the flow characteristics and generate the compensation coefficient matrix of each phase flow .

6. A multiphase flow measurement method according to claim 5, characterized in that: The S42 includes: S421, calculate the flow compensation factor: the interface coupling factor and flow parameters Perform weighted calculation to obtain the compensation factor for each phase ; S422, generate compensation coefficient matrix: the compensation factors of all phases Combined into compensation coefficient matrix .

7. A multiphase flow measurement method according to claim 6, characterized in that: The S5 includes: S51, according to the compensation coefficient matrix , calculate the weighted fusion coefficient of each sensor; S52, adjust the multi-sensor data fusion weights and build the optimal measurement model.

8. A multiphase flow measurement method according to claim 7, characterized in that: The weighted fusion coefficient of each sensor in S51 is expressed as: ； in, It is Phase compensation factor, It is The raw measurement data of each sensor, is the weighted fusion coefficient.

9. A multiphase flow measurement method according to claim 8, characterized in that: The optimal measurement model constructed in S52 is expressed as: ； in, It is The weighting coefficients of the sensors are It is The measurement data of the sensors, It is the optimal measurement model generated based on weighted fusion data.

10. A multiphase flow metering system, used to implement a multiphase flow metering method as claimed in any one of claims 1 to 9, characterized in that: Includes the following modules: Multi-source heterogeneous data acquisition module: used to synchronously acquire acoustic emission signals, microwave attenuation signals and differential pressure pulsation signals at the pipe cross section through a distributed sensor array to form original sensing data in the three-dimensional space-time domain; Time-frequency joint analysis module: used to perform time-frequency joint analysis on the original sensor data and extract flow pattern feature vectors, including flow pattern identification code, interphase slip rate and interface fluctuation intensity coefficient; Flow pattern evolution prediction module: used to input the flow pattern feature vector into the long short-term memory network model optimized by transfer learning, and output the predicted flow pattern parameter set within the next 3 seconds; Multiphase coupling compensation calculation module: used to generate the compensation coefficient matrix of each phase flow based on the predicted flow state parameter set and the preset interface coupling factor database; Dynamic weighted metering model building module: used to adaptively adjust the fusion weights of multi-sensor data according to the compensation coefficient matrix to build the optimal metering model under the current flow state.

Images