Well cementation quality determination method and device, storage medium and electronic equipment

Abstract

The invention discloses a well cementation quality determination method and device, a storage medium and electronic equipment. The well cementation evaluation method comprises the steps that a well cementation evaluation request carrying various types of well logging data for a target well is obtained, and the various types of well logging data are used for describing the strength between a well wall and a well cementation cemented body from different angles; initial logging characteristics corresponding to the multiple types of logging data are determined; determining a request type of the well cementation evaluation request, and obtaining a fusion feature model corresponding to the request type; inputting the initial logging features corresponding to the multiple types of logging data into a fusion feature model to obtain fusion features of the target well; and obtaining target well cementation quality corresponding to the target well based on the fusion features. The technical problem that the treatment efficiency for determining the well cementation quality is not ideal in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of geophysical application technology, and in particular to a method, device, storage medium and electronic equipment for determining cementing quality. Background Art

[0002] Cementing quality plays an important role in ensuring wellbore stability, increasing production, reducing risks and improving drilling efficiency. At present, the number of wells that need to determine cementing quality can reach tens of thousands per year, and the length of the well section that needs to be processed is also very long. In addition, it is difficult to evaluate the interface processing in the well to be tested, and the interactive modification leads to a huge workload, which makes it difficult to meet the needs of rapid processing and interpretation of geological data. Related technologies rely on manual cementing quality evaluation. For the evaluation of basic logging data, the data sources are diverse, the data types are complicated and the dimensions are different. On the one hand, it has high subjective technical requirements for personnel and great processing limitations. On the other hand, it leads to low efficiency and large errors in cementing quality processing.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiments of the present invention provide a cementing quality determination method, device, storage medium and electronic device to at least solve the technical problem of unsatisfactory processing efficiency for cementing quality determination in the related art.

[0005] According to one aspect of an embodiment of the present invention, a method for determining cementing quality is provided, comprising: obtaining a cementing evaluation request carrying multiple types of logging data for a target well, wherein the multiple types of logging data are used to describe the strength between the well wall and the cementing bond from different angles; determining initial logging features corresponding to the multiple types of logging data; determining a request type of the cementing evaluation request, and obtaining a fusion feature model corresponding to the request type; inputting the initial logging features corresponding to the multiple types of logging data into the fusion feature model to obtain the fusion features of the target well; and obtaining a target cementing quality corresponding to the target well based on the fusion features.

[0006] Optionally, determining the initial well logging features corresponding to the multiple types of well logging data respectively includes: determining the neural network models corresponding to the multiple types of well logging data respectively, wherein the multiple types of well logging data respectively correspond to different neural network models, wherein the multiple types of well logging data include two-dimensional well logging data, one-dimensional well logging data, and discrete well logging data, the two-dimensional well logging data include at least any one of the following: a well logging image for the target well, and an acoustic variable density curve, the one-dimensional well logging data include at least any one of the following: acoustic time difference and acoustic amplitude obtained by acoustic logging of the target well, and the discrete well logging data include at least any one of the following: interface grade evaluation data obtained by dividing the interface in the target well, and interface marking data; the multiple types of well logging data are respectively input into the corresponding neural network models for processing to obtain the initial well logging features corresponding to the multiple types of well logging data respectively.

[0007] Optionally, when the number of output units of the neural network models corresponding to the multiple types of well logging data are equal, the initial well logging features corresponding to the multiple types of well logging data are input into the fusion feature model to obtain the fusion feature of the target well, including: determining the neural network models corresponding to the multiple types of well logging data, and the addition layer between the fusion feature model, wherein the number of input units of the addition layer is equal to the number of output units, and the addition layer is provided with a linear rectification function and / or a residual function; the initial well logging features corresponding to the multiple types of well logging data are input into the fusion feature model through the addition layer to obtain the fusion feature.

[0008] Optionally, the request type includes a qualitative request, the fusion feature model includes a first fusion sub-model, the first fusion sub-model is used to process the qualitative request, and obtaining the fusion feature model corresponding to the request type includes: determining that the first loss function corresponding to the qualitative request is a cross entropy loss function; obtaining training data whose data type matches the multiple types of logging data; inputting the training data into a predetermined first initial sub-model and using the first loss function to obtain a first output error value; when the first output error value is less than a predetermined first error value threshold, determining that the first initial sub-model is the first fusion sub-model.

[0009] Optionally, the request type includes a quantitative request, the fusion feature model includes a second fusion sub-model, the second fusion sub-model is used to process the quantitative request, and obtaining the fusion feature model corresponding to the request type includes: determining that the second loss function corresponding to the quantitative request is a mean absolute error function and / or a mean square error function; obtaining training data whose data type matches the multiple types of logging data; inputting the training data into a second initial sub-model and using the second loss function to obtain a second output error value; and determining that the second initial sub-model is the second fusion sub-model when the second output error value is less than a predetermined second error value threshold.

[0010] Optionally, the request type is a qualitative request for the target well, the fusion feature model includes a first fusion sub-model, the first fusion sub-model is used to process the qualitative request, and the initial well logging features corresponding to the multiple types of well logging data are input into the fusion feature model to obtain the fusion feature of the target well, including: determining a first output layer corresponding to the qualitative request, wherein the first output layer is provided with a predetermined activation function; when the first fusion sub-model includes a predetermined first input layer, a predetermined first hidden layer, and the first output layer, the initial well logging features corresponding to the multiple types of well logging data are input into the predetermined first input layer included in the fusion feature model to obtain the first well logging features corresponding to the multiple types of well logging data; the first well logging features corresponding to the multiple types of well logging data are input into the predetermined first hidden layer included in the first fusion sub-model to obtain the second well logging features corresponding to the multiple types of well logging data; the second well logging features corresponding to the multiple types of well logging data are input into the first output layer included in the first fusion sub-model to obtain the discrete qualitative features corresponding to the target well as the fusion feature.

[0011] Optionally, the request type is a quantitative request for the target well, the fusion feature model includes a second fusion sub-model, the second fusion sub-model is used to process the quantitative request, and the initial logging features corresponding to the multiple types of logging data are input into the fusion feature model to obtain the fusion feature of the target well, including: determining the second output layer corresponding to the quantitative request, wherein the second output layer is prohibited from setting a predetermined activation function; when the second fusion sub-model includes a predetermined second input layer, a predetermined second hidden layer, and the second output layer, the initial logging features corresponding to the multiple types of logging data are input into the predetermined second input layer included in the second fusion sub-model to obtain the third logging features corresponding to the multiple types of logging data; the third logging features corresponding to the multiple types of logging data are input into the predetermined second hidden layer included in the second fusion sub-model to obtain the fourth logging features corresponding to the multiple types of logging data; the fourth logging features corresponding to the multiple types of logging data are input into the second output layer included in the second fusion sub-model to obtain the cementation index corresponding to the target well as the fusion feature.

[0012] According to another aspect of an embodiment of the present invention, a cementing quality determination device is provided, comprising: an acquisition module, used to acquire a cementing evaluation request carrying multiple types of logging data for a target well, wherein the multiple types of logging data are used to describe the strength between the well wall and the cementing binder from different angles; a first feature module, used to determine initial logging features corresponding to the multiple types of logging data; a model determination module, used to determine the request type of the cementing evaluation request and acquire a fusion feature model corresponding to the request type; a second feature module, used to input the initial logging features corresponding to the multiple types of logging data into the fusion feature model to obtain the fusion feature of the target well; and a cementing quality determination module, which obtains a target cementing quality corresponding to the target well based on the fusion feature.

[0013] According to another aspect of an embodiment of the present invention, a non-volatile storage medium is provided, wherein the non-volatile storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing any one of the cementing quality determination methods.

[0014] According to another aspect of an embodiment of the present invention, there is provided an electronic device, comprising: one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement any one of the cementing quality determination methods.

[0015] In the embodiment of the present invention, a method combining machine learning is adopted to obtain a cementing evaluation request carrying multiple types of logging data for a target well, wherein the multiple types of logging data are used to describe the strength between the well wall and the cementing binder from different angles; determine the initial logging features corresponding to the multiple types of logging data; determine the request type of the cementing evaluation request, and obtain a fusion feature model corresponding to the request type; input the initial logging features corresponding to the multiple types of logging data into the fusion feature model to obtain the fusion features of the target well; based on the fusion features, obtain the target cementing quality corresponding to the target well. The purpose of automatically determining the cementing quality in the target well section by using multiple types of logging data is achieved, and the technical effect of improving the efficiency of processing cementing quality is achieved, thereby solving the technical problem of unsatisfactory processing efficiency for cementing quality determination in the related art. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0017] Figure 1 is a flow chart of an optional cementing quality determination method provided according to an embodiment of the present invention;

[0018] Figure 2 is a schematic flow chart of an optional cementing quality determination method provided according to an embodiment of the present invention;

[0019] Figure 3 is a schematic diagram of an optional cementing quality determination method provided according to an embodiment of the present invention;

[0020] Figure 4 is a schematic diagram of an optional cementing quality determination device provided according to an embodiment of the present invention; DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only 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 should fall within the scope of protection of the present invention.

[0022] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] The cementing scheme currently used in related technologies is designed by relevant technicians based on years of rich experience and a large number of calculations. The traditional method is based on statistical principles and has poor fault tolerance, anti-interference and adaptability. The logging data relied on in the cementing process has a huge volume of data, which leads to complex calculations in the evaluation process. In addition, due to the manual error factors in the manual evaluation stage, it will take too long to determine the cementing quality of a well, and there will be accuracy problems.

[0024] In response to the above problems, an embodiment of the present invention provides an embodiment of a method for determining cementing quality. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0025] Figure 1 is a flow chart of a method for determining cementing quality according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0026] Step S102, obtaining a cementing evaluation request carrying multiple types of logging data for a target well, wherein the multiple types of logging data are used to describe the strength between the well wall and the cementing bond from different angles;

[0027] It can be understood that the various types of logging data for the target well describe the strength between the wellbore and the cementing bond from different angles according to the corresponding types. The cementing bond can be pumped or pressure injected into the wellbore to fill the gap between the wellbore and the casing to form a strong, sealed bond to enhance the stability of the wellbore and prevent well fluid leakage. The purpose of obtaining a cementing evaluation request is to determine the cementing quality of the target well, which plays an important role in the oil and gas production process, and can ensure the integrity of the wellbore of the target well and increase the wellhead production.

[0028] Optionally, the above-mentioned cementing binder refers to a material used to fix casing and prevent well wall collapse and leakage in oil wells, gas wells or water wells, and is a slurry mixed with cement, sand and other additives.

[0029] Step S104, determining initial well logging features corresponding to various types of well logging data;

[0030] It can be understood that various types of well logging data are multi-source data, from which corresponding initial well logging features can be obtained. The above initial well logging features are features characterized from the well logging data of this type.

[0031] In an optional embodiment, initial logging features corresponding to multiple types of logging data are determined, including: determining neural network models corresponding to multiple types of logging data, wherein the multiple types of logging data correspond to different neural network models, wherein the multiple types of logging data include two-dimensional logging data, one-dimensional logging data, and discrete logging data, the two-dimensional logging data includes at least any one of the following: a logging image for a target well, and an acoustic variable density curve, the one-dimensional logging data includes at least any one of the following: acoustic time difference and acoustic amplitude obtained by acoustic logging of the target well, and the discrete logging data includes at least any one of the following: interface grade evaluation data obtained by dividing the interface in the target well, and interface marking data; the multiple types of logging data are respectively input into the corresponding neural network models for processing to obtain initial logging features corresponding to the multiple types of logging data.

[0032] It can be understood that various types of well logging data may be data of different dimensions and have different processing requirements, so a neural network model is determined to process them respectively. In other words, each type of well logging data has a corresponding neural network model for processing. The above-mentioned various types of well logging data include two-dimensional well logging data, one-dimensional well logging data, and discrete well logging data. The various types of well logging data are input into the corresponding neural network model for processing, and the initial well logging features corresponding to each type of well logging data in the various types of well logging data can be obtained.

[0033] Optionally, the neural network model corresponding to the above two-dimensional logging data is determined to be at least one of the following: Convolutional Neural Network (CNN), Capsule Neural Network (CapsNet). Convolutional neural networks mainly extract image features through convolution layers, reduce the dimension of features through pooling layers, and finally classify through fully connected layers. The basic unit of capsule neural networks is "capsule", and each capsule is represented by a vector. This vector contains two parts of information: one represents the existence probability of the entity represented by the capsule (such as the existence probability of an object or parameter), and the other represents the attributes of the entity (such as the posture and features of the object). The connection weights between capsules represent the relationship between the two entities. These weights will be automatically adjusted according to the training data, so that the capsule neural network can learn the hierarchical relationship between entities. The above convolutional neural networks and capsule neural networks can effectively process the logging images and acoustic variable density curves in the two-dimensional logging data.

[0034] Optionally, the above-mentioned variable density log (VDL) can provide the acoustic wave reflection information between the well wall and the cementing binder. By analyzing the acoustic variable density log data, the density change of the cementing binder can be evaluated, including the quality uniformity and density of the cementing binder.

[0035] Optionally, the neural network model corresponding to the one-dimensional well logging data is determined to be at least one of the following: Fully Connected Neural Network (FCN), Residual Neural Network (ResNet), Recurrent Neural Network (RNN). Fully connected neural network is also called multi-layer perceptron. The main feature of the network is that it is composed of multiple neurons arranged in layers. Each neuron is connected to all neurons in the previous layer. It is calculated through weights and activation functions, and then the calculation results are passed to the neurons in the next layer. It is suitable for processing feature extraction and classification tasks of one-dimensional well logging data. Residual neural network is a deep neural network architecture. It introduces skip connection to solve the problems of gradient disappearance and gradient explosion in the training process of deep network. When processing one-dimensional well logging data, residual neural network can better capture the details and local structure of input data, and improve the expression ability and learning ability of the network. Recurrent neural network is mainly used to process sequence data. When processing one-dimensional well logging data, recurrent neural network can model the temporal dependency of data, capture the time correlation in data, and share the same set of parameters at each time step, so that the network can keep memory and process sequence data of variable length.

[0036] Optionally, the above-mentioned Acoustic Time of Flight Difference (AC), as a logging parameter, can provide information on the propagation velocity of acoustic waves between the well wall and the cementing bond. By analyzing the acoustic time difference data, the quality of the cementing bond can be evaluated, including the strength and density of the cementing bond. The acoustic amplitude (CementBond Log, CBL) can provide information on the attenuation of acoustic wave propagation between the well wall and the cementing bond. By analyzing the acoustic amplitude data, the integrity of the cementing bond can be evaluated, including cracks and defects in the cementing bond.

[0037] Optionally, the neural network model corresponding to the above discrete logging data is determined to be at least one of the following: a fully connected neural network, a residual neural network. For discrete logging data, a fully connected neural network can automatically learn the association and nonlinear relationship between features, thereby extracting important features in the data. And residual connections can help the network learn the residual part in the discrete logging data, thereby more accurately finding the features and changes in the discrete logging data.

[0038] Optionally, the target well is evaluated according to the interface cementing, which is to evaluate and determine the cementing quality between different strata in the wellbore. In oil and gas drilling, the wellbore of the target well usually passes through different strata, and cementing operations are required between each stratum to prevent wellbore collapse and well fluid leakage. Therefore, discrete logging data such as interface grade evaluation data and interface marking data are required to be expressed.

[0039] Optionally, in order to obtain the acoustic time difference, acoustic amplitude and other parameters of the target well, acoustic logging can be performed. Acoustic logging is a technology that uses the characteristics of acoustic wave propagation in the formation to obtain formation information. In the process of acoustic logging, an acoustic wave transmitter is generally used to send the acoustic wave signal into the well, and then the reflected acoustic wave signal is received by the receiver. According to the time delay and amplitude of the received acoustic wave signal, the time difference and acoustic amplitude of the acoustic wave can be calculated. The specific processing steps include: using an acoustic wave transmitter to send the acoustic wave signal into the well, using a receiver to receive the reflected acoustic wave signal, and processing the received acoustic wave signal, including time domain and frequency domain analysis. By comparing the time delay of transmitting and receiving the acoustic wave signal, the time difference of the acoustic wave can be calculated to determine the depth and interface position of the formation. According to the amplitude of the received acoustic wave signal, the attenuation of the acoustic wave can be calculated to evaluate the parameters such as the lithology and porosity of the formation. The results of acoustic logging are affected by many factors such as formation media and wellbore conditions. Therefore, these factors need to be considered during data processing and interpretation to obtain accurate parameters such as acoustic time difference and amplitude.

[0040] Optionally, there are multiple ways to obtain the above-mentioned logging images, which are obtained by lowering logging instruments in the well to measure the physical properties and geological information of the formation. Different types of logging instruments and processing methods will produce different types of logging images, such as electrical logging, acoustic logging, natural gamma logging, etc. For example, a logging instrument is used to measure in the well and record various physical parameters, such as resistivity, natural gamma, acoustic wave velocity, density, etc. The corrected logging data is processed, including filtering, inversion, interpolation and other operations, to extract the characteristic information of the formation where the target well is located, and the formation interface, lithology, porosity, oil and gas content and other information can be identified. On this basis, according to the results of data interpretation, the formation information is presented in the form of images, such as logging curves, profiles, logging cross-sections, etc.

[0041] In an optional embodiment, the two-dimensional logging data is preprocessed by at least one of the following to obtain preprocessed data of the two-dimensional logging data, including: missing value completion, data normalization, data standardization, outlier processing, grayscale, and opening and closing operations.

[0042] In an optional embodiment, the one-dimensional logging data is preprocessed by performing at least one of the following preprocessing to obtain preprocessed data of the one-dimensional logging data, including: missing value completion, data normalization, data standardization, and outlier processing.

[0043] It should be noted that the difference between data normalization and data standardization is that data normalization is to scale the data to a specific range, usually by linearly mapping the data to [0,1] or [-1,1], such as Min-Max Normalization and Feature Scaling. Data normalization can eliminate the dimensional differences between features, so that different features have the same scale range, which is conducive to the convergence and performance of the model. Data standardization is to convert the data into a standard normal distribution with a mean of 0 and a standard deviation of 1. For example, Z-score standardization and mean variance standardization, data standardization can eliminate the deviation and scale differences between features, making the distribution of features closer to the normal distribution, which is conducive to the robustness and stability of the model to outliers. In other words, data normalization focuses on the range of data, and data standardization focuses on the distribution form of data.

[0044] In an optional embodiment, the discrete logging data is preprocessed by at least one of the following to obtain preprocessed data of one-dimensional logging data, including: missing value completion, data normalization, data standardization, outlier processing, numerical representation, feature encoding, and one-hot encoding.

[0045] Optionally, missing value filling is to adopt certain methods to fill or estimate missing values. Missing value filling methods include mean, median, mode filling, regression model prediction, interpolation method, etc.

[0046] Optionally, outlier processing is to take certain methods to detect and process these outliers. Outliers may be caused by measurement errors, input errors or the particularity of the data itself. Outlier processing methods include outlier correction based on statistics, Euclidean distance and other methods.

[0047] Optionally, grayscale refers to converting a color image into a grayscale image. In a grayscale image, each pixel has only one brightness value, and no longer has color information of three channels: red, green, and blue. Grayscale can simplify the image processing and calculation process, and retain the feature information in the logging image.

[0048] Optionally, the opening and closing operations can be used to process the logging images in the two-dimensional logging data, and perform image segmentation, edge detection, image noise removal and other processing.

[0049] In an optional embodiment, when the number of output units of the neural network models corresponding to the multiple types of well logging data are equal, the initial well logging features corresponding to the multiple types of well logging data are input into the fusion feature model to obtain the fusion features of the target well, including: determining the neural network models corresponding to the multiple types of well logging data, and the addition layer between the fusion feature model, wherein the number of input units of the addition layer is equal to the number of output units, and the addition layer is provided with a linear rectification function and / or a residual function; the initial well logging features corresponding to the multiple types of well logging data are input into the fusion feature model through the addition layer to obtain the fusion features.

[0050] It can be understood that the addition layer refers to an operation of adding multiple input features, which is used to fuse and integrate the initial logging features corresponding to various types of logging data, and then input them into the input layer of the fusion feature model to obtain fusion features. The above-mentioned addition layer can choose to use a linear rectification function (i.e., ReLU function) and / or a residual function to process and activate features, so as to enhance the representation ability and learning ability in the process of feature fusion and feature transfer.

[0051] Optionally, a linear rectifier function (Rectified Linear Unit, ReLU) is defined as outputting 0 when the input is less than 0, and outputting the input value itself when the input is greater than 0. In the addition layer, the ReLU function can be used to perform nonlinear activation on the added features to increase the nonlinear capability of the network.

[0052] Optionally, a residual function refers to an operation that adds the input directly to the output, that is, the output is equal to the input plus the residual term. In the addition layer, the residual connection can be realized by adding the input features directly to the summed features. The residual connection can help the rapid transmission of information and the direct feedback of gradients, which helps to alleviate the gradient vanishing problem and accelerate the convergence speed of the network.

[0053] It should be noted that the above-mentioned addition layer itself does not involve parameter change. When the numbers of output units of the neural network models corresponding to the various types of well logging data are not equal, the above-mentioned addition layer is not set.

[0054] Step S106, determining the request type of the cementing evaluation request, and acquiring a fusion feature model corresponding to the request type;

[0055] It can be understood that according to different request types, the corresponding fusion feature models to be processed are different.

[0056] In an optional embodiment, the request type includes a qualitative request, the fusion feature model includes a first fusion sub-model, the first fusion sub-model is used to process the qualitative request, and the fusion feature model corresponding to the request type is obtained, including: determining that the first loss function corresponding to the qualitative request is a cross entropy loss function; obtaining training data whose data type matches multiple types of logging data; inputting the training data into a predetermined first initial sub-model and using a first loss function to obtain a first output error value; when the first output error value is less than a predetermined first error value threshold, determining that the first initial sub-model is a first fusion sub-model.

[0057] It can be understood that the qualitative request is to obtain the discrete qualitative characteristics of the target well, which affects the setting of the first loss function in the first fusion sub-model. The trained first fusion sub-model is used to perform classification tasks. By setting the first loss function to the cross entropy loss function, the model accuracy of the first fusion sub-model can be effectively determined. According to the first loss function, the optimizer is used to iterate the parameters. The smaller the first output error value of the output result, that is, less than the first error value threshold, the more the first fusion sub-model can meet the expected effect. At this time, the training of the first fusion sub-model ends. If the first output error value of the output result of the first fusion sub-model is still large, the first fusion sub-model has not yet achieved the expected result. At this time, the first fusion sub-model should be adjusted and re-trained.

[0058] In an optional embodiment, the request type includes a quantitative request, the fusion feature model includes a second fusion sub-model, the second fusion sub-model is used to process the quantitative request, and the fusion feature model corresponding to the request type is obtained, including: determining that the second loss function corresponding to the quantitative request is a mean absolute error function and / or a mean square error function; obtaining training data whose data type matches multiple types of logging data; inputting the training data into the second initial sub-model and using the second loss function to obtain a second output error value; when the second output error value is less than a predetermined second error value threshold, determining that the second initial sub-model is the second fusion sub-model.

[0059] It can be understood that the quantitative request is to obtain the cementation index of the target well, which affects the setting of the second loss function in the second fusion sub-model. The trained second fusion sub-model is used to perform regression tasks. By setting the first loss function to the mean absolute error function and / or the mean square error function, the model accuracy of the second fusion sub-model can be effectively determined. The optimizer is used to iterate the parameters according to the second loss function. The smaller the second output error value of the output result, that is, less than the second error value threshold, the more the second fusion sub-model can meet the expected effect. At this time, the training of the second fusion sub-model ends. If the second output error value of the output result of the second fusion sub-model is still large, the second fusion sub-model has not yet achieved the expected result. At this time, the second fusion sub-model should be adjusted and re-trained.

[0060] Step S108, inputting the initial well logging features corresponding to the various types of well logging data into the fusion feature model to obtain the fusion features of the target well;

[0061] It can be understood that the initial logging features corresponding to various types of logging data are respectively input into the fusion feature model for processing, and the output result is the fusion feature of the target well.

[0062] In an optional embodiment, the request type is a qualitative request for a target well, and the fusion feature model includes a first fusion sub-model, which is used to process the qualitative request, and the initial well logging features corresponding to the various types of well logging data are respectively input into the fusion feature model to obtain the fusion features of the target well, including: determining a first output layer corresponding to the qualitative request, wherein the first output layer is provided with a predetermined activation function; when the first fusion sub-model includes a predetermined first input layer, a predetermined first hidden layer, and a first output layer, the initial well logging features corresponding to the various types of well logging data are respectively input into the predetermined first input layer included in the fusion feature model to obtain the first well logging features corresponding to the various types of well logging data; the first well logging features corresponding to the various types of well logging data are respectively input into the predetermined first hidden layer included in the first fusion sub-model to obtain the second well logging features corresponding to the various types of well logging data; the second well logging features corresponding to the various types of well logging data are respectively input into the first output layer included in the first fusion sub-model to obtain the discrete qualitative features corresponding to the target well as the fusion features.

[0063] It can be understood that different request types require different fusion sub-models to be input, and the expected output results are also different. For qualitative requests, the first fusion sub-model needs to be used for processing, and the first output layer is set with a predetermined activation function. If the activation function is set in the output layer, an appropriate activation function is usually used to limit the range of the output value, so that it is suitable for the discrete qualitative features required by the qualitative request, and suitable for the first fusion sub-model to perform the classification task. Through the processing of the predetermined first input layer, the predetermined first hidden layer, and the first output layer in the first fusion sub-model, the output result can be obtained as a discrete qualitative feature as a fusion feature.

[0064] Optionally, the above-mentioned predetermined activation function can be set to a softmax function, or a sigmoid function, etc., and can be set according to specific needs. The function of the softmax function is to convert a set of real numbers into a probability distribution, and the input real number can be converted into a probability, so that the sum of the probabilities of all outputs is 1, and the probability of each output is between 0 and 1. The function of the sigmoid function is to map real numbers to the range of 0 to 1. It is often used in the output layer of a binary classification problem, and can convert the input real number into a probability, indicating the probability of an event occurring.

[0065] In an optional embodiment, the request type is a quantitative request for a target well, and the fusion feature model includes a second fusion sub-model, which is used to process the quantitative request, and the initial well logging features corresponding to the various types of well logging data are respectively input into the fusion feature model to obtain the fusion feature of the target well, including: determining the second output layer corresponding to the quantitative request, wherein the second output layer is prohibited from setting a predetermined activation function; when the second fusion sub-model includes a predetermined second input layer, a predetermined second hidden layer, and a second output layer, the initial well logging features corresponding to the various types of well logging data are respectively input into the predetermined second input layer included in the second fusion sub-model to obtain the third well logging features corresponding to the various types of well logging data; the third well logging features corresponding to the various types of well logging data are respectively input into the predetermined second hidden layer included in the second fusion sub-model to obtain the fourth well logging features corresponding to the various types of well logging data; the fourth well logging features corresponding to the various types of well logging data are respectively input into the second output layer included in the second fusion sub-model to obtain the cementation index corresponding to the target well as the fusion feature.

[0066] It can be understood that the second fusion sub-model needs to be used for processing the quantitative request, and the second output layer cannot be set with a predetermined activation function. If the activation function is not set in the output layer, the output value is linear, that is, equal to the weighted sum of the input layer, which is suitable for regression tasks, where the output value can be any real number, and the bonding index required for the quantitative request can be output. Through the processing of the predetermined second input layer, the predetermined second hidden layer, and the second output layer in the second fusion sub-model, the output result can be obtained as the bonding index as a fusion feature.

[0067] It should be noted that the model structures of the first fusion sub-model and the second fusion sub-model are different. Whether to set the activation function depends on the output requirements corresponding to the first fusion sub-model and the second fusion sub-model.

[0068] Optionally, the above-mentioned bonding index is a physical property and mechanical property characterizing the cementing bond, including bonding strength, bonding density, bonding hardness and other indicators. The bonding index can reflect the quality and performance of the cementing bond. A high bonding index indicates that the cementing bond is of good quality, has high bonding strength and density, and can effectively fix the casing and prevent wellbore collapse. A low bonding index may mean that there are quality problems in the cementing bond, such as poor bonding, low bonding strength, etc., which may lead to problems such as wellbore instability and well fluid leakage. Therefore, the evaluation and analysis of the bonding index of the target well is conducive to ensuring the quality of cementing, ensuring the stability of the wellbore, and facilitating well fluid control and other treatments.

[0069] Step S110, obtaining the target cementing quality corresponding to the target well based on the fusion features.

[0070] It can be understood that based on the fusion features, the target cementing quality of the target well can be characterized. Cementing is to stabilize the well wall and prevent the well wall from collapsing or caving in. By ensuring the quality of cementing, sufficient support force and the ability to prevent the well wall from collapsing can be provided, thereby protecting the stability of the target wellbore.

[0071] Optionally, the above-mentioned fused features are subjected to data processing, and then the target cementing quality is obtained. When the fused features are discrete cementing features, the fused features are subjected to numerical representation, feature coding and other processing. When the fused features are cementation indexes, the fused features are subjected to data normalization, data standardization and other processing.

[0072] It should be noted that the target cementing quality can also be used in a variety of application scenarios such as preventing well fluid leakage, increasing oil and gas production, reducing operational risks, and improving drilling efficiency. The sealing performance of the cementing binder can prevent the leakage of well fluid in the formation, thereby ensuring the pressure and fluid control inside the wellbore. Good cementing quality can avoid well fluid spillage, environmental pollution, and waste of resources. The quality of cementing has an important impact on oil and gas production. High-quality cementing binders can ensure the effective sealing between the wellbore and the formation, reduce the leakage and loss of oil and gas, and increase production. Insufficient cementing quality may lead to wellbore instability, wellbore collapse, well fluid leakage, and oil and gas well spillage, all of which will increase operational risks and safety hazards. By ensuring cementing quality, these risks can be reduced and the safety of operations can be ensured. Good cementing quality can reduce the time and cost of cementing operations and improve drilling efficiency. If the cementing quality does not meet the standards, cementing repair or re-construction may be required, which will increase operation time and cost. Therefore, determining the cementing quality of the target well can be applied to many geological engineering scenarios to ensure wellbore stability, increase production, reduce risks, and improve drilling efficiency.

[0073] Through the above step S102, a cementing evaluation request carrying multiple types of logging data for a target well is obtained, wherein the multiple types of logging data are used to describe the strength between the well wall and the cementing binder from different angles; step S104, initial logging features corresponding to the multiple types of logging data are determined; step S106, the request type of the cementing evaluation request is determined, and a fusion feature model corresponding to the request type is obtained; step S108, the initial logging features corresponding to the multiple types of logging data are input into the fusion feature model to obtain the fusion features of the target well; step S110, based on the fusion features, the target cementing quality corresponding to the target well is obtained. The purpose of automatically determining the cementing quality in the target well section by using multiple types of logging data can be achieved, and the technical effect of improving the efficiency of processing cementing quality is achieved, thereby solving the technical problem of unsatisfactory processing efficiency for cementing quality determination in the related art.

[0074] Based on the above embodiments and optional embodiments, the present invention proposes an optional implementation mode: Figure 2 Detailed description is given below of an optional method for determining cementing quality according to an embodiment of the present invention.

[0075] Step S1, determine the target well that needs to be evaluated for cementing quality, and the cementing evaluation request carries two-dimensional logging data, one-dimensional logging data, and discrete logging data. The request types of the above cementing evaluation request include qualitative request and quantitative request, that is, the discrete cementing characteristics and cementation index of the target well are expected to be output. The above discrete cementing characteristics can be set as three levels of "good", "medium" and "poor" as cementing conclusions.

[0076] Step S2, pre-process the two-dimensional logging data, the one-dimensional logging data, and the discrete logging data, and then input them into the corresponding neural network models for processing, to obtain the initial logging features corresponding to the two-dimensional logging data, the one-dimensional logging data, and the discrete logging data. The two-dimensional logging data corresponds to the neural network model 1, the one-dimensional logging data corresponds to the neural network model 2, and the discrete logging data corresponds to the neural network model 3.

[0077] Step S3, in order to process the qualitative request, the initial logging features corresponding to the two-dimensional logging data, the one-dimensional logging data, and the discrete logging data are respectively input into the first fusion sub-model in the fusion feature model for processing to obtain the discrete qualitative features of the target well.

[0078] Step S4, in order to process the quantitative request, the initial logging features corresponding to the two-dimensional logging data, the one-dimensional logging data, and the discrete logging data are respectively input into the second fusion sub-model in the fusion feature model for processing to obtain the cementation index of the target well.

[0079] The first fusion sub-model and the second fusion sub-model are models trained using training data, and have different model structure settings respectively.

[0080] Step S5, after obtaining the fusion feature, the cementation index is subjected to data normalization or data standardization, and the discrete qualitative feature is subjected to numerical representation or feature coding, and finally the target cementing quality of the target well is obtained based on the processed cementation index and discrete qualitative feature.

[0081] To facilitate understanding of the algorithm flow, Figure 3 is a schematic diagram of an optional cementing quality determination method provided according to an embodiment of the present invention, such as Figure 3 As shown, after preprocessing, the two-dimensional logging data, the one-dimensional logging data, and the discrete logging data are respectively input into the neural network model 1, the neural network model 2, and the neural network model 3 for processing, and the initial logging features corresponding to the two-dimensional logging data, the one-dimensional logging data, and the discrete logging data are obtained. According to the difference between qualitative requests and quantitative requests, they are respectively input into the first fusion sub-model and the second fusion sub-model, and the discrete cementing features and the cementation index are obtained as fusion features. After data processing of the output results, the target cementing quality of the target well is obtained.

[0082] The above optional implementation method achieves at least the following effects: the data used for cementing quality evaluation are divided into three categories according to their characteristics: two-dimensional logging data, one-dimensional logging data, and discrete logging data, and the neural network models are determined according to the characteristics of the data. Then, through the request type, the feature model is fused with the above multiple neural network models to enable the three types of logging data with different structures, characteristics, and dimensions to jointly evaluate the cementing quality. The hybrid model is used to analyze multi-source data, evaluate the cementing quality, and improve the efficiency and accuracy of cementing quality evaluation.

[0083] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0084] In this embodiment, a cementing quality determination device is also provided, which is used to implement the above embodiments and preferred implementations, and the descriptions that have been made will not be repeated. As used below, the terms "module" and "device" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0085] According to an embodiment of the present invention, there is also provided an embodiment of a device for implementing a method for determining cementing quality. Figure 4 is a schematic diagram of a cementing quality determination device according to an embodiment of the present invention. Figure 4 As shown, the above-mentioned cementing quality determination device includes: an acquisition module 402, a first feature module 404, a model determination module 406, a second feature module 408, and a cementing quality determination module 410. The device is described below.

[0086] An acquisition module 402 is used to acquire a cementing evaluation request carrying multiple types of logging data for a target well, wherein the multiple types of logging data are used to describe the strength between the well wall and the cementing bond from different angles;

[0087] A first feature module 404, connected to the acquisition module 402, is used to determine initial well logging features corresponding to various types of well logging data;

[0088] The model determination module 406 is connected to the first feature module 404 and is used to determine the request type of the cementing evaluation request and obtain a fusion feature model corresponding to the request type;

[0089] The second feature module 408 is connected to the model determination module 406 and is used to input the initial well logging features corresponding to the various types of well logging data into the fusion feature model to obtain the fusion features of the target well;

[0090] The cementing quality determination module 410 is connected to the second feature module 408, and obtains the target cementing quality corresponding to the target well based on the fusion feature.

[0091] In a cementing quality determination device provided by an embodiment of the present invention, an acquisition module 402 is set to obtain a cementing evaluation request carrying multiple types of logging data for a target well, wherein the multiple types of logging data are used to describe the strength between the well wall and the cementing binder from different angles; a first feature module 404 is connected to the acquisition module 402, and is used to determine the initial logging features corresponding to the multiple types of logging data; a model determination module 406 is connected to the first feature module 404, and is used to determine the request type of the cementing evaluation request and obtain a fusion feature model corresponding to the request type; a second feature module 408 is connected to the model determination module 406, and is used to input the initial logging features corresponding to the multiple types of logging data into the fusion feature model to obtain the fusion feature of the target well; a cementing quality determination module 410 is connected to the second feature module 408, and obtains the target cementing quality corresponding to the target well based on the fusion feature. The purpose of automatically determining the cementing quality in the target well section by using various types of logging data is achieved, and the technical effect of improving the efficiency of processing cementing quality is realized, thereby solving the technical problem of unsatisfactory processing efficiency for cementing quality determination in related technologies.

[0092] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0093] It should be noted that the acquisition module 402, the first feature module 404, the model determination module 406, the second feature module 408, and the cementing quality determination module 410 correspond to steps S102 to S110 in the embodiment, and the examples and application scenarios implemented by the modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above embodiment. It should be noted that the modules as part of the device can be run in a computer terminal.

[0094] It should be noted that the optional or preferred implementation of this embodiment can refer to the relevant description in the embodiment, which will not be repeated here.

[0095] The above-mentioned cementing quality determination device may also include a processor and a memory, and the acquisition module 402, the first feature module 404, the model determination module 406, the second feature module 408, the cementing quality determination module 410, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize the corresponding functions.

[0096] The processor includes a kernel, which retrieves the corresponding program unit from the memory. There can be one or more kernels. The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one storage chip.

[0097] An embodiment of the present invention provides a non-volatile storage medium on which a program is stored. When the program is executed by a processor, a method for determining cementing quality is implemented.

[0098] An embodiment of the present invention provides an electronic device, which includes a processor, a memory, and a program stored in the memory and can be run on the processor. When the processor executes the program, the following steps are implemented: obtaining a cementing evaluation request carrying multiple types of logging data for a target well, wherein the multiple types of logging data are used to describe the strength between the well wall and the cementing cement body from different angles; determining initial logging features corresponding to the multiple types of logging data; determining the request type of the cementing evaluation request, and obtaining a fusion feature model corresponding to the request type; inputting the initial logging features corresponding to the multiple types of logging data into the fusion feature model to obtain the fusion feature of the target well; based on the fusion feature, obtaining the target cementing quality corresponding to the target well. The device in this article can be a server, a PC, etc.

[0099] Optionally, the memory included in the electronic device may be independent or integrated with the processor. The electronic device also includes a communication component, wherein the processor, the memory and the communication component are connected via a bus.

[0100] The present invention also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes the following method steps: obtaining a cementing evaluation request carrying multiple types of logging data for a target well, wherein the multiple types of logging data are used to describe the strength between the well wall and the cementing bond from different angles; determining initial logging features corresponding to the multiple types of logging data; determining the request type of the cementing evaluation request, and obtaining a fusion feature model corresponding to the request type; inputting the initial logging features corresponding to the multiple types of logging data into the fusion feature model to obtain the fusion features of the target well; and obtaining the target cementing quality corresponding to the target well based on the fusion features.

[0101] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0102] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0103] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0104] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0105] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0106] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0107] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0108] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0109] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0110] The above are only embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A method for determining cementing quality, characterized in that: include: Acquire a cementing evaluation request carrying multiple types of logging data for a target well, wherein the multiple types of logging data are used to describe the strength between the well wall and the cementing bond from different angles; Determining initial well logging features respectively corresponding to the multiple types of well logging data; Determining a request type of the cementing evaluation request, and acquiring a fusion feature model corresponding to the request type; Inputting the initial well logging features respectively corresponding to the multiple types of well logging data into the fusion feature model to obtain the fusion feature of the target well; Based on the fusion features, a target cementing quality corresponding to the target well is obtained.

2. The method according to claim 1, characterized in that The determining of initial well logging features respectively corresponding to the multiple types of well logging data includes: Determine the neural network models corresponding to the multiple types of logging data, respectively, wherein the multiple types of logging data correspond to different neural network models, respectively, wherein the multiple types of logging data include two-dimensional logging data, one-dimensional logging data, and discrete logging data, wherein the two-dimensional logging data include at least any one of the following: a logging image for the target well, and an acoustic variable density curve, wherein the one-dimensional logging data include at least any one of the following: an acoustic time difference and an acoustic amplitude obtained by acoustic logging of the target well, and wherein the discrete logging data include at least any one of the following: interface grade evaluation data obtained by dividing interfaces in the target well, and interface marking data; The multiple types of well logging data are respectively input into corresponding neural network models for processing to obtain initial well logging features respectively corresponding to the multiple types of well logging data.

3. The method according to claim 2, characterized in that In the case where the numbers of output units of the neural network models corresponding to the multiple types of well logging data are equal, inputting the initial well logging features corresponding to the multiple types of well logging data into the fusion feature model to obtain the fusion features of the target well includes: Determine the neural network models corresponding to the multiple types of well logging data, respectively, and the addition layer between the fusion feature model, wherein the number of input units of the addition layer is equal to the number of output units, and the addition layer is provided with a linear rectification function and / or a residual function; The initial well logging features respectively corresponding to the multiple types of well logging data are input into the fusion feature model through the addition layer to obtain the fusion feature.

4. The method according to claim 1, characterized in that: The request type includes a qualitative request, the fusion feature model includes a first fusion sub-model, the first fusion sub-model is used to process the qualitative request, and the acquiring of the fusion feature model corresponding to the request type includes: Determine that a first loss function corresponding to the qualitative request is a cross entropy loss function; Acquire training data whose data types match the multiple types of well logging data; Inputting the training data into a predetermined first initial sub-model and using the first loss function to obtain a first output error value; When the first output error value is less than a predetermined first error value threshold, the first initial sub-model is determined to be the first fusion sub-model.

5. The method according to claim 1, characterized in that The request type includes a quantitative request, the fusion feature model includes a second fusion sub-model, the second fusion sub-model is used to process the quantitative request, and the acquiring of the fusion feature model corresponding to the request type includes: Determine that a second loss function corresponding to the quantitative request is a mean absolute error function and / or a mean square error function; Acquire training data whose data types match the multiple types of well logging data; Input the training data into a second initial sub-model and use the second loss function to obtain a second output error value; When the second output error value is less than a predetermined second error value threshold, the second initial sub-model is determined to be the second fusion sub-model.

6. The method according to any one of claims 1 to 5, characterized in that The request type is a qualitative request for the target well, the fusion feature model includes a first fusion sub-model, and the first fusion sub-model is used to process the qualitative request. The initial well logging features corresponding to the multiple types of well logging data are input into the fusion feature model to obtain the fusion features of the target well, including: Determining a first output layer corresponding to the qualitative request, wherein the first output layer is provided with a predetermined activation function; In the case where the first fusion sub-model includes a predetermined first input layer, a predetermined first hidden layer, and the first output layer, the initial well logging features respectively corresponding to the multiple types of well logging data are input into the predetermined first input layer included in the fusion feature model to obtain the first well logging features respectively corresponding to the multiple types of well logging data; Inputting the first well logging features respectively corresponding to the multiple types of well logging data into a predetermined first hidden layer included in the first fusion sub-model to obtain the second well logging features respectively corresponding to the multiple types of well logging data; The second well logging features respectively corresponding to the multiple types of well logging data are input into the first output layer included in the first fusion sub-model to obtain discrete qualitative features corresponding to the target well as the fusion features.

7. The method according to any one of claims 1 to 5, characterized in that The request type is a quantitative request for the target well, the fusion feature model includes a second fusion sub-model, and the second fusion sub-model is used to process the quantitative request. The initial well logging features corresponding to the multiple types of well logging data are input into the fusion feature model to obtain the fusion features of the target well, including: Determining a second output layer corresponding to the quantitative request, wherein the second output layer is prohibited from setting a predetermined activation function; In the case where the second fusion sub-model includes a predetermined second input layer, a predetermined second hidden layer, and the second output layer, the initial well logging features respectively corresponding to the multiple types of well logging data are input into the predetermined second input layer included in the second fusion sub-model to obtain third well logging features respectively corresponding to the multiple types of well logging data; Inputting the third well logging features respectively corresponding to the multiple types of well logging data into a predetermined second hidden layer included in the second fusion sub-model to obtain fourth well logging features respectively corresponding to the multiple types of well logging data; The fourth well logging features respectively corresponding to the multiple types of well logging data are input into the second output layer included in the second fusion sub-model to obtain the cementation index corresponding to the target well as the fusion feature.

8. A cementing quality determination device, characterized in that: include: An acquisition module, used to acquire a cementing evaluation request carrying multiple types of logging data for a target well, wherein the multiple types of logging data are used to describe the strength between the well wall and the cementing bond from different angles; A first feature module is used to determine initial well logging features corresponding to the multiple types of well logging data respectively; A model determination module, used to determine the request type of the cementing evaluation request and obtain a fusion feature model corresponding to the request type; A second feature module is used to input the initial well logging features corresponding to the multiple types of well logging data into the fusion feature model to obtain the fusion feature of the target well; The cementing quality determination module obtains the target cementing quality corresponding to the target well based on the fusion feature.

9. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the cementing quality determination method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: include: One or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the cementing quality determination method described in any one of claims 1 to 7.

Images