Wellhead high temperature early warning method, device and equipment for fire flooding production of heavy oil reservoir and medium

Abstract

The invention relates to the field of fireflooding production well mouth temperature prediction, and discloses a heavy oil reservoir fireflooding production well mouth high temperature early warning method, device, equipment and medium, which can obtain production data of a fireflooding production well in a heavy oil reservoir at a first time point and obtain a trained high temperature early warning model. The production data of the fireflooding production well at the first time point are preprocessed to obtain preprocessed production data, the preprocessed production data are input into a high-temperature early warning model, the wellhead high-temperature probability of the fireflooding production well at the second time point is predicted, whether the second time point is a high-temperature early warning time point or not is judged according to the wellhead high-temperature probability, and if yes, the fireflooding production well is subjected to early warning. And intelligent identification on whether the wellhead temperature of the fireflood production well is too high or not is realized.

Description

Technical Field

[0001] The present invention relates to the field of fire drive production well head temperature prediction, and in particular to a fire drive production well head high temperature early warning method, device, equipment and medium for a heavy oil reservoir. Background Art

[0002] Currently, most of the crude oil in the world is heavy oil. Heavy oil has high viscosity, low fluidity, and is difficult to extract. Less than half of the reserves are recoverable.

[0003] Fire flooding is an important means to improve the recovery rate of heavy oil, which has the advantages of strong reservoir adaptability and high recovery rate. During the fire flooding production process, the temperature at the wellhead of the production well will affect the fire flooding effect.

[0004] The relevant technology is generally used by technical personnel to determine whether the wellhead temperature of the production well is too high, and the degree of intelligence is relatively low. Summary of the invention

[0005] The present invention provides a method, device, equipment and medium for early warning of high temperature at the wellhead of a fire-driven production well in a heavy oil reservoir, which are used to solve the defect in the related technology that a technician has to judge whether the wellhead temperature of a production well is too high, and the degree of intelligence is low, and realizes intelligent identification of whether the wellhead temperature of a fire-driven production well is too high.

[0006] In a first aspect, the present invention provides a method for early warning of high temperature at a fire-driven production wellhead in a heavy oil reservoir, comprising:

[0007] Obtaining production data of a fire-driven production well in a heavy oil reservoir at a first time point, and obtaining a trained high-temperature early warning model; wherein the high-temperature early warning model includes a residual network, a coding layer, a feature weighted fusion layer, and a fully connected layer;

[0008] Preprocessing the production data of the fire drive production well at a first time point to obtain preprocessed production data;

[0009] Inputting the preprocessed production data into the residual network of the high temperature warning model to extract spatial features and obtain spatial feature data; and inputting the preprocessed production data into the encoding layer of the high temperature warning model to extract temporal features and obtain temporal feature data;

[0010] Inputting the spatial feature data and the temporal feature data into the feature weighted fusion layer for weighted fusion to obtain fused data;

[0011] Inputting the fused data into the fully connected layer for feature mapping to obtain the probability of high temperature at the wellhead of the fire drive production well at a second time point; wherein the second time point is after the first time point;

[0012] According to the probability of high temperature at the wellhead of the fire drive production well at the second time point, it is determined whether the second time point is a high temperature warning time point.

[0013] Preferably, the first time point is a plurality of consecutive days, and the production data of the fire drive production well at the first time point includes the production data of each day in the plurality of consecutive days;

[0014] The preprocessing of the production data of the fire drive production well at the first time point to obtain the preprocessed production data includes:

[0015] According to the chronological order, the missing value processing is performed on the production data of each day of the fire drive production well in the consecutive days to obtain a plurality of corresponding processed production data;

[0016] Based on the order of the processed production data, normalizing each of the processed production data in turn to obtain a corresponding plurality of normalized production data;

[0017] Based on the order of the normalized production data, all the normalized production data are arranged to obtain a data sequence as the preprocessed production data.

[0018] Preferably, the residual network includes a convolutional layer, a pooling layer and a plurality of residual modules connected in sequence;

[0019] The step of inputting the pre-processed production data into the residual network of the high temperature warning model to extract spatial features and obtain spatial feature data includes:

[0020] Inputting the preprocessed production data into the convolution layer and the pooling layer of the residual network for forward propagation to obtain output data of the pooling layer;

[0021] Inputting the output data of the pooling layer into a first residual module of the plurality of residual modules for residual processing to obtain output data of the first residual module;

[0022] The output data of the first residual module is input into the second residual module of the multiple residual modules for residual processing until the output data of the last residual module of the multiple residual modules is obtained and used as the spatial feature data.

[0023] Preferably, the first residual module includes a first convolutional layer, a first activation function, a second convolutional layer, a second activation function and a residual connection layer;

[0024] The step of inputting the output data of the pooling layer into a first residual module of the plurality of residual modules for residual processing to obtain the output data of the first residual module includes:

[0025] Input the output data of the pooling layer into the first convolution layer, the first activation function, the second convolution layer and the second activation function of the first residual module for forward propagation to obtain the output data of the second activation function;

[0026] The output data of the second activation function and the output data of the pooling layer are input into the residual connection layer of the first residual module for residual connection, and the connection data is obtained as the output data of the first residual module.

[0027] Preferably, the encoding layer includes a plurality of encoders connected in sequence;

[0028] The step of inputting the pre-processed production data into the coding layer of the high temperature warning model to extract time features and obtain time feature data includes:

[0029] Inputting the preprocessed production data into a first encoder of the plurality of encoders for feature extraction to obtain output data of the first encoder;

[0030] The output data of the first encoder is input into a second encoder of the multiple encoders for feature extraction, until the output data of the last encoder of the multiple encoders is obtained and used as the time feature data.

[0031] Preferably, the first encoder includes a multi-head attention layer, a feedforward neural network, a first processing layer and a second processing layer, and the first processing layer and the second processing layer both include a residual connection layer and a normalization layer;

[0032] The step of inputting the pre-processed production data into a first encoder of the plurality of encoders for feature extraction to obtain output data of the first encoder comprises:

[0033] Inputting the preprocessed production data into the multi-head attention layer of the first encoder for feature extraction, and obtaining feature extraction data output by the multi-head attention layer;

[0034] Input the feature extraction data output by the multi-head attention layer and the pre-processed production data into the residual connection layer of the first processing layer for residual connection to obtain first connection data;

[0035] Inputting the first connection data into the normalization layer of the first processing layer for normalization to obtain first normalized data;

[0036] Inputting the first normalized data into the feedforward neural network for feature extraction to obtain feature extraction data output by the feedforward neural network;

[0037] Inputting the feature extraction data output by the feedforward neural network and the first normalized data into the residual connection layer of the second processing layer for residual connection to obtain second connection data;

[0038] The second connection data is input into the normalization layer of the second processing layer for normalization to obtain second normalized data as output data of the first encoder.

[0039] Preferably, judging whether the second time point is a high temperature warning time point according to the probability of a wellhead high temperature of the fire drive production well at the second time point includes:

[0040] Determine whether the probability of high temperature at the wellhead of the fire drive production well at the second time point is greater than a preset threshold;

[0041] If the probability of high temperature at the wellhead of the fire drive production well at the second time point is greater than the preset threshold, determining the second time point as a high temperature warning time point;

[0042] If the probability of high temperature at the wellhead of the fire drive production well at the second time point is not greater than the preset threshold, the second time point is determined to be a non-high temperature warning time point.

[0043] In a second aspect, the present invention provides a high temperature early warning device for a fire-driven production wellhead in a heavy oil reservoir, comprising:

[0044] An acquisition unit is used to acquire production data of a fire-driven production well in a heavy oil reservoir at a first time point, and to acquire a trained high-temperature early warning model; wherein the high-temperature early warning model includes a residual network, a coding layer, a feature weighted fusion layer, and a fully connected layer;

[0045] A preprocessing unit, used for preprocessing the production data of the fire drive production well at a first time point to obtain preprocessed production data;

[0046] A first extraction unit, used for inputting the preprocessed production data into the residual network of the high temperature warning model to extract spatial features and obtain spatial feature data;

[0047] A second extraction unit is used to input the pre-processed production data into the coding layer of the high temperature warning model to extract time features and obtain time feature data;

[0048] A feature weighting unit, used for inputting the spatial feature data and the temporal feature data into the feature weighted fusion layer for weighted fusion to obtain fused data;

[0049] A feature mapping unit, used for inputting the fused data into the fully connected layer for feature mapping, so as to obtain the probability of high temperature at the wellhead of the fire drive production well at a second time point; wherein the second time point is after the first time point;

[0050] The early warning judgment unit is used to judge whether the second time point is a high temperature early warning time point according to the high temperature probability of the wellhead of the fire drive production well at the second time point.

[0051] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to execute the high temperature early warning method for a fire-driven production wellhead in a heavy oil reservoir according to the first aspect or any corresponding embodiment thereof.

[0052] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method for high temperature early warning at a fire-driven production wellhead in a heavy oil reservoir according to the first aspect or any corresponding embodiment thereof.

[0053] The method, device, equipment and medium for early warning of high temperature at the wellhead of a fire-driven production in a heavy oil reservoir provided by the present invention can obtain production data of a fire-driven production well in a heavy oil reservoir at a first time point, and obtain a trained high temperature early warning model, pre-process the production data of the fire-driven production well at the first time point to obtain pre-processed production data, input the pre-processed production data into the high temperature early warning model, predict the probability of high temperature at the wellhead of the fire-driven production well at a second time point, judge whether the second time point is a high temperature early warning time point according to the high temperature probability at the wellhead, and realize intelligent identification of whether the wellhead temperature of the fire-driven production well is too high. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0055] Figure 1 A flowchart of a high temperature early warning method for a fire-driven production wellhead in a heavy oil reservoir provided by an embodiment of the present invention;

[0056] Figure 2 A schematic diagram of the relationship between a fire-driven gas injection well and a production well provided in an embodiment of the present invention;

[0057] Figure 3A flow chart of another method for early warning of high temperature at a fire-driven production wellhead in a heavy oil reservoir provided by an embodiment of the present invention;

[0058] Figure 4 A Transformer residual network model diagram provided by an embodiment of the present invention;

[0059] Figure 5 A prediction result diagram of the probability of high temperature at the wellhead of a production well over time provided by an embodiment of the present invention;

[0060] Figure 6 A curve diagram of the change of the wellhead temperature of a production well over time provided by an embodiment of the present invention;

[0061] Figure 7 A schematic structural diagram of a high-temperature early warning device for a fire-driven production wellhead in a heavy oil reservoir provided by an embodiment of the present invention;

[0062] Figure 8 A schematic diagram of the structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] Combine the following Figure 1-Figure 6 The invention describes a high temperature early warning method for a fire-driven production wellhead in a heavy oil reservoir.

[0065] like Figure 1 As shown, this embodiment proposes a first method for early warning of high temperature at a fire-driven production wellhead in a heavy oil reservoir, which method may include the following steps:

[0066] S101. Obtaining production data of a fire-driven production well in a heavy oil reservoir at a first time point, and obtaining a trained high-temperature warning model; wherein the high-temperature warning model includes a residual network, a coding layer, a feature weighted fusion layer, and a fully connected layer.

[0067] The fire drive production well may be a fire drive production well in a normal production state.

[0068] The first time point may be one time point or a plurality of consecutive time points. It should be noted that each time point may be a unit of time, such as one day, 36 hours or multiple days.

[0069] Specifically, the production data at the first time point may include production data at each time point in the first time point.

[0070] Among them, the high temperature warning model can be used to predict the probability that the wellhead temperature of the fire drive production well is high temperature at a future time point based on the production data of the fire drive production well at a first time point. It can be understood that the future time point can be a time point after the first time point, for example, it can be a time point located after the first time point and adjacent to the first time point.

[0071] Specifically, the high temperature warning model may be composed of a residual network, a coding layer, a feature weighted fusion layer, and a fully connected layer connected in sequence. Optionally, the high temperature warning model may be a trained transformer residual network model.

[0072] It should be noted that, in this embodiment, production data of the fire-driven production well at a historical time point can be first obtained, the production data can be preprocessed, and the preprocessed production data can be used as sample production data to remove abnormal data, fill in normal data, and reduce the dimension of the data, and the label of the sample production data can be determined according to the wellhead temperature of the fire-driven production well at the next time point at the historical time point. When the wellhead temperature is high temperature, the label can be set to 1, and when the wellhead temperature is not high temperature, the label can be set to 0. The pre-trained time series prediction model is trained using the sample production data with labels until a trained time series prediction model is obtained and used as a high temperature warning model.

[0073] S102, preprocessing the production data of the fire drive production well at the first time point to obtain preprocessed production data.

[0074] Specifically, in this embodiment, after obtaining the production data of the production well at the first time point, the production data of the fire drive production well at the first time point may be preprocessed to obtain the preprocessed production data.

[0075] Optionally, in other heavy oil reservoir fire drive production wellhead high temperature early warning methods proposed in this embodiment, the first time point is a continuous number of days, and the production data of the fire drive production well at the first time point includes the production data of each day in the continuous number of days. In this case, step S102 may include:

[0076] According to the chronological order, the missing value processing is performed on the daily production data of the fire drive production wells in a number of consecutive days, and the corresponding multiple processed production data are obtained;

[0077] Based on the order of the processed production data, each processed production data is normalized in turn to obtain a corresponding plurality of normalized production data;

[0078] Based on the order of the normalized production data, all normalized production data are arranged to obtain a data sequence as the preprocessed production data.

[0079] Optionally, the first time point may be 30 consecutive days. In this case, the production data of the fire drive production well at the first time point may include the production data of the production well obtained every day in the 30 days.

[0080] The production data may include parameter values ​​of multiple parameters related to the wellhead temperature of the fire drive production well.

[0081] Optionally, the daily production data includes oil pressure, casing pressure, daily oil production, daily gas production, water content, wellhead temperature and daily gas injection volume of the gas injection well.

[0082] The daily gas injection volume of the gas injection well can be the daily gas injection volume of the gas injection well corresponding to the fire drive production well. The gas injection well is used to inject gas into the fire drive production well. The relationship between the gas injection well and the fire drive production well can be as follows: Figure 2 As shown, both the production wells in the normal production state and the production wells in the shut-in state are fire-driven production wells, and the gas injection wells are connected to the fire-driven production wells connected by the dotted lines. One gas injection well can be used to inject gas into multiple fire-driven production wells.

[0083] It should be noted that the production data of the fire drive production wells obtained on site may have missing values ​​due to problems with the collector, and the missing values ​​are abnormal values. In this embodiment, the missing values ​​can be processed by filling or interpolation to improve the accuracy of subsequent high temperature warnings.

[0084] Specifically, this embodiment may use a minimum-maximum normalization algorithm for normalization processing.

[0085] Optionally, this embodiment can first perform missing value processing on the production data of the first day of the fire drive production well at the first time point to obtain processed production data. After that, the production data of the second day of the fire drive production well at the first time point is processed for missing values ​​to obtain processed production data, until the production data of the last day of the fire drive production well at the first time point is processed for missing values ​​to obtain all processed production data. Finally, all processed production data are uniformly normalized to obtain multiple corresponding normalized production data, and each normalized production data is arranged in the order in which each normalized production data is obtained to obtain a corresponding data sequence as pre-processed production data.

[0086] Specifically, this embodiment may also first perform missing value processing on the production data of the first day of the fire drive production well at the first time point to obtain processed production data, normalize the processed production data, and obtain corresponding normalized production data. Afterwards, the production data of the second day of the fire drive production well at the first time point may be processed for missing values ​​to obtain processed production data, normalize the processed production data, and obtain corresponding normalized production data, until the production data of the last day of the fire drive production well at the first time point is preprocessed to obtain corresponding normalized production data. Afterwards, this embodiment may arrange all normalized production data in the order in which all normalized production data are obtained, obtain a corresponding data sequence, and use the data sequence as preprocessed production data.

[0087] S103, inputting the preprocessed production data into the residual network of the high temperature warning model to extract spatial features and obtain spatial feature data.

[0088] Specifically, after obtaining the preprocessed production data, this embodiment can input the preprocessed production data into the residual network of the high temperature warning model to extract spatial features, and obtain the spatial feature data extracted and output by the residual network from the preprocessed production data.

[0089] Optionally, the residual network includes a convolutional layer, a pooling layer and a plurality of residual modules connected in sequence. In this case, step S103 may include:

[0090] The preprocessed production data is input into the convolution layer and pooling layer of the residual network for forward propagation to obtain the output data of the pooling layer;

[0091] Input the output data of the pooling layer into the first residual module of the multiple residual modules for residual processing to obtain the output data of the first residual module;

[0092] The output data of the first residual module is input into the second residual module of the multiple residual modules for residual processing until the output data of the last residual module of the multiple residual modules is obtained and used as the spatial feature data.

[0093] Specifically, the convolution layer may be a convolution layer with a 7*7 convolution kernel, and the pooling layer may be a pooling layer with a kernel size of 3*3. The number of residual modules may be 34. At this time, the convolution layer, the pooling layer, the first residual module, the second residual module ... the 34th residual module are sequentially communicated and connected.

[0094] Specifically, in the process of forward propagating the preprocessed production data in the convolution layer, pooling layer and multiple residual modules of the residual network, the preprocessed production data can be input into the convolution layer of the residual network for convolution processing to obtain the output data of the convolution layer. Afterwards, the output data of the convolution layer is input into the pooling layer for pooling processing to obtain the output data of the pooling layer. Afterwards, the output data of the pooling layer is input into the first residual module of the multiple residual modules for residual processing to obtain the output data of the first residual module, and the output data of the first residual module is input into the second residual module for residual processing until the output data of the last residual module is obtained, and the output data of the last residual module is used as the aforementioned spatial feature data.

[0095] Optionally, the first residual module includes a first convolution layer, a first activation function, a second convolution layer, a second activation function and a residual connection layer. At this time, the output data of the pooling layer is input into the first residual module of the plurality of residual modules for residual processing to obtain the output data of the first residual module, including:

[0096] The output data of the pooling layer is input into the first convolution layer, the first activation function, the second convolution layer and the second activation function of the first residual module for forward propagation to obtain the output data of the second activation function;

[0097] The output data of the second activation function and the output data of the pooling layer are input into the residual connection layer of the first residual module for residual connection, and the connection data is obtained as the output data of the first residual module.

[0098] Wherein, the first activation function and the second activation function can both be rectified linear unit (ReLU) activation functions.

[0099] Specifically, the first residual module includes a first convolution layer, a first activation function, a second convolution layer, a second activation function and a residual connection layer. In this embodiment, when the output data of the pooling layer is input into the first convolution layer, the first activation function, the second convolution layer and the second activation function of the first residual module for forward propagation, the output data of the pooling layer can be input into the first convolution layer for convolution processing to obtain the output data of the first convolution layer, and then the output data of the first convolution layer is input into the first activation function for operation to obtain the output data of the first activation function, the output data of the first activation function is input into the second convolution layer for convolution processing to obtain the output data of the second convolution layer, and the output data of the second convolution layer is input into the second activation function for operation to obtain the output data of the second activation function. Specifically, in this embodiment, the output data of the second activation function and the output data of the pooling layer can be input into the residual connection layer for residual connection to obtain the connection data, i.e., the output data of the residual connection layer, and used as the output data of the first residual module.

[0100] It is understandable that the structural composition of each other residual module is consistent with the structural composition of the first residual module. In this embodiment, the output data of the first residual module can be forward propagated in each other residual module, and the processing flow of each other residual module on the input data can refer to the processing flow of the first residual module on the input data.

[0101] Specifically, in this embodiment, the output data of the first residual module can be input into the first convolution layer, the first activation function, the second convolution layer, and the second activation function of the second residual module for forward propagation to obtain the output data of the second residual module. Afterwards, in this embodiment, the output data of the second residual module can be input into the first convolution layer, the first activation function, the second convolution layer, and the second activation function of the third residual module for forward propagation until the output data of the last residual module is obtained.

[0102] S104, inputting the pre-processed production data into the coding layer of the high temperature warning model to extract time features and obtain time feature data.

[0103] Specifically, after obtaining the preprocessed production data, this embodiment can input the preprocessed production data into the coding layer of the high temperature warning model to extract time features, and obtain the time feature data extracted by the coding layer from the preprocessed production data.

[0104] Optionally, the coding layer includes multiple encoders connected in sequence. In this case, step S104 may include:

[0105] Inputting the preprocessed production data into a first encoder of the plurality of encoders for feature extraction to obtain output data of the first encoder;

[0106] The output data of the first encoder is input into the second encoder of the multiple encoders for feature extraction, until the output data of the last encoder of the multiple encoders is obtained and used as the time feature data.

[0107] It is understandable that each encoder can perform feature extraction on its input data to obtain and output feature extracted data, which is the output data of the encoder.

[0108] Optionally, the first encoder includes a multi-head attention layer, a feedforward neural network, a first processing layer and a second processing layer, and the first processing layer and the second processing layer both include a residual connection layer and a normalization layer. At this time, the pre-processed production data is input into the first encoder of the multiple encoders for feature extraction, and the output data of the first encoder is obtained, including:

[0109] The preprocessed production data is input into the multi-head attention layer of the first encoder for feature extraction, and the feature extraction data output by the multi-head attention layer is obtained;

[0110] Input the feature extraction data output by the multi-head attention layer and the pre-processed production data into the residual connection layer of the first processing layer for residual connection to obtain the first connection data;

[0111] Inputting the first connection data into the normalization layer of the first processing layer for normalization to obtain first normalized data;

[0112] Inputting the first normalized data into a feedforward neural network for feature extraction to obtain feature extraction data output by the feedforward neural network;

[0113] Inputting the feature extraction data output by the feedforward neural network and the first normalized data into the residual connection layer of the second processing layer for residual connection to obtain second connection data;

[0114] The second connection data is input into the normalization layer of the second processing layer for normalization to obtain second normalized data, which is used as the output data of the first encoder.

[0115] Specifically, in this embodiment, the preprocessed production data can be input into the multi-head attention layer of the first encoder among the above-mentioned multiple encoders for feature extraction, and feature extraction data extracted and output by the multi-head attention layer from the preprocessed production data is obtained. The feature extraction data output by the multi-head attention layer and the preprocessed production data are input into the residual connection layer of the first processing layer for residual connection, and the first connection data output by the residual connection layer is obtained. The first connection data is input into the normalization layer of the first processing layer for normalization, and the first normalized data output by the normalization layer is obtained. The first normalized data is input into the feedforward neural network for feature extraction, and the feature extraction data extracted and output by the feedforward neural network in the first normalized data is obtained. The feature extraction data output by the feedforward neural network and the first normalized data are input into the residual connection layer of the second processing layer for residual connection, and the second connection data is obtained. The second connection data is input into the normalization layer of the second processing layer for normalization, and the second normalized data output by the normalization layer of the second processing layer is obtained, and used as the output data of the first encoder.

[0116] Optionally, the feedforward neural network may include a first linear transformation layer, a ReLU activation function, and a second linear transformation layer. Specifically, after obtaining the first normalized data output by the normalization layer, the present embodiment may input the first normalized data into a linear transformation layer of the feedforward neural network for linear transformation to obtain first transformed data output by the first linear transformation layer, then input the first transformed data into the ReLU activation function for operation to obtain output data of the ReLU activation function, then input the output data of the ReLU activation function into the second linear transformation layer for linear transformation to obtain second transformed data output by the second linear transformation layer, and the second transformed data is the feature extraction data extracted and output by the feedforward neural network from the first normalized data.

[0117] Specifically, in this embodiment, the output data of the first encoder can be used as the input data of the second encoder, and input into the second encoder for processing to obtain the output data of the second encoder. The output data of the second encoder is input into the third encoder for processing until the output data of the last encoder is obtained and used as the above-mentioned time feature data. It can be understood that the structural composition of other encoders is consistent with that of the first encoder, and the processing flow of other encoders for their input data can refer to the processing flow of the first encoder for input data, which will not be repeated.

[0118] S105: Input the spatial feature data and the temporal feature data into a feature weighted fusion layer for weighted fusion to obtain fused data.

[0119] Specifically, this embodiment can input the spatial feature data and time feature data extracted from the preprocessed production data into the feature weighted fusion layer of the high temperature warning model, and the feature weighted fusion layer performs weighted summation, i.e., weighted fusion, on the spatial feature data and the time feature data according to the set weight coefficient to obtain the summed data, i.e., fused data, output by the feature weighted fusion layer.

[0120] It should be noted that this embodiment can optimize the weight coefficient to be used in the feature weighted fusion layer during the training of the high temperature warning model, and after the high temperature warning model is trained, the currently used weight coefficient is determined as the optimal weight coefficient, and applied to the actual high temperature prediction.

[0121] S106. Input the fused data into a fully connected layer for feature mapping to obtain the probability of high temperature at the wellhead of the fire drive production well at a second time point; wherein the second time point is after the first time point.

[0122] The second time point may include one or a plurality of consecutive time points. The second time point may be after the first time point, for example, the second time point may be after the first time point and adjacent to the first time point.

[0123] Specifically, the probability of a high wellhead temperature of the fire drive production well at the second time point is the probability that the wellhead temperature of the production well is high at the second time point.

[0124] S107. Determine whether the second time point is a high temperature warning time point according to the probability of high temperature at the wellhead of the fire drive production well at the second time point.

[0125] Among them, the high temperature warning time point is the time point when a high temperature warning is required.

[0126] Specifically, after obtaining the probability of a high temperature at the wellhead of the fire drive production well at the second time point, this embodiment can determine whether the second time point is a high temperature warning time point according to the probability of a high temperature at the wellhead.

[0127] Optionally, step S107 may include:

[0128] Determine whether the probability of high temperature at the wellhead of the fire drive production well at the second time point is greater than a preset threshold;

[0129] If the probability of high temperature at the wellhead of the fire drive production well at the second time point is greater than a preset threshold, the second time point is determined as a high temperature warning time point;

[0130] If the probability of high temperature at the wellhead of the fire drive production well at the second time point is not greater than a preset threshold, a non-high temperature warning time point at the second time point is determined.

[0131] The preset threshold value may be set by a technician according to actual conditions, for example, it may be set to 0.5, or it may be set to 0.6, and this embodiment does not limit this.

[0132] Specifically, in this embodiment, when it is determined that the probability of high temperature at the wellhead of the fire-driven production well at the second time point is greater than a preset threshold, the second time point is determined to be a high temperature warning time point. When it is determined that the probability of high temperature at the wellhead of the fire-driven production well at the second time point is not greater than the preset threshold, the second time point is determined not to be a high temperature warning time point.

[0133] The high temperature early warning method for a fire-driven production wellhead in a heavy oil reservoir proposed in this embodiment can obtain production data of a fire-driven production well in a heavy oil reservoir at a first time point, and obtain a trained high temperature early warning model, pre-process the production data of the fire-driven production well at the first time point to obtain pre-processed production data, input the pre-processed production data into the high temperature early warning model, predict the probability of a wellhead high temperature in the fire-driven production well at a second time point, determine whether the second time point is a high temperature early warning time point according to the wellhead high temperature probability, and realize intelligent identification of whether the wellhead temperature of the fire-driven production well is too high.

[0134] like Figure 3 As shown, this embodiment proposes another high temperature warning method for fire-driven production wellheads in heavy oil reservoirs. In this method, this embodiment can perform model training based on sample production data to obtain a trained high temperature warning model, and use the high temperature warning model to predict the high temperature warning time point of the fire-driven production well.

[0135] Specifically, this embodiment can collect the initial production data of the on-site fire drive production well at multiple historical time points, and each initial production data may include the corresponding gas injection well number and the daily oil production, wellhead temperature and the corresponding daily gas injection volume of the gas injection well that change with time. There are many characteristic parameter values ​​in the initial production data, and not all characteristic parameter values ​​have an impact on the wellhead temperature. This embodiment can perform feature selection based on correlation analysis to perform data dimension reduction, and select the parameter values ​​of 9 characteristic parameters such as oil pressure, casing pressure and daily gas injection volume of the corresponding gas injection well that have a high correlation with the wellhead temperature as the production data after data dimension reduction.

[0136] It should be noted that the production data may have missing values ​​due to problems with the collector. Missing values ​​are abnormal values, and failure to process them may significantly affect the accuracy of the model. Optionally, this embodiment can use filling and interpolation methods to process missing values ​​to obtain processed production data. Among them, each characteristic parameter value of this embodiment can be filled or interpolated according to actual conditions. Afterwards, this embodiment can use the minimum-maximum normalization algorithm to normalize each characteristic parameter value in the processed production data to obtain normalized production data.

[0137] Specifically, this embodiment can obtain the normalized production data of 14,500 consecutive historical time points and arrange them in the order of time points to obtain the corresponding data sequence, and divide the data sequence into 80% training set and 20% test set, that is, the training set has the normalized production data of 10,150 consecutive historical time points, and the test set has the normalized production data of 2,900 historical time points. Afterwards, in the training set, this embodiment takes the normalized production data of 30 consecutive historical time points as a sample data, and sets the sliding window to 1 to continuously determine new sample data. For any sample data, this embodiment can set the label of the sample data according to the wellhead temperature of the next time point of the 30 consecutive historical time points in the sample data. When the wellhead temperature is higher than 80 degrees Celsius, the label of the sample data is set to 1, otherwise it is set to 0. The sample data with the label is input into the high temperature warning model to be trained to determine the probability of high temperature at the wellhead at the next time point, calculate the loss function value according to the difference between the probability of high temperature at the wellhead and 1, and update the parameters of the high temperature warning model to be trained according to the loss function value.

[0138] Specifically, this embodiment can continuously generate new sample data in the training set, and use the new sample data to update the parameters of the high temperature warning model to be trained, until a trained high temperature warning model is obtained.

[0139] The high temperature warning model to be trained can be a Transformer residual network model. The processing flow of the Transformer residual network model on sample data can be as follows: Figure 4 As shown in the figure, the Transformer residual network model includes a residual network, an encoding layer, a feature weighted fusion layer, and a fully connected layer.

[0140] like Figure 4 As shown in the left half, the residual network includes a convolution layer, a pooling layer and 34 residual modules connected in sequence. The convolution layer can be a convolution layer with a 7*7 convolution kernel, and the pooling layer can be a maximum pooling layer with a kernel size of 3*3. Each residual module includes a first convolution layer, a first activation function, a second convolution layer, a second activation function and a residual connection layer. The first convolution layer and the second convolution layer are convolution layers with a 3*3 convolution kernel. The first activation function and the second activation function are both ReLU activation functions. The sample data, i.e. the input data, passes through a convolution layer with a 7*7 convolution kernel, a maximum pooling layer with a kernel size of 3*3, and then enters 34 stacked residual modules to obtain a feature vector v rse The ReLU activation function formula is:

[0141]

[0142] Here, x represents the input.

[0143] Specifically, in this embodiment, the sample data, i.e., the input data, can be input into the convolution layer, the pooling layer, and the multiple residual modules of the residual network for forward propagation to extract spatial features of the sample data, and obtain the output data of the last residual module in the multiple residual modules, i.e., the feature vector v rse .

[0144] Among them, for any residual module, this embodiment can input the input data of the residual module into the first convolution layer, the first activation function, the second convolution layer and the second activation function of the residual module for forward propagation to obtain the output data of the second activation function. The output data of the second activation function and the input data of the residual module are input into the residual connection layer for residual connection to obtain the connection data as the output data of the residual module. The output data of the residual module is input into the next residual module as the input data of the next residual module for corresponding processing.

[0145] like Figure 4 As shown in the right half, the encoding layer includes 6 Transformer encoders. Each Transformer encoder includes a multi-head attention layer, a feedforward neural network, two residual connections and layer normalization (Add&Norm). The two Add&Norms can be the first processing layer and the second processing layer. The feedforward neural network includes two linear transformation layers and a ReLU activation function. The input first passes through a linear transformation layer, then a ReLU activation function, and then a linear transformation layer. The input data enters the 6 stacked Transformer encoders for forward propagation to obtain the feature vector v tra That is, the output data of the last Transformer encoder.

[0146] like Figure 4 As shown in the figure, the feature weighted fusion layer can perform weighted fusion on the feature vectors output by the residual network and the Transformer encoding layer according to the given weight w. The formula is:

[0147] w=[w rse w tra ];

[0148] v=w rse ·v rse +w tra ·v tra ;

[0149] Among them, v represents the feature vector obtained by weighted fusion of the feature weighted fusion layer, w represents the weight, and v rse Represents the output feature vector of the residual network, v traRepresents the output feature vector of the Transformer encoding layer. rse v rse The weight, w tra v tra The weight of .

[0150] like Figure 4 As shown, in this embodiment, the output vector of the feature weighted fusion layer can be input into the fully connected layer for feature mapping to obtain the probability of high temperature at the wellhead output by the fully connected layer.

[0151] It should be noted that in this embodiment, during the process of using the training set to train the model, the parameters in each network layer can be updated, and the weights w in the feature weighted fusion layer can also be updated. rse and w tra After training with the training set, you can use the test set to test whether the prediction accuracy of the model meets the requirements, that is, to determine whether the model is an ideal model. If so, it can be directly applied to actual production for prediction. Otherwise, the model needs to be retrained until the prediction accuracy of the model meets the requirements.

[0152] Specifically, in this embodiment, the weight w can be finally determined by the accuracy of the test set. rse and w tra The accuracy is calculated as:

[0153]

[0154] Among them, p acc represents accuracy, n represents the number of samples predicted correctly, and N represents the total number of samples.

[0155] Specifically, w rse and w tra The determination formula can be:

[0156]

[0157] Among them, p accres is the accuracy of the residual network, p acctra Represents the accuracy of the Transformer encoding layer.

[0158] Specifically, this embodiment can use the training set training model to obtain a trained high temperature warning model, and after determining that the high temperature warning model is an ideal model, the residual network and Transformer coding layer in the high temperature warning model can be extracted respectively, and a fully connected layer can be connected to the tail of the residual network and the Transformer coding layer respectively. The connected residual network and fully connected layer are regarded as an independent model. This embodiment can use the sample data in the test set to input into the residual network and the fully connected layer for forward propagation to obtain the corresponding prediction results, determine whether the prediction results are correct based on the prediction results and the labels of the sample data, and count the prediction accuracy of the residual network. Similarly, the prediction accuracy of the Transformer coding layer can be determined.

[0159] Optionally, if the label of the test set sample data is 1, and the probability that the model output result, i.e., the label is 1, is greater than 50%, then the model prediction is considered correct. If the label of the test set sample data is 0, and the probability that the model output result, i.e., the label is 1, is less than 50%, then the model prediction is considered correct, otherwise the model prediction is considered wrong.

[0160] It should be noted that, in this embodiment, when the high temperature warning model is used to predict whether the production time point of the fire drive production well is the high temperature warning time point, if the judgment result is yes, that is, it is determined that the time point is the high temperature warning time point, then production can be stopped and adjusted. If the judgment result is no, that is, it is determined that the time point is not the high temperature warning time point, then production can be maintained and continued.

[0161] like Figure 5 The high temperature probability-time scatter plot of the fire drive production well head temperature, and Figure 6 As shown in the wellhead temperature-time curve of the fire-driven production well, the high temperature warning model of this embodiment predicts that the time when the high temperature probability of the wellhead temperature of the production well is greater than 50% is the time when the wellhead temperature of the production well is higher than 80 degrees. This verifies the accuracy of the time point when the high temperature probability of the wellhead temperature of the production well is greater than 50% as the high temperature warning point. Therefore, the time point when the high temperature probability of the wellhead temperature of the production well predicted by the Transformer residual network model with feature weighted fusion is greater than 50% can be considered as the high temperature warning time point.

[0162] It should be noted that this embodiment can determine the high temperature warning time point of the fire drive production well by analyzing the model prediction results. For the existing fire drive production wells and the adjacent wells of the fire drive production wells to be produced that have similar geological structures, sedimentary characteristics, rock properties, aquifer characteristics and possible geological effects in adjacent or close areas, under the same construction parameters such as but not limited to gas injection rate, ignition method, formation temperature and other conditions, the high temperature warning time point of the adjacent wells can refer to the obtained high temperature warning time point. This embodiment predicts the high temperature warning time of the production of the fire drive production well by using a machine learning method. The design of the adjacent wells refers to the situation of the analyzed production wells and the high temperature warning time as a reference, which can guide the production operation parameters of the adjacent wells.

[0163] There are few studies on the wellhead temperature of fire-driven production wells in related technologies. On-site judgment of whether the wellhead temperature of the production well is too high can only be made based on the production data of the day, which has a lag and high comprehensive cost, and low on-site application efficiency. Failure to accurately predict the wellhead temperature of the production well may lead to unsatisfactory production results and even safety problems.

[0164] The high temperature early warning method for the fire-driven production wellhead of heavy oil reservoirs proposed in this embodiment can solve the problem of lag in the existing wellhead temperature prediction method, thereby guiding field engineers to perform operations such as shutting in production wells, reducing production costs, and improving production safety. In addition, the machine learning prediction method adopted in this embodiment can learn the potential patterns and similarities of data, so as to intelligently identify the time of the high temperature early warning point at the wellhead of the production well, thereby improving the prediction accuracy.

[0165] like Figure 7 As shown, this embodiment proposes a high temperature early warning device for a fire-driven production wellhead in a heavy oil reservoir, which may include:

[0166] The acquisition unit 701 is used to acquire the production data of the fire drive production well in the heavy oil reservoir at the first time point, and to acquire the trained high temperature warning model; wherein the high temperature warning model includes a residual network, a coding layer, a feature weighted fusion layer and a fully connected layer;

[0167] A preprocessing unit 702 is used to preprocess the production data of the fire drive production well at the first time point to obtain preprocessed production data;

[0168] The first extraction unit 703 is used to input the pre-processed production data into the residual network of the high temperature warning model to extract spatial features and obtain spatial feature data;

[0169] The second extraction unit 704 is used to input the pre-processed production data into the coding layer of the high temperature warning model to extract time features and obtain time feature data;

[0170] The feature weighting unit 705 is used to input the spatial feature data and the temporal feature data into the feature weighted fusion layer for weighted fusion to obtain fused data;

[0171] The feature mapping unit 706 is used to input the fused data into the fully connected layer for feature mapping to obtain the probability of high temperature at the wellhead of the fire drive production well at the second time point; wherein the second time point is after the first time point;

[0172] The early warning judgment unit 707 is used to judge whether the second time point is a high temperature early warning time point according to the probability of high temperature at the wellhead of the fire drive production well at the second time point.

[0173] It should be noted that the processing of the acquisition unit 701, the preprocessing unit 702, the first extraction unit 703, the second extraction unit 704, the feature weighting unit 705, the feature mapping unit 706 and the early warning judgment unit 707 and the beneficial effects thereof can be respectively referred to in Figure 1 Steps S101 to S107 in the above are not described in detail.

[0174] Optionally, the first time point is a plurality of consecutive days, and the production data of the fire drive production well at the first time point includes the production data of each day of the plurality of consecutive days;

[0175] The preprocessing unit 702 is further used for:

[0176] The production data of the fire drive production wells on each of the consecutive days are traversed in chronological order;

[0177] For the production data of any day traversed, the missing values ​​of the production data are processed to obtain the processed production data, and the processed production data are normalized to obtain the normalized production data;

[0178] Arrange each normalized production data in the order in which it is obtained to obtain a data sequence which is used as the preprocessed production data.

[0179] Optionally, the residual network includes a convolutional layer, a pooling layer, and a plurality of residual modules connected in sequence;

[0180] The first extraction unit 703 is further used for:

[0181] The preprocessed production data is input into the convolution layer and pooling layer of the residual network for forward propagation to obtain the output data of the pooling layer;

[0182] Input the output data of the pooling layer into the first residual module of the multiple residual modules for residual processing to obtain the output data of the first residual module;

[0183] The output data of the first residual module is input into the second residual module of the multiple residual modules for residual processing until the output data of the last residual module of the multiple residual modules is obtained and used as the spatial feature data.

[0184] Optionally, the first residual module includes a first convolutional layer, a first activation function, a second convolutional layer, a second activation function and a residual connection layer;

[0185] The first extraction unit 703 is further used for:

[0186] The output data of the pooling layer is input into the first convolution layer, the first activation function, the second convolution layer and the second activation function of the first residual module for forward propagation to obtain the output data of the second activation function;

[0187] The output data of the second activation function and the output data of the pooling layer are input into the residual connection layer of the first residual module for residual connection, and the connection data is obtained as the output data of the first residual module.

[0188] Optionally, the encoding layer includes a plurality of encoders connected in sequence;

[0189] The second extraction unit 704 is further used for:

[0190] Inputting the preprocessed production data into a first encoder of the plurality of encoders for feature extraction to obtain output data of the first encoder;

[0191] The output data of the first encoder is input into the second encoder of the multiple encoders for feature extraction, until the output data of the last encoder of the multiple encoders is obtained and used as the time feature data.

[0192] Optionally, the first encoder includes a multi-head attention layer, a feedforward neural network, a first processing layer and a second processing layer, and both the first processing layer and the second processing layer include a residual connection layer and a normalization layer;

[0193] The second extraction unit 704 is further used for:

[0194] The preprocessed production data is input into the multi-head attention layer of the first encoder for feature extraction, and the feature extraction data output by the multi-head attention layer is obtained;

[0195] Input the feature extraction data output by the multi-head attention layer and the pre-processed production data into the residual connection layer of the first processing layer for residual connection to obtain the first connection data;

[0196] Inputting the first connection data into the normalization layer of the first processing layer for normalization to obtain first normalized data;

[0197] Inputting the first normalized data into a feedforward neural network for feature extraction to obtain feature extraction data output by the feedforward neural network;

[0198] Inputting the feature extraction data output by the feedforward neural network and the first normalized data into the residual connection layer of the second processing layer for residual connection to obtain second connection data;

[0199] The second connection data is input into the normalization layer of the second processing layer for normalization to obtain second normalized data, which is used as the output data of the first encoder.

[0200] Optionally, the early warning judgment unit 707 is further used to:

[0201] Determine whether the probability of high temperature at the wellhead of the fire drive production well at the second time point is greater than a preset threshold;

[0202] If the probability of high temperature at the wellhead of the fire drive production well at the second time point is greater than a preset threshold, the second time point is determined as a high temperature warning time point;

[0203] If the probability of high temperature at the wellhead of the fire drive production well at the second time point is not greater than a preset threshold, a non-high temperature warning time point at the second time point is determined.

[0204] The high-temperature early warning device for a fire-driven production wellhead in a heavy oil reservoir proposed in this embodiment can obtain production data of a fire-driven production well in a heavy oil reservoir at a first time point, and obtain a trained high-temperature early warning model, pre-process the production data of the fire-driven production well at the first time point to obtain pre-processed production data, input the pre-processed production data into the high-temperature early warning model, predict the probability of a high temperature at the wellhead of the fire-driven production well at a second time point, determine whether the second time point is a high-temperature early warning time point according to the high-temperature probability at the wellhead, and realize intelligent identification of whether the wellhead temperature of the fire-driven production well is too high.

[0205] The high temperature early warning device for the fire-driven production wellhead of the heavy oil reservoir in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0206] The embodiment of the present invention also provides a computer device having the above Figure 7 The high temperature early warning device at the fire drive production wellhead of heavy oil reservoir is shown.

[0207] See also Figure 8, a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components are connected to each other using different buses for communication, and can be installed on a common motherboard or installed in other ways as needed. The processor can process instructions executed in the computer device, including instructions stored in or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 8 A processor 10 is taken as an example.

[0208] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0209] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0210] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function. The data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage devices. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0211] The memory 20 may include a volatile memory, such as a random access memory. The memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive. The memory 20 may also include a combination of the above-mentioned types of memory.

[0212] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0213] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0214] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A high temperature early warning method for a fire-driven production wellhead in a heavy oil reservoir, characterized in that: include: Obtaining production data of a fire-driven production well in a heavy oil reservoir at a first time point, and obtaining a trained high-temperature early warning model; wherein the high-temperature early warning model includes a residual network, a coding layer, a feature weighted fusion layer, and a fully connected layer; Preprocessing the production data of the fire drive production well at a first time point to obtain preprocessed production data; Inputting the preprocessed production data into the residual network of the high temperature warning model to extract spatial features and obtain spatial feature data; and inputting the preprocessed production data into the encoding layer of the high temperature warning model to extract temporal features and obtain temporal feature data; Inputting the spatial feature data and the temporal feature data into the feature weighted fusion layer for weighted fusion to obtain fused data; Inputting the fused data into the fully connected layer for feature mapping to obtain the probability of high temperature at the wellhead of the fire drive production well at a second time point; wherein the second time point is after the first time point; According to the probability of high temperature at the wellhead of the fire drive production well at the second time point, it is determined whether the second time point is a high temperature warning time point.

2. The method according to claim 1, characterized in that The first time point is a plurality of consecutive days, and the production data of the fire drive production well at the first time point includes the production data of each day in the plurality of consecutive days; The preprocessing of the production data of the fire drive production well at the first time point to obtain the preprocessed production data includes: According to the chronological order, the missing value processing is performed on the production data of each day of the fire drive production well in the consecutive days to obtain a plurality of corresponding processed production data; Based on the order of the processed production data, normalizing each of the processed production data in turn to obtain a corresponding plurality of normalized production data; Based on the order of the normalized production data, all the normalized production data are arranged to obtain a data sequence as the preprocessed production data.

3. The method according to claim 1, characterized in that The residual network includes a convolutional layer, a pooling layer and a plurality of residual modules connected in sequence; The step of inputting the pre-processed production data into the residual network of the high temperature warning model to extract spatial features and obtain spatial feature data includes: Inputting the preprocessed production data into the convolution layer and the pooling layer of the residual network for forward propagation to obtain output data of the pooling layer; Inputting the output data of the pooling layer into a first residual module of the plurality of residual modules for residual processing to obtain output data of the first residual module; The output data of the first residual module is input into the second residual module of the multiple residual modules for residual processing until the output data of the last residual module of the multiple residual modules is obtained and used as the spatial feature data.

4. The method according to claim 3, characterized in that The first residual module includes a first convolutional layer, a first activation function, a second convolutional layer, a second activation function and a residual connection layer; The step of inputting the output data of the pooling layer into a first residual module of the plurality of residual modules for residual processing to obtain the output data of the first residual module includes: Input the output data of the pooling layer into the first convolution layer, the first activation function, the second convolution layer and the second activation function of the first residual module for forward propagation to obtain the output data of the second activation function; The output data of the second activation function and the output data of the pooling layer are input into the residual connection layer of the first residual module for residual connection, and the connection data is obtained as the output data of the first residual module.

5. The method according to claim 1, characterized in that The coding layer includes a plurality of encoders connected in sequence; The step of inputting the pre-processed production data into the coding layer of the high temperature warning model to extract time features and obtain time feature data includes: Inputting the preprocessed production data into a first encoder of the plurality of encoders for feature extraction to obtain output data of the first encoder; The output data of the first encoder is input into a second encoder of the multiple encoders for feature extraction, until the output data of the last encoder of the multiple encoders is obtained and used as the time feature data.

6. The method according to claim 5, characterized in that The first encoder includes a multi-head attention layer, a feedforward neural network, a first processing layer and a second processing layer, and the first processing layer and the second processing layer both include a residual connection layer and a normalization layer; The step of inputting the pre-processed production data into a first encoder of the plurality of encoders for feature extraction to obtain output data of the first encoder comprises: Inputting the preprocessed production data into the multi-head attention layer of the first encoder for feature extraction, and obtaining feature extraction data output by the multi-head attention layer; Input the feature extraction data output by the multi-head attention layer and the pre-processed production data into the residual connection layer of the first processing layer for residual connection to obtain first connection data; Inputting the first connection data into the normalization layer of the first processing layer for normalization to obtain first normalized data; Inputting the first normalized data into the feedforward neural network for feature extraction to obtain feature extraction data output by the feedforward neural network; Inputting the feature extraction data output by the feedforward neural network and the first normalized data into the residual connection layer of the second processing layer for residual connection to obtain second connection data; The second connection data is input into the normalization layer of the second processing layer for normalization to obtain second normalized data as output data of the first encoder.

7. The method according to any one of claims 1 to 6, characterized in that The determining, based on the probability of a high temperature at the wellhead of the fire drive production well at the second time point, whether the second time point is a high temperature warning time point includes: Determine whether the probability of high temperature at the wellhead of the fire drive production well at the second time point is greater than a preset threshold; If the probability of high temperature at the wellhead of the fire drive production well at the second time point is greater than the preset threshold, determining the second time point as a high temperature warning time point; If the probability of high temperature at the wellhead of the fire drive production well at the second time point is not greater than the preset threshold, the second time point is determined to be a non-high temperature warning time point.

8. A high temperature early warning device for a fire-driven production wellhead in a heavy oil reservoir, characterized in that: include: An acquisition unit is used to acquire production data of a fire-driven production well in a heavy oil reservoir at a first time point, and to acquire a trained high-temperature early warning model; wherein the high-temperature early warning model includes a residual network, a coding layer, a feature weighted fusion layer, and a fully connected layer; A preprocessing unit, used for preprocessing the production data of the fire drive production well at a first time point to obtain preprocessed production data; A first extraction unit, used for inputting the preprocessed production data into the residual network of the high temperature warning model to extract spatial features and obtain spatial feature data; A second extraction unit is used to input the pre-processed production data into the coding layer of the high temperature warning model to extract time features and obtain time feature data; A feature weighting unit, used for inputting the spatial feature data and the temporal feature data into the feature weighted fusion layer for weighted fusion to obtain fused data; A feature mapping unit, used for inputting the fused data into the fully connected layer for feature mapping, so as to obtain the probability of high temperature at the wellhead of the fire drive production well at a second time point; wherein the second time point is after the first time point; The early warning judgment unit is used to judge whether the second time point is a high temperature early warning time point according to the high temperature probability of the wellhead of the fire drive production well at the second time point.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the high temperature early warning method for a fire-driven production wellhead in a heavy oil reservoir according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the high-temperature early warning method for a fire-driven production wellhead in a heavy oil reservoir according to any one of claims 1 to 7.

Images