Wellhead three-phase holdup metering method based on multi-parameter fusion

Abstract

The invention discloses a wellhead three-phase holdup metering method based on multi-parameter fusion, and a temperature and pressure multi-parameter fusion network comprises a temperature and pressure NN-LSTM module, a microwave Transform module, an attention multi-parameter aggregation module and a prediction output module. The input samples of the temperature-pressure electric multi-parameter fusion network are temperature-related signals, pressure-related signals, microwave difference frequency phase-related signals and microwave difference frequency amplitude-related signals, and the output samples are moisture content predicted values of the samples. The wellhead three-phase holdup metering method comprises the following steps: obtaining a to-be-predicted sample; and predicting a to-be-predicted sample by using the trained temperature piezoelectric multi-parameter fusion network to obtain a water content predicted value of the to-be-predicted sample. According to the wellhead three-phase volume fraction metering method, the water content of the petroleum wellhead production liquid is accurately predicted.

Description

Technical Field

[0001] The invention belongs to the field of oil well parameter measurement, and in particular relates to a wellhead three-phase content measurement method based on multi-parameter fusion. Background Art

[0002] In the process of oilfield exploitation, accurate measurement of the water content of the three-phase fluid at the wellhead is of great significance to production scheduling, equipment maintenance and optimization management. At present, the traditional manual sampling and testing methods have high labor intensity. Although the commonly used sensors based on electrical conductivity, capacitance and radioactive sources can provide a certain monitoring accuracy, their cost is relatively high. In addition, in practical applications, measurement methods that rely solely on artificial intelligence often face the problem of insufficient measurement accuracy.

[0003] With the continuous development of sensor technology, the application of microwave sensors in wellhead fluid monitoring has gradually been promoted. By measuring the amplitude and phase changes of microwave signals, the composition of the fluid can be indirectly reflected. However, a single sensor often cannot meet the high-precision requirements, and for complex wellhead fluid state changes, the traditional data processing method based on a single sensor cannot fully tap the potential of multi-source information.

[0004] Therefore, a high-precision, real-time wellhead three-phase content measurement method is urgently needed. Summary of the invention

[0005] In view of the deficiencies in the prior art, an object of the present invention is to provide a thermo-piezoelectric multi-parameter fusion network.

[0006] Another object of the present invention is to provide a wellhead three-phase content measurement method based on multi-parameter fusion, which enables the temperature-piezoelectric multi-parameter fusion network to accurately predict the water content of the oil wellhead produced fluid, thereby realizing high-precision prediction and real-time monitoring of the water content of the oil wellhead produced fluid.

[0007] The present invention is achieved through the following technical solutions.

[0008] A temperature-pressure-electricity multi-parameter fusion network, comprising: a temperature-pressure CNN-LSTM module, a microwave Transformer module, an attention multi-parameter aggregation module and a prediction output module;

[0009] The temperature-pressure CNN-LSTM module includes: a first CNN-LSTM module and a second CNN-LSTM module, the input signal of the first CNN-LSTM module is a temperature-related signal, and the output is a feature of the temperature-related signal, and the input signal of the second CNN-LSTM module is a pressure-related signal, and the output is a feature of the pressure-related signal;

[0010] The microwave Transformer module includes: a first Transformer model and a second Transformer model, wherein the input signal of the first Transformer model is a microwave difference frequency phase-related signal, and the output is a feature of the microwave difference frequency phase-related signal; the input signal of the second Transformer model is a microwave difference frequency amplitude-related signal, and the output is a feature of the microwave difference frequency amplitude-related signal;

[0011] The attention multi-parameter aggregation module uses the attention fusion mechanism to perform weighted fusion on the features of temperature-related signals, pressure-related signals, microwave difference frequency phase-related signals, and microwave difference frequency amplitude-related signals to obtain fused features.

[0012] The temperature-related signal, the pressure-related signal, the microwave difference frequency phase-related signal and the microwave difference frequency amplitude-related signal are respectively the signal after the temperature signal is standardized, the signal after the pressure signal is standardized, the signal after the microwave difference frequency phase signal is standardized and the signal after the microwave difference frequency amplitude signal is standardized, wherein the temperature signal, the pressure signal, the microwave difference frequency phase signal and the microwave difference frequency amplitude signal are obtained by the corresponding sensors collecting the oil wellhead production liquid in the same time period;

[0013] The input of the prediction output module is the fusion feature, and the output value is the predicted value of the water content of the oil wellhead production liquid collected in this time period. The prediction output module uses Sigmoid as the activation function.

[0014] A wellhead three-phase content measurement method based on multi-parameter fusion, comprising:

[0015] Step 1, obtaining a temperature-related signal, a pressure-related signal, a microwave difference frequency phase-related signal, and a microwave difference frequency amplitude-related signal of a sample to be predicted;

[0016] Step 2: Use the trained temperature-piezoelectric multi-parameter fusion network to predict the sample to be predicted and obtain the moisture content prediction value of the sample to be predicted.

[0017] In the above technical scheme, the method for obtaining a trained thermo-piezoelectric multi-parameter fusion network includes: preparing multiple samples as training sets, each sample includes: temperature-related signals, pressure-related signals, microwave difference frequency phase-related signals, and microwave difference frequency amplitude-related signals, substituting the samples into the thermo-piezoelectric multi-parameter fusion network for training, substituting the moisture content prediction value of the sample and the moisture content label value of the sample into the loss function to obtain the loss function value, and using the AMSGrad optimization algorithm to calculate the gradient according to the loss function value, and updating the weights of the thermo-piezoelectric multi-parameter fusion network through the gradient back propagation method.

[0018] In the above technical solution, the water content label value is the true value of the water content of the oil wellhead production fluid within the sample collection time period.

[0019] In the above technical solution, the loss function is the mean square error.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] 1. The wellhead three-phase content measurement method of the present invention integrates the signals of multiple sensors, overcomes the limitations of a single sensor, improves the accuracy and reliability of prediction, and accurately predicts the water content of the oil wellhead production fluid with high prediction accuracy.

[0022] 2. The wellhead three-phase content measurement method of the present invention realizes the prediction of the water content of the oil wellhead production fluid through the temperature-piezoelectric multi-parameter fusion network, effectively improves the accuracy and real-time performance of wellhead fluid monitoring, and is of great significance to oilfield production management and equipment maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is the structural diagram of the temperature-piezoelectric multi-parameter fusion network;

[0024] Figure 2 It is the structural diagram of the multi-sensor measurement system;

[0025] Figure 3 Schematic diagram of the structure of the multi-sensor measurement system.

[0026] 1: Oil inlet pipe section, 2: Opposing microwave sensor, 3: Temperature sensor, 4: Pressure sensor, 5: Oil outlet pipe section, 6: Casing. DETAILED DESCRIPTION

[0027] The wellhead three-phase content measurement method of the present invention is described in detail below with reference to the accompanying drawings.

[0028] Example 1

[0029] A thermo-piezoelectric multi-parameter fusion network, such as Figure 1 As shown, it includes: warm compression CNN-LSTM module, microwave Transformer module, attention multi-parameter aggregation module and prediction output module;

[0030] The temperature-pressure CNN-LSTM module includes: a first CNN-LSTM module and a second CNN-LSTM module (the first CNN-LSTM module and the second CNN-LSTM module refer to: Li Mei, Ning Dejun, Guo Jiacheng. CNN-LSTM model based on attention mechanism and its application [J]. Computer Engineering and Applications, 2019, 55(13): 20-27.), the input signal of the first CNN-LSTM module is a temperature-related signal, and the output is the feature of the temperature-related signal, the input signal of the second CNN-LSTM module is a pressure-related signal, and the output is the feature of the pressure-related signal, the first CNN-LSTM module and the second CNN-LSTM module are both used to extract local spatial features and temporal features of the input signal;

[0031] Specifically, the first CNN-LSTM module uses CNN (convolutional neural network) to extract local features of temperature-related signals, capture local patterns and rules in temperature-related signals, and obtain feature X1. The Relu activation function is then used to perform nonlinear transformation on feature X1 to obtain feature X2. Finally, LSTM (long short-term memory network) is used to capture the temporal relationship in feature X2. LSTM updates the hidden state through a gating mechanism to capture the changes of temperature-related signals over time and obtain the characteristics of temperature-related signals. The second CNN-LSTM module uses CNN to extract local features of pressure-related signals, capture local patterns and rules in pressure-related signals, and obtain feature X3. The Relu activation function is then used to perform nonlinear transformation on feature X3 to obtain feature X4. Finally, LSTM is used to capture the temporal relationship in feature X4. LSTM updates the hidden state through a gating mechanism to capture the changes of pressure-related signals over time and obtain the characteristics of pressure-related signals.

[0032] The microwave Transformer module includes: a first Transformer model and a second Transformer model (for the first Transformer model and the second Transformer model, see: Fu Yan, Yang Xu, Ye Ou. Smoke recognition method based on CNN and Transformer feature fusion [J]. Computer Engineering and Science, 2024, 46(11): 2045-2052.), the input signal of the first Transformer model is a microwave difference frequency phase correlation signal, and the output is the feature of the microwave difference frequency phase correlation signal; the input signal of the second Transformer model is a microwave difference frequency amplitude correlation signal, and the output is the feature of the microwave difference frequency amplitude correlation signal; the first Transformer model and the second Transformer model are both used to perform sequence modeling and feature extraction on the input signal;

[0033] The attention multi-parameter aggregation module adopts the attention fusion mechanism (Wang Xiaolan, Zhang Weidong, Wang Huizhong. CNN-LSTM short-term load forecasting based on attention mechanism [J]. Computers and Digital Engineering, 2024, 52(10): 3014-3018.) to perform weighted fusion on the features of the temperature-related signal and the pressure-related signal output by the temperature-pressure CNN-LSTM module, as well as the features of the microwave difference frequency phase-related signal and the microwave difference frequency amplitude-related signal output by the microwave Transformer module to obtain the fusion feature;

[0034] The temperature-related signal, the pressure-related signal, the microwave difference frequency phase-related signal and the microwave difference frequency amplitude-related signal are respectively the signal after the temperature signal is standardized, the signal after the pressure signal is standardized, the signal after the microwave difference frequency phase signal is standardized and the signal after the microwave difference frequency amplitude signal is standardized, wherein the temperature signal, the pressure signal, the microwave difference frequency phase signal and the microwave difference frequency amplitude signal are obtained by the corresponding sensors collecting the oil wellhead production liquid in the same time period;

[0035] The input of the prediction output module is the fusion feature, and the output value is the predicted value of the water content of the oil wellhead production liquid collected in this time period. The prediction output module uses Sigmoid as the activation function.

[0036] Example 2

[0037] On the basis of Example 1, the temperature signal, pressure signal, microwave difference frequency phase signal and microwave difference frequency amplitude signal are collected by a multi-sensor measurement system, such as Figure 2 and Figure 3 As shown, the multi-sensor measurement system includes: an oil inlet pipe section 1, an opposing microwave sensor 2, a temperature sensor 3, a pressure sensor 4, an oil outlet pipe section 5 and a casing 6. One end of the oil inlet pipe section 1 is connected to the oil wellhead, and the casing 6 is sleeved outside one end of the oil inlet pipe section 1. The opposing microwave sensor 2, the temperature sensor 3, and the pressure sensor 4 are installed on the oil inlet pipe section 1 and are located on one side outside the casing 6. One end of the oil outlet pipe section 5 is connected to the casing, and the other end of the oil outlet pipe section 5 is connected to the oil inlet pipe section 1. The oil wellhead produced liquid (three-phase fluid) enters the multi-sensor measurement system through the oil inlet pipe section 1; the opposing microwave sensor collects the microwave difference frequency phase signal and the microwave difference frequency amplitude signal of the oil wellhead produced liquid; the temperature sensor collects the temperature signal of the oil wellhead produced liquid; the pressure sensor collects the pressure signal of the oil wellhead produced liquid, and the oil wellhead produced liquid flows into the casing 6 through the oil outlet pipe section and is discharged.

[0038] The method for preparing a multi-parameter data set and obtaining the multi-parameter data set comprises the following steps:

[0039] S1, the multi-sensor measurement system collects sample data once every T1 interval, and collects N times in total, and the time of each collection is T2. The sample data collected each time includes: microwave difference frequency phase signal, microwave difference frequency amplitude signal, temperature signal and pressure signal; within the time T2 of each sample data collection by the multi-sensor measurement system, the oil wellhead production liquid (three-phase fluid) is manually sampled M times at a time interval T3, and the total number of manual samplings is M×N times. All the manually sampled oil wellhead production liquid (three-phase fluid) is manually tested to detect the water content of the oil wellhead production liquid, and the water content is used as the test value, and M×N test values ​​are obtained;

[0040] Among them, the acquisition start time of the opposing microwave sensor, temperature sensor and pressure sensor in the multi-sensor measurement system is the same, and the sampling frequency is F = 1000 Hz; T1 = 1h, T2 = 300s, T3 = 60s, N = 100, M = 5, M × N = 500, and the number of test values ​​is 500;

[0041] S2, taking the microwave difference frequency phase signal, microwave difference frequency amplitude signal, temperature signal and pressure signal collected by the multi-sensor measurement system as collection signals, and preprocessing each collection signal:

[0042] S2-1, using deviation normalization (see: Xu Yifang, Chen Jin, Li Lin, et al. Credit prediction based on Isomap fusion naive Bayes classifier [J]. Computer Knowledge and Technology, 2021, 17(35): 125-126+139.) to scale the collected signal to the range of [0, 1] to obtain a standardized signal;

[0043] S2-2, the sliding window method (Zhao Yuxuan, Gong Jianning, Liu Han, et al. Research on front-end processing of mobile robots based on multi-source sensor information fusion [J]. Manufacturing Automation, 2023, 45(04): 217-220.) is used to segment the standardized signal according to the acquisition time sequence, and the sliding window length is set to L = 6 × 10 4 And there is no overlapping segmentation, then each standardized signal is divided into Related signals.

[0044] When the collected signal is a microwave difference frequency phase signal, the correlation signal is a microwave difference frequency phase correlation signal;

[0045] When the collected signal is a microwave difference frequency amplitude signal, the correlation signal is a microwave difference frequency amplitude correlation signal;

[0046] When the collected signal is a temperature signal, the related signal is a temperature related signal;

[0047] When the collected signal is a pressure signal, the related signal is a pressure related signal.

[0048] The temperature-related signal, pressure-related signal, microwave difference frequency phase-related signal and microwave difference frequency amplitude-related signal obtained in the same time period are taken as a sample. All samples constitute a multi-parameter data set, that is, the multi-parameter data set in this embodiment has samples.

[0049] The M×N test values ​​are taken in the order of manual sampling. The water content label value of the oil wellhead production fluid corresponding to the sample.

[0050] Example 3

[0051] On the basis of Example 2, a method for training a temperature-piezoelectric multi-parameter fusion network includes: using a ten-fold cross-validation method to train and validate the temperature-piezoelectric multi-parameter fusion network for a multi-parameter data set to obtain a trained temperature-piezoelectric multi-parameter fusion network. 9 copies are used as training sets, and 1 copy is used as a validation set, so that all samples in the multi-parameter data set are used as validation sets.

[0052] Each fold is trained for 100 rounds. During each fold training, the thermo-piezoelectric multi-parameter fusion network is used to predict the moisture content of the training set and output the moisture content prediction value; the loss function is calculated based on the moisture content prediction value and the moisture content label value to obtain the loss function value, and the AMSGrad optimization algorithm (Li Manyuan, Luo Fei, Gu Chunhua, et al. Adams algorithm based on adaptive momentum update strategy [J]. Journal of University of Shanghai for Science and Technology, 2023, 45(02): 112-119.) is used to calculate the gradient according to the loss function value, and the weights of the thermo-piezoelectric multi-parameter fusion network are updated by the gradient back propagation method;

[0053] Among them, the loss function is the mean square error (MSE), which measures the gap between the moisture content prediction value and the moisture content label value output by the thermo-piezoelectric multi-parameter fusion network; the gradient back propagation method uses minimizing the loss function value as the criterion to guide the training and optimization of the thermo-piezoelectric multi-parameter fusion network in the right direction.

[0054] Example 4

[0055] A wellhead three-phase content measurement method based on multi-parameter fusion, comprising:

[0056] Step 1, obtaining temperature-related signals, pressure-related signals, microwave difference frequency phase-related signals, microwave difference frequency amplitude-related signals and moisture content label values ​​of 50 samples to be predicted;

[0057] Step 2: Use the thermo-piezoelectric multi-parameter fusion network trained in Example 3 to predict the sample to be predicted and obtain the moisture content prediction value of the sample to be predicted.

[0058] Example 5

[0059] A wellhead three-phase content measurement method based on multi-parameter fusion is basically the same as Example 4, with the only difference being that the temperature-pressure CNN-LSTM module in the temperature-pressure-electricity multi-parameter fusion network used in Example 4 is removed in this embodiment.

[0060] Example 6

[0061] A wellhead three-phase content measurement method based on multi-parameter fusion is basically the same as Example 4, with the only difference being that the microwave Transformer module in the temperature-piezoelectric multi-parameter fusion network used in Example 4 is removed in this example.

[0062] The prediction results of Example 4, Example 5, and Example 6 were evaluated by mean square error (MSE), mean absolute error (MAE), and mean absolute percentage error (MAPE), and the indicators are as follows:

[0063] Table 1

[0064] Model MSE MAE MAPE Example 5 0.00428 0.0637 0.1086 Example 6 0.0334 0.437 0.856 Example 4 0.00206 0.0389 0.0593

[0065] It can be seen from Table 1 that the water content error predicted by the wellhead three-phase content measurement method in Example 4 is low, and has a high prediction accuracy.

[0066] The present invention is described above by way of example. It should be noted that, without departing from the core of the present invention, any simple deformation, modification or other equivalent replacement that can be made by those skilled in the art without inventive effort falls within the protection scope of the present invention.

Claims

1. A temperature-piezoelectric multi-parameter fusion network, characterized in that: include: Warm pressure CNN-LSTM module, microwave Transformer module, attention multi-parameter aggregation module and prediction output module; The temperature-pressure CNN-LSTM module includes: a first CNN-LSTM module and a second CNN-LSTM module, the input signal of the first CNN-LSTM module is a temperature-related signal, and the output is a feature of the temperature-related signal, and the input signal of the second CNN-LSTM module is a pressure-related signal, and the output is a feature of the pressure-related signal; The microwave Transformer module includes: a first Transformer model and a second Transformer model, wherein the input signal of the first Transformer model is a microwave difference frequency phase-related signal, and the output is a feature of the microwave difference frequency phase-related signal; the input signal of the second Transformer model is a microwave difference frequency amplitude-related signal, and the output is a feature of the microwave difference frequency amplitude-related signal; The attention multi-parameter aggregation module uses the attention fusion mechanism to perform weighted fusion on the features of temperature-related signals, pressure-related signals, microwave difference frequency phase-related signals, and microwave difference frequency amplitude-related signals to obtain fused features. The temperature-related signal, the pressure-related signal, the microwave difference frequency phase-related signal and the microwave difference frequency amplitude-related signal are respectively the signal after the temperature signal is standardized, the signal after the pressure signal is standardized, the signal after the microwave difference frequency phase signal is standardized and the signal after the microwave difference frequency amplitude signal is standardized, wherein the temperature signal, the pressure signal, the microwave difference frequency phase signal and the microwave difference frequency amplitude signal are obtained by the corresponding sensors collecting the oil wellhead production liquid in the same time period; The input of the prediction output module is the fusion feature, and the output value is the predicted value of the water content of the oil wellhead production liquid collected in this time period. The prediction output module uses Sigmoid as the activation function.

2. A wellhead three-phase content measurement method based on multi-parameter fusion, characterized in that: include: Step 1, obtaining a temperature-related signal, a pressure-related signal, a microwave difference frequency phase-related signal, and a microwave difference frequency amplitude-related signal of a sample to be predicted; Step 2, using the trained temperature-piezoelectric multi-parameter fusion network in claim 1, predict the sample to be predicted to obtain the moisture content prediction value of the sample to be predicted.

3. The wellhead three-phase content measurement method based on multi-parameter fusion according to claim 2 is characterized in that: The method for obtaining the trained temperature-piezoelectric multi-parameter fusion network includes: Prepare multiple samples as training sets, each sample includes: temperature-related signals, pressure-related signals, microwave difference frequency phase-related signals, and microwave difference frequency amplitude-related signals. Substitute the samples into the temperature-piezoelectric multi-parameter fusion network for training, substitute the moisture content prediction value of the sample and the moisture content label value of the sample into the loss function to obtain the loss function value, and calculate the gradient using the AMSGrad optimization algorithm based on the loss function value, and update the weights of the temperature-piezoelectric multi-parameter fusion network by the gradient back propagation method.

4. The wellhead three-phase content measurement method based on multi-parameter fusion according to claim 3 is characterized in that: The water content label value is the true value of the water content of the oil wellhead production fluid during the sample collection period.

5. The wellhead three-phase content measurement method based on multi-parameter fusion according to claim 3 is characterized in that: The loss function is a mean square error function.

Images