Intelligent identification method and system for scaling state of mechanical production well
By combining causal reasoning with visualization dimensionality reduction technology, a CNN-LSTM neural network model was constructed, which solved the problem of rapid and accurate identification of scaling status in machine-operated wells, improved detection efficiency and accuracy, and reduced equipment failure and maintenance costs.
Patent Information
- Application Number
- CN202511131950.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-13
- Publication Date
- 2025-11-18
AI Technical Summary
Existing technologies struggle to quickly and accurately identify the scaling status of machine-produced wells. Traditional detection methods are inefficient and cannot acquire scaling status in real time. Existing data-driven methods are ill-suited to handle high-dimensionality, complexity, and diversity.
By combining causal reasoning and visualization dimensionality reduction techniques, outliers and noise in machine-produced well data are filtered out, scale-related features are selected, a CNN-LSTM neural network model is constructed, and an SE attention mechanism is introduced. Multi-class cross-entropy is used for training and prediction.
It enables efficient identification and prediction of scaling conditions in machine-operated wells, improves detection efficiency and accuracy, reduces equipment failure and maintenance costs, and has strong universality and practical value.
Smart Images
Figure CN120974278A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of oilfield production technology, and in particular relates to an intelligent identification method and system for scaling status in mechanically operated wells. Background Technology
[0002] With the continuous development of the oil and gas industry, the level of intelligence and automation of oilfield production equipment has gradually improved. Mechanically operated wells, as one of the important pieces of equipment in oilfield production, are widely used in oil and gas extraction. However, with long-term use, scaling has gradually become one of the main factors restricting the efficient operation of mechanically operated wells. Scaling not only reduces the working efficiency of mechanically operated wells but may also cause equipment failure, increase maintenance costs, and even lead to oilfield production interruptions. Timely and accurate identification of scaling in mechanically operated wells and prevention of its impact on equipment has become an important task in modern oilfield management.
[0003] Traditional methods for detecting scale buildup in mechanized wells largely rely on physical monitoring and manual inspections, which suffer from long detection cycles, low efficiency, and the inability to obtain real-time scale status information. While some advanced sensing technologies and monitoring methods are applied in scale detection, the complexity and noise of the data often make it difficult to obtain accurate and real-time scale status information. Therefore, how to utilize big data and intelligent algorithms to achieve rapid prediction and accurate identification of scale status through in-depth analysis of mechanized well operation data has become a pressing technical challenge for oilfield equipment management.
[0004] In recent years, with the rapid development of machine learning and deep learning technologies, data-driven scaling identification methods have gradually gained attention. By collecting and analyzing multi-dimensional monitoring data from mechanically operated wells, combined with intelligent algorithms, early warning and accurate identification of scaling can be achieved. However, existing methods mostly rely on specific data models or empirical rules, making it difficult to cope with the high dimensionality, complexity, and diversity of data, resulting in certain limitations in practical applications. Summary of the Invention
[0005] The technical problem to be solved by this invention is to provide an intelligent identification method and system for the scaling state of machine-operated wells. By combining causal reasoning with visualization dimensionality reduction technology, the scaling characteristics can be accurately identified, providing a scientific basis for decision-making.
[0006] To achieve the above objectives, the present invention adopts the following technical solution:
[0007] A method for intelligent identification of scaling conditions in mechanically operated wells, comprising:
[0008] Step 1: Filter out outliers and noise from the machine-produced well dataset;
[0009] Step 2: Obtain reliable scaling labels based on the filtered machine-produced well dataset;
[0010] Step 3: Filter out the influencing factors that are highly correlated with scaling in machine-produced wells from the raw data and determine the scaling association characteristics; the raw dataset includes machine-produced well datasets and reliable scaling labels;
[0011] Step 4: Use the t-SNE algorithm to reduce the dimensionality of the original data and visualize it, and construct a separable pattern vector regression model;
[0012] Step 5: Construct a CNN-LSTM neural network model, introduce the SE attention mechanism to build the SE-LSTM algorithm, and use the Adam optimizer to optimize the model parameters;
[0013] Step 6: Based on the separable variables and scaling-related features, input them into a CNN-LSTM neural network model for training, use multi-class cross-entropy as the loss function for training, and output the scaling type;
[0014] Step 7: Using the trained CNN-LSTM neural network model, predict the scaling data of unknown machine-operated wells and generate a visualization of the scaling status.
[0015] As a preferred option, in step one, an outlier detection method based on standard deviation and a Gaussian filtering method are used to filter out outliers and noise information in the machine-produced well dataset.
[0016] As a preferred option, in step three, based on the positive causal logic analysis method, factors highly correlated with scaling in mechanically operated wells are screened from the original data to determine the scaling correlation characteristics.
[0017] This invention also provides an intelligent identification system for scaling status in mechanically operated wells, comprising:
[0018] The first processing module is used to filter out outliers and noise information in the machine-produced well dataset;
[0019] The second processing module is used to obtain reliable scaling labels based on the filtered machine-produced well dataset;
[0020] The third processing module is used to filter out influencing factors that are highly correlated with scaling in mechanically operated wells from the raw data and determine scaling association characteristics; the raw dataset includes mechanically operated well datasets and reliable scaling labels;
[0021] The fourth processing module is used to reduce the dimensionality of the original data and visualize it using the t-SNE algorithm, and to construct a separable pattern vector regression model.
[0022] The fifth processing module is used to construct a CNN-LSTM neural network model, introduce the SE attention mechanism to build the SE-LSTM algorithm, and use the Adam optimizer to optimize the model parameters.
[0023] The sixth processing module is used to input the separable variables and scaling-related features into a CNN-LSTM neural network model for training, using multi-class cross-entropy as the loss function for training, and outputting the scaling type.
[0024] The seventh processing module is used to predict the scaling data of unknown mechanically operated wells using a trained CNN-LSTM neural network model and generate a visualization of the scaling status.
[0025] As a preferred option, the first processing module uses an outlier detection method based on standard deviation and a Gaussian filtering method to filter out outliers and noise information in the machine-produced well data set.
[0026] As a preferred option, the third processing module uses a positive causal logic analysis method to screen out influencing factors that are highly correlated with scaling in machine-operated wells from the raw data and determine the scaling correlation characteristics.
[0027] This invention integrates causal analysis and data visualization techniques. Based on large-scale mechanized well operation data, it uses positive causal logic analysis to screen out key factors highly correlated with scaling, and employs the t-SNE algorithm to reduce the dimensionality of the data for visualization, improving its interpretability and operability. Furthermore, it combines deep learning models to construct a scaling prediction model for mechanized wells, thereby achieving efficient identification and prediction of scaling occurrence. Attached Figure Description
[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0029] Figure 1 This is a flowchart of the intelligent identification method for scaling status in mechanically operated wells according to an embodiment of the present invention;
[0030] Figure 2 This is a schematic diagram illustrating the use of the standard difference constant value detection method to identify outliers in the scaling data of a mechanically operated well according to the present invention.
[0031] Figure 3 This is a schematic diagram of the positive causal logic analysis process of the present invention;
[0032] Figure 4 This is a schematic diagram illustrating the dimensionality reduction visualization and regression performance of the original data according to the present invention;
[0033] Figure 5 This is a schematic diagram of the SE-LSTM architecture used in this invention;
[0034] Figure 6 This is a schematic diagram illustrating the loss function of the CNN-LSTM model during the optimization process in this invention.
[0035] Figure 7 This is a schematic diagram illustrating the visualization of unknown well scaling information according to the present invention. Detailed Implementation
[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0037] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0038] Example 1:
[0039] like Figure 1 As shown, this embodiment of the invention provides a method for intelligent identification of scaling status in mechanically operated wells, including:
[0040] Step 1: Filter out outliers and noise from the machine-produced well dataset;
[0041] Step 2: Obtain reliable scaling labels based on the filtered machine-produced well dataset;
[0042] Step 3: Filter out the influencing factors that are highly correlated with scaling in machine-produced wells from the raw data and determine the scaling association characteristics; the raw dataset includes machine-produced well datasets and reliable scaling labels;
[0043] Step 4: Use the t-SNE algorithm to reduce the dimensionality of the original data and visualize it, and construct a separable pattern vector regression model;
[0044] Step 5: Construct a CNN-LSTM neural network model, introduce the SE attention mechanism to build the SE-LSTM algorithm, and use the Adam optimizer to optimize the model parameters;
[0045] Step 6: Based on the separable variables and scaling-related features, input them into a CNN-LSTM neural network model for training, use multi-class cross-entropy as the loss function for training, and output the scaling type;
[0046] Step 7: Using the trained CNN-LSTM neural network model, predict the scaling data of unknown machine-operated wells and generate a visualization of the scaling status.
[0047] As one embodiment of the present invention, the method for filtering outliers and noise information in the machine-produced well dataset in step one is as follows:
[0048] First, by calculating the mean and standard deviation of the dataset, an outlier detection method based on standard deviation is used to identify and remove outliers in the machine-produced well data that exceed a specific range.
[0049]
[0050] like:
[0051] |x i -μ N |>n*σ (3)
[0052] Then determine x i This is an outlier. Figure 2 An example graph showing anomalies identified in a mechanically operated well is provided.
[0053] Where N represents the total number of data points in the machine-produced well dataset, x i This represents the value of a data point in a machine-operated well, μ represents the mean of the logging dataset, σ represents the standard deviation of the machine-operated well dataset, and n represents a multiple, which is usually a positive integer.
[0054] After removing outliers, Gaussian filtering is used to process the noise in the machine-produced well dataset. Convolution operations are used to reduce high-frequency noise in the data while preserving the main features.
[0055]
[0056] Where G(x,y) represents the output value of the Gaussian filter, σ represents the standard deviation of the machine-produced well dataset, and x and y represent the corresponding position coordinates in the machine-produced well dataset. Specifically, x represents the horizontal coordinate and y represents the vertical coordinate.
[0057] As one embodiment of the present invention, the method for obtaining reliable scaling labels in step two is as follows:
[0058] For a single-well sample sequence, the scale labeling decision rule formula (5) is as follows:
[0059]
[0060] Where i is the data index number of the valid scale identification sample data. This is the index number of the last data point that remained unscaled. This is the data index number for the first instance of scaling, i.e., the "first" scaling sample. The index number of the last data entry with a scaling count identifier. To fit the local mean of the pH value sequence of the produced liquid analysis, T is a threshold constant.
[0061] As one implementation method of this invention, step three, based on the positive causal logic analysis method, is as follows:
[0062] Considering that traditional scaling analysis of produced fluid from mechanized wells is a "post-hoc analysis" based on inverse causal logic, lacking the predictive power of "pre-analysis" and the timeliness of scaling status perception, and also incurring high testing costs, this paper proposes a novel "pre-analysis" strategy for timely scaling status identification, utilizing well information and production dynamics information with a positive causal relationship to scaling. Specifically, based on all well information and production dynamics information, and excluding factors severely affected by missing data, the Spearman rank correlation coefficient is first used to calculate the Spearman correlation coefficient between each candidate variable and the scaling degree label. This initially identifies features with a strong monotonic trend in their relationship with scaling. The calculation formula is as follows:
[0063]
[0064] Where, d i This represents the difference in rank between the samples corresponding to two influencing factors, where n represents the total number of data points in the machine-produced well dataset, and r represents the difference in rank between the samples corresponding to the two influencing factors. s This represents the Spearman correlation coefficient.
[0065] Then, feature selection methods such as random forest regression, logistic regression, and chi-square test are used to automatically select influencing factors. Finally, by combining expert experience and the causal logic of the scaling process, positive causal factors related to scaling are selected and integrated with the label data to form the original scaling prediction dataset. Figure 3 A flowchart based on the positive causal logic analysis method is provided.
[0066] As one embodiment of the present invention, the method for constructing the separable pattern vector regression model in step four is as follows:
[0067] Let X = {x1, x2, ..., x} n}, x i ∈R d×1 Given a high-dimensional scaling-related feature dataset consisting of n d-dimensional samples, Y = {y1, y2, ..., y3}, y i Let ∈{1,2,...,u} represent the label information for u categories, and let D = [X; L] ∈ R. n×(d+1)This is for a new high-dimensional dataset. Specifically, based on the Gaussian kernel function, the data points x in the high-dimensional space... j For x i The conditional probability of a neighboring point is:
[0068]
[0069] Where, σ i It is used to adjust the width of the kernel function relative to the data point x. i One relevant parameter, the value of which can be determined based on the perplexity related to the number of neighboring points of the data point, is set to 30 in the t-SNE dimensionality reduction process of this patent. The conditional probability given in equation (7) can be obtained by performing the following symmetric processing: i and x j The joint probability distribution, that is, the probability that the two are neighbors and appear together, is:
[0070]
[0071] Correspondingly, in low-dimensional space, the t-distribution is used to calculate the joint probability of the sum:
[0072]
[0073]
[0074] Where, f(y) i ,y j Let be the joint probability density function of the t-distribution. Due to the long-tailed nature of the t-distribution, it can better capture the global structure of data in low-dimensional space and make the data more clustered in that space. Therefore, gradient descent can be used, with the following low-dimensional data point update formula:
[0075]
[0076] in, y represents the previous iteration, the current iteration, and the next iteration, respectively. i The value of .
[0077] Finally, after the fluctuation of the objective function is less than a certain threshold or the maximum number of iterations is reached, the near-optimal low-dimensional space data mapping Y corresponding to the high-dimensional space dataset D can be obtained. * .
[0078] Constructing a separable variable regression model, once the low-dimensional space of separable variables Y is obtained... * Combining the scaling-related features X selected in step three, we use them as input to the random forest regression model, employing 200 decision trees to predict separable variables. The calculation formula is as follows:
[0079]
[0080] in, T represents the final predicted value of the separable variables in the random forest model, N represents the total number of decision trees in the random forest, and T represents the total number of decision trees in the random forest. i (S) represents the prediction result of the i-th decision tree in the random forest for the input data S. Figure 4 A schematic diagram illustrating the dimensionality reduction visualization of the original data and the regression performance is provided.
[0081] As one embodiment of the present invention, the method for constructing the SE-LSTM algorithm based on the CNN-LSTM model by introducing the SE module in step five is as follows:
[0082] The CNN+LSTM model is a deep learning model that combines a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) network. After CNN feature extraction and before LSTM processing, an SE module is inserted to construct the SE-LSTM algorithm. This SE module performs channel attention on the features extracted by the CNN. Figure 5 A schematic diagram of the SE-LSTM architecture is provided.
[0083] The LSTM structure is as follows:
[0084] The LSTM structure consists of three parts: an input gate, a forget gate, and an output gate. It processes data and retains the setting information using the following formula:
[0085] f t =sigmoid(W f h t-1 +W f x t +b f (14)
[0086] i t =sigmoid(W i h t-1 +W i x t +b i (15)
[0087] o t =sigmoid(W o h t-1 +W o x t +b o (16)
[0088]
[0089]
[0090] h t =o t ×tanh(C t (19)
[0091] Where t represents the current discrete time step, t-1 represents the time step t before t, and f t i t o t , C t h t x t Let W represent the activation vectors of the forget gate, input gate, and output gate, as well as the latent cell state, cell state, hidden state, and input activation vector, respectively, at a discrete time step t. f h is the forget gate weight matrix. t-1 Let b be the hidden state at the previous time step t. f W is the bias function for the forget gate, sigmoid is the logistic function, and W is the bias function for the forget gate. i Let b be the input gate weight matrix. i W is the input gate bias function. o Let b be the output gate weight matrix. o W is the output gate bias function. C Update the weight matrix for the state. For the state update bias function, C t-1 Let t represent the cell state at the previous time step, and tanh be the hyperbolic tangent function.
[0092] The Adam optimizer is used to optimize the parameters, which include the weight matrix of the CNN convolutional kernel and the forget gate weight matrix W in the LSTM. f Input gate weight matrix, output gate weight matrix, and state update matrix.
[0093] As one embodiment of the present invention, the method for evaluating the trained model using set evaluation metrics in step six is as follows:
[0094] Combining the separable variables obtained in step four and the scaling-related features from step three, the dataset is divided into training and testing sets. During training, multi-class cross-entropy is used as the loss function, calculated as follows:
[0095]
[0096] Where C represents the total number of categories, y i p represents the true label of the sample. i This represents the probability that the model predicts the sample belongs to the i-th class. Figure 6A diagram illustrating the loss function during the optimization process using the CNN-LSTM model is provided.
[0097] To ensure the model's generalization ability on unknown wells, multiple metrics, including accuracy, precision, recall, and F1 score, are used to comprehensively evaluate the trained model using a test set. The calculation formulas are as follows:
[0098]
[0099] Among them, TP indicates that both the predicted result and the actual result are positive, FN indicates that the actual result is positive but the predicted result is negative, FP indicates that the actual result is negative but the predicted result is positive, and TN indicates that both the actual result and the predicted result are negative.
[0100] As one embodiment of the present invention, the method for visualizing scaling information in step seven is as follows:
[0101]
[0102] Among them, X unknown This represents the scaling characteristic data of unknown mechanically operated wells; LSTM_CNN represents the scaling prediction model obtained in step S5; and TSNE represents the dimensionality reduction function. Z represents the predicted scaling state of an unknown oil well, where Z represents the dimension-reduced features. Figure 7 A schematic diagram is provided to visualize the scaling information of unknown wells.
[0103] This invention combines causal reasoning with visualization and dimensionality reduction techniques to accurately identify scaling characteristics in mechanically operated wells, avoiding the experience-based errors inherent in traditional methods. Visualizing the results allows operators to intuitively understand the scaling situation, enabling rapid response and decision-making, and improving work efficiency. This method not only adapts to complex oil well environments but also possesses strong universality, allowing for continuous optimization based on data. Furthermore, by predicting scaling conditions in advance, this invention can effectively reduce equipment failures and repair costs, improving the economic benefits of oil wells.
[0104] To verify the effectiveness of the method of this invention, production data from five mechanized wells in the Qijia-Gulong Depression in the western part of Daqing Changyuan in the northern Songliao Basin were selected as the research object, and the proposed method for predicting scaling in mechanized wells was verified through practical application. Experimental results show that the method of this invention, while integrating the structural information of mechanized well production data, achieves high accuracy in scaling prediction and effective identification of scaling trends, demonstrating significant practical value.
[0105] In selecting production parameters, four main production indicators were chosen for analysis: daily liquid production, water content, motor power, and dynamic liquid level depth. However, the experimental results also reflect that the method of this invention still needs improvement. Since the weight coefficients of the separable pattern vector are closely related to the t-SNE dimensionality reduction parameters, their current values are not optimal, which may lead to some misjudgments in the final scaling prediction results. Therefore, in subsequent studies, the embodiments of this invention will further optimize the calculation method of the weight coefficients to reduce the overall prediction error of the model, thereby further improving the accuracy of scaling prediction.
[0106] This invention aims to address the common problem in classic mechanically operated well scaling prediction studies that insufficiently consider reservoir heterogeneity and the temporal correlation of production data. This invention proposes a scaling prediction method based on multi-view information integration. Through scaling correlation analysis experiments on different production parameters of different oil wells, this invention found that the correlation of most production parameters is not particularly significant. Therefore, this invention mainly uses positive causal logic analysis to mine the correlation features in the production data. After feature screening of the original data, this invention obtains a feature set that combines production data structure information and has a strong correlation with the degree of scaling. Considering the large dimensionality differences of these features, a nonlinear dimensionality reduction model is constructed. To fully utilize the rich high-dimensional contextual information in the production data at specific time points, this invention adopts a multi-view, multi-scale nonlinear mapping approach, directly mapping features at each scale to the target scaling prediction result. These multi-scale features are fused and optimized using a deep convolutional neural network to finally obtain the scaling prediction result for mechanically operated wells.
[0107] Example 2:
[0108] This invention also provides an intelligent identification system for scaling status in mechanically operated wells, comprising:
[0109] The first processing module is used to filter out outliers and noise information in the machine-produced well dataset;
[0110] The second processing module is used to obtain reliable scaling labels based on the filtered machine-produced well dataset;
[0111] The third processing module is used to filter out influencing factors that are highly correlated with scaling in mechanically operated wells from the raw data and determine scaling association characteristics; the raw dataset includes mechanically operated well datasets and reliable scaling labels;
[0112] The fourth processing module is used to reduce the dimensionality of the original data and visualize it using the t-SNE algorithm, and to construct a separable pattern vector regression model.
[0113] The fifth processing module is used to construct a CNN-LSTM neural network model, introduce the SE attention mechanism to build the SE-LSTM algorithm, and use the Adam optimizer to optimize the model parameters.
[0114] The sixth processing module is used to input the separable variables and scaling-related features into a CNN-LSTM neural network model for training, using multi-class cross-entropy as the loss function for training, and outputting the scaling type.
[0115] The seventh processing module is used to predict the scaling data of unknown mechanically operated wells using a trained CNN-LSTM neural network model and generate a visualization of the scaling status.
[0116] As one embodiment of the present invention, the first processing module uses an outlier detection method based on standard deviation and a Gaussian filtering method to filter out outliers and noise information in the machine-produced well data set.
[0117] As one embodiment of the present invention, the third processing module uses a positive causal logic analysis method to filter out influencing factors that are highly correlated with scaling in machine-produced wells from the original data and determine the scaling correlation characteristics.
[0118] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.
Claims
1. A method for intelligent identification of scaling status in mechanically operated wells, characterized in that, include: Step 1: Filter out outliers and noise from the machine-produced well dataset; Step 2: Obtain reliable scaling labels based on the filtered machine-produced well dataset; Step 3: Filter out the influencing factors that are highly correlated with scaling in machine-produced wells from the raw data and determine the scaling association characteristics; the raw dataset includes machine-produced well datasets and reliable scaling labels; Step 4: Use the t-SNE algorithm to reduce the dimensionality of the original data and visualize it, and construct a separable pattern vector regression model; Step 5: Construct a CNN-LSTM neural network model, introduce the SE attention mechanism to build the SE-LSTM algorithm, and use the Adam optimizer to optimize the model parameters; Step 6: Based on the separable variables and scaling-related features, input them into a CNN-LSTM neural network model for training, use multi-class cross-entropy as the loss function for training, and output the scaling type; Step 7: Using the trained CNN-LSTM neural network model, predict the scaling data of unknown machine-operated wells and generate a visualization of the scaling status.
2. The intelligent identification method for scaling status in machine-produced wells as described in claim 1, characterized in that, In step one, outlier detection based on standard deviation and Gaussian filtering are used to filter out outliers and noise in the machine-produced well dataset.
3. The intelligent identification method for scaling status in machine-produced wells as described in claim 2, characterized in that, In step three, based on the positive causal logic analysis method, factors highly correlated with scaling in machine-operated wells are screened from the original data to determine the scaling correlation characteristics.
4. An intelligent identification system for scaling status in mechanically operated wells, characterized in that, include: The first processing module is used to filter out outliers and noise information in the machine-produced well dataset; The second processing module is used to obtain reliable scaling labels based on the filtered machine-produced well dataset; The third processing module is used to filter out influencing factors that are highly correlated with scaling in mechanically operated wells from the raw data and determine scaling association characteristics; the raw dataset includes mechanically operated well datasets and reliable scaling labels; The fourth processing module is used to reduce the dimensionality of the original data and visualize it using the t-SNE algorithm, and to construct a separable pattern vector regression model. The fifth processing module is used to construct a CNN-LSTM neural network model, introduce the SE attention mechanism to build the SE-LSTM algorithm, and use the Adam optimizer to optimize the model parameters. The sixth processing module is used to input the separable variables and scaling-related features into a CNN-LSTM neural network model for training, using multi-class cross-entropy as the loss function for training, and outputting the scaling type. The seventh processing module is used to predict the scaling data of unknown mechanically operated wells using a trained CNN-LSTM neural network model and generate a visualization of the scaling status.
5. The intelligent identification system for scaling status in machine-produced wells as described in claim 4, characterized in that, The first processing module uses a standard deviation-based outlier detection method and a Gaussian filtering method to filter out outliers and noise in the machine-produced well dataset.
6. The intelligent identification system for scaling status in machine-produced wells as described in claim 5, characterized in that, The third processing module uses positive causal logic analysis to filter out influencing factors that are highly correlated with scaling in machine-operated wells from the raw data and determine the scaling correlation characteristics.