Intelligent real-time monitoring method for liquid production state of oil well

Abstract

The invention discloses an intelligent real-time monitoring method for the liquid production state of an oil well. The intelligent real-time monitoring method comprises the steps that S1, a to-be-predicted sample is obtained; and S2, predicting a to-be-predicted sample by using the trained OilWellNet network to obtain a prediction result and a water content prediction value of the to-be-predicted sample. The OilWellNet network comprises an indicator diagram working condition judgment network and a multi-working-condition moisture content prediction network, the indicator diagram working condition judgment network comprises a convolutional layer module, a pooling layer and a full connection layer which are connected in sequence, and the multi-working-condition moisture content prediction network comprises a feature coding layer, a multi-scale pooling layer, a dynamic division layer, a feature fusion layer and a parameter prediction layer. According to the intelligent real-time monitoring method for the liquid production state of the oil well, real-time response is made according to changes in different working states through the trained OilWellNet network, so that management and decision making in the oil well production process are optimized, and the water content of the oil well is accurately predicted.

Description

Technical Field

[0001] The invention belongs to the field of oil and gas production engineering and intelligent monitoring technology, and specifically relates to an intelligent real-time monitoring method for oil well fluid production status. Background Art

[0002] At present, the monitoring of the production status of oil wells mainly relies on manual inspections and conventional sensor monitoring. The manual inspection method usually judges the status by regularly recording and analyzing the production parameters of the oil wells, such as pressure, temperature and liquid level depth, but it has the problems of slow data update, difficulty in meeting real-time requirements, and reliance on manual experience resulting in large data collection errors. On the other hand, conventional sensor monitoring technology usually uses a single type of sensor, such as a conductivity sensor or a density meter, to monitor the production status, but since a single type of sensor can only provide a single physical parameter, it is difficult to fully characterize the complex production status of the oil well, and the detection accuracy is low under complex working conditions such as high water content or heavy oil.

[0003] Although traditional technologies can monitor the status of liquid production to a certain extent, they have significant limitations:

[0004] 1. The production status of oil wells is complex and changeable, especially under special conditions such as high water content, gas invasion and leakage. Traditional methods are difficult to accurately distinguish different conditions.

[0005] 2. In high water-cut wells, due to the great changes in the water-oil ratio, the traditional model cannot capture the dynamics of water-oil distribution in real time, resulting in high water-cut prediction errors.

[0006] 3. Traditional monitoring systems mostly use offline analysis methods and cannot achieve real-time monitoring and early warning functions.

[0007] 4. Existing technologies usually only process single signals and lack the ability to fuse multimodal data, making it difficult to utilize the synergistic effects of multi-source data to improve the accuracy and robustness of monitoring.

[0008] In recent years, with the rapid development of artificial intelligence technology and sensor technology, intelligent monitoring technology has gradually become the research focus of oilfield production optimization. The successful application of deep learning algorithms in image classification and time series analysis has provided new ideas for intelligent diagnosis of oil well production status, while multimodal data fusion technology can integrate data from different sensors to explore potential laws in complex production environments. However, existing intelligent monitoring technologies still have problems such as insufficient classification of working conditions, data processing and model inference lagging behind production dynamics, and insufficient applicability. Especially under extreme working conditions such as high water content and heavy oil thermal recovery, its performance still needs to be optimized and improved. Summary of the invention

[0009] In view of the deficiencies in the prior art, an object of the present invention is to provide an OilWellNet network.

[0010] Another object of the present invention is to provide an intelligent real-time monitoring method for the oil well fluid production status. The intelligent real-time monitoring method for the oil well fluid production status is based on a trained OilWellNet network, combined with a dynamometer image and a standardized microwave difference frequency signal, to achieve intelligent identification of the oil well working status and water content prediction, thereby improving the intelligence and accuracy of oil well fluid production monitoring.

[0011] The purpose of the present invention is achieved through the following technical solutions.

[0012] An OilWellNet network includes: a dynamometer diagram working condition judgment network and a multi-working condition water content prediction network, each sample input to the OilWellNet network includes: a pump dynamometer diagram image collected at time T and a standardized microwave difference frequency signal in a time period of Tt to T, wherein the standardized microwave difference frequency signal is obtained by standardizing the microwave difference frequency signal; T is the collection time, and the value range of t is 0.5min to 1min;

[0013] The dynamometer working condition judgment network includes: a convolutional layer module, a pooling layer and a fully connected layer connected in sequence. The convolutional layer module is used to input the pump dynamometer image of the sample and extract the local features of the pump dynamometer image at different scales. The convolutional layer module outputs the feature X1.

[0014] The pooling layer is used to input feature X1 and output feature X2;

[0015] The fully connected layer inputs feature X2 and outputs the prediction result, which is used to predict the working condition of the oil well, including sufficient liquid supply, insufficient liquid supply, heavy oil thermal recovery, gas impact or leakage impact.

[0016] The multi-condition water content prediction network includes a sufficient fluid supply water content prediction network, an insufficient fluid supply water content prediction network, a heavy oil thermal recovery water content prediction network, a gas-affected water content prediction network, or a leakage-affected water content prediction network;

[0017] Select the multi-condition water content prediction network according to the prediction result of the working condition judgment network of the indicator diagram: when the prediction result is sufficient fluid supply, the multi-condition water content prediction network is a sufficient fluid supply water content prediction network; when the prediction result is insufficient fluid supply, the multi-condition water content prediction network is an insufficient fluid supply water content prediction network; when the prediction result is heavy oil thermal recovery, the multi-condition water content prediction network is a heavy oil thermal recovery water content prediction network; when the prediction result is gas influence, the multi-condition water content prediction network is a gas influence water content prediction network; when the prediction result is leakage influence, the multi-condition water content prediction network is a leakage influence water content prediction network;

[0018] The structures of the water cut prediction network for sufficient fluid supply, the water cut prediction network for insufficient fluid supply, the water cut prediction network for heavy oil thermal recovery, the water cut prediction network affected by gas, or the water cut prediction network affected by leakage are the same. Each multi-condition water cut prediction network includes: a feature encoding layer, a multi-scale pooling layer, a dynamic partitioning layer, a feature fusion layer, and a parameter prediction layer.

[0019] The feature coding layer includes: three convolution layers with different convolution kernel sizes and a nonlinear activation function ReLU. The inputs of the three convolution layers in the feature coding layer are the same. Each convolution layer in the feature coding layer is used to input the standardized microwave difference frequency signal of the sample. The outputs of the three convolution layers in the feature coding layer are feature X3, feature X4 and feature X5.

[0020] The input of the nonlinear activation function ReLU is feature X3, feature X4 and feature X5, and the nonlinear activation function ReLU obtains output X3, output X4 and output X5;

[0021] The multi-scale pooling layer uses the pyramid pooling method to aggregate the features of output X3, output X4, and output X5 to obtain feature X6;

[0022] The dynamic partitioning layer partitions the feature X6 into multiple patches and maps each patch to d model dimension, and obtain multiple patch features;

[0023] The feature fusion layer uses a multi-layer perceptron MLP to extract and fuse high-order features from multiple patch features to obtain feature X7;

[0024] The parameter prediction layer uses the Sigmoid activation function to perform nonlinear transformation on feature X7 and outputs the predicted value of moisture content.

[0025] In the above technical solution, the convolution layer module is three convolution layers connected in sequence, and the convolution kernel sizes of the three convolution layers connected in sequence from front to back are 50×50, 30×30 and 10×10 respectively.

[0026] In the above technical solution, the fully connected layer uses the Softmax activation function to flatten the feature X2.

[0027] In the above technical solution, the convolution kernel sizes of the three convolution layers in the feature coding layer are 50×1, 30×1 and 10×1 respectively.

[0028] In the above technical solution, the pooling scales of the multi-scale pooling layer are 25×1, 50×1, 75×1 and 100×1 respectively.

[0029] In the above technical solution, the length of each patch is 50, d model is 256.

[0030] In the above technical solution, the dimension of the standardized microwave difference frequency signal is L×1, where L is the length of the standardized microwave difference frequency signal.

[0031] In the above technical solution, the microwave difference frequency signal is collected by installing a high-frequency microwave sensor at the wellhead of the oil well.

[0032] In the above technical solution, the pump dynamometer image is acquired by collecting data from a dynamometer installed at the wellhead of the oil well.

[0033] In the above technical solution, the microwave difference frequency signal is standardized by using deviation standardization.

[0034] An intelligent real-time monitoring method for oil well fluid production status, comprising:

[0035] S1, obtaining samples to be predicted, the samples to be predicted include: a pump dynamometer image acquired at time T′ and a standardized microwave difference frequency signal in a time period of T′-t to T′;

[0036] S2, using the trained OilWellNet network to predict the sample to be predicted, and obtaining the prediction result and water content prediction value of the sample to be predicted.

[0037] In the above technical solution, the method for obtaining the trained OilWellNet network includes:

[0038] Step 1, prepare multiple samples for input into the OilWellNet network as a training set; each sample in the training set corresponds to a label value and a working condition label, wherein the label value is the true value of the water content of the oil wellhead produced liquid, and the working condition label is the actual working condition obtained according to the working state of the oil well, and the working condition label is sufficient liquid supply, insufficient liquid supply, heavy oil thermal recovery, gas impact or leakage impact;

[0039] Among them, sufficient fluid supply refers to the condition in which the liquid supply in the formation is sufficient to meet the production demand; insufficient fluid supply refers to the insufficient supply of formation liquid. Under this condition, the wellbore liquid level may be too low, the pump may be empty or the production efficiency may be reduced; heavy oil thermal recovery refers to the condition in which thermal recovery technology (such as steam stimulation, steam drive or burning oil layer) is used to improve the fluidity of heavy oil, reduce viscosity and increase recovery rate during the heavy oil production process; gas impact refers to the interference with fluid flow caused by the presence of gas during the production process, usually including the impact of dissolved gas, free gas or external injected gas; leakage impact refers to the escape of liquid or gas due to formation leakage or wellbore leakage. The leakage impact condition may significantly affect the wellbore pressure, production and recovery rate, and needs to be diagnosed and repaired in time.

[0040] Step 2, substitute the training set into the OilWellNet network for training. During training, the OilWellNet network outputs the prediction results and the water content prediction values. The loss function of the multi-condition water content prediction network is calculated according to the water content prediction values ​​and label values. The loss function value of the dynamometer diagram working condition judgment network is calculated according to the prediction results and working condition labels. The total loss function of the OilWellNet network is the sum of the loss function of the multi-condition water content prediction network and the loss function of the dynamometer diagram working condition judgment network. The parameters of the OilWellNet network are updated by the gradient back propagation method.

[0041] In the above technical solution, the loss function of the dynamometer diagram working condition judgment network adopts the cross entropy loss function.

[0042] In the above technical solution, the loss function of the multi-condition moisture content prediction network adopts the mean square error function.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. The OilWellNet network proposed in the present invention combines the dynamometer diagram working condition judgment network and the multi-condition water content prediction network. The dynamometer diagram working condition judgment network analyzes the pump dynamometer diagram image through a deep convolutional neural network (CNN) to accurately judge the working status of the oil well and provide accurate working condition information for subsequent water content prediction. The multi-condition water content prediction network accurately predicts the water content of the oil well based on the standardized microwave difference frequency signal under different working conditions.

[0045] 2. The intelligent real-time monitoring method of oil well production status of the present invention utilizes the OilWellNet network to select and activate the multi-condition water content prediction network of the corresponding working condition according to the real-time monitored working status of the oil well. This selection mechanism effectively improves the accuracy of the prediction and can make real-time responses to changes under different working conditions, thereby optimizing the management and decision-making in the oil well production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic diagram of the OilWellNet network of the present invention. DETAILED DESCRIPTION

[0047] The intelligent real-time monitoring method for oil well fluid production status of the present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0048] Example 1

[0049] An OilWellNet network, such as Figure 1As shown, the OilWellNet network includes: a dynamometer diagram working condition judgment network and a multi-working condition water content prediction network. Each sample input to the OilWellNet network includes: a pump dynamometer diagram image collected at time T and a standardized microwave difference frequency signal in the time period of Tt to T. The standardized microwave difference frequency signal is obtained by standardizing the microwave difference frequency signal. T is the collection time.

[0050] The dynamometer working condition judgment network includes: a convolutional layer module, a pooling layer and a fully connected layer connected in sequence. The convolutional layer module is used to input the pump dynamometer image of the sample and extract the local features of the pump dynamometer image at different scales. The convolutional layer module outputs the feature X1. The convolutional layer module is three convolutional layers connected in sequence. The convolutional kernel sizes of the three convolutional layers connected in sequence from front to back are 50×50, 30×30 and 10×10 respectively.

[0051] The pooling layer is used to input feature X1 and output feature X2. The pooling layer uses the maximum pooling operation to reduce the spatial size of feature X1;

[0052] The fully connected layer inputs feature X2 and outputs the prediction result. The fully connected layer uses the Softmax activation function to flatten feature X2. The prediction result is used to predict the working conditions of the oil well, which include sufficient fluid supply, insufficient fluid supply, heavy oil thermal recovery, gas impact or leakage impact.

[0053] The multi-condition water content prediction network is a water content prediction network for sufficient fluid supply, a water content prediction network for insufficient fluid supply, a water content prediction network for heavy oil thermal recovery, a water content prediction network for gas-affected water content prediction network or a water content prediction network for leakage-affected water content prediction network. The multi-condition water content prediction network is selected according to the prediction result of the indicator diagram working condition judgment network to predict the water content of the oil well production fluid under the working condition. When the prediction result is sufficient fluid supply, the multi-condition water content prediction network is a water content prediction network for sufficient fluid supply; when the prediction result is insufficient fluid supply, the multi-condition water content prediction network is a water content prediction network for insufficient fluid supply; when the prediction result is heavy oil thermal recovery, the multi-condition water content prediction network is a water content prediction network for heavy oil thermal recovery; when the prediction result is gas influence, the multi-condition water content prediction network is a water content prediction network for gas influence; when the prediction result is leakage influence, the multi-condition water content prediction network is a water content prediction network for leakage-affected water content prediction;

[0054] The structures of the water cut prediction network for sufficient fluid supply, the water cut prediction network for insufficient fluid supply, the water cut prediction network for heavy oil thermal recovery, the water cut prediction network affected by gas, or the water cut prediction network affected by leakage are the same. Each multi-condition water cut prediction network includes: a feature encoding layer, a multi-scale pooling layer, a dynamic partitioning layer, a feature fusion layer, and a parameter prediction layer.

[0055] The feature coding layer includes: three convolution layers with different convolution kernel sizes and a nonlinear activation function ReLU. The inputs of the three convolution layers in the feature coding layer are the same. Each convolution layer in the feature coding layer is used to input the standardized microwave difference frequency signal of the sample. The outputs of the three convolution layers in the feature coding layer are feature X3, feature X4 and feature X5. The convolution kernel sizes of the three convolution layers in the feature coding layer are 50×1, 30×1 and 10×1 respectively. The feature coding layer extracts the local time series features of the input signal through three convolution layers, and each convolution layer captures different local features of the input signal.

[0056] The input of the nonlinear activation function ReLU is feature X3, feature X4 and feature X5. The nonlinear activation function ReLU processes feature X3, feature X4 and feature X5, and the nonlinear activation function ReLU obtains output X3, output X4 and output X5;

[0057] The multi-scale pooling layer uses the pyramid pooling method to aggregate the features of output X3, output X4 and output X5 to obtain feature X6. The pooling scales of the multi-scale pooling layer are 25×1, 50×1, 75×1 and 100×1. The pyramid pooling method enhances the multi-scale perception ability of the water content prediction network with multiple working conditions.

[0058] The dynamic partitioning layer partitions the feature X6 into multiple patches and maps each patch to d model Dimension, multiple patch features are obtained. The dynamic partitioning layer captures the local pattern of feature X6 and preserves the global relationship. The length of each patch is 50, d model is 256.

[0059] The feature fusion layer uses a multi-layer perceptron MLP to extract and fuse high-order features of multiple patch features to obtain feature X7.

[0060] The parameter prediction layer uses the Sigmoid activation function to perform nonlinear transformation on feature X7 and outputs the predicted value of moisture content.

[0061] The dimension of the standardized microwave difference frequency signal is L×1; the dimension of feature X6 is L×1; the dimension of multiple patch features is (L / 50)×50×256; the dimension of feature X7 is 256×1, where L is the length of the standardized microwave difference frequency signal.

[0062] Example 2

[0063] The microwave difference frequency signal is acquired by collecting a high-frequency microwave sensor installed at the wellhead of the oil well. The high-frequency microwave sensor is not limited here. For example, the double-helix high-frequency microwave sensor in CN202311628698.7 can be used. The pump dynamometer image is acquired by collecting a dynamometer installed at the wellhead of the oil well.

[0064] The method of preparing a multimodal dataset and obtaining the multimodal dataset comprises the following steps:

[0065] S1, the high-frequency microwave sensor collects data 100 times, with a collection time of 1 hour, and each collection time is within 5 minutes of microwave difference frequency signal;

[0066] Within 5 minutes of data collection by the high-frequency microwave sensor, the dynamometer collects a pump dynamometer image every 1 minute (the size of each pump dynamometer image is 224×224), and a total of 500 collections are performed (500 pump dynamometer images); each time the dynamometer collects a pump dynamometer image, the oil wellhead production fluid is sampled (a total of 500 manual samplings are performed), and all manually sampled oil wellhead production fluids are manually tested to detect the water content of the oil wellhead production fluid, and the water content is used as the test value, 500 test values ​​are obtained, and one test value is used as a label value.

[0067] Among them, the sampling frequency of the high-frequency microwave sensor is 1000Hz, and the sampling frequency of the dynamometer is 1Hz.

[0068] S2, using synchronous alignment technology (Wang Guojing, Wang Weidong. Research on data synchronization acquisition method between polysomnography monitoring equipment and bowel sound equipment [J]. Chinese Medical Equipment, 2024, 39(05): 9-14.) to time align the collected microwave difference frequency signal with the collected pump dynamometer image;

[0069] S3, denoising and filtering out abnormalities of the time-aligned pump dynamometer images, and converting the image format into RGB, labeling the RGB format pump dynamometer images according to the working status of the oil well, and obtaining the working condition label of each pump dynamometer image, the working condition labels include: "sufficient fluid supply", "insufficient fluid supply", "heavy oil thermal recovery", "gas influence" and "loss influence";

[0070] Deviation standardization (deviation standardization, 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.) is used to standardize the time-aligned microwave difference frequency signal, that is, scale it to the range of [0, 1]. The standardized microwave difference frequency signal is segmented according to the acquisition time sequence by 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.), and the sliding window length is set to L = 6×10 4 The segmentation is non-overlapping, so that each time the microwave difference frequency signal is collected within 5 minutes, 5 standardized microwave difference frequency signals are obtained;

[0071] That is, 500 samples can be obtained through the above method, each sample includes: the pump dynamometer image collected at time T and the standardized microwave difference frequency signal in the time period of Tt~T (t=1min), each sample corresponds to a working condition label and a label value. 500 samples constitute a multimodal data set.

[0072] Example 3

[0073] Based on Example 2, a method for training and evaluating an OilWellNet network includes the following steps: using a ten-fold cross-validation method to train and validate the OilWellNet network on a multimodal data set to obtain a trained OilWellNet network. Among them, 9 of the multimodal data sets are used as training sets and 1 is used as a validation set, so that all samples in the multimodal data set are used as validation sets.

[0074] Each fold is trained for 100 rounds. During each fold training, the OilWellNet network is used to predict the working conditions and water content of the oil wells on the training set, and the prediction results and water content prediction values ​​are output; the loss function of the multi-condition water content prediction network is calculated based on the water content prediction value and the label value, and the loss function value of the dynamometer diagram working condition judgment network is calculated based on the prediction result and the working condition label. The loss function value of the multi-condition water content prediction network and the loss function value of the dynamometer diagram working condition judgment network are added together to obtain the total loss function value, and the parameters of the OilWellNet network are updated by the gradient back propagation method;

[0075] Among them, the loss function of the working condition judgment network of the indicator diagram adopts the cross entropy loss function (Ren Jinjun, Wang Ning. Research on loss function in artificial neural network[J]. Journal of Gansu Normal University, 2018, 23(02):61-63.);

[0076] The loss function of the multi-condition moisture content prediction network adopts the mean square error function (MSE);

[0077] The total loss function of the OilWellNet network is the sum of the cross entropy loss function and the mean square error function (MSE).

[0078] Example 4

[0079] An intelligent real-time monitoring method for oil well fluid production status, comprising:

[0080] S1, obtaining 50 samples to be predicted, each sample to be predicted includes: a pump dynamometer image collected at time T' and a standardized microwave difference frequency signal in a time period of T'-t to T', t = 1 min;

[0081] S2, using the OilWellNet network trained in Example 3, predicting the sample to be predicted, and obtaining the prediction result and water content prediction value of the sample to be predicted.

[0082] Example 5

[0083] An intelligent real-time monitoring method for oil well production status is basically the same as Example 4, with the only difference that only the water content prediction network in Example 1 is retained. In this embodiment, each sample to be predicted is a standardized microwave difference frequency signal in the time period T'-t~T'.

[0084] Example 6

[0085] An intelligent real-time monitoring method for oil well fluid production status is basically the same as Example 4, with the only difference being that the convolution layer module of the dynamometer diagram working condition judgment network in Example 1 is replaced by a convolution layer with a convolution kernel size of 50×50, and the three convolution layers of the feature coding layer of the water content prediction network in Example 1 are replaced by a convolution layer with a convolution kernel size of 50×1.

[0086] Example 7

[0087] An intelligent real-time monitoring method for the production status of oil wells is basically the same as Example 4, with the only difference being that the multi-scale pooling layer of the water cut prediction network in Example 1 is replaced by a pooling layer, and the pooling scale of the pooling layer is 50×1.

[0088] The mean square error (MSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) were used to evaluate Example 4, Example 5, Example 6 and Example 7, and the results are as follows:

[0089] Table 1

[0090]

[0091]

[0092] As can be seen from Table 1, the OilWellNet network of the present invention effectively judges the working condition of the oil well by combining the dynamometer diagram working condition judgment network with the multi-working condition water cut prediction network, provides accurate working condition information for subsequent water cut prediction, and integrates and extracts multi-scale features through multi-scale analysis technology, thereby significantly improving the accuracy of water cut prediction.

[0093] 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. An OilWellNet network, characterized in that: include: The dynamometer diagram working condition judgment network and the multi-condition water content prediction network, each sample input by the OilWellNet network includes: the pump dynamometer diagram image collected at time T and the standardized microwave difference frequency signal in the time period of Tt~T, the standardized microwave difference frequency signal is obtained after standardizing the microwave difference frequency signal; T is the collection time; The dynamometer working condition judgment network includes: a convolutional layer module, a pooling layer and a fully connected layer connected in sequence. The convolutional layer module is used to input the pump dynamometer image of the sample and extract the local features of the pump dynamometer image at different scales. The convolutional layer module outputs the feature X1. The pooling layer is used to input feature X1 and output feature X2; The fully connected layer inputs feature X2 and outputs prediction results, which are used to predict the working conditions of the oil well, including sufficient liquid supply, insufficient liquid supply, thermal recovery of heavy oil, gas impact or leakage impact; The multi-condition water content prediction network includes a sufficient fluid supply water content prediction network, an insufficient fluid supply water content prediction network, a heavy oil thermal recovery water content prediction network, a gas-affected water content prediction network, or a leakage-affected water content prediction network; Select the multi-condition water content prediction network according to the prediction result of the working condition judgment network of the indicator diagram: when the prediction result is sufficient fluid supply, the multi-condition water content prediction network is a sufficient fluid supply water content prediction network; when the prediction result is insufficient fluid supply, the multi-condition water content prediction network is an insufficient fluid supply water content prediction network; when the prediction result is heavy oil thermal recovery, the multi-condition water content prediction network is a heavy oil thermal recovery water content prediction network; when the prediction result is gas influence, the multi-condition water content prediction network is a gas influence water content prediction network; when the prediction result is leakage influence, the multi-condition water content prediction network is a leakage influence water content prediction network; The structures of the water cut prediction network for sufficient fluid supply, the water cut prediction network for insufficient fluid supply, the water cut prediction network for heavy oil thermal recovery, the water cut prediction network affected by gas, or the water cut prediction network affected by leakage are the same. Each multi-condition water cut prediction network includes: a feature encoding layer, a multi-scale pooling layer, a dynamic partitioning layer, a feature fusion layer, and a parameter prediction layer. The feature coding layer includes: three convolution layers with different convolution kernel sizes and a nonlinear activation function ReLU. The inputs of the three convolution layers in the feature coding layer are the same. Each convolution layer in the feature coding layer is used to input the standardized microwave difference frequency signal of the sample. The outputs of the three convolution layers in the feature coding layer are feature X3, feature X4 and feature X5. The input of the nonlinear activation function ReLU is feature X3, feature X4 and feature X5, and the nonlinear activation function ReLU obtains output X3, output X4 and output X5; The multi-scale pooling layer uses the pyramid pooling method to aggregate the features of output X3, output X4, and output X5 to obtain feature X6; The dynamic partitioning layer partitions the feature X6 into multiple patches and maps each patch to d model dimension, and obtain multiple patch features; The feature fusion layer uses a multi-layer perceptron MLP to extract and fuse high-order features from multiple patch features to obtain feature X7; The parameter prediction layer uses the Sigmoid activation function to perform nonlinear transformation on feature X7 and outputs the predicted value of moisture content.

2. The OilWellNet network according to claim 1, characterized in that: The convolution layer module consists of three convolution layers connected in sequence, and the convolution kernel sizes of the three convolution layers connected in sequence from front to back are 50×50, 30×30 and 10×10 respectively.

3. The OilWellNet network according to claim 1, characterized in that: The fully connected layer uses the Softmax activation function to flatten the feature X2.

4. The OilWellNet network according to claim 1, characterized in that: The convolution kernel sizes of the three convolution layers in the feature encoding layer are 50×1, 30×1 and 10×1 respectively.

5. The OilWellNet network according to claim 1, characterized in that: The pooling scales of the multi-scale pooling layer are 25×1, 50×1, 75×1 and 100×1, respectively.

6. The OilWellNet network according to claim 1, characterized in that: The length of each patch output by the dynamic partitioning layer is 50, d model is 256.

7. The OilWellNet network according to claim 1, characterized in that: The microwave difference frequency signal is acquired by a high-frequency microwave sensor installed at the wellhead of the oil well, and the pump dynamometer image is acquired by a dynamometer installed at the wellhead of the oil well. The microwave difference frequency signal is standardized by deviation standardization, and the dimension of the standardized microwave difference frequency signal is L×1, where L is the length of the standardized microwave difference frequency signal.

8. An intelligent real-time monitoring method for oil well fluid production status, characterized in that: include: S1, obtaining samples to be predicted, the samples to be predicted include: a pump dynamometer image collected at time T′ and a standardized microwave difference frequency signal in a time period of T′-t to T; S2, using the trained OilWellNet network of any one of claims 1 to 7 to predict the sample to be predicted, and obtaining the prediction result and water content prediction value of the sample to be predicted.

9. The intelligent real-time monitoring method for oil well production status according to claim 8, characterized in that: Methods for obtaining the trained OilWellNet network include: Step 1, prepare multiple samples for input into the OilWellNet network as a training set; each sample in the training set corresponds to a label value and a working condition label, wherein the label value is the true value of the water content of the oil wellhead produced liquid, and the working condition label is the actual working condition obtained according to the working state of the oil well, and the working condition label is sufficient liquid supply, insufficient liquid supply, heavy oil thermal recovery, gas impact or leakage impact; Step 2, substitute the training set into the OilWellNet network for training. During training, the OilWellNet network outputs the prediction results and the water content prediction values. The loss function of the multi-condition water content prediction network is calculated according to the water content prediction values ​​and label values. The loss function value of the dynamometer diagram working condition judgment network is calculated according to the prediction results and working condition labels. The total loss function of the OilWellNet network is the sum of the loss function of the multi-condition water content prediction network and the loss function of the dynamometer diagram working condition judgment network. The parameters of the OilWellNet network are updated by the gradient back propagation method.

10. The method for obtaining a trained OilWellNet network according to claim 9, characterized in that: The loss function of the dynamometer diagram working condition judgment network adopts the cross entropy loss function, and the loss function of the multi-working condition moisture content prediction network adopts the mean square error function.

Images