An intelligent management system and method for oil well data

Abstract

The present invention provides an intelligent management system and method for oil well data, relating to the field of oil well production technology. The system comprises an edge layer and a platform layer; the edge layer is used to perform anomaly analysis on oil well data; the platform layer is used to formulate intervention control strategies based on abnormal oil well data; the edge computing node is used to extract features from the original oil well data; the model building module is used to establish an anomaly detection model in combination with a historical oil well database; and the anomaly recognition module is used to input the features of the oil well data into the anomaly detection model for anomaly recognition. The present invention solves the high latency problem of centralized cloud processing in traditional solutions by combining edge computing nodes with a lightweight LSTM model, enabling sudden equipment failures to be accurately captured within a low response cycle; by adopting a dynamically optimized intervention strategy generation mechanism, the core parameters are adaptively adjusted for different well conditions, and in conjunction with a continuously iterative diagnostic knowledge base on the cloud, a closed-loop control link is formed, reducing the frequency of manual inspections and operation and maintenance costs.

Description

Technical Field

[0001] The present invention relates to the technical field of oil well production, and in particular to an intelligent management system and method for oil well data. Background Art

[0002] With global energy demand continuing to grow and the increasing difficulty of oil and gas resource development, oil well production management is facing unprecedented technical challenges. Traditional oil well development models incur an average daily production cost of over 20,000 RMB per well, while unplanned downtime due to equipment failures results in industry losses exceeding 10 billion RMB annually. In complex operating environments such as shale gas fields and high-sulfur oil fields, existing production management systems are unable to meet the core requirements of safe production, cost reduction, and efficiency improvement. The International Energy Agency (IEA) 2023 report indicates that the global oil and gas industry has a digital penetration rate of less than 35% and data utilization rates below 20%, necessitating the development of a next-generation intelligent production management system. The continued growth of global energy demand and the increasing difficulty of oil and gas resource development are driving technological innovation in oil well production management. According to the IEA's 2023 World Energy Outlook, global crude oil demand is expected to grow at an average annual rate of 0.8% over the next decade, while oil and gas resource recovery rates are projected to decline to 45%-50% over the same period due to increasing reservoir complexity and lagging development technologies. This contradiction is particularly acute in the development of unconventional resources such as deepwater oil and gas, shale oil, and tight gas. The average burial depth of newly discovered oil fields has increased from 2,000 meters in 2000 to 3,500 meters in 2023. The cost of drilling a single well has soared by 40%. High-temperature, high-pressure wells (temperatures > 150°C, pressures > 70 MPa) account for over 25% of all drilling, making it difficult for traditional monitoring technologies to accurately perceive downhole conditions. For example, an ultra-deepwater oil field in the Middle East boasts a daily production of 80,000 barrels per well, but unplanned downtime due to equipment failures results in direct economic losses exceeding $200 million annually. The International Energy Agency predicts that without intelligent transformation, the global oil and gas industry will face an average annual loss of $150 billion in production efficiency by 2030.

[0003] Conventional oil well data acquisition only captures surface parameters such as wellhead pressure and flow rate, lacking real-time monitoring of key data such as wellbore dynamics and formation seepage. Furthermore, linear analysis models are often used, making them difficult to handle high-dimensional, nonlinear problems. Equipment maintenance often relies on a "failure repair" model, resulting in low predictive maintenance coverage. Summary of the Invention

[0004] The present invention provides an intelligent oil well data management system and method, which are used to solve the defects in the prior art.

[0005] In one aspect, the present invention provides an intelligent oil well data management system, comprising:

[0006] Edge layer and platform layer; the edge layer is used to perform abnormal analysis on oil well data and output abnormal oil well data; the platform layer is used to formulate intervention control strategies based on abnormal oil well data; the edge layer includes:

[0007] Downhole sensor arrays for real-time acquisition of raw well data;

[0008] Edge computing nodes are used to extract features from raw oil well data and output oil well data features;

[0009] Model building module, used to build anomaly detection model by combining historical oil well database;

[0010] The anomaly recognition module is used to input the oil well data features into the anomaly detection model for anomaly recognition and output abnormal oil well data.

[0011] According to an intelligent oil well data management system provided by the present invention, the edge computing node includes a data preprocessing unit and a feature extraction unit. The preprocessing unit is used to preprocess the original oil well data and generate preprocessed oil well data; the feature extraction unit is used to extract features from the preprocessed oil well data and output the oil well data features.

[0012] According to an intelligent oil well data management system provided by the present invention, the steps of establishing an anomaly detection model include:

[0013] Normalize and serialize the data in the historical oil well database to construct training sets, test sets, and validation sets;

[0014] Build an LSTM model network;

[0015] According to the training set, the LSTM model is trained using the cross entropy loss function and the Adam optimization algorithm, and the training results are output;

[0016] The training results are verified using the test set until the loss of the validation set does not improve over multiple consecutive validations and the training is terminated.

[0017] According to the oil well data intelligent management system provided by the present invention, the specific steps of normalization and serialization processing include:

[0018] Using the continuity of time series, data in the historical oil well database is filled with missing data;

[0019] Perform outlier correction on data that exceeds the reasonable value range;

[0020] Calculate the mean and standard deviation of the data.

[0021] According to an intelligent management system for oil well data provided by the present invention, the anomaly identification module includes a window division unit and an anomaly determination unit; the window division unit is used to divide the oil well data features into multiple detection windows of fixed lengths, and the anomaly determination unit is used to input the oil well data features within the multiple detection windows into an anomaly detection model for detection and output abnormal oil well data.

[0022] According to the present invention, an intelligent oil well data management system is provided, wherein the platform layer includes:

[0023] Cloud data processing module, used to analyze the abnormal causes of abnormal oil well data and generate abnormal diagnosis reports;

[0024] The intelligent decision-making module is used to generate intervention control strategies based on abnormal diagnosis reports.

[0025] According to the intelligent oil well data management system provided by the present invention, the specific steps of analyzing the cause of the abnormality include:

[0026] Perform spatiotemporal alignment of abnormal oil well data with similar operating condition data in the historical oil well database, and output abnormal parameter combinations;

[0027] Based on the association rule mining algorithm and combined with abnormal parameter combinations, a structured fault mapping table is generated;

[0028] Based on the abnormal parameter combination and the equipment fault database, the abnormal propagation path and potential interference factors are located;

[0029] Generate abnormal diagnosis reports based on abnormal propagation paths and potential interference factors.

[0030] According to an intelligent oil well data management system provided by the present invention, the specific steps of generating abnormal parameter combinations include:

[0031] According to the abnormal parameter combination, set the parameters of the association rule mining algorithm and generate high-frequency parameter combinations;

[0032] Generate candidate rules based on high-frequency parameter combinations, and calculate the screening rules of candidate rules to obtain a list of highly relevant rules;

[0033] A triplet mapping is constructed based on a list of high-correlation rules to generate a structured fault mapping table.

[0034] According to an intelligent oil well data management system provided by the present invention, the specific steps of outputting the abnormal parameter combination include:

[0035] Extracting the operating condition characteristics of the abnormal oil well data and the historical oil well data, and outputting an operating condition characteristic matrix;

[0036] The abnormal parameters of the working condition feature matrix are clustered based on the random forest algorithm, and the abnormal parameter combinations are output.

[0037] On the other hand, the present invention also provides an oil well data intelligent management method, comprising:

[0038] Real-time collection of raw oil well data;

[0039] Extract features from original oil well data and output oil well data features;

[0040] Based on the LSTM algorithm model, abnormal feature detection is performed on oil well data features, and abnormal oil well data is output;

[0041] Analyze the abnormal causes of abnormal oil well data and generate abnormal diagnosis reports;

[0042] Generate intervention control strategies based on abnormal diagnosis reports.

[0043] The present invention provides an intelligent management system and method for oil well data. Through the embedded edge computing nodes of the edge layer combined with a lightweight LSTM model, it can independently complete data cleaning, feature extraction and initial screening of anomalies at the well site, effectively solving the high latency problem of centralized cloud processing in traditional solutions, so that sudden equipment failures can be accurately captured within a millisecond response cycle. Through spatiotemporal alignment technology and association rule mining algorithms, it can not only automatically decouple the nonlinear correlation patterns between key parameters from massive historical data, but also intuitively present the abnormal conduction path through a three-dimensional fault mapping table, significantly improving the root cause diagnosis accuracy under complex working conditions. By adopting an intervention strategy generation mechanism based on dynamic optimization, core parameters such as water injection pressure and oil production rate can be adaptively adjusted for different well conditions. In conjunction with the continuously iterative diagnostic knowledge base on the cloud, a closed-loop control link of "perception-analysis-decision-execution" is formed, which greatly reduces the frequency of manual inspections and operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 This is a schematic diagram of the structure of an intelligent oil well data management system provided by the first embodiment of the present invention;

[0046] Figure 2 This is a diagram of the steps for establishing an anomaly detection model in Example 1 of the present invention;

[0047] Figure 3This is a step diagram of an intelligent oil well data management method provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0048] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. The embodiments of the present invention and all other embodiments obtained by persons of ordinary skill in the art without inventive effort are within the scope of protection of the present invention.

[0049] The following combination Figure 1-Figure 3 An intelligent oil well data management system and method of the present invention are described.

[0050] Example 1:

[0051] Figure 1 It is a structural diagram of an intelligent management system for oil well data provided by an embodiment of the present invention.

[0052] Figure 2 This is a diagram of the steps for establishing an anomaly detection model in Example 1 of the present invention.

[0053] like Figure 1-Figure 2 As shown, an embodiment of the present invention provides an intelligent oil well data management system and method, the execution subject of which can be a cloud computing big data analysis system, including:

[0054] Edge layer and platform layer. The edge layer is used to perform anomaly analysis on oil well data and output abnormal oil well data. The edge layer includes:

[0055] Downhole sensor arrays collect raw oil well data in real time. These include new IoT devices such as triaxial accelerometers (for vibration monitoring), fiber Bragg grating sensors (for strain measurement), multi-parameter probes (for temperature, pressure, and flow), and acoustic sensors (for leak detection). These sensors monitor key parameters such as oil well vibration, strain, temperature, pressure, flow, and leak signals.

[0056] Edge computing nodes are used to extract features from raw oil well data and output well data features. Edge computing nodes include a data preprocessing unit and a feature extraction unit. The preprocessing unit preprocesses raw oil well data to generate preprocessed oil well data. Data preprocessing is a critical step in data analysis, improving data availability and reliability and providing a sound data foundation for subsequent feature extraction and model building. Data cleaning is typically achieved using a sliding window mechanism, with cubic spline interpolation used to fill time series gaps. K-means clustering is used to identify outliers, and median substitution is employed for data preprocessing. The feature extraction unit extracts features from preprocessed oil well data and outputs well data features. This is achieved by converting complex raw data into analyzable numerical features. By integrating time-frequency analysis techniques, an improved wavelet packet transform is used to extract energy entropy features in the 0.1-100 kHz frequency band. Principal component analysis is then used to reduce the dimensionality to a 20-dimensional feature vector.

[0057] The model building module is used to combine the historical oil well database and take a fixed number of random samples from the operating data of 100,000 oil wells covering a 5-year period. The steps to build the anomaly detection model include:

[0058] Normalize and serialize the data in the historical oil well database to construct the training set, test set, and validation set. The specific steps for normalizing and serializing the data in the historical oil well database include:

[0059] Leveraging the continuity of time series, we fill missing data in the historical oil well database. We use a bidirectional LSTM network for cross-period data completion, and initiate multi-source data fusion for sequences with a missing rate exceeding 15%. This approach effectively utilizes information in time series and improves the accuracy of data completion.

[0060] Outlier correction is performed on data that falls outside the reasonable range. A modified Grubbs test method combined with Markov Chain Monte Carlo simulation is used to determine confidence intervals. This allows for accurate identification and correction of outliers, improving data reliability.

[0061] Calculate the mean and standard deviation of the data. Robust statistics are introduced when calculating the mean and standard deviation of the data, using the Huber loss function instead of the traditional variance calculation to reduce the impact of outliers on the statistical results and improve the robustness of the statistical results. A hybrid neural network architecture consisting of two bidirectional LSTM layers (128 units) and an attention mechanism layer is constructed to effectively extract features from time series and improve the model's predictive performance.

[0062] Build an LSTM model network. Using a deep learning framework such as Keras or TensorFlow, construct a hybrid neural network architecture consisting of two bidirectional LSTM layers (128 units) and an attention mechanism layer. The LSTM layer effectively extracts features from time series, while the attention mechanism further highlights important features, improving the model's predictive performance.

[0063] Based on the training set, the LSTM model was trained using an improved Focal Loss function (with adjustment factors γ = 2 and α = 0.85) and the AdamW optimizer (with weight decay 0.01), and the training results were output. The improved Focal Loss function can effectively address class imbalance and improve the model's generalization ability. The AdamW optimizer can effectively improve the model's convergence speed and performance.

[0064] The training results are validated using the test set, and training is terminated when the validation set loss does not improve over multiple consecutive validation runs. This involves using the dynamic time warping algorithm to perform variable-length alignment on the test set. Training is terminated when the validation set MAE decreases by less than 0.1% over three consecutive epochs to ensure the model's generalization and improve its predictive performance.

[0065] The anomaly identification module inputs oil well data features into the anomaly detection model for anomaly identification and outputs abnormal well data. The anomaly identification module consists of a window partitioning unit and an anomaly determination unit. The window partitioning unit uses a variable-length window strategy (basic window of 128 sampling points, expansion factor 1.5) in conjunction with a dynamic threshold adjustment mechanism. The anomaly determination unit incorporates a multimodal decision fusion mechanism, integrating LSTM prediction probability (threshold 0.8), reconstruction error (threshold 3σ), and SHAP value interpretation results to output abnormal well data.

[0066] The platform layer is used to formulate intervention control strategies based on abnormal oil well data. The platform layer includes:

[0067] The cloud data processing module is used to analyze the causes of abnormal oil well data and generate abnormal diagnosis reports. The specific steps for analyzing the causes of abnormalities include:

[0068] The module performs spatiotemporal alignment of the abnormal well data with similar operating condition data in the historical oil well database, uses a dynamic time warping algorithm to eliminate clock bias, and outputs the abnormal parameter combination. Specifically, the module performs spatiotemporal alignment of the abnormal well data with similar operating condition data in the historical oil well database. This step ensures comparability between the data. Spatiotemporal alignment can help us more accurately identify and analyze anomalies. During the spatiotemporal alignment process, the module uses a dynamic time warping algorithm to eliminate clock bias. The dynamic time warping algorithm is a nonlinear time series analysis method that can effectively handle nonlinear changes in time series, thereby improving the accuracy of the alignment.

[0069] The specific steps for outputting abnormal parameter combinations include:

[0070] The operating characteristics of abnormal oil well data and historical oil well data are extracted, and an operating characteristic matrix is ​​output. This is to convert complex oil well data into analyzable numerical features. The extracted operating characteristics may include key parameters such as oil well pressure, flow rate, and temperature. These characteristics can more accurately analyze and identify abnormal conditions. VMD decomposition is used to obtain the intrinsic mode function, and a 32×32-dimensional operating characteristic matrix is ​​constructed using the correlation coefficient matrix. VMD is a non-recursive, adaptive signal processing method that can effectively decompose complex signals into multiple intrinsic mode functions (IMFs). Through VMD decomposition, we can obtain IMF components that reflect the operating characteristics.

[0071] We cluster abnormal parameters within the operating condition feature matrix using the random forest algorithm. We use the silhouette coefficient to optimize the number of clusters and output abnormal parameter combinations that include key parameters such as excessive pressure fluctuations and abnormal pump and valve stroke frequency. The silhouette coefficient is a comprehensive measure of clustering effectiveness, reflecting both the compactness within clusters and the degree of separation between clusters. By optimizing the number of clusters, we can obtain more reasonable abnormal parameter clustering results.

[0072] Based on the association rule mining algorithm and combined with abnormal parameter combinations, a structured fault mapping table is generated. Abnormal parameter combinations can help us more accurately identify and analyze abnormal situations, providing a basis for subsequent fault diagnosis and intervention control. The specific steps for generating a structured fault mapping table include:

[0073] Extract the core elements from the list of highly correlated rules and output a preliminary rule collection. Use the FP-Growth algorithm to mine strong association rules between parameters (minimum support 0.05, confidence 0.9), and extract a list of highly correlated rules through frequent item set analysis. Specific methods for obtaining highly correlated rules include:

[0074] The data in the abnormal parameter combination are discretized into continuous parameters (quantile classification) and encoded (one-hot / binary identification).

[0075] The FP-Growth algorithm is used to extract the parameter combination that meets the minimum support.

[0076] Generate candidate rules A→B for each frequent item set. Specific method: Generate all non-empty true subsets A of X based on the frequent item set X={a,b,c}, and the remaining part is B=X / A. For each subset A, form a rule A→B. If the support of A or B is lower than the preset threshold (such as min support =0.05), then discard the rule.

[0077] For example: the support of the frequent item set X={a,b,c} is 0.10, and the generated candidate rules include: A→BC (support must be ≥0.05), AB→C (support must be ≥0.05), AC→B, B→AC, BC→A, C→AB.

[0078] Calculate confidence, lift, and leverage, and filter out low-strength rules (e.g., lift < 1.2). Confidence indicates the probability that rule B will also occur when rule A occurs; lift indicates the ratio of the probability of rules A and B occurring simultaneously to the probability of A and B occurring independently; and leverage reflects the difference between the occurrence of rules A and B occurring simultaneously and the occurrence of A and B occurring independently. The formula for calculating confidence is:

[0079]

[0080] It represents the probability that event B will occur when time A occurs.

[0081] The formula for calculating lift is:

[0082]

[0083] Used to measure the correlation between events A and B. If Lift > 1, there is a positive correlation (A and B tend to occur at the same time). If Lift = 1, there is independence (no correlation). If Lift < 1, there is a negative correlation (A and B tend to be mutually exclusive).

[0084] The formula for calculating leverage is:

[0085]

[0086] Used to measure the deviation between the actual co-occurrence frequency of event A and event B and the independent case: Leverage > 0, there is a positive correlation (the co-occurrence frequency is higher than the independent expectation). Leverage = 0, there is independence. Leverage < 0, there is a negative correlation.

[0087] Filter out low-strength rules with lift less than 1.2, retain rules with lift ≥ 1.2 (to avoid weak correlations), and retain rules with confidence ≥ 0.85 (to avoid low confidence). For rules with lift close to the threshold, further filter them based on leverage (e.g., leverage > 0.05).

[0088] Then sort the rules in descending order of lift (prioritizing strongly correlated rules). If the lifts are the same, sort them in descending order of confidence.

[0089] Rules that meet preset weights (such as the confidence threshold for high-risk faults) are retained. Rules containing relationships are pruned, and semantic duplicates are merged (such as A→B and A∧C→B).

[0090] Statistical tests (chi-square / Fisher test) confirm significance. Prioritize based on domain knowledge (e.g., oil production decline rules take precedence over equipment efficiency rules) and output a list of highly relevant rules.

[0091] Semantic analysis of the rules is performed, using natural language processing technology (BERT model) to identify core fault factors (such as the combination of "abnormal pump power ↑ + oil pressure fluctuation ↓"). This outputs a preliminary set of rules with weight coefficients. Association rule mining is a data mining method that can discover potential relationships and rules in data.

[0092] The specific steps of outputting a structured fault mapping table include:

[0093] Map the preliminary rule collection to the ontology predefined in the knowledge graph, map the parameters in the preliminary rule collection to ontology attributes, standardize the relationship types into ontology relationships such as isCausedBy and affects, and annotate the rule confidence through RDF-star, that is, based on the success rate of historical case verification, output standardized rule data.

[0094] Convert each rule in the standardized rule data into a triple and output a list of standardized triples. Design a multi-hop relation encoding strategy to convert the standardized rules into a <subject, predicate, object> triple form.

[0095] Generate a structured fault mapping table based on a standardized list of triples, including table fields. Develop a dynamic field mapping engine to automatically match database fields based on device type. Use a columnar database (ClickHouse) to build a structured fault mapping table, supporting millisecond-level query responses.

[0096] Based on abnormal parameter combinations and in conjunction with the equipment fault database, the anomaly propagation path and potential interference factors were located. Based on the equipment fault database (including a knowledge graph of over 3,000 fault cases), a graph convolutional network (GCN) was used to rank node importance. A random walk algorithm (with a restart probability of 0.15) was designed to identify abnormal parameter association paths within the knowledge graph (e.g., "abnormal downhole vibration → drive shaft eccentricity → surface torque fluctuation"). An LSTM-Autoencoder model was introduced to analyze time-series interference signals and detect potential interference factors such as sudden changes in formation pressure and equipment aging trends.

[0097] Generate anomaly diagnosis reports based on anomaly propagation paths and potential interference factors. Build a Bayesian network inference engine to calculate root cause probability rankings (e.g., bearing wear probability reaches 87.3%). Monte Carlo simulations are used to predict production losses (with a 95% confidence interval) and environmental impact levels. The final output includes an interactive diagnostic report with a fault tree analysis diagram and a 3D heat map.

[0098] The intelligent decision-making module generates intervention control strategies based on abnormality diagnosis reports. It includes an adaptive control strategy library (pre-configured solutions for various typical operating conditions) and a digital twin simulation verification module (using a physical information neural network to accelerate simulation). It dynamically optimizes control parameters through a reinforcement learning framework and outputs a final control strategy that includes throttle valve adjustment instructions, water injection plan adjustments, and equipment start-up and shutdown plans.

[0099] In summary, this embodiment provides a cloud computing big data analysis system. By combining embedded edge computing nodes at the edge layer with a lightweight LSTM model, it can autonomously complete data cleaning, feature extraction, and initial anomaly screening at the well site. This effectively solves the high latency problem of centralized cloud processing in traditional solutions, allowing sudden equipment failures to be accurately captured within a millisecond response cycle. Through spatiotemporal alignment technology and association rule mining algorithms, it can not only automatically decouple nonlinear correlation patterns between key parameters from massive amounts of historical data, but also intuitively present anomaly transmission paths through a three-dimensional fault mapping table, significantly improving the accuracy of root cause diagnosis under complex working conditions. By adopting an intervention strategy generation mechanism based on dynamic optimization, core parameters such as injection pressure and oil production rate can be adaptively adjusted for different well conditions. Combined with the cloud-based continuously iterative diagnostic knowledge base, a closed-loop control chain of "perception-analysis-decision-execution" is formed, significantly reducing the frequency of manual inspections and operation and maintenance costs.

[0100] Example 1:

[0101] Data acquisition revealed that the downhole triaxial accelerometer detected an abnormal vibration frequency of 25kHz with a sustained pulse. The fiber Bragg grating sensor indicated that the pump barrel strain exceeded the threshold by 15%.

[0102] After edge computing processing, the data was cleaned using a sliding window (128 points / window), and two time gaps were repaired using cubic spline interpolation. An improved wavelet packet was used to extract the 0.5-45kHz energy entropy features, which were then reduced to 20 dimensions using PCA.

[0103] The trained LSTM prediction model achieved a probability of 0.83, a reconstruction error of 3.2σ, and SHAP analysis showed a 62% contribution from the vibration spectrum features.

[0104] The system's proposed solution: Digital twin simulation indicated an urgent 30% speed reduction. Reinforcement learning outputted a compensation plan of increasing water injection by 15%.

[0105] Example 2:

[0106] With the same general inventive concept, the present invention also protects an intelligent management method for oil well data. The intelligent management method for oil well data provided by the present invention is described below. The intelligent management method for oil well data described below and the cloud computing big data analysis system described above can be referenced to each other.

[0107] Figure 3 This is a step diagram of an oil well data intelligent management method provided by an embodiment of the present invention.

[0108] like Figure 3 As shown, an embodiment of the present invention provides an intelligent management method for oil well data, including:

[0109] Real-time collection of raw oil well data. Multi-parameter monitoring devices (including pressure transmitters, temperature sensors, flow meters, and vibration probes) deployed in a downhole sensor network capture dynamic data from the entire oil well production process at a millisecond sampling rate. Data transmission is achieved using an Industrial Internet of Things (IIoT) architecture, integrating ZigBee, LoRaWAN, and 5G hybrid networking technologies to ensure real-time data. A distributed data cache pool is established to perform preliminary verification of raw data, using a sliding window algorithm to eliminate transient noise interference, while simultaneously completing timestamp alignment and unit normalization. For oil wells in remote areas, satellite communication modules are deployed as redundant links to ensure data integrity and continuity.

[0110] Feature extraction is performed on raw oil well data, and the data features are output. A multidimensional feature space is constructed based on digital signal processing technology, encompassing time domain features (mean, variance, peak-to-peak value), frequency domain features (FFT spectrum peak, harmonic distortion), and nonlinear dynamic features (Lyapunov exponent, fractal dimension). Peak-to-peak value refers to the difference between the maximum and minimum values ​​of a signal within a cycle in signal processing. It is often used to describe the signal's fluctuation range or amplitude. An improved wavelet packet transform algorithm is used to extract the energy entropy index of fault-sensitive frequency bands, combined with principal component analysis (PCA) for feature dimensionality reduction. An adaptive feature selection model is developed, and reinforcement learning mechanisms are used to dynamically optimize feature subset combinations. A three-dimensional feature mapping matrix is ​​constructed, encompassing geological structural parameters, equipment operating parameters, and production operation parameters, enabling correlation and mapping between reservoir seepage characteristics and surface production data.

[0111] Anomaly detection is performed on oil well data features based on the LSTM algorithm model, and abnormal well data is output. A deep long short-term memory neural network architecture is constructed, with three layers of bidirectional LSTM units forming the time series prediction network. The number of nodes in the input layer corresponds to the optimized feature dimensions, and an attention mechanism module is introduced in the hidden layer to enhance the feature response at key time steps. The Adam optimizer is used with Dropout regularization to prevent overfitting. The training set uses labeled normal operating condition samples from historical data, and the validation set incorporates the synthetic minority oversampling technique (SMOTE) to balance the class distribution. A dynamic threshold judgment mechanism is designed, combining the Mahalanobis distance and the Grubbs test statistic to determine the multivariate joint anomaly judgment criteria. An incremental learning framework is developed to support online model updates, automatically retraining model parameters every 24 hours to adapt to changing production conditions.

[0112] Analyze the causes of abnormal oil well data and generate anomaly diagnosis reports. Build an expert system based on a knowledge graph, integrating equipment manuals, fault case libraries, and worker experience. Use natural language processing to parse unstructured documents and construct causal reasoning chains. Apply Bayesian networks to perform multi-source heterogeneous data fusion reasoning and quantify the conditional probability relationships between different factors. Develop a hierarchical diagnostic module: Primary diagnosis uses pattern matching to identify typical failure modes (such as pump and valve failure and tubing leakage); advanced diagnosis uses a physical simulation engine to build a digital twin model to simulate the evolution of production parameters under different operating conditions.

[0113] Generate intervention control strategies based on abnormal diagnosis reports. Establish a rule-based strategy library containing three types of knowledge units: equipment maintenance plans, production parameter optimization plans, and emergency response measures. Develop a multi-objective optimization engine that comprehensively considers economic benefits, safety risks, and environmental constraints, and uses the NSGA-II genetic algorithm to solve the optimal control sequence. For deterministic fault types, trigger the preset standard operating procedure (SOP) and issue control instructions to the RTU unit through the Industrial Internet of Things; for complex working conditions, call the digital twin platform for virtual verification and generate a decision proposal that includes parameter adjustment range, implementation time, and expected results. Introduce a human-machine collaboration mechanism, and key decisions must be reviewed and confirmed by experts before they can be executed. Blockchain technology is used to record operation logs to ensure traceability.

[0114] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0115] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. With this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes instructions for enabling a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or portions thereof.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An intelligent management system for oil well data, characterized in that: include: edge layer and platform layer; The edge layer is used to perform abnormal analysis on oil well data and output abnormal oil well data; The platform layer is used to formulate an intervention control strategy based on the abnormal oil well data; the platform layer includes: A cloud data processing module is used to analyze the abnormal causes of the abnormal oil well data and generate an abnormality diagnosis report; An intelligent decision-making module, configured to generate an intervention control strategy based on the abnormal diagnosis report; The specific steps of analyzing the cause of the abnormality include: Performing spatiotemporal alignment on the abnormal oil well data and similar operating condition data in a historical oil well database, and outputting an abnormal parameter combination; Based on the association rule mining algorithm, combined with the abnormal parameter combination, a structured fault mapping table is generated; Based on the abnormal parameter combination and in combination with the equipment fault database, locate the abnormal propagation path and potential interference factors; Producing the abnormality diagnosis report based on the abnormality propagation path and potential interference factors; The specific steps of outputting the abnormal parameter combination include: Extracting the operating condition characteristics of the abnormal oil well data and the historical oil well data, and outputting an operating condition characteristic matrix; Based on the random forest algorithm, abnormal parameters of the working condition feature matrix are clustered and abnormal parameter combinations are output; The edge layer comprises: Downhole sensor arrays for real-time acquisition of raw well data; An edge computing node is used to extract features from the raw oil well data and output features of the oil well data; Model building module, used to build anomaly detection model by combining historical oil well database; The anomaly identification module is used to input the oil well data features into the anomaly detection model to perform anomaly identification and output the abnormal oil well data.

2. The intelligent oil well data management system according to claim 1, characterized in that: The edge computing node includes a data preprocessing unit and a feature extraction unit. The preprocessing unit is used to perform data preprocessing on the original oil well data to generate preprocessed oil well data; the feature extraction unit is used to perform feature extraction on the preprocessed oil well data and output oil well data features.

3. The intelligent oil well data management system according to claim 1, characterized in that: The steps of establishing the anomaly detection model include: Normalize and serialize the data in the historical oil well database to construct training sets, test sets, and validation sets; Build an LSTM model network; According to the training set, the LSTM model is trained using a cross entropy loss function and an Adam optimization algorithm, and a training result is output; The training results are verified using the test set until the training is terminated when the loss of the verification set does not improve in multiple consecutive verifications.

4. The intelligent oil well data management system according to claim 3, characterized in that: The specific steps of performing the normalization and serialization processing include: Using the continuity of time series, data in the historical oil well database is filled with missing data; Perform outlier correction on the data that exceeds the reasonable value range; Calculate the mean and standard deviation of the data.

5. The intelligent oil well data management system according to claim 1, characterized in that: The anomaly identification module includes a window division unit and an anomaly determination unit; the window division unit is used to divide the oil well data features into multiple detection windows of fixed lengths, and the anomaly determination unit is used to input the oil well data features within the multiple detection windows into the anomaly detection model for detection and output the abnormal oil well data.

6. The intelligent oil well data management system according to claim 1, characterized in that: The specific steps of generating the structured fault mapping table include: According to the abnormal parameter combination, the parameters of the association rule mining algorithm are set, and a high-frequency parameter combination is generated; Generating candidate rules according to the high-frequency parameter combination, and calculating screening rules of the candidate rules to obtain a high-relevance rule list; A triplet mapping is constructed according to the high-correlation rule list to generate the structured fault mapping table.

7. An intelligent management method for oil well data, an intelligent management system for oil well data according to any one of claims 1 to 6, characterized in that: include: Real-time collection of raw oil well data; performing feature extraction on the original oil well data and outputting oil well data features; Perform abnormal feature detection on the oil well data features based on the LSTM algorithm model, and output abnormal oil well data; Analyze the abnormal causes of the abnormal oil well data and generate an abnormality diagnosis report; An intervention control strategy is generated based on the abnormal diagnosis report.

Images