Oilfield safety environmental protection abnormal information dynamic monitoring method and device and related products

Abstract

The application provides an oilfield safety and environmental protection abnormal information dynamic monitoring method and device and related products, wherein the oilfield safety and environmental protection abnormal information dynamic monitoring method comprises the following steps: acquiring information data of oilfield safety and environmental protection, preprocessing the information data of oilfield safety and environmental protection to obtain preprocessing data, extracting features from the preprocessing data to obtain feature data, identifying abnormalities of the feature data based on a discriminant model to obtain problem data, and constructing a dynamic monitoring model based on the problem data to analyze to-be-detected data and obtain a monitoring result. The dynamic monitoring model can be used for real-time identification and positioning of information data of oilfield safety and environmental protection, ensuring instant understanding and rapid response to the state of the oilfield, improving identification accuracy, and solving the lag problem of traditional monitoring methods.

Description

Technical Field

[0001] This application belongs to the field of oilfield safety and environmental protection, specifically involving a method, device and related products for dynamic monitoring of abnormal information in oilfield safety and environmental protection. Background Technology

[0002] Oilfield development is a crucial pillar of global energy supply. With continuously growing energy demand, oilfield exploration and development technologies are constantly advancing to extract oil resources in more efficient and environmentally friendly ways. Oilfield development not only relates to national energy security and economic development but also places higher demands on environmental protection and sustainable development. With technological advancements, oilfield development is moving towards intelligent and green development to address the dual challenges of global energy demand and environmental protection.

[0003] The field of oilfield monitoring encompasses a comprehensive technological system, ranging from basic data acquisition to advanced data analysis and processing, and from real-time monitoring to remote monitoring. Firstly, advanced monitoring sensors and data transmission technologies enable the transmission and storage of real-time monitoring data on oil wells, reservoirs, and the environment. Secondly, big data analytics and artificial intelligence technologies are used to rapidly process and analyze monitoring data, extracting useful information to support decision-making. Furthermore, oilfield monitoring also involves regular inspections and remote monitoring to ensure the stability and safety of various oilfield indicators.

[0004] However, while some monitoring systems exist in the field of oilfield environmental safety standards, they typically employ static monitoring methods and are unable to identify and locate abnormal information related to environmental safety standards in real time. These systems often suffer from problems such as low identification accuracy and delayed monitoring response. Summary of the Invention

[0005] The purpose of this application is to provide a method, device, and related products for dynamic monitoring of abnormal information in oilfield safety and environmental protection, so as to solve or alleviate the problems existing in the prior art.

[0006] To achieve the above objectives, this application provides the following technical solution: a method for dynamic monitoring of abnormal information on oilfield safety and environmental protection, comprising: acquiring information data on oilfield safety and environmental protection; preprocessing the information data to obtain preprocessed data; extracting features from the preprocessed data to obtain feature data; identifying anomalies in the feature data based on a discriminant model to obtain problem data; and constructing a dynamic monitoring model based on the problem data to analyze the data to be detected and obtain monitoring results.

[0007] Optionally, the preprocessing of the oilfield safety and environmental protection information data to obtain preprocessed data includes: removing duplicate data, deleting missing and outlier values, and performing data transformation and data reduction operations on the oilfield safety and environmental protection information data to obtain preprocessed data.

[0008] Optionally, the step of extracting features from the preprocessed data to obtain feature data includes: extracting features from the preprocessed data based on the random forest algorithm to obtain feature data.

[0009] Optionally, the step of extracting features from the preprocessed data based on the random forest algorithm to obtain feature data includes: initializing the oilfield safety and environmental protection information data and setting the number of decision trees in the random forest algorithm; performing data augmentation on the oilfield safety and environmental protection information data based on a generative adversarial network to obtain a dataset; sampling the dataset based on a bagging algorithm to form a training dataset; and constructing the decision tree based on the training dataset to obtain feature data.

[0010] Optionally, anomaly identification is performed on the feature data based on a discriminant model to obtain problematic data, including: calculating the feature standard deviation and feature probability density based on the discriminant model to identify anomalies in the feature data to obtain problematic data.

[0011] Optionally, the standard deviation of the features is calculated based on the discriminant model defined by the following formula: The standard deviation of the c-th feature is The preprocessed data for the m-th and c-th features is The amount of preprocessed data is q. The feature probability density is calculated based on a discriminant model defined by the following formula: The probability density of the preprocessed data r is h(r), and the exponential function with base e is given, with an anomaly threshold exp(·). When the probability density is less than the anomaly threshold, the feature data is problematic data.

[0012] Optionally, the step of constructing a dynamic monitoring model based on the problem data to analyze the data to be detected and obtain monitoring results includes: constructing a dynamic monitoring model based on the problem data and optimizing the dynamic monitoring model to obtain an optimized dynamic monitoring model; and analyzing the data to be detected based on the optimized dynamic monitoring model to obtain the monitoring results.

[0013] This application also provides a dynamic monitoring device for abnormal information on oilfield safety and environmental protection, comprising: a data acquisition module for acquiring information data on oilfield safety and environmental protection, wherein the information data is preprocessed to obtain preprocessed data; a data processing module for extracting features from the preprocessed data to obtain feature data; and anomaly identification of the feature data based on a discriminant model to obtain problem data; a model building module for building a dynamic monitoring model based on the problem data; and a monitoring module for analyzing the data to be detected based on the dynamic monitoring model to obtain monitoring results.

[0014] This application also provides an electronic device, including: a memory and a processor, wherein the memory stores a computer-executable program, and the processor is used to execute the computer-executable program to implement the dynamic monitoring method for abnormal information on oilfield safety and environmental protection as described in any one of the claims.

[0015] This application also provides a computer storage medium storing a computer-executable program, which, when run, implements any of the oilfield safety and environmental protection anomaly monitoring methods described above.

[0016] This invention provides a method, device, and related products for dynamic monitoring of abnormal information related to oilfield safety and environmental protection. The method includes: acquiring oilfield safety and environmental protection information data; preprocessing the information data to obtain preprocessed data; extracting features from the preprocessed data to obtain feature data; identifying anomalies in the feature data based on a discriminant model to obtain problematic data; and constructing a dynamic monitoring model based on the problematic data to analyze the data to be monitored and obtain monitoring results. By establishing a dynamic monitoring model, oilfield safety and environmental protection information data can be identified and located in real time, ensuring immediate understanding and rapid response to the oilfield status, improving identification accuracy, and solving the lag problem of traditional monitoring methods. Furthermore, through preprocessing and feature extraction, this method improves the accuracy and effectiveness of the data, providing a more reliable foundation for subsequent anomaly identification. In addition, anomaly identification based on the discriminant model can accurately identify potential problematic data, providing strong support for the prevention and control of safety risks. Finally, the dynamic monitoring model built on the problem data can analyze the data to be detected in real time, provide real-time monitoring results, help oilfield managers to discover and solve problems in a timely manner, and ensure the safe production and environmental protection of the oilfield. Attached Figure Description

[0017] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. Wherein:

[0018] Figure 1 This is a flowchart illustrating a method for dynamic monitoring of abnormal information related to oilfield safety and environmental protection, as described in an embodiment of this application.

[0019] Figure 2 This is a schematic diagram of an electronic device for dynamic monitoring of abnormal information in oilfield safety and environmental protection, as described in an embodiment of this application. Detailed Implementation

[0020] The present application will now be described in detail with reference to the accompanying drawings and embodiments. Various examples are provided by way of explanation and not by way of limitation. In fact, those skilled in the art will recognize that modifications and variations can be made to the present application without departing from the scope or spirit thereof. For example, a feature shown or described as part of one embodiment may be used in another embodiment to produce yet another embodiment. Therefore, it is desirable that the present application encompass such modifications and variations that fall within the scope of the appended claims and their equivalents.

[0021] In the description of this application, the terms "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," and "bottom," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this application and do not require that this application be constructed and operated in a specific orientation, and therefore should not be construed as limiting this application. The terms "connected," "linked," and "set up" used in this application should be interpreted broadly. For example, they can refer to a fixed connection or a detachable connection; they can refer to a direct connection or an indirect connection through intermediate components. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0022] Example 1

[0023] Figure 1 This is a flowchart illustrating a method for dynamically monitoring abnormal information related to oilfield safety and environmental protection, as described in an embodiment of this application. Figure 1The present invention discloses a dynamic monitoring method for abnormal information on oilfield safety and environmental protection, comprising: acquiring oilfield safety and environmental protection information data; preprocessing the oilfield safety and environmental protection information data to obtain preprocessed data; extracting features from the preprocessed data to obtain feature data; identifying anomalies in the feature data based on a discriminant model to obtain problem data; and constructing a dynamic monitoring model based on the problem data to analyze the data to be detected and obtain monitoring results. By establishing a dynamic monitoring model, oilfield safety and environmental protection information data can be identified and located in real time, ensuring immediate understanding and rapid response to the oilfield status, improving identification accuracy, and solving the lag problem of traditional monitoring methods. Furthermore, through preprocessing and feature extraction, this method improves the accuracy and effectiveness of the data, providing a more reliable foundation for subsequent anomaly identification. In addition, anomaly identification based on the discriminant model can accurately identify potential problem data, providing strong support for the prevention and control of safety risks. Finally, the dynamic monitoring model constructed based on the problem data can analyze the data to be detected in real time, providing real-time monitoring results, helping oilfield managers to promptly identify and solve problems, and ensuring safe production and environmental protection in the oilfield.

[0024] Optionally, the information data on oilfield safety and environmental protection includes, but is not limited to, smoke and dust emission concentration values, sulfur dioxide emission concentration values, oxide emission concentration values, chemical oxygen demand emission concentration values, ammonia nitrogen emission concentration values, phosphorus emission concentration values, total nitrogen emission concentration values, noise decibel values, and heavy metal content values ​​in the soil.

[0025] Optionally, the oilfield safety and environmental protection information data includes: standard information and historical dynamic data; the standard information refers to the standard values ​​set for evaluating oilfield safety and environmental protection; the historical dynamic data refers to real-time data generated during oilfield operation. In this embodiment, combining standard information and historical dynamic data provides a comprehensive perspective for assessing the oilfield's safety and environmental protection status. Standard information provides preset compliance requirements, while historical dynamic data reflects real-time changes during actual operation. This combination makes monitoring more comprehensive, avoiding the bias that may result from relying on a single data source. Furthermore, historical dynamic data reflects the real-time status during oilfield operation, enabling the monitoring system to promptly capture any abnormal changes, improving the timeliness of anomaly identification, and facilitating rapid response and handling of potential safety and environmental protection issues. In addition, by comparing standard information and historical dynamic data, it is possible to clearly identify which indicators deviate from the established standards, thereby enabling targeted problem investigation and rectification, and improving the accuracy of problem location. Moreover, by combining standard information and historical dynamic data, the monitoring system can generate more accurate analysis reports, providing management with decision-making basis and helping to formulate more effective safety and environmental protection strategies and measures to ensure the sustainable and safe operation of the oilfield.

[0026] Optionally, the historical dynamic data can be preprocessed to obtain preprocessed data. In this embodiment, preprocessing can reduce unnecessary data, decrease data complexity, reduce the use of computing resources, and improve processing speed, making real-time dynamic monitoring possible.

[0027] Optionally, the preprocessing of the oilfield safety and environmental protection information data to obtain preprocessed data includes: removing duplicate data, deleting missing and outliers, data transformation, and data reduction operations to obtain preprocessed data. In this embodiment, removing duplicate data ensures the independence and accuracy of the analysis, avoiding misleading results caused by duplicate data. Deleting missing and outliers reduces data noise, improving the training effect and prediction accuracy of the model. Furthermore, data transformation ensures that all data are on the same scale, facilitating comparison and analysis, and avoiding confusion between data of different units or scales. For example, converting all numerical data to the same unit, or encoding categorical data. In addition, data reduction operations, such as principal component analysis (PCA) or feature selection, can reduce the dimensionality of the data, reduce computational complexity, and improve processing speed, while retaining most of the information in the data, which is especially important for processing large datasets. Moreover, preprocessed data is more conducive to the training of machine learning models, improving the model's generalization ability, reducing the risk of overfitting, making the model's performance on unknown data more stable, and thus more effectively identifying abnormal information.

[0028] Optionally, the feature extraction of the preprocessed data to obtain feature data includes: performing feature extraction on the preprocessed data based on the random forest algorithm to obtain feature data. In this embodiment, the random forest algorithm improves prediction accuracy and reduces the overfitting problem that may exist in a single decision tree by constructing multiple decision trees and integrating their prediction results. Furthermore, the random forest algorithm can evaluate the importance of each feature to the prediction results during training. This helps identify which features contribute most to the model's predictions, thus providing a powerful tool for feature selection and data understanding. Additionally, since the random forest is an ensemble model composed of multiple decision trees, it is highly robust to noise and outliers in the data. Even if some data has problems, the performance of the entire model will not be significantly affected. Finally, the random forest algorithm is suitable for classification and regression problems and can handle various types of datasets, including complex data in the field of oilfield safety and environmental protection. It does not require assumptions about the distribution of the data, thus it can be flexibly applied to different practical problems.

[0029] Optionally, the step of extracting features from the preprocessed data using the random forest algorithm to obtain feature data includes: initializing the oilfield safety and environmental protection information data; setting the number of decision trees in the random forest algorithm; performing data augmentation on the oilfield safety and environmental protection information data using a generative adversarial network (GAN) to obtain a dataset; sampling the dataset using a bagging algorithm to form a training dataset; and constructing the decision trees based on the training dataset to obtain feature data. In this embodiment, using a GAN to augment the oilfield safety and environmental protection information data can generate more training samples, which helps improve the model's generalization ability, especially when the amount of data is limited, and can significantly improve the model's performance. Furthermore, by setting the number of decision trees in the random forest algorithm, the diversity of the model can be increased. Each decision tree uses a different subset of data during training, which helps reduce the model's variance and improve the overall prediction stability. In addition, the bagging algorithm, by sampling the dataset to form multiple training datasets and then constructing multiple decision trees based on these datasets, can effectively reduce the overfitting problem that may occur with a single decision tree and improve the model's prediction ability on unknown data. Finally, during the construction of the decision tree, the random forest algorithm can evaluate the contribution of each feature to the prediction results, thereby performing feature selection. This not only removes unimportant features but also helps identify the most critical features for oilfield safety and environmental monitoring, simplifying the model and improving interpretability.

[0030] Optionally, the data augmentation of the oilfield safety and environmental protection information data based on generative adversarial networks (GANs) to obtain a dataset includes: augmenting the oilfield safety and environmental protection information data based on GANs to obtain a first dataset; and fitting outliers to the first dataset using a nonlocal mean filtering algorithm to obtain the final dataset. In this embodiment, generative adversarial networks are powerful generative models capable of generating high-quality, realistic data samples. In the field of oilfield safety and environmental protection, this means that data close to real-world conditions can be created, thereby expanding the training set and improving the model's generalization ability. Furthermore, the data generated by GANs not only increases the size of the dataset but also enhances its diversity. This is crucial for oilfield safety and environmental monitoring because it helps the model learn a wider range of environmental conditions and potential anomalies. Additionally, the nonlocal mean filtering algorithm is an effective denoising and feature-preserving algorithm that can process the first dataset generated by the GAN to identify and fit outliers. This helps reduce the negative impact of outliers on model performance while preserving important features in the data. Moreover, through data augmentation and outlier processing, the resulting dataset is more complete and accurate, which helps in building a more robust monitoring model. Robustness is particularly important for oilfield safety and environmental monitoring because it ensures that the model can still work stably when faced with complex and ever-changing real-world environments.

[0031] Optionally, constructing the decision tree based on the training dataset to obtain feature data includes: calculating the decision tree accuracy; sorting the decision tree accuracy in descending order, and successively removing the last two decision trees to obtain optimized decision tree accuracy; training the decision tree based on the optimized decision tree accuracy to obtain feature data. In this embodiment, by calculating and evaluating the accuracy of the decision trees, it can be ensured that the selected decision trees have high accuracy on the prediction task. This method helps to identify and retain the most effective models, thereby improving the overall quality of feature extraction. Moreover, by sorting in descending order and successively removing decision trees with lower accuracy, the set of decision trees can be optimized. This method is similar to boosting techniques, which can reduce noise and overfitting in the model set while enhancing the model's generalization ability. In addition, during the optimization of the decision trees, the contribution of each feature to improving the accuracy of the decision trees can be evaluated. This helps to identify key features and give these features higher weights during feature extraction. Finally, by removing some inefficient decision trees, the complexity of the model can be reduced. This not only improves the interpretability of the model but also reduces the demand for computing resources, making the model more efficient and practical.

[0032] Optionally, the step of sorting the decision trees in descending order of accuracy and successively removing the last two decision trees to obtain optimized decision tree accuracy includes: stopping the step of sorting the decision trees in descending order of accuracy and successively removing the last two decision trees to obtain optimized decision tree accuracy when the optimized decision tree accuracy is lower than a set accuracy value. In this embodiment, by setting an accuracy threshold and stopping the optimization process when the optimized decision tree accuracy is lower than this threshold, the accuracy of the model can be dynamically controlled. This method avoids over-optimization and ensures that the model maintains high accuracy without wasting resources on inefficient decision trees. Moreover, successively removing the last two decision trees with lower accuracy helps optimize resource allocation. This method ensures that computational resources are concentrated on decision trees that are more critical to improving the overall model performance. In addition, setting a stopping condition helps prevent model overfitting. When the decision tree accuracy is already high enough, further optimization may not bring significant performance improvement, but may instead lead to the model overfitting the training data. Furthermore, removing inefficient decision trees can improve the stability of the model. The optimized decision tree ensemble is more concise, reducing the model's dependence on specific training samples and thus improving its predictive stability on new data. Finally, the reduced number of decision trees in the optimized ensemble decreases the computational load during model training and prediction, thereby improving overall computational efficiency. Fewer decision trees also make the model structure simpler, contributing to improved interpretability and making it easier to analyze and understand the model's decision-making process.

[0033] Optionally, training the decision tree based on the optimized decision tree accuracy to obtain feature data includes: determining the classification accuracy and weights of the decision tree based on the optimized decision tree accuracy; calculating the feature importance metric in the decision tree based on the classification accuracy and the decision tree weights; calculating the feature importance metric based on the feature importance metric and the decision tree weights; and extracting features from the preprocessed data based on the feature importance metric to obtain feature data. In this embodiment, by calculating the classification accuracy and weights of the optimized decision tree, the contribution of each feature to the classification result can be evaluated more accurately. This helps to identify the most critical features for model prediction, thereby enabling effective feature selection. Furthermore, by extracting features based on the feature importance metric, features that contribute little to model prediction can be removed, reducing model complexity, avoiding overfitting, and enhancing the model's generalization ability on unseen data. In addition, the feature importance metric provides a method to quantify the impact of different features on the model's predictive ability. This not only helps to understand the model's decision-making process but also improves the model's interpretability, making the results easier for domain experts to understand and trust. Furthermore, considering the weights of decision trees during feature extraction ensures that features significantly impacting multiple decision trees are retained, while those important only in a few trees may be excluded, thus optimizing the efficiency of the decision-making process. This method allows for dynamic adjustment of feature weights, adjusting their importance in the overall model based on their performance in different decision trees, enabling greater flexibility to adapt to data variations and different scenario requirements. Finally, by comprehensively considering the classification accuracy and weights of multiple decision trees, this method can more comprehensively evaluate feature importance, contributing to the construction of a higher-performing predictive model.

[0034] Optionally, the accuracy of the decision tree can be calculated based on the following formula:

[0035] The number of training datasets is H, and the number of samples in the dataset that are classified as u by the i-th decision tree is m. i，uThe formula denoted by u, where u represents the class that matches the true classification, and q represents the number of classes that match the true classification. In this embodiment, the formula provides a clear and quantitative way to measure the prediction accuracy of a decision tree. By comparing the classification results of the decision tree with the true classification of the samples, the number of correctly classified samples for each decision tree can be accurately calculated. Furthermore, the calculation method is logically clear, easy to understand, and easy to implement. Based on the intuitive concept of correct or incorrect classification, it allows even non-experts to understand the process of evaluating decision tree accuracy. In addition, the unified accuracy calculation formula facilitates the comparison of the performance of different decision trees or different models. This provides a fair benchmark for model selection and optimization. Moreover, the calculated decision tree accuracy can be used to guide model optimization. For example, by identifying decision trees with low accuracy, the reasons can be further analyzed and measures can be taken for improvement, such as adjusting model parameters or reselecting features. Furthermore, this method emphasizes the consistency between the decision tree classification results and the true labels of the samples, which helps improve the predictive reliability of the model. This is particularly important in the dynamic monitoring of abnormal information in oilfield safety and environmental protection, where accurate prediction is crucial for preventing potential risks. As part of model evaluation, this accuracy calculation method supports the assessment of the overall model performance, helping to determine whether the model meets specific performance requirements or needs further training and adjustments. Finally, high-precision decision trees can enhance decision-makers' confidence in the model's predictions, promoting data-driven decision-making, especially in the field of oilfield safety and environmental protection, where accurate monitoring results are crucial for ensuring operational safety and environmental protection.

[0036] Optionally, the weights of the decision tree can be calculated based on the following formula:

[0037] The weight of the i-th decision tree is τ. i The number of datasets for which the i-th decision tree classifies the training dataset the same as the datasets for which the random forest decision tree classifies the dataset is m. i,kThe adjustment coefficient is υ, and the number of trees with the same classification is a. In this embodiment, by assigning weights to each decision tree, their influence in the random forest can be adjusted based on their consistency in classifying the training dataset. This helps enhance the overall predictive performance of the model. Furthermore, the calculation of weights allows the model to dynamically adjust the contribution of each decision tree to the final prediction result. This dynamic adjustment helps balance the predictive power of different decision trees, thereby improving the stability and accuracy of the overall model. In addition, since the weights consider the consistency between decision trees, those trees that are consistent with the majority of decision trees in classification will receive higher weights. This helps reduce the negative impact of outlier or misclassified decision trees on the overall model performance, improving the model's robustness. Moreover, during model training and prediction, computational resources can be allocated according to the weights of the decision trees, prioritizing those with higher weights. This not only improves computational efficiency but also ensures that key decision trees are fully considered. The introduction of weights provides an additional interpretive dimension to the model. By analyzing the weights of different decision trees, we can better understand the model's decision-making process and which features and patterns have a greater impact on the model's prediction results. When constructing a random forest, the best-performing trees can be selected for ensemble based on their weights, or trees with lower weights and smaller contributions can be removed during model optimization. Finally, by adjusting the coefficients, the model can adapt to different data distributions and classification problems. This adaptability allows the model to better handle imbalanced datasets or datasets with complex patterns.

[0038] Optionally, the feature importance metric in the decision tree can be calculated based on the following formula:

[0039] The importance metric of the d-th feature in the i-th decision tree is and the average classification accuracy after adding Gaussian noise to the d-th feature in the dataset is . The number of features with added Gaussian noise is e. In this embodiment, the method quantifies the impact of each feature on the model's predictive ability by comparing the changes in classification accuracy before and after adding noise, thereby identifying the features most critical to model performance. Furthermore, introducing Gaussian noise into the features tests the model's sensitivity to data perturbations, which helps assess the model's robustness, especially in applications like oilfield safety and environmental monitoring where accuracy and stability are paramount. Additionally, identifying and removing unimportant features reduces model complexity, lowers the risk of overfitting, and improves the model's generalization ability on new data. Moreover, based on feature importance metrics, effective feature selection can be performed, retaining features that contribute most to model performance while removing those with little impact on prediction results, thus optimizing the model structure. Clearly defining the contribution of features to the model's predictive ability improves interpretability, enabling domain experts and decision-makers to understand the model's working principles and key influencing factors. Finally, this method allows the model to adaptively adjust based on the different importance of data features, enabling the model to better adapt to various complex environments and situations that may arise in oilfield safety and environmental monitoring. By identifying and removing unimportant features, we can simplify the model structure, reduce the computational burden during model training and prediction, and improve the efficiency of model operation.

[0040] Optionally, the feature importance metric is calculated based on the following formula:

[0041] The number of decision trees is b, and the importance metric of the d-th feature is W. d In this embodiment, by considering the importance assessment of each feature by all decision trees, this method provides a comprehensive measure of feature importance. This helps to balance the potential biases of individual decision trees, resulting in a more comprehensive and reliable feature importance index. Furthermore, this method allows for the overall identification of the features most critical to the performance of the random forest model, thereby making more rational decisions during feature selection and improving the model's predictive accuracy. Additionally, by identifying and emphasizing features that show importance in most decision trees, this method helps to build a model that performs well under various data conditions, enhancing the model's generalization ability. Moreover, by focusing on features considered important by most decision trees, this method simplifies the model's interpretation process, making the model's behavior and decisions easier to understand and interpret.

[0042] Optionally, the step of extracting features from the preprocessed data based on the feature importance metric to obtain feature data includes: using a backward search method to remove features with the lowest feature importance metric value one by one, and outputting the remaining features to obtain the feature data. In this embodiment, the backward search method ensures that the model retains only the features that have the greatest impact on the prediction results by removing the lowest-ranking features one by one based on the feature importance metric value, thereby improving the accuracy of feature selection. Moreover, removing unimportant features helps reduce the risk of overfitting while maintaining or improving the model's prediction accuracy, because the model can focus more on those features that substantially contribute to the results. In addition, reducing the number of features directly reduces the computational complexity of model training and prediction, which is especially important when dealing with large-scale datasets, and can significantly save computational resources and time costs. Finally, when the number of features is reduced, the contribution of each feature to the model output becomes more obvious, making the model's decision-making process more transparent and easier for experts and users to understand and interpret the model's behavior.

[0043] Optionally, anomaly identification is performed on the feature data based on a discriminant model to obtain problematic data. This includes calculating the feature standard deviation and feature probability density based on the discriminant model to identify anomalies in the feature data and obtain problematic data. In this embodiment, feature standard deviation and probability density are fundamental tools in statistics for describing data distribution. Using these statistics for anomaly identification ensures a solid theoretical foundation and reliable results. Furthermore, by combining the feature standard deviation and probability density, abnormal patterns in the data can be captured more comprehensively. Standard deviation reflects the volatility of the data, while probability density provides information about the shape of the data distribution; combining the two helps to accurately identify data points that do not conform to normal patterns. In addition, this method does not rely on specific data distribution assumptions, thus possessing good adaptability. It can handle various types of data distributions, including those that do not conform to a normal distribution, which is crucial in practical applications because real-world data often has complex and variable distribution characteristics. Finally, anomaly identification based on discriminant models can be quickly applied to real-time data streams, enabling the system to detect and respond to anomalies in oilfield safety and environmental protection in a timely manner, which is essential for accident prevention and timely response measures.

[0044] Optionally, the standard deviation of the features is calculated based on the discriminant model defined by the following formula:

[0045]

[0046] The standard deviation of the c-th feature is The preprocessed data for the m-th and c-th features is The amount of preprocessed data is q. In this embodiment, by calculating the standard deviation of the features, the fluctuation of the feature data can be quantitatively analyzed, providing an important numerical basis for subsequent anomaly detection and data analysis. Furthermore, the standard deviation is an effective tool for identifying outliers in a dataset. A larger standard deviation indicates that the data points are more dispersed, potentially containing outliers, which helps in the timely detection of potential problems in oilfield safety and environmental monitoring. Conversely, a smaller standard deviation indicates that the data points are closely distributed around the mean, suggesting a relatively stable dataset and enhancing confidence in the normality of the oilfield's safety and environmental status. Finally, the magnitude of the standard deviation can serve as a measure of feature importance, as features with larger standard deviations may contain more useful information, which helps in retaining features more critical to the prediction task during feature selection.

[0047] Optionally, the feature probability density is calculated based on a discriminant model defined by the following formula:

[0048]

[0049] The probability density of the preprocessed data r is h(r), and the exponential function with base e is given, with an anomaly threshold exp(·). In this embodiment, by calculating the probability density of features, the distribution characteristics of the data can be better understood. This helps to identify patterns and anomalies in the data, as outliers often correspond to low probability density regions. Furthermore, the probability density function can be used to determine the degree of anomaly of data points. By setting an anomaly threshold, anomalous data points with probability densities below the threshold can be effectively identified, thereby enabling timely responses to potential safety and environmental issues. Additionally, using an exponential function with base e to calculate the probability density provides a flexible method to adapt to different data distributions. This method does not rely on the data conforming to a specific distribution pattern, enhancing the model's adaptability. Moreover, the probability density function provides a mathematical basis for quantitative risk assessment. By analyzing the probability density of different data points, their potential risks to the safety and environmental status of the oilfield can be assessed, providing support for decision-making. This method supports statistically based decision-making. The calculation results of the probability density can be used to support or reject assumptions about the normality of the data, thereby helping to formulate data-based decisions. Furthermore, the calculation of feature probability density can guide model optimization. When building predictive models, probability density information can be used to adjust model parameters, improving the model's ability and accuracy in identifying anomalies. Finally, the calculated probability density can guide data preprocessing strategies, such as improving dataset quality by identifying and processing data points in low probability density regions.

[0050] Optionally, when the probability density is less than the anomaly threshold, the feature data is considered problematic data. In this embodiment, setting an anomaly threshold provides a clear standard for determining which data points are abnormal or problematic. This clear definition helps to quickly identify potential problem areas and accelerate response time. This method allows the system to automatically and efficiently filter out abnormal data without complex analysis, thereby improving the speed and efficiency of the anomaly detection process. Furthermore, the anomaly threshold can be adjusted according to the specific requirements and risk tolerance of oilfield safety and environmental protection, providing a customized option for anomaly detection sensitivity, enabling the monitoring system to adapt to different monitoring standards and environments. Finally, a reasonably set anomaly threshold helps to balance false alarms and missed alarms. By adjusting the threshold, false alarms can be reduced without missing genuine anomalies, improving the reliability of the monitoring system.

[0051] Optionally, the step of constructing a dynamic monitoring model based on the problem data to analyze the data to be detected and obtain monitoring results includes: constructing a dynamic monitoring model based on the problem data and optimizing the dynamic monitoring model to obtain an optimized dynamic monitoring model; and analyzing the data to be detected based on the optimized dynamic monitoring model to obtain the monitoring results. In this embodiment, the dynamic monitoring model can be updated and adjusted in real time according to changes in oilfield safety and environmental protection data, ensuring that the monitoring system can adapt to the dynamic characteristics of the data and respond promptly to newly emerging problems. Furthermore, optimizing the dynamic monitoring model can improve the model's operating efficiency and predictive ability, reduce the consumption of computing resources, and maintain or improve the accuracy of monitoring results. In addition, the optimized dynamic monitoring model can analyze problem data more quickly, accelerate the identification process of abnormal situations, and enable oilfield managers to take timely countermeasures to reduce potential risks and losses. Finally, through continuous model optimization, the monitoring model can be continuously improved based on new data and experience, ensuring that the monitoring system remains efficient and effective over time.

[0052] Optionally, constructing a dynamic monitoring model based on the problem data includes: classifying and predicting the problem data using a support vector machine (SVM) algorithm to obtain a training set and a test set of problem data; training the dynamic monitoring model based on the training set of problem data to obtain a trained dynamic monitoring model; testing the trained dynamic monitoring model based on the test set to evaluate its accuracy and obtain a tested dynamic monitoring model; and parsing the data to be detected based on the tested dynamic monitoring model to obtain the monitoring result. In this embodiment, a support vector machine (SVM) is a supervised learning algorithm known for its high accuracy in classification and regression problems. By using SVM to classify and predict problem data, a high-performance dynamic monitoring model can be constructed. Furthermore, SVM defines the decision boundary by selecting support vectors, giving it excellent generalization ability in the feature space. It maintains high prediction accuracy even when facing unseen data. Additionally, the SVM algorithm excels at handling high-dimensional datasets; it effectively handles nonlinear relationships in data through kernel tricks, making it suitable for complex and high-dimensional datasets that may be encountered in the field of oilfield safety and environmental protection. In addition, the Support Vector Machine (SVM) algorithm exhibits robustness to noise and outliers in the data. Through appropriate parameter tuning and kernel function selection, SVM can provide stable monitoring results even in the presence of noise. By constructing separate training and test sets, the dynamic monitoring model can be systematically evaluated. This method allows model developers to test and verify the model's accuracy and reliability. The concept of support vectors in the SVM algorithm provides interpretability for the model's decision-making process. This helps oilfield managers understand the model's predictive logic and increases their trust in the monitoring results. Finally, the SVM algorithm allows the use of different kernel functions to adapt to different data characteristics, providing high flexibility and enabling the adjustment of model parameters according to the specific needs of oilfield safety and environmental monitoring.

[0053] Optionally, constructing a dynamic monitoring model based on the problem data further includes: using a random forest algorithm to classify and predict the data to be monitored by constructing multiple decision trees and obtaining the average of their outputs, thereby analyzing the data to be detected and obtaining the monitoring results. In this embodiment, random forest is an ensemble learning method that can significantly improve the stability and accuracy of the model by constructing multiple decision trees and averaging or majority voting on their prediction results. Furthermore, since the random forest algorithm uses different subsets of data when constructing each decision tree, this helps reduce the risk of overfitting and enhances the model's generalization ability on new data. In addition, the random forest algorithm can evaluate the importance of each feature to the prediction results during training, which helps identify key features, optimize the model structure, and improve model interpretability. Moreover, the random forest algorithm has good robustness to noise and outliers in the data. Even if some data has problems, the overall performance of the model will not be significantly affected, making the model more reliable in practical applications. Finally, each decision tree in the random forest algorithm can be trained independently, meaning that the model training process can be easily parallelized, thereby accelerating the model training speed.

[0054] Optionally, the construction of a dynamic monitoring model based on the problem data further includes: learning the problem data using a neural network algorithm to obtain a learned neural network algorithm, which is then used to analyze the data to be detected and obtain monitoring results. In this embodiment, neural networks, especially deep neural networks, are capable of learning and simulating very complex nonlinear relationships, making them well-suited for handling complex data patterns that may exist in the field of oilfield safety and environmental protection. Furthermore, neural networks can automatically extract features during the learning process, meaning they can process raw data without manual feature engineering, reducing preprocessing workload and potentially revealing useful information that traditional methods cannot capture. Additionally, by learning from large amounts of problem data, neural networks can gradually adapt to new or changing data patterns, thus providing continuous improvement and adaptability in oilfield safety and environmental monitoring. Properly trained neural networks can provide high-precision prediction results, especially on large-scale datasets, where they often outperform traditional algorithms. Neural network algorithms are highly flexible, allowing for the design of different network structures and learning strategies to adapt to different monitoring tasks, and they can also be easily scaled to handle larger datasets or more complex models. Modern neural network training can utilize parallel computing resources such as GPUs, which significantly accelerates model training and makes it possible to handle large-scale datasets. Finally, through proper regularization and network design, neural networks can have good generalization ability, which means that they not only perform well on training data, but are also able to effectively predict unseen data.

[0055] Optionally, the step of constructing a dynamic monitoring model based on the problem data further includes: calculating a target function based on a loss function of the given dynamic monitoring model; and constructing the dynamic monitoring model based on the target function.

[0056] Optionally, the loss function can be constructed based on the following formula:

[0057]

[0058] The objective function for the problem data r is Ξ(r), the features are c, the number of features is b, and the problem data for the c-th feature is r. c The weight of feature c is λ. c The standard value of the c-th feature is e c In this embodiment, by defining and minimizing a loss function to build the model, it is ensured that the model focuses on reducing prediction error during training. This approach has a clear objective: to optimize the model to minimize the difference between the actual and predicted outputs. Furthermore, the loss function can be customized according to specific monitoring needs and data characteristics. For example, different loss functions can be chosen to handle different types of problems, such as mean squared error for regression problems and cross-entropy loss for classification problems, providing high flexibility. Additionally, the loss function provides a means of quantifying model performance. During model training and testing, the model's effectiveness can be evaluated by monitoring the value of the loss function, thereby determining whether the model has sufficiently learned the features of the data. Moreover, minimizing the loss function, especially the regularized form, can help the model avoid overfitting, thereby improving its generalization ability on unseen data. Problem-specific constraints and requirements can be considered when constructing the objective function, making the final monitoring model more closely aligned with the actual monitoring needs of oilfield safety and environmental protection. Furthermore, the loss function-based approach is compatible with various optimization algorithms, such as gradient descent and its variants, providing multiple options for model training and helping to find better solutions. Finally, minimizing the loss function is usually achieved through an iterative process, which means that the model can be gradually improved in each iteration until the predetermined accuracy or number of iterations is reached.

[0059] Optionally, the objective function can be constructed based on the following formula:

[0060]

[0061] The objective function for the problem data r is Ξ(r), the features are c, the number of features is b, and the problem data for the c-th feature is r. c The weight of feature c is given by , and the standard value of the c-th feature is e. cIn this embodiment, the objective function assigns weights to each feature, allowing the model to encode the importance of different features during training. This means that features with a greater impact on the prediction results can receive more attention, thereby improving the model's prediction accuracy. Furthermore, by incorporating the comparison of feature values ​​with standard values ​​into the objective function, the data can be standardized. This method helps the model better understand and learn the distribution of the data, especially when features have different dimensions or distribution ranges, ensuring the effectiveness of model training. Additionally, the objective function typically incorporates bias and variance terms, allowing the model to balance the relationship between the two during training. Appropriate balance helps the model avoid overfitting or underfitting, improving its generalization ability. Moreover, since the objective function can include specific weights and standard values ​​for different features, this allows the model to adapt to various data characteristics and monitoring needs. The model can be customized according to the specific characteristics of the problem data, improving the relevance and effectiveness of the monitoring task. The objective function has a clear structure, making it easy to understand and interpret. The use of feature weights and standard values ​​helps explain the model's prediction decision-making process, increasing the model's interpretability and contributing to oilfield managers' trust and understanding of the monitoring results. The objective function provides the model with a clear optimization objective. By minimizing the objective function, the model can automatically adjust its parameters during training to achieve optimal predictive performance. Finally, the objective function can be adjusted or extended as needed to include additional constraints or penalty terms, such as regularization terms, thus providing additional control to prevent overfitting or enhance certain characteristics of the model.

[0062] Optionally, optimizing the dynamic monitoring model to obtain an optimized dynamic monitoring model includes: optimizing the dynamic monitoring model based on a particle swarm optimization algorithm to obtain the optimized dynamic monitoring model. In this embodiment, particle swarm optimization is a swarm-based optimization method that mimics the social behavior of flocks of birds or schools of fish to perform the search. This algorithm can effectively perform a global search within the solution space, increasing the probability of finding the global optimum, thereby improving the model's performance.

[0063] Furthermore, the particles in the Particle Swarm Optimization (PSO) algorithm are processed in parallel. Each particle represents a potential solution in the solution space, and updating multiple particles simultaneously can significantly accelerate the search speed, which is particularly important for oilfield safety and environmental monitoring that requires rapid response. In addition, the PSO algorithm has relatively few parameters, mainly including the number of particles, inertia weight, and learning factor. Adjusting these parameters is relatively simple and intuitive, making it easy to implement and adjust, thus facilitating the algorithm's understanding and application. PSO guides the search direction of particles by tracking individual optimal solutions and the global optimal solution, which helps particles escape local optima, further improving the algorithm's search capability and the model's optimization effect. Moreover, the PSO algorithm is insensitive to the choice of initial parameters and has strong robustness. This means that even under different initial conditions, the algorithm can work effectively and find excellent solutions. Finally, the PSO algorithm can be applied to different types of optimization problems, including continuous, nonlinear, nonconvex, and multi-peak function optimization problems, making it very suitable for optimizing oilfield safety and environmental monitoring models. By optimizing model parameters, the PSO algorithm helps improve the performance of dynamic monitoring models, including increasing prediction accuracy and reducing prediction errors, thereby making monitoring results more reliable.

[0064] Optionally, optimizing the dynamic monitoring model based on the particle swarm optimization algorithm to obtain the optimized dynamic monitoring model includes: initializing the dynamic monitoring model to obtain an initial population; iterating the initial population to obtain updated follower positions; and determining the updated population position based on the iterated follower positions to obtain the optimized dynamic monitoring model. In this embodiment, the population initialization step ensures the diversity of the particle swarm, thereby covering a wider area in the search space and increasing the probability of finding the optimal solution. Furthermore, the updating of particle positions during iteration is based on individual and group experience, enabling particles to dynamically adjust their search strategies according to historical information. This learning mechanism helps to quickly converge to the optimal solution. Additionally, particles in the particle swarm optimization algorithm adjust their search direction based on their individual best position and the global best position. This adaptive search strategy helps the algorithm achieve a balance between global and local searches. Finally, the particle swarm optimization algorithm does not depend on the specific structure of the problem, has good robustness and versatility, and can adapt to various types of dynamic monitoring model optimization problems, including nonlinear, nonconvex, and multivariable cases.

[0065] Optionally, the initial population is calculated based on the following formula:

[0066]

[0067] The number of iterations is t, and the maximum number of iterations is t. maxThe relation parameter is δ, and the position information of the i-th individual in the population in the j-th dimension during the t-th iteration is... The random numbers are distributed normally, H; the matrix of all ones is Y; and the safety threshold is V. T The warning value for the location of an individual in the population is E. S The exponential function with the natural constant e as base is exp(·), and the position information of the i-th individual in the j-th dimension in the (t+1)-th iteration is... In this embodiment, by combining normally distributed random numbers H with deterministic relational parameters, the formula introduces deterministic factors while maintaining a certain degree of randomness, which contributes to population diversity and prevents the initial population from getting trapped in local optima. Furthermore, the formula includes a safety threshold, ensuring that the initial population's positional information is within an acceptable safety range, providing a robust starting point for subsequent optimization. In addition, by setting a warning value, potential risks and anomalies can be considered during the population initialization phase, providing preventative measures for subsequent iterative updates. Finally, using an exponential function with the natural constant e as its base helps control the update magnitude of individual population positions, avoiding excessive jumps in the early stages of iteration, thus contributing to the stable convergence of the algorithm.

[0068] Optionally, the updated follower position is calculated based on the following formula:

[0069]

[0070] The worst position of the sub-global exploration in the t-th iteration is: The optimal position of the explorer in the (t+1)th iteration is The population size is m, and the matrix B contains elements with absolute values ​​less than or equal to 1. 13% of the population individuals are randomly selected as observers. In this embodiment, by considering both the worst and best global positions, the algorithm achieves a balance between exploration (global search) and exploitation (local search), increasing the probability of finding the global optimum. Furthermore, the matrix included in the formula allows for dynamic adjustment of the particle's search direction, enabling the particle to adjust its flight path based on information from the current iteration and historical best results. Additionally, randomly selecting 13% of the population individuals as observers helps maintain population diversity and avoids premature convergence, where the entire population prematurely clusters near non-optimal solutions. Finally, the matrix containing elements with absolute values ​​less than or equal to 1 allows the algorithm to control the magnitude of particle position updates, achieving adaptive exploration—that is, adjusting the search intensity and direction according to the needs of the search process.

[0071] Optionally, the updated population position can be calculated based on the following formula:

[0072]

[0073] The optimal population individual position in the t-th iteration is . Let p be the random parameter, and g be the fitness value of sparrow individual i. i The global optimal fitness value is r g The worst global fitness function value is r. w The constant is ρ, and iteration stops when the fitness value reaches its minimum. In this embodiment, updating the population position based on the fitness value ensures that the optimization process is driven by the quality of the solution, thereby guiding the particle swarm to move towards higher fitness regions and improving optimization efficiency. Furthermore, the introduction of a random parameter adds exploratory nature to the algorithm, allowing particles to not only follow the optimal solution but also explore the search space to a certain extent, increasing the chance of finding the global optimum. In addition, by calculating the global best fitness value and the global worst fitness function value, the algorithm can evaluate the performance of the current solution in each iteration and compare it with historical best solutions, providing a basis for population updates. Finally, the condition of stopping iteration when the fitness value reaches its minimum provides a clear termination point for the algorithm, which helps avoid endless iterations, saves computational resources, and ensures that the algorithm stops promptly after finding a satisfactory solution.

[0074] Specifically, in the actual assessment, two sets of standard information and dynamic data are provided:

[0075] Group 1:

[0076] Standard value for smoke and dust emission concentration: 20 mg / m³ 3 Actual value: 25mg / m³ 3 ;

[0077] Sulfur dioxide emission concentration standard: 35 mg / m³ 3 Actual value: 42 mg / m³ 3 ;

[0078] Nitrogen oxide emission concentration standard value: 80 mg / m³ 3 Actual value: 95mg / m³ 3 ;

[0079] Chemical oxygen demand (COD) emission concentration standard: 100 mg / L, actual value: 120 mg / L;

[0080] Ammonia nitrogen emission concentration standard: 15 mg / L, actual value: 20 mg / L;

[0081] Total phosphorus emission concentration standard: 1 mg / L, actual value: 1.5 mg / L;

[0082] Total nitrogen emission concentration standard: 20 mg / L, actual value: 25 mg / L;

[0083] Standard noise level in decibels: 70 dB; Actual value: 75 dB.

[0084] Standard value for heavy metal content in soil: 0.5 mg / kg; Actual value: 1.0 mg / kg;

[0085] Group 2:

[0086] Standard value for smoke and dust emission concentration: 30 mg / m³ 3 Actual value: 35mg / m³ 3 ;

[0087] Sulfur dioxide emission concentration standard value: 50 mg / m³ 3 Actual value: 55mg / m³ 3 ;

[0088] Nitrogen oxide emission concentration standard value: 120 mg / m³ 3 Actual value: 130 mg / m³ 3 ;

[0089] Chemical oxygen demand (COD) emission concentration standard: 150 mg / L, actual value: 160 mg / L;

[0090] Ammonia nitrogen emission concentration standard: 20 mg / L, actual value: 22 mg / L;

[0091] Total phosphorus emission concentration standard: 2 mg / L, actual value: 2.5 mg / L;

[0092] Total nitrogen emission concentration standard: 30 mg / L, actual value: 32 mg / L;

[0093] Standard noise level in decibels: 80 dB; Actual value: 85 dB.

[0094] Standard value for heavy metal content in soil: 1.0 mg / kg; Actual value: 1.5 mg / kg;

[0095] The dynamic data is subjected to standard information feature extraction, and the standard information is used to identify anomalies by a discriminant model to obtain problematic data;

[0096] In the actual assessment, the characteristic extracted from the first set of data was a smoke and dust emission concentration of 25 mg / m³. 3 Sulfur dioxide emission concentration 42 mg / m³ 3 Nitrogen oxide emission concentration 95 mg / m³ 3 The first set of data showed the following emission concentrations: Chemical Oxygen Demand (COD) 120 mg / L, Ammonia Nitrogen 20 mg / L, Total Phosphorus 1.5 mg / L, Total Nitrogen 25 mg / L, Noise Level 75 dB, and Heavy Metal Content in Soil 1.0 mg / kg. The second set of data extracted a characteristic of particulate matter emission concentration of 35 mg / m³. 3 Sulfur dioxide emission concentration 55 mg / m³3 Nitrogen oxide emission concentration 130 mg / m³ 3 Chemical oxygen demand (COD) emission concentration: 160 mg / L; ammonia nitrogen emission concentration: 22 mg / L; total phosphorus emission concentration: 2.5 mg / L; total nitrogen emission concentration: 32 mg / L; noise level: 85 dB; heavy metal content in soil: 1.5 mg / kg.

[0097] The selected standard information feature is that the first group has a smoke and dust emission concentration of 25 mg / m³. 3 Sulfur dioxide emission concentration 42 mg / m³ 3 Nitrogen oxide emission concentration 95 mg / m³ 3 The emission concentrations for the first group are: Chemical Oxygen Demand (COD) 120 mg / L, Ammonia Nitrogen 20 mg / L, Total Phosphorus 1.5 mg / L, Total Nitrogen 25 mg / L, and Heavy Metals in Soil 1.0 mg / kg; the second group is for particulate matter emissions of 35 mg / m³. 3 Sulfur dioxide emission concentration 55 mg / m³ 3 The concentrations of ammonia nitrogen emissions were 22 mg / L, total phosphorus emissions were 2.5 mg / L, and heavy metal content in the soil was 1.5 mg / kg.

[0098] Based on the aforementioned construction of the dynamic monitoring model, the dynamic monitoring model is optimized.

[0099] In the actual assessment, the dynamic monitoring result for Group 1 was 0.932662, and the dynamic monitoring result for Group 2 was 0.3243.

[0100] The data to be detected is input into the optimized dynamic monitoring model, and the monitoring results are output.

[0101] In the actual evaluation, the input test data is as follows:

[0102] Standard value for smoke and dust emission concentration: 40 mg / m³ 3 Actual value: 48 mg / m³ 3 ;

[0103] Sulfur dioxide emission concentration standard: 60 mg / m³ 3 Actual value: 65mg / m³ 3 ;

[0104] Nitrogen oxide emission concentration standard: 160 mg / m³ 3 Actual value: 170 mg / m³ 3 ;

[0105] Chemical oxygen demand (COD) emission concentration standard: 200 mg / L, actual value: 220 mg / L;

[0106] Ammonia nitrogen emission concentration standard: 25 mg / L, actual value: 30 mg / L;

[0107] Total phosphorus emission concentration standard: 3 mg / L, actual value: 3.5 mg / L;

[0108] Total nitrogen emission concentration standard: 40 mg / L, actual value: 43 mg / L;

[0109] Standard noise level in decibels: 90 dB; Actual value: 95 dB.

[0110] Standard value for heavy metal content in soil: 1.5 mg / kg; Actual value: 2.0 mg / kg;

[0111] The dynamic monitoring result of the data to be monitored is 0.228886.

[0112] Second Embodiment

[0113] This application also provides a dynamic monitoring device for abnormal information on oilfield safety and environmental protection, comprising: a data acquisition module for acquiring information data on oilfield safety and environmental protection, wherein the information data is preprocessed to obtain preprocessed data; a data processing module for extracting features from the preprocessed data to obtain feature data; and anomaly identification of the feature data based on a discriminant model to obtain problem data; a model building module for building a dynamic monitoring model based on the problem data; and a monitoring module for analyzing the data to be detected based on the dynamic monitoring model to obtain monitoring results.

[0114] Third Embodiment

[0115] Figure 2 This is a schematic diagram of an electronic device for dynamic monitoring of abnormal information in oilfield safety and environmental protection, as described in an embodiment of this application. Figure 2 As shown, this application also provides an electronic device, including: a memory and a processor, wherein the memory stores a computer-executable program, and the processor is used to execute the computer-executable program to implement the dynamic monitoring method for abnormal information on oilfield safety and environmental protection as described in any one of the claims.

[0116] At the hardware level, the electronic device includes a processor, and optionally also an internal bus, network interface, and memory. The memory may include main memory, such as high-speed random-access memory (RAM), or non-volatile memory, such as at least one disk drive. Of course, the electronic device may also include other hardware required for other business operations.

[0117] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 2 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0118] Memory is used to store programs. Specifically, programs may include program code, which includes computer operation instructions. Memory may include main memory and non-volatile memory, and provides instructions and data to the processor.

[0119] The processor reads the corresponding computer program from non-volatile memory into main memory and then runs it, forming a dynamic monitoring device for oilfield environmental safety standards based on data analysis at the logical level. The processor executes the program stored in memory and specifically performs any of the aforementioned dynamic monitoring methods for oilfield environmental safety standards based on data analysis.

[0120] The method for dynamic monitoring of abnormal information in oilfield safety and environmental protection as described in Embodiment 1 can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by the integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps disclosed in Embodiment 1 of this application can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0121] Fourth embodiment

[0122] This application also provides a computer storage medium storing a computer-executable program, which, when run, implements any of the oilfield safety and environmental protection anomaly monitoring methods described above.

Claims

1. A method for dynamic monitoring of abnormal information on oilfield safety and environmental protection, characterized in that, include: Obtain information data on oilfield safety and environmental protection, and preprocess the oilfield safety and environmental protection information data to obtain preprocessed data; Feature extraction is performed on the preprocessed data to obtain feature data; Anomaly identification is performed on the feature data based on a discriminant model to obtain problematic data; A dynamic monitoring model is constructed based on the problem data to analyze the data to be detected and obtain monitoring results.

2. The method for dynamic monitoring of abnormal information in oilfield safety and environmental protection according to claim 1, characterized in that, The preprocessing of the oilfield safety and environmental protection information data to obtain preprocessed data includes: removing duplicate data, deleting missing and outlier values, and performing data transformation and data reduction operations on the oilfield safety and environmental protection information data to obtain preprocessed data.

3. The method for dynamic monitoring of abnormal information in oilfield safety and environmental protection according to claim 1, characterized in that, The step of extracting features from the preprocessed data to obtain feature data includes: extracting features from the preprocessed data based on the random forest algorithm to obtain feature data.

4. The method for dynamic monitoring of abnormal information in oilfield safety and environmental protection according to claim 3, characterized in that, The feature extraction process based on the random forest algorithm, which extracts features from the preprocessed data to obtain feature data, includes: Initialize the information data on oilfield safety and environmental protection, and set the number of decision trees in the random forest algorithm; Data augmentation is performed on the oilfield safety and environmental protection information data based on generative adversarial networks to obtain a dataset; The dataset is sampled based on a bagging algorithm to form a training dataset; The decision tree is constructed based on the training dataset to obtain feature data.

5. The method for dynamic monitoring of abnormal information in oilfield safety and environmental protection according to claim 1, characterized in that, The method for identifying anomalies in the feature data based on a discriminant model to obtain problematic data includes: calculating the feature standard deviation and feature probability density based on the discriminant model to identify anomalies in the feature data to obtain problematic data.

6. The method for dynamic monitoring of abnormal information in oilfield safety and environmental protection according to claim 5, characterized in that, The standard deviation of the features is calculated based on the discriminant model defined by the following formula: The standard deviation of the c-th feature is η. c The preprocessed data for the m-th and c-th features is The amount of preprocessed data is q; The feature probability density is calculated based on the discriminant model defined by the following formula: The probability density of the preprocessed data r is h(r), the exponential function with the natural base e is given, and the anomaly threshold exp(·) is given. When the probability density is less than the anomaly threshold, the feature data is considered problematic data.

7. The method for dynamic monitoring of abnormal information on oilfield safety and environmental protection according to claim 1, characterized in that, The process of constructing a dynamic monitoring model based on the problem data to analyze the data to be detected and obtain monitoring results includes: A dynamic monitoring model is constructed based on the problem data, and the dynamic monitoring model is optimized to obtain an optimized dynamic monitoring model. The monitoring results are obtained by analyzing the data to be detected based on the optimized dynamic monitoring model.

8. A dynamic monitoring device for abnormal information on oilfield safety and environmental protection, characterized in that, include: Data acquisition module: used to acquire information data on oilfield safety and environmental protection, and to preprocess the information data to obtain preprocessed data; Data processing module: used to extract features from the preprocessed data to obtain feature data; Anomaly identification is performed on the feature data based on a discriminant model to obtain problematic data; Model building module: used to build a dynamic monitoring model based on the problem data; Monitoring module: Used to analyze the data to be detected based on a dynamic monitoring model to obtain monitoring results.

9. An electronic device, characterized in that, include: A memory and a processor, wherein the memory stores a computer-executable program, and the processor is configured to execute the computer-executable program to implement the method of any one of claims 1-7.

10. A computer storage medium, characterized in that, The computer storage medium stores a computer-executable program, which, when run, implements the method described in any one of claims 1-7.

Images