An oil drilling and production data analysis and prediction system

Abstract

The application relates to the technical field of petroleum drilling engineering, and particularly discloses a petroleum drilling and production data analysis and prediction system, which realizes virtual reconstruction of a physical drilling system by constructing a digital mapping body, adopts a multi-source sensor network to collect drill string vortex frequency and wellbore temperature field gradient data, and forms a time sequence feature set through time sequence alignment and fusion processing; through differential transformation and modal decomposition on the feature set, a transient response characteristic vector is extracted and a risk distribution map is generated; a parameter-risk correlation model is established, mechanical vibration and thermal stress interference are decoupled by solving a physical equation, and a dynamic risk index is obtained; the risk index and process parameters are mapped to a three-dimensional grid to construct a multi-dimensional feature space, information aggregation and reconstruction are realized by adopting a feature fusion grid, and a comprehensive risk assessment index is generated; a dynamic early warning threshold is established according to the assessment index, and drilling parameters are optimized in real time through closed-loop control.

Description

Technical Field

[0001] This invention relates to the field of oil drilling engineering technology, and specifically to a system for analyzing and predicting oil drilling and production data. Background Technology

[0002] As oil drilling extends into deeper and more complex formations, downhole condition monitoring and risk early warning face unprecedented technical challenges. During drilling in deep, ultra-deep, and complex geological structures, the drill string system is subjected to intense mechanical vibration and thermal stress coupling. This multi-physics coupling effect results in downhole risks exhibiting complex characteristics such as nonlinearity, time-varying nature, and uneven spatial distribution. Traditional drilling monitoring methods primarily rely on single-point parameter threshold alarms, which struggle to effectively capture the dynamic evolution of risks under multi-parameter coupling. This leads to insufficient accuracy and timeliness in early warning of complex downhole conditions, severely hindering the safe and efficient development of deep oil and gas resources.

[0003] Current methods monitor mechanical vibration and thermal stress as independent parameters, neglecting their interaction mechanism in the high-temperature and high-pressure environment of deep wells. This limitation prevents the system from accurately analyzing the causes of combined faults. For example, when mechanical vibration and thermal stress jointly cause drill string fatigue damage, traditional monitoring methods often only detect abnormal surface parameters but cannot identify the underlying coupling mechanism, leading to delayed warnings or even misjudgments. Especially during deep formation drilling, due to the drastic changes in downhole temperature gradients and the complex coupling with drill string dynamics, conventional monitoring systems struggle to accurately locate risk sources and predict their evolution trends, severely impacting the timeliness and effectiveness of drilling safety control under complex conditions. Summary of the Invention

[0004] The purpose of this invention is to provide a system for analyzing and predicting oil drilling and production data to solve the problems mentioned above.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] A system for analyzing and predicting oil drilling and production data includes:

[0007] The data acquisition module is used to apply oil drilling excitation in a preset mode and simultaneously acquire multi-source dynamic response signals of the downhole system during the operation cycle, and construct a time-series feature set including drill string vortex frequency sequence and wellbore temperature field gradient distribution map.

[0008] The feature extraction module is used to perform differential transformation and mode decomposition on the time series feature set, extract transient response feature vectors that characterize the stability of the drilling system, and generate corresponding risk distribution maps;

[0009] The risk modeling module is used to establish a parameter-risk correlation model based on the risk distribution map. By decoupling mechanical vibration response and thermal stress interference, dynamic risk indicators reflecting the true risk status downhole are obtained.

[0010] The risk assessment module is used to construct a multi-dimensional feature space by combining dynamic risk indicators and drilling process parameters. It uses a feature fusion grid to aggregate information and reconstruct features of feature nodes to generate comprehensive risk assessment indicators.

[0011] The early warning and control module is used to generate risk level determination results based on comprehensive risk assessment indicators, analyze the temporal evolution characteristics of the risk distribution map, establish dynamic risk early warning thresholds, and adjust drilling control parameters in real time according to the early warning results.

[0012] As a further aspect of the present invention: the construction of a time-series feature set including the drill string eddy frequency sequence and the wellbore temperature field gradient distribution map specifically includes:

[0013] Construct a dynamic mapping of a three-dimensional drilling system, and synchronize the structural parameters and operating parameters of the physical drilling system through a real-time data interface to establish a digital mapping body that includes drill string geometry, wellbore structural dimensions and formation physical property parameters.

[0014] A dynamic virtual sensor network is deployed at key dynamic nodes of the digital mapping body, including setting up eddy monitoring points at drill string connections and setting up temperature gradient monitoring arrays in the wellbore annulus. High-frequency sampling data of drill string eddy frequency and spatial distribution data of wellbore temperature field are obtained by solving fluid-structure coupling equations.

[0015] The actual data collected by physical sensors and the simulation data generated by virtual sensors are fused from multiple sources. The two types of data sources are processed synchronously to construct a comprehensive feature set, which includes the time-series variation curve of drill string vortex frequency and the gradient distribution map of wellbore temperature field. This is denoted as the time-series feature set.

[0016] As a further aspect of the present invention: the differential transformation and mode decomposition of the time-series feature set specifically includes:

[0017] Point-by-point difference calculations were performed on the drill string eddy frequency sequence to obtain the first and second order differential sequences of the eddy frequency, thereby capturing the instantaneous change characteristics of the drill string's dynamic behavior.

[0018] Multi-scale mode decomposition is performed on the sequence after differential processing. The intrinsic mode functions characterizing different physical mechanisms are separated through an iterative screening process, and the energy entropy value of each mode is calculated to quantify the mode contribution.

[0019] The spatial topology of the virtual sensor network in the digital mapping volume is coupled with the modal decomposition results. By projecting the modal energy into the three-dimensional wellbore mesh, a multi-dimensional modal feature space of the drill string dynamic state is constructed.

[0020] Cluster analysis is performed using data points in the modal feature space to identify abnormal modal combination patterns and update the modal decomposition parameters.

[0021] As a further aspect of the present invention: the generation of the corresponding risk distribution map specifically includes:

[0022] From the modal decomposition results, intrinsic mode functions with energy entropy values ​​exceeding a threshold are selected, and their instantaneous amplitude, frequency, and phase parameters are extracted to construct a multi-dimensional transient response feature vector characterizing the stability of the drilling system.

[0023] The feature vectors are correlated and mapped with the spatial coordinates of the wellbore to generate a continuous risk index field, and the distribution density of the feature vectors in the three-dimensional grid is optimized.

[0024] Based on the data points of the risk indicator field, a risk level surface is constructed, and different risk areas are visualized through color coding and transparency rendering to form a risk distribution map.

[0025] As a further aspect of the present invention: the establishment of the parameter-risk correlation model based on the risk distribution map specifically includes:

[0026] Extract the coordinates of high-risk areas and their corresponding risk levels from the risk distribution map, and simultaneously collect the drilling process parameters corresponding to the high-risk areas, including drilling pressure, rotation speed and drilling fluid performance parameters.

[0027] A parameter-risk mapping table was constructed, with drilling process parameters as input variables and risk level data as output variables. A quantitative correlation between parameter changes and risk levels was established through multiple regression analysis.

[0028] In the digital mapping body, a parameter sensitivity analysis node is set up. By controlling the change of a single variable, the response characteristics of the risk distribution map are observed, and the influence weight of each process parameter on the risk status is determined.

[0029] Based on the influence weight optimization parameter-risk correlation, a parameter-risk correlation model that can reflect the dynamic correspondence between drilling parameters and risk status is established.

[0030] As a further aspect of the present invention: the dynamic risk index reflecting the true downhole risk state is obtained by decoupling mechanical vibration response and thermal stress interference, specifically including:

[0031] Physical equations for the propagation path of mechanical vibration and the path of heat conduction are established separately. The theoretical values ​​of the pure mechanical vibration response and the pure thermal stress distribution are obtained by solving the partial differential equation system.

[0032] The actual collected composite response data is compared and analyzed with the theoretical values, and the mechanical vibration component and thermal stress component are separated by residual calculation method;

[0033] The energy of the separated mechanical vibration response is integrated to calculate the vibration risk index, and the gradient analysis of the thermal stress distribution is performed to calculate the thermal stress risk index.

[0034] By weighting and fusing the vibration risk index and the thermal stress risk index, a dynamic risk index that reflects the true risk state downhole is generated.

[0035] As a further aspect of the present invention: the construction of a multi-dimensional feature space by combining dynamic risk indicators and drilling process parameters specifically includes:

[0036] A three-dimensional spatial grid coordinate system is established, and the wellbore space is divided into several grid units according to geometric features. Each grid unit serves as a basic dimension of the feature space.

[0037] Dynamic risk indicators are mapped to the grid coordinate system according to their corresponding spatial locations, while drilling process parameters are allocated to the corresponding grid nodes according to their range of action and radius of influence.

[0038] Data is filled into the grid areas that are not directly covered to ensure the continuity and integrity of the feature space in the 3D wellbore environment;

[0039] Through parameter density analysis and spatial distribution verification of grid nodes, a multidimensional feature space containing dynamic risk indicators and drilling process parameters was constructed.

[0040] As a further aspect of the present invention: the method of using a feature fusion grid to aggregate information and reconstruct features of feature nodes to generate a comprehensive risk assessment index specifically includes:

[0041] Based on the constructed multidimensional feature space, a hierarchical structure of the feature fusion mesh is designed, in which the size of the coarse mesh is set to 2m×2m×4m, and each coarse mesh node governs 64 basic mesh cells to capture global features;

[0042] The fine grid retains its original size of 0.5m × 0.5m × 1m and is used to extract local detail features. Through information transmission and interaction between adjacent grid nodes, a weighted average algorithm is used to calculate the feature representative value of the area under the jurisdiction of each coarse grid node. The weight is determined according to the distance of the grid node to the center point of the coarse grid. The closer the distance, the higher the weight.

[0043] Based on the spatial correlation of grid nodes, the distribution pattern of feature vectors is reconstructed. Specifically, this includes calculating the variance of each feature parameter in the entire feature space and selecting the top 30% of feature parameters by variance based on feature importance ranking as key risk factors.

[0044] The reconstructed feature vectors are normalized to generate a comprehensive risk assessment index that fully reflects the downhole risk situation.

[0045] As a further aspect of the present invention: the establishment of a dynamic risk warning threshold specifically includes:

[0046] Establish a risk level determination rule base, and divide the comprehensive risk assessment indicators into multiple risk levels according to their numerical ranges. Each level corresponds to a different early warning status and response strategy.

[0047] Extract historical evolution data of risk distribution maps, analyze the statistical distribution characteristics of risk indicators, and determine the benchmark thresholds for each risk level;

[0048] By combining real-time drilling condition changes, the risk warning threshold is dynamically adjusted in the digital mapping system.

[0049] Optimize threshold parameters based on the frequency and severity of recent risks.

[0050] As a further aspect of the present invention: the real-time adjustment of drilling control parameters based on the early warning results specifically includes:

[0051] Establish a mapping table between early warning levels and control parameters, and pre-set corresponding parameter adjustment schemes for different risk levels;

[0052] When a risk warning is triggered, a control parameter adjustment sequence is automatically generated based on the warning level. The system response after parameter adjustment is simulated through a digital mapping body to verify the effectiveness of the adjustment plan.

[0053] Balancing drilling efficiency and safety requirements, and optimizing the combination of control parameters, including the coordinated adjustment of drilling pressure, rotational speed and drilling fluid circulation rate, while ensuring that risks are controllable;

[0054] The optimized parameters are sent to the physical drilling system through the real-time control interface, while the changes in the adjusted risk status are monitored to form a closed-loop control.

[0055] The beneficial effects of this invention are:

[0056] (1) By constructing a digital mapping volume and a sensor network that integrates virtual and real data, the system can simultaneously collect dynamic parameters from multiple sources, such as mechanical vibration and thermal stress distribution. Combined with differential transformation and mode decomposition techniques, the system can analyze the drilling system status from multiple dimensions, including the time domain, frequency domain, and spatial domain. This multi-dimensional analysis method can effectively identify hidden risks that are difficult to detect by traditional single-parameter monitoring, thus improving the early warning capability for complex downhole conditions.

[0057] (2) Based on dynamic risk assessment indicators and adaptive early warning thresholds, the system can intelligently adjust the combination of drilling parameters according to real-time operating conditions. The effectiveness of the adjustment scheme is verified through virtual tests in the digital mapping body, and then the optimized parameters are sent to the physical equipment in real time through a closed-loop control mechanism. This intelligent control method reduces unnecessary downtime while ensuring operational safety, and improves the mechanical drilling rate through parameter co-optimization, thus achieving a balance between safety and drilling efficiency. Attached Figure Description

[0058] The invention will now be further described with reference to the accompanying drawings.

[0059] Figure 1 This is a flowchart of the system of the present invention. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] Please see Figure 1 As shown, this invention is a system for analyzing and predicting oil drilling and production data, comprising:

[0062] The data acquisition module is used to apply oil drilling excitation in a preset mode and simultaneously acquire multi-source dynamic response signals of the downhole system during the operation cycle, and construct a time-series feature set including drill string vortex frequency sequence and wellbore temperature field gradient distribution map.

[0063] The feature extraction module is used to perform differential transformation and mode decomposition on the time series feature set, extract transient response feature vectors that characterize the stability of the drilling system, and generate corresponding risk distribution maps;

[0064] The risk modeling module is used to establish a parameter-risk correlation model based on the risk distribution map. By decoupling mechanical vibration response and thermal stress interference, dynamic risk indicators reflecting the true risk status downhole are obtained.

[0065] The risk assessment module is used to construct a multi-dimensional feature space by combining dynamic risk indicators and drilling process parameters. It uses a feature fusion grid to aggregate information and reconstruct features of feature nodes to generate comprehensive risk assessment indicators.

[0066] The early warning and control module is used to generate risk level determination results based on comprehensive risk assessment indicators, analyze the temporal evolution characteristics of the risk distribution map, establish dynamic risk early warning thresholds, and adjust drilling control parameters in real time according to the early warning results.

[0067] In the data acquisition module, during oil drilling operations, a three-dimensional digital mapping of the drilling system is first constructed. This digital mapping is synchronized with the physical drilling system through a real-time data interface, receiving and integrating structural and operational parameters from the actual drilling site. Structural parameters include the drill string's geometry, connection methods, and material properties; operational parameters cover the current drilling depth, drill bit type, and formation lithology data. By reconstructing these parameters in a three-dimensional virtual environment, a high-precision digital mapping accurately reflects the real-time status of the physical drilling system.

[0068] Within the constructed digital mapping system, a dynamic virtual sensor network is deployed for key dynamic nodes. Specifically, eddy current monitoring points are set at each connection joint of the drill string to capture the lateral vibration characteristics during drill string rotation; simultaneously, temperature gradient monitoring arrays are deployed at certain intervals in the annulus region of the wellbore, forming a three-dimensional monitoring network for the wellbore temperature field. By solving the fluid-structure coupling equations describing the interaction between the drilling fluid and the drill string structure, continuous sampling data of the drill string eddy current frequency are obtained; at the same time, the temperature distribution at different locations within the wellbore is calculated, generating spatially continuous temperature field data.

[0069] The system integrates measured data from physical sensors actually installed on drilling rigs with simulated data generated by a virtual sensor network. A time-series alignment algorithm based on timestamp matching is employed to precisely synchronize the two data sources, eliminating time deviations caused by acquisition delays. For areas with insufficient physical sensor coverage, spatial interpolation algorithms are used to supplement virtual sensor data, ensuring the integrity of the monitoring range. The final time-series feature set comprises two main data sequences: one is a curve showing the change in drill string eddy frequency over time, recording the continuous evolution of the drill string's dynamic state; the other is a gradient distribution map of the wellbore temperature field, presenting the spatiotemporal distribution characteristics of temperature changes within the wellbore in the form of a thermogram. This time-series feature set provides a complete data foundation for subsequent feature extraction and risk analysis.

[0070] During data processing, missing data segments are supplemented using linear interpolation based on neighboring node data. Abnormal data points significantly exceeding the normal range are filtered and removed by setting reasonable data validity thresholds to ensure the reliability and accuracy of the feature set data. The entire data acquisition and processing process continues throughout the drilling operation cycle, enabling comprehensive monitoring and recording of the dynamic response of the drilling system.

[0071] In the feature extraction module, the drill string eddy frequency sequence in the time-series feature set undergoes differential transformation. The central difference method is used to calculate the frequency sequence point by point. First, the first derivative of each data point is calculated, representing the rate of change between adjacent data points. Then, based on the first derivative, the second derivative is calculated to characterize the acceleration feature of the rate of change. A sliding window of three data points is used to ensure the stability of the differential calculation. This differential processing effectively captures the instantaneous changes in the dynamic behavior of the drill string, providing fundamental data for subsequent analysis.

[0072] Multi-scale mode decomposition is performed on the differentiated sequence. An adaptive local mean decomposition method is employed, using an iterative selection process to decompose the signal into multiple intrinsic mode functions (EMFs). Each iterative step includes: identifying all extreme points of the signal, constructing upper and lower envelopes through interpolation, calculating the envelope mean, subtracting the envelope mean from the original signal to obtain candidate components, and repeating this process until the EMF criteria are met. For each decomposed EMF, its energy entropy is calculated by taking the logarithm of the ratio of the mode's energy to the total signal energy, which quantifies the contribution of each mode to the overall signal.

[0073] The spatial topology of the virtual sensor network in the digital mapping volume is coupled with the modal decomposition results for analysis. A three-dimensional wellbore grid coordinate system is established based on the deployment locations of the virtual sensors, mapping the energy value of each mode to the corresponding spatial grid node. A complete modal energy distribution field is constructed using a three-dimensional interpolation algorithm. Based on this, the energy distributions of different modes are superimposed to form a multi-dimensional modal feature space of the drill string dynamics state, which contains the multi-modal energy feature vector of each grid node.

[0074] Cluster analysis is performed using data points in the modal feature space. A density-based spatial clustering method is employed, with the neighborhood radius set to 1.5 times the grid spacing and a minimum number of points to be included set to 5. Spatial regions with similar modal features are identified by calculating the Euclidean distance between data points. For identified anomalous modal combinations—i.e., feature vectors that deviate from the main cluster centers—their spatial location and feature parameters are recorded. Based on the cluster analysis results, the subsequent decomposition effect is optimized by adjusting the number of iterations and the selection criteria parameters of the modal decomposition.

[0075] The intrinsic mode functions (EMFs) with energy entropy values ​​exceeding a preset threshold are selected from the modal decomposition results. This threshold is determined through analysis of historical data, typically taking the 70th percentile of the energy entropy values ​​across all modes. For each selected EMF, three key parameters are extracted: instantaneous amplitude obtained through Hilbert transform, instantaneous frequency calculated through phase differentiation, and instantaneous phase obtained from the phase angle of the analytic signal. These parameters are then combined in a time series to construct a multi-dimensional transient response feature vector.

[0076] The feature vectors are mapped to the spatial coordinates of the wellbore. A correspondence between the feature vectors and 3D grid nodes is established based on the deployment locations of the virtual sensors. For grid areas without direct monitoring data, an inverse distance-weighted interpolation algorithm is used for data filling, where the weight coefficients are the reciprocal of the square of the distance. By adjusting the search radius and minimum neighbor number of the interpolation algorithm, the distribution density of the feature vectors in the 3D grid is optimized to ensure the spatial continuity of the risk indicator field.

[0077] A risk level surface is constructed based on data points from a risk indicator field. A smooth surface is fitted using the moving least squares method, with the smoothness controlled by setting different fitting precisions. Risk values ​​are divided into five levels, represented by blue, green, yellow, orange, and red from low to high. During visualization, different transparency levels are set according to the risk level, with higher opacity for high-risk areas and lower opacity for low-risk areas. The resulting risk distribution map visually displays the risk status at different locations within the wellbore, providing support for drilling operation decisions.

[0078] In the risk modeling module, the spatial coordinates and corresponding risk level data of high-risk areas are extracted from the generated risk distribution map. High-risk areas are defined as continuous areas with a risk level of 4 or 5, and an area of ​​at least 1 square meter. Simultaneously, real-time drilling process parameters corresponding to these high-risk areas are collected, including instantaneous values ​​of drilling pressure, fluctuation ranges of drilling speed, and real-time measurements of drilling fluid density and viscosity. These parameters are synchronously recorded 10 times per second by the data acquisition system to ensure accurate correspondence over time.

[0079] A parameter-risk mapping table was constructed, using collected drilling process parameters as input variables and risk level data as output variables. Multiple regression analysis was employed to establish a quantitative correlation between parameter changes and risk levels. The specific process included: first, standardizing the input variables to eliminate the influence of dimensions; then, solving for the regression coefficients using the least squares method to establish a linear regression equation; and finally, verifying the goodness of fit of the regression equation by calculating the coefficient of determination. This quantitative correlation reflects the degree of influence of different combinations of drilling process parameters on risk levels.

[0080] In the digital mapping system, parameter sensitivity analysis nodes are set up at key locations in high-risk areas. Analysis is conducted by controlling for changes in single variables: keeping other process parameters constant, parameters such as drilling pressure, drilling speed, and drilling fluid properties are adjusted one by one, and the response characteristics of the risk distribution map are observed. The magnitude of the risk level change caused by each parameter change is recorded, and the influence weight of each process parameter on the risk status is calculated through normalization. The weight calculation uses the ratio of the change magnitude to the parameter change amount, and is normalized to ensure that the sum of all weights is 1.

[0081] Based on the calculated influence weights, the parameter-risk correlation is optimized. The regression coefficients are recalculated using weighted least squares, and the influence weights are incorporated as weighting factors into the regression analysis. The optimized parameter-risk correlation model more accurately reflects the dynamic correspondence between drilling parameters and risk status, providing a theoretical basis for subsequent risk assessment.

[0082] Physical equations for the propagation paths of mechanical vibration and the heat conduction paths were established separately. Mechanical vibration propagation was described using the wave equation, considering the elastic modulus and damping characteristics of the drill string; heat conduction was described using the Fourier heat conduction equation, considering the thermal conductivity and specific heat capacity of the drilling fluid. These two sets of partial differential equations were solved using the finite difference method to obtain the theoretical values ​​of the pure mechanical vibration response and the pure thermal stress distribution, respectively. A time step of 0.01 seconds and a spatial grid size of 0.1 meters were used in the solution process.

[0083] The actual collected composite response data was compared and analyzed with the theoretical calculation values. The composite response data includes the coupling effect of mechanical vibration and thermal stress. The contributions of the two physical effects were separated by residual calculation method: first, the theoretical value of pure mechanical vibration response was subtracted from the composite response data to obtain a preliminary estimate of the thermal stress component; then, the theoretical value of pure thermal stress distribution was subtracted from the composite response data to obtain a preliminary estimate of the mechanical vibration component. The residual was minimized through iterative calculation, and finally, the two components were effectively separated.

[0084] Energy integration was performed on the separated mechanical vibration response. The vibration acceleration signal was integrally squared over the time domain, with the integration interval covering the entire drilling cycle. The integration result was divided by the time span to obtain the vibration risk index. This index reflects the time average of the mechanical vibration energy. Simultaneously, gradient analysis was performed on the separated thermal stress distribution, calculating the temperature gradient vector magnitude at each point in space. The maximum gradient value across the entire monitoring area was taken as the thermal stress risk index.

[0085] The vibration risk index and the thermal stress risk index are weighted and fused. The weighting coefficients are determined based on the results of previous parameter sensitivity analysis, with mechanical vibration weighted at 0.6 and thermal stress weighted at 0.4. The weighting formula is: the dynamic risk index equals the vibration risk index multiplied by 0.6 plus the thermal stress risk index multiplied by 0.4. Through this weighted fusion process, a dynamic risk index that comprehensively reflects the actual downhole risk state is generated. This index is updated every 5 seconds to ensure timely reflection of changes in the downhole risk state.

[0086] In the risk assessment module, when establishing a three-dimensional spatial grid coordinate system, the origin is the center of the wellhead, the positive Z-axis is vertically downward, and the X and Y axes are established in the horizontal plane. Based on the actual geometric dimensions of the wellbore, the wellbore space is divided into regular grid cells, each with dimensions of 0.5m × 0.5m × 1m (length × width × height). The grid division considers the wellbore's tilt angle and azimuth variations to ensure that the grid cells completely cover the entire wellbore space. Each grid cell serves as a basic dimension of the feature space, recording the characteristic parameters of that spatial location.

[0087] Dynamic risk indicators are mapped to a grid coordinate system according to their spatial location. The grid cell to which each risk indicator belongs is determined based on the spatial coordinates used in its calculation. For multiple risk indicators within the same grid cell, the arithmetic mean is taken as the representative value for that cell. Simultaneously, drilling process parameters are allocated to corresponding grid nodes according to their range of action and radius of influence. The specific allocation principles are as follows: drilling pressure parameters primarily affect grid cells near the drill bit and are allocated according to a distance decay function; rotational speed parameters affect grid cells traversed by the entire drill string; drilling fluid performance parameters affect grid cells in the wellbore annulus region based on the circulation path. This mapping method ensures that each grid node contains complete characteristic information.

[0088] Data filling is performed on grid areas not directly covered. An inverse distance weighted interpolation algorithm is used, searching for neighboring known data points within a spherical region with a radius of 2 meters, centered on the node to be filled. The weight of each known data point is inversely proportional to the square of its distance to the node to be filled. When the number of known data points in the search area is less than 3, the search radius is expanded to 4 meters. This interpolation method ensures the continuity and integrity of the feature space in the 3D wellbore environment, eliminating data gaps.

[0089] A multidimensional feature space was constructed through parameter density analysis and spatial distribution verification of grid nodes. Parameter density analysis employed a kernel density estimation algorithm, calculating the distribution density of feature parameters within a 1-meter radius of each grid node. Spatial distribution verification checked the spatial continuity of feature parameters, ensuring that the feature values ​​of adjacent grid nodes changed smoothly without abrupt changes. For regions with abnormal density or discontinuous distribution, data mapping and filling were performed again until the spatial continuity requirements were met.

[0090] Based on the constructed multidimensional feature space, a hierarchical structure for the feature fusion mesh is designed. The coarse mesh is set to a size of 2m × 2m × 4m, with each coarse mesh node governing 64 basic mesh cells, used to capture global features. The fine mesh retains its original size of 0.5m × 0.5m × 1m, used to extract local detailed features. The two mesh levels are linked through spatial inclusion relationships to ensure the transferability of feature information across different scales.

[0091] Spatial aggregation of feature data is achieved through information transmission and interaction between adjacent grid nodes. Within the area governed by each coarse grid node, a weighted average algorithm is used to calculate the representative value of the feature. The weights are determined based on the distance from the grid node to the center point of the coarse grid; the closer the distance, the higher the weight. Specifically, the feature value of each basic grid node is first multiplied by its corresponding weight coefficient, then summed and divided by the sum of the weight coefficients to obtain the representative value of the coarse grid node. This aggregation method preserves the main information of the regional features while reducing data dimensionality.

[0092] Based on the spatial correlation of grid nodes, the distribution pattern of feature vectors is reconstructed. First, the variance of each feature parameter in the entire feature space is calculated; a larger variance indicates stronger spatial variability of the feature parameter. Then, by ranking the features by importance, the top 30% of feature parameters with the highest variance are selected as key risk factors. For the selected key risk factors, their spatial distribution patterns are analyzed to identify high-risk clustering areas and abnormal distribution patterns.

[0093] The reconstructed feature vectors are normalized to generate a comprehensive risk assessment index. The normalization process uses a min-max standardization method, linearly transforming the value of each feature parameter to the range of 0-1. Then, based on the feature importance ranking, weights are assigned to each normalized feature parameter, with higher importance resulting in greater weights. Finally, the weighted feature parameter values ​​are summed to obtain the comprehensive risk assessment index within the range of 0-100. This index is updated every 10 seconds to ensure timely reflection of changes in downhole risk conditions.

[0094] In the early warning and control module, when establishing the risk level determination rule base, the numerical range of the comprehensive risk assessment indicators is divided into 5 levels. The specific classification criteria are: 0-20 is Level 1 (Safe), 21-40 is Level 2 (Caution), 41-60 is Level 3 (Warning), 61-80 is Level 4 (Danger), and 81-100 is Level 5 (High Risk). Each level corresponds to a different early warning status and response strategy: Level 1 requires only routine monitoring, Level 2 requires enhanced observation, Level 3 requires initiating preventative measures, Level 4 requires implementing control measures, and Level 5 requires immediate cessation of operations. These rules are based on historical accident data and expert experience and are continuously updated according to actual operational conditions.

[0095] When analyzing historical evolution data of the risk distribution map, a sliding window technique is used to obtain the risk indicator sequence over the most recent 24 hours. A quantile regression algorithm is then used to calculate the baseline threshold for each risk level, specifically taking the 95th quantile of the risk indicator value corresponding to each level as the initial threshold. For example, for a Level 3 alert status, the 95th quantile of all historically identified Level 3 risk indicator values ​​is used as the baseline threshold for that level. This method ensures that the threshold setting considers both normal fluctuations and effectively identifies abnormal situations.

[0096] By combining real-time drilling condition trends, risk warning thresholds are dynamically adjusted within the digital mapping system. When a significant change in drilling conditions is detected, such as encountering different formations or changing the drilling method, the thresholds for each level are recalculated based on the risk indicator data from the most recent four hours. The adjustment range is determined according to the degree of change in conditions, and an exponentially weighted moving average method is used for a smooth transition to avoid abrupt threshold changes. The impact of environmental factors such as diurnal temperature variations and seasonal changes on the thresholds is also considered to ensure the accuracy of the warnings.

[0097] The threshold parameters are optimized based on the recent frequency and severity of risks. The frequency and duration of each risk level over the past 7 days are statistically analyzed. For risk levels occurring more than 10 times per day on average, the threshold is appropriately increased by 5%; for risk levels that have not occurred for 3 consecutive days, the threshold is appropriately decreased by 3%. Simultaneously, feedback on the effectiveness of risk event handling is used to fine-tune the thresholds, forming a closed-loop mechanism for continuous optimization.

[0098] A mapping table between early warning levels and control parameters was established. This table clearly specifies the adjustment range and magnitude of each control parameter under different early warning levels. For example, when a Level 4 early warning is triggered, the drilling pressure adjustment range is 80%–90% of the current value, the drilling speed adjustment range is 85%–95% of the current value, and the drilling fluid circulation rate adjustment range is 105%–115% of the current value. These mapping relationships were determined based on a large amount of field test data and numerical simulation results, ensuring the scientific validity and feasibility of the adjustment plan.

[0099] When a risk warning is triggered, a control parameter adjustment sequence is automatically generated based on the warning level. First, a baseline adjustment plan is retrieved from the mapping table, and then modified according to the current specific operating conditions. After generating the adjustment sequence, a virtual test is conducted in the digital mapping system to simulate the drilling system's response over the next 30 minutes. The effectiveness of the adjustment plan is verified by comparing the simulation results with safety standards. If the simulation results show that the risk index reduction does not reach the expected target of 70%, the adjustment plan is regenerated.

[0100] Balancing drilling efficiency and safety requirements, the optimal combination of control parameters is employed. A multi-objective optimization method is used, simultaneously considering three objectives: mechanical drilling rate, cost control, and risk control. By establishing an objective function, the optimal parameter combination is found while ensuring controllable risk. The optimization process focuses on addressing the coupling relationship between drilling pressure, drilling speed, and drilling fluid circulation rate, ensuring the synergy of parameter adjustments. For example, appropriately increasing the drilling speed while reducing drilling pressure can reduce risk levels while maintaining drilling efficiency.

[0101] The optimized parameters are transmitted to the physical drilling system via a real-time control interface. A dual verification mechanism is employed for parameter transmission to ensure accuracy and integrity. The transmission process consists of three stages: first, a preparatory command is sent to confirm equipment readiness; then, parameter data is transmitted for numerical verification; and finally, parameter adjustments are executed, with real-time monitoring of equipment response. Throughout the adjustment process, changes in risk status are continuously monitored, with risk indicator data collected every 5 seconds. If the risk level does not decrease to the target level within 30 minutes of adjustment, a secondary adjustment procedure is automatically initiated, forming a complete closed-loop control system.

[0102] The working principle of this invention is as follows: Virtual reconstruction of the physical drilling system is achieved by constructing a digital mapping body. First, a multi-source sensor network is used to collect drill string eddy frequency and wellbore temperature field gradient data, which are then time-series aligned and fused to form a time-series feature set. Subsequently, the feature set undergoes differential transformation and mode decomposition to extract transient response feature vectors and generate a risk distribution map. Based on the risk distribution, a parameter-risk correlation model is established. By solving physical equations, mechanical vibration and thermal stress interference are decoupled to obtain dynamic risk indicators. Next, the risk indicators and process parameters are mapped to a three-dimensional mesh to construct a multi-dimensional feature space. A hierarchical feature fusion mesh is used to achieve information aggregation and reconstruction, generating a comprehensive risk assessment index. Finally, risk levels are classified according to the assessment index. Through dynamic early warning thresholds and closed-loop control mechanisms, drill pressure, rotational speed, and drilling fluid parameters are optimized in real time to achieve risk early warning and adaptive control of the drilling process.

[0103] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A system for analyzing and predicting oil drilling and production data, characterized in that, include: The data acquisition module is used to apply oil drilling excitation in a preset mode and simultaneously acquire multi-source dynamic response signals of the downhole system during the operation cycle, and construct a time-series feature set including drill string vortex frequency sequence and wellbore temperature field gradient distribution map. The construction of the time-series feature set, which includes the drill string eddy frequency sequence and the wellbore temperature field gradient distribution map, specifically includes: Construct a dynamic mapping of a three-dimensional drilling system, and synchronize the structural parameters and operating parameters of the physical drilling system through a real-time data interface to establish a digital mapping body that includes drill string geometry, wellbore structural dimensions and formation physical property parameters. A dynamic virtual sensor network is deployed at key dynamic nodes of the digital mapping body, including setting up eddy monitoring points at drill string connections and setting up temperature gradient monitoring arrays in the wellbore annulus. High-frequency sampling data of drill string eddy frequency and spatial distribution data of wellbore temperature field are obtained by solving fluid-structure coupling equations. The actual data collected by physical sensors and the simulation data generated by virtual sensors are fused from multiple sources. The two types of data sources are processed synchronously to construct a comprehensive feature set, which includes the time-series variation curve of drill string eddy frequency and the gradient distribution map of wellbore temperature field. This is denoted as the time-series feature set. The feature extraction module is used to perform differential transformation and mode decomposition on the time series feature set, extract transient response feature vectors that characterize the stability of the drilling system, and generate corresponding risk distribution maps; The risk modeling module is used to establish a parameter-risk correlation model based on the risk distribution map. By decoupling mechanical vibration response and thermal stress interference, dynamic risk indicators reflecting the true risk status downhole are obtained. The risk assessment module is used to construct a multi-dimensional feature space by combining dynamic risk indicators and drilling process parameters. It uses a feature fusion grid to aggregate information and reconstruct features of feature nodes to generate comprehensive risk assessment indicators. The early warning and control module is used to generate risk level determination results based on comprehensive risk assessment indicators, analyze the temporal evolution characteristics of the risk distribution map, establish dynamic risk early warning thresholds, and adjust drilling control parameters in real time according to the early warning results.

2. The oil drilling and production data analysis and prediction system according to claim 1, characterized in that, The differential transformation and mode decomposition of the time-series feature set specifically include: Point-by-point difference calculations were performed on the drill string eddy frequency sequence to obtain the first and second order differential sequences of the eddy frequency, thereby capturing the instantaneous change characteristics of the drill string's dynamic behavior. Multi-scale mode decomposition is performed on the sequence after differential processing. The intrinsic mode functions characterizing different physical mechanisms are separated through an iterative screening process, and the energy entropy value of each mode is calculated to quantify the mode contribution. The spatial topology of the virtual sensor network in the digital mapping volume is coupled with the modal decomposition results. By projecting the modal energy into the three-dimensional wellbore mesh, a multi-dimensional modal feature space of the drill string dynamic state is constructed. Cluster analysis is performed using data points in the modal feature space to identify abnormal modal combination patterns and update the modal decomposition parameters.

3. The oil drilling and production data analysis and prediction system according to claim 1, characterized in that, The generation of the corresponding risk distribution map specifically includes: From the modal decomposition results, intrinsic mode functions with energy entropy values ​​exceeding a threshold are selected, and their instantaneous amplitude, frequency, and phase parameters are extracted to construct a multi-dimensional transient response feature vector characterizing the stability of the drilling system. The feature vectors are correlated and mapped with the spatial coordinates of the wellbore to generate a continuous risk index field, and the distribution density of the feature vectors in the three-dimensional grid is optimized. Based on the data points of the risk indicator field, a risk level surface is constructed, and different risk areas are visualized through color coding and transparency rendering to form a risk distribution map.

4. The oil drilling and production data analysis and prediction system according to claim 1, characterized in that, The establishment of the parameter-risk correlation model based on the risk distribution map specifically includes: Extract the coordinates of high-risk areas and their corresponding risk levels from the risk distribution map, and simultaneously collect the drilling process parameters corresponding to the high-risk areas, including drilling pressure, rotation speed and drilling fluid performance parameters. A parameter-risk mapping table was constructed, with drilling process parameters as input variables and risk level data as output variables. A quantitative correlation between parameter changes and risk levels was established through multiple regression analysis. In the digital mapping body, a parameter sensitivity analysis node is set up. By controlling the change of a single variable, the response characteristics of the risk distribution map are observed, and the influence weight of each process parameter on the risk status is determined. Based on the influence weight optimization parameter-risk correlation, a parameter-risk correlation model that can reflect the dynamic correspondence between drilling parameters and risk status is established.

5. The oil drilling and production data analysis and prediction system according to claim 1, characterized in that, The process of decoupling mechanical vibration response and thermal stress interference to obtain dynamic risk indicators reflecting the true downhole risk state specifically includes: Physical equations for the propagation path of mechanical vibration and the path of heat conduction are established separately. The theoretical values ​​of the pure mechanical vibration response and the pure thermal stress distribution are obtained by solving the partial differential equation system. The actual collected composite response data is compared and analyzed with the theoretical values, and the mechanical vibration component and thermal stress component are separated by residual calculation method; The energy of the separated mechanical vibration response is integrated to calculate the vibration risk index, and the gradient analysis of the thermal stress distribution is performed to calculate the thermal stress risk index. By weighting and fusing the vibration risk index and the thermal stress risk index, a dynamic risk index that reflects the true risk state downhole is generated.

6. The oil drilling and production data analysis and prediction system according to claim 1, characterized in that, The construction of a multi-dimensional feature space by combining dynamic risk indicators and drilling process parameters specifically includes: A three-dimensional spatial grid coordinate system is established, and the wellbore space is divided into several grid units according to geometric features. Each grid unit serves as a basic dimension of the feature space. Dynamic risk indicators are mapped to the grid coordinate system according to their corresponding spatial locations, while drilling process parameters are allocated to the corresponding grid nodes according to their range of action and radius of influence. Data is filled into the grid areas that are not directly covered to ensure the continuity and integrity of the feature space in the 3D wellbore environment; Through parameter density analysis and spatial distribution verification of grid nodes, a multidimensional feature space containing dynamic risk indicators and drilling process parameters was constructed.

7. The oil drilling and production data analysis and prediction system according to claim 1, characterized in that, The method of using a feature fusion grid to aggregate information and reconstruct features from feature nodes to generate a comprehensive risk assessment index specifically includes: Based on the constructed multidimensional feature space, a hierarchical structure of the feature fusion mesh is designed, in which the size of the coarse mesh is set to 2m×2m×4m, and each coarse mesh node governs 64 basic mesh cells to capture global features; The fine grid retains its original size of 0.5m × 0.5m × 1m and is used to extract local detail features. Through information transmission and interaction between adjacent grid nodes, a weighted average algorithm is used to calculate the feature representative value of the area under the jurisdiction of each coarse grid node. The weight is determined according to the distance of the grid node to the center point of the coarse grid, with higher weights for closer nodes. Based on the spatial correlation of grid nodes, the distribution pattern of feature vectors is reconstructed. Specifically, this includes calculating the variance of each feature parameter in the entire feature space and selecting the top 30% of feature parameters with the highest variance as key risk factors by ranking features by importance. The reconstructed feature vectors are normalized to generate a comprehensive risk assessment index that fully reflects the downhole risk situation.

8. The oil drilling and production data analysis and prediction system according to claim 1, characterized in that, The establishment of dynamic risk early warning thresholds specifically includes: Establish a risk level determination rule base, and divide the comprehensive risk assessment indicators into multiple risk levels according to their numerical ranges. Each level corresponds to a different early warning status and response strategy. Extract historical evolution data of risk distribution maps, analyze the statistical distribution characteristics of risk indicators, and determine the benchmark thresholds for each risk level; By combining real-time drilling condition changes, the risk warning threshold is dynamically adjusted in the digital mapping system. Optimize threshold parameters based on the frequency and severity of recent risks.

9. The oil drilling and production data analysis and prediction system according to claim 1, characterized in that, The real-time adjustment of drilling control parameters based on early warning results specifically includes: Establish a mapping table between early warning levels and control parameters, and pre-set corresponding parameter adjustment schemes for different risk levels; When a risk warning is triggered, a control parameter adjustment sequence is automatically generated based on the warning level. The system response after parameter adjustment is simulated through a digital mapping body to verify the effectiveness of the adjustment plan. Balancing drilling efficiency and safety requirements, and optimizing the combination of control parameters, including the coordinated adjustment of drilling pressure, rotational speed and drilling fluid circulation rate, while ensuring that risks are controllable; The optimized parameters are sent to the physical drilling system through a real-time control interface, while the changes in the adjusted risk status are monitored to form a closed-loop control.

Images