A drilling complex working condition early warning and perception method based on space-time data fusion

Abstract

The present application relates to the technical field of petroleum and natural gas drilling complex working condition early warning, and particularly relates to a drilling complex working condition early warning and sensing method based on space-time data fusion, comprising setting a working condition set covering different drilling complex working condition types; constructing a space risk probability prediction module; constructing a space risk indication sample set based on known drilling observation point coordinates and indication variables; obtaining the coordinates of a well to be measured, calculating the risk prior probability of the coordinates of the well to be measured by using a three-dimensional Kriging model; and / or constructing a real-time working condition sensing early warning module, outputting a trend description of the working condition. The present application solves the problem that the existing technology cannot realize the unification of prospective prediction and instant early warning of drilling risks.

Description

Technical Field

[0001] This invention relates to the field of early warning technology for complex working conditions in oil and gas drilling, and in particular to an early warning and perception method for complex working conditions in drilling based on spatiotemporal data fusion. Background Technology

[0002] Drilling is one of the most expensive and risky processes in oil and gas exploration and development. During the drilling process, due to the complex and variable geological conditions, it is easy for complex working conditions and accidents such as well leakage, overflow, stuck drill bit, and drill string leakage to occur, which can not only cause huge economic losses, but may also lead to serious safety and environmental accidents.

[0003] Currently, the main technologies for monitoring and early warning of drilling risks include the following: 1. Macro-prediction based on static geological data: In the pre-drilling design stage, macro-regional predictions of the risks that new wells may encounter are usually made based on data from adjacent wells and regional geological understanding. This method provides a reference for long-term planning, but its prediction results are static, have low resolution, and cannot effectively monitor the immediate risks caused by dynamic changes in engineering parameters during the drilling process.

[0004] 2. Traditional alarm systems based on a single parameter threshold typically set fixed high and low thresholds for engineering parameters in the field instruments and software systems. Once the parameter value exceeds the limit, an alarm is triggered. This method has a rigid and single-dimensional perception logic, completely ignoring the fact that most complex operating conditions are evolved by the coordinated interaction of multiple parameters. It is very easy to generate a large number of false alarms and fatal missed alarms that fail to be triggered at critical moments. 3. In recent years, some studies have begun to apply machine learning or deep learning models to analyze drilling data. These models can learn complex nonlinear relationships from massive amounts of historical data for working condition identification or anomaly detection. However, this method lacks interpretability in practical engineering applications and is difficult to guide the next step of operation. Moreover, due to differences in geological and engineering conditions, the model is often difficult to apply directly to another block.

[0005] In summary, existing technologies lack a comprehensive solution that can organically combine static, macroscopic geological spatial risk prediction with dynamic, microscopic engineering parameter real-time perception, thus failing to achieve a unified approach to forward-looking prediction and immediate early warning of drilling risks. Summary of the Invention

[0006] To address the shortcomings of existing methods, this invention integrates geological spatial data and real-time drilling time series data to achieve real-time accurate perception, thereby significantly improving the accuracy and timeliness of risk warning and ensuring the safety and efficiency of drilling operations.

[0007] The technical solution adopted in this invention is: a method for early warning and perception of complex drilling conditions based on spatiotemporal data fusion, comprising the following steps: Set up a set of operating conditions that cover different types of complex drilling conditions; As a preferred embodiment of the present invention, the operating conditions include: well leakage, overflow, stuck drill bit, and drill string leakage.

[0008] A spatial risk probability prediction module is constructed, including: constructing a spatial risk indicator sample set based on the coordinates of observation points and indicator variables of known wells; and calculating the prior risk probability of the well to be logged using a three-dimensional kriging model based on the spatial risk indicator sample set and the coordinates of the well to be logged. and / or Set up a set of working conditions covering different types of complex drilling conditions, build a real-time working condition perception and early warning module, and output a trend description of the working conditions.

[0009] As a preferred embodiment of the present invention, the three-dimensional kriging model includes: Calculate the risk probability of the target point ; As a working condition indicator variable, These are the weighting coefficients; The weighting coefficients satisfy the unbiased constraint and the Kriging equations that minimize the estimated variance. , It is the spatial variability function. For Lagrange multipliers, Given the spatial coordinates of the well sample points; These are the spatial coordinates of the well sample points to be logged.

[0010] In a preferred embodiment of the present invention, when the prior probability of risk exceeds the risk threshold, the risk trajectory point is output.

[0011] As a preferred embodiment of the present invention, a continuous risk probability profile of the well to be measured is constructed using the prior risk probability and the well depth.

[0012] In a preferred embodiment of the present invention, the real-time operating condition perception and early warning module includes: Real-time acquisition of engineering parameter data for the well to be logged; As a preferred embodiment of the present invention, the engineering parameter data include: mud pit volume, outlet flow rate, riser pressure, total hydrocarbon value, turntable speed, turntable torque, hook load, pump flushing, and outlet temperature.

[0013] Calculate the slope using the change sequence of engineering parameters and time index ; Based on the range of slope values, trends are defined, and an expert knowledge rule base is constructed based on engineering parameters, trends, and trend weights. At any point in time t， The collected engineering parameters are compared with the rule set of the expert knowledge rule base to obtain the matching indicator function. ; Calculation of working condition type t Risk matching score ; Trend weight.

[0014] In a preferred embodiment of the present invention, when the risk matching degree value exceeds the trigger threshold, a trend description is output.

[0015] As a preferred embodiment of the present invention, a drilling complex condition early warning and perception system based on spatiotemporal data fusion includes: a memory for storing instructions executable by a processor; and a processor for executing the instructions to implement the drilling complex condition early warning and perception method based on spatiotemporal data fusion.

[0016] As a preferred embodiment of the present invention, a computer-readable medium storing computer program code implements a method for early warning and perception of complex drilling conditions based on spatiotemporal data fusion when executed by a processor.

[0017] The beneficial effects of this invention are: 1. This invention combines pre-drilling geological spatial probability prediction with real-time trend diagnosis during drilling, breaking down the technical barriers between long-term planning and instantaneous early warning; through spatial probability profiles, drilling personnel can avoid high-risk areas in advance; using real-time dynamic diagnosis, the system can achieve 24 / 7 uninterrupted risk precursor capture. 2. This invention employs a multi-parameter dynamic trend fusion algorithm, introducing feature importance weights. It elevates the diagnostic dimension from instantaneous parameter values ​​to a multidimensional evolutionary feature space; through comprehensive judgment of quantified risk matching degree, it effectively eliminates redundant signals caused by the random fluctuations of a single sensor, and significantly reduces the false alarm rate and false negative rate of the system. 3. This invention is based on an expert rule base for diagnosis and can directly generate structured descriptions of abnormal operating conditions through semantic mapping. This transparent diagnostic logic ensures that on-site engineers can understand the causes of risks in a timely manner, providing a clear basis for efficient decision-making. 4. By introducing a dynamic suppression mechanism for early warning, the system effectively filters redundant alarms under the same continuous risk event, solving the long-standing problem of "alarm fatigue" on site; combined with incremental data processing and adaptive anomaly elimination algorithm, the system achieves the optimal balance between computational efficiency and early warning quality, and has excellent industrial application value and economic benefits. Attached Figure Description

[0018] Figure 1This is a flowchart of the early warning and perception method for complex drilling conditions based on spatiotemporal data fusion according to the present invention. Figure 2 This is a schematic diagram illustrating the implementation effect of the spatial risk probability prediction module of the present invention; Figure 3 This invention relates to a complex operating condition early warning record and its corresponding qualitative trend description. Detailed Implementation

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments. The drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0020] like Figure 1 As shown, a method for early warning and perception of complex drilling conditions based on spatiotemporal data fusion includes the following steps: This invention focuses on four common and highly dangerous complex conditions in drilling operations: well leakage, overflow, stuck pipe, and drill string leakage. This invention innovatively integrates risk probability prediction in the geological spatial dimension with real-time operational trend perception in the engineering time dimension; it is achieved by executing the following two interrelated and complementary technical modules.

[0021] The preset risk condition type set is ;in, These correspond to well leakage, overflow, stuck pipe, and drill string leakage conditions, and can be used to address these issues. The prediction and monitoring process can be executed independently for any operating condition type.

[0022] Module 1: Spatial Risk Probability Prediction Module Based on Geostatistics; The core function of this module is to provide a forward-looking, quantitative risk probability profile of the wellbore trajectory before or during drilling operations. This is achieved by following these technical steps: 1. Construct a sample database of geological spatial risk indicators; First, extract and integrate multi-source data from the historical database of the designated block; multi-source data includes, but is not limited to: the actual wellbore trajectories of all known wells. It also records well section information for various working conditions throughout history; and maps all well trajectory points to a unified three-dimensional geographic coordinate system. Define complex operating condition indicator variables For complex working condition type R, if the coordinate points If the well section is located within a known complex working condition, then (Risk sample); otherwise, (Safety Sample); Using this indicator variable, construct a spatial risk indicator sample set containing N observation points. ;in, Represents the spatial coordinates of the sample points; This step transforms discrete engineering risk events into quantified spatial coordinate sequences, enabling a digital representation of the distribution patterns of complex working conditions in the block, and providing a high-dimensional, refined sample foundation for subsequent geostatistical modeling.

[0023] 2. Generate a set of target trajectory points to be predicted; This step obtains the designed wellbore trajectory data of the well to be logged, and maps each trajectory point to a unified three-dimensional geographic coordinate system, forming a precise sample point set for risk assessment. Where M is the total number of samples of the target trajectory; this set of points constitutes the basic coordinate space for risk assessment.

[0024] 3. Spatial probability interpolation calculation based on the Kriging model; This is the core technology of this module. For each predefined complex working condition type R, a three-dimensional ordinary kriging geostatistical model is established using the generated indicator sample library as input data.

[0025] For each type of complex working condition, the spatial variability function is calculated using the sample set S. Quantitatively describe the autocorrelation of risk in space; targeting specific points. The Kriging interpolation algorithm is used to perform unbiased optimal estimation and calculate the risk probability of the target point. Its mathematical model is defined as a linear weighted sum of the observations: ; Among them, the weighting coefficient Satisfying unbiasedness constraints And it satisfies the Kriging equations that minimize the estimated variance:

[0026] In the formula, For Lagrange multipliers, This represents the spatial variability function value between sample points.

[0027] The optimal weights are obtained by solving this system of linear equations. Thus, the target point is obtained. Prior probability of risk at the location ; This step involves analyzing the spatial variability function of sample points to quantitatively learn the spatial autocorrelation and distribution patterns of this type of risk in a three-dimensional geological body.

[0028] 4. Generate risk profiles and store high-risk warnings; Will With the target point By correlating information such as well depth and coordinates, a continuous risk probability profile along the trajectory of the wellbore to be drilled is generated. ;in, For the first j Well depth at each sampling point; automatic selection to meet the requirements. ;in, For trajectory points with preset high-risk thresholds, their spatial coordinates, well depths, and predicted probabilities are extracted and stored in a database to form a high-risk early warning dataset. This data provides a forward-looking quantitative decision-making basis for drilling design optimization and construction early warning.

[0029] Module 2: Real-time Operating Condition Perception and Early Warning Module Based on Dynamic Trend Analysis; The core function of this module is to provide real-time and interpretable risk perception and early warning during drilling operations. Its technical advantage lies in capturing risk precursors by analyzing the dynamic evolution trends of multiple parameters rather than static instantaneous values. The implementation method follows these technical steps: 1. Incremental time-series data stream construction; To achieve low-latency real-time processing, this module employs a timestamp-based breakpoint resume mechanism. The system persistently records the latest time point of the processed data for each well. , This refers to the calendar date corresponding to the last logging data record in the previous processing batch where the system successfully completed risk analysis and model prediction. With the last processing date ( Correspondingly, this refers to the precise time point of the last logging data record successfully processed by the system in the previous processing cycle; during module operation, only incremental data is retrieved from the data source that meets the requirements. The new data stream, This refers to the sampling timestamp corresponding to each data record in the logging data sequence; and automatically piggybacks a length of... The system uses a context window to ensure seamless operation of the sliding window at the beginning of the new data stream. For historical data being processed for the first time, the system automatically switches to a block processing mode, dividing the full historical data into multiple data blocks for iterative processing to ensure the processing capacity for massive amounts of data and the stability of memory usage.

[0030] 2. Data stream preprocessing and cleaning; The acquired raw data stream is cleaned in real time to ensure the accuracy of subsequent analysis; a dynamic sliding window-based approach is employed. The criteria for outlier removal involve dynamically calculating the mean and standard deviation of the local neighborhood for each data point, thus effectively adapting to normal parameter fluctuations under different geological formations and operating conditions; for engineering parameter sequences... any data point in In length of L Calculate its mean within a local window with standard deviation : ； ； If satisfied Then determine Outliers are identified and removed. Then, a forward filling algorithm is used to interpolate and correct outliers or missing values ​​to ensure the continuity and physical consistency of the data sequence on the time axis.

[0031] 3. Quantitative extraction of dynamic trend characteristics of drilling parameters; This is the core engineering step of this module. For the time series of each key engineering parameter after preprocessing, such as mud pit volume, outlet flow rate, riser pressure, etc. (as shown in Table 1), a sliding window analysis is applied. Within each window, the slope of the data series is calculated using a linear regression algorithm. This slope accurately quantifies the direction and rate of change of the parameter within that time window. The least squares method is used for linear regression fitting to calculate the parameter change sequence. Time Index slope : ; The slope The system precisely quantifies the direction and rate of change of parameters within the window, and uses a preset stability threshold. With rapidly changing thresholds , slope Discretize into trend classification code C; when When C is 2 (increasing rapidly), When C is 1 (increasing slowly); when When, C is 0 (unchanged); when When C is -1 (decreasing slowly); when When C is -2 (decreasing rapidly), the original time series data is transformed into feature data that contains engineering significance and can be directly used by the system through this mapping.

[0032] 4. Risk pattern matching based on a weighted rule base; Intelligent perception is achieved through a built-in configurable expert knowledge rule base, which defines various complex operating conditions as combinations of multi-parameter trends. For a specific complex operating condition type R, its diagnostic rule set consists of... n Rules of the strip Composition, each sub-rule This corresponds to the trend requirement of an engineering parameter within a specific time window; and is assigned an importance weight coefficient. The system sets the weights of all sub-rules to satisfy the normalization condition, that is... This weighting coefficient quantitatively characterizes the sub-rules. Contribution to perceived risk R; at any point in time. t The diagnostic engine will extract the trend codes and rule sets of various parameters in real time. Perform a one-by-one comparison and define a matching indicator function. If the time point t The working conditions satisfy the sub-rules but ,otherwise Subsequently, the risk R at time point is calculated. t Risk matching degree : ; In this formula, The range of its value is [0, 1]; Used to determine whether the current operating condition meets the sub-rule. ; By weighted summation of the matched rules, the similarity between the current real-time operating condition and the standard risk model R is directly and quantitatively characterized. The closer the value is to 1, the higher the certainty that the risk will occur in the operating condition.

[0033] 5. Intelligent early warning generation and dynamic suppression; when Exceeding the preset trigger threshold That is, satisfying When this occurs, the system automatically generates an alert; this alert is highly interpretable, and its content includes not only the risk type R and the matching degree. (which can be converted into risk probability), and also clearly lists all factors that make it possible to... The sub-rules describe the trends; for example, "Pump flow drops rapidly; outlet flow increases slowly," providing clear decision-making support for field engineers.

[0034] Before the warning data is stored in the database, the diagnostic engine uses the current measured depth. Based on this, search the historical early warning database for in-depth records of the same type of risk R. If the discrimination condition is met: ; in, The system uses a preset deep suppression window. If the current warning is determined to be within the coverage of recorded risk events, the suppression logic is triggered, which terminates the storage operation of this warning. This mechanism ensures that the system only pushes warnings at the initial stage of a risk event, effectively filtering out redundant interference and ensuring the high effectiveness and criticality of risk warnings in decision-making.

[0035] Ultimately, all structured early warning records that pass the suppression verification are officially stored in the risk early warning database, completing the real-time perception and closed-loop early warning of complex drilling conditions.

[0036] Using "Well XX-1" in a certain block as the target well, this demonstration shows how the system performs pre-drilling spatial risk probability prediction and real-time drilling condition diagnosis for that well; focusing on the control of four complex conditions: well leakage, overflow, stuck pipe, and drill string leakage; the overall physical architecture of the system is as follows. Figure 1 As shown, it includes a data input layer, a dual-core processing engine layer, and an output layer.

[0037] Example 1: During the pre-drilling planning phase of "Well XX-1", the system invokes the spatial risk probability prediction module based on geostatistics. Taking the prediction of "well leakage" risk as an example, the specific execution process is as follows: 1. Sample library construction: The system automatically extracts the 3D drilling trajectories and complex operating condition records of adjacent wells in the block from the historical database. It iterates through all trajectory coordinates; if a coordinate point has experienced well leakage in the historical records, a well leakage indicator variable is defined for that point. Conversely, it is set to 0. This constructs a three-dimensional spatial risk indicator sample set containing tens of thousands of discrete points. .

[0038] 2. Target point mapping: Input the designed well trajectory of "XX-1 well", discretize it in 30-meter increments to generate a set of spatial sampling points for the target to be predicted. .

[0039] 3. Solving and interpolating the Kriging equation: For the sample set The system fits a spherical variogram model to calculate spatial autocorrelation. For any target point on the designed trajectory The system automatically constructs a system of Kriging linear equations:

[0040] The optimal weights are obtained by matrix inversion. And calculate the target point Prior probability of well leakage .

[0041] 4. Profile generation and early warning: The system sets a high-risk threshold. Extract trajectory points with a calculated probability greater than or equal to 0.8. For example... Figure 2 As shown in the implementation effect diagram, the system marks the high-probability points of "XX-1 well" with red dots in the three-dimensional visualization interface, automatically outputs a quantitative risk probability profile, and prompts the drilling personnel that there is an extremely high probability of well leakage in this depth.

[0042] Example 2: When well XX-1 enters the actual drilling phase, the real-time operating condition diagnosis and early warning module is activated, processing the sensor data stream at a frequency of 0.2Hz. The following uses the diagnosis of a complex condition such as overflow as an example to explain the dynamic configuration in detail: 1. Incremental fetching and cleaning: Read the latest timestamp Real-time parameters (such as mud pit volume, outlet flow rate, total hydrocarbon value, etc.); setting the length of the local sliding window. Seconds; if the outlet flow rate suddenly reaches a minimum value due to sensor jitter. The system calculates the local mean in real time. and standard deviation ,determination The data is identified as abnormal noise and removed, and then padded forward with the previous valid value to ensure a smooth data flow.

[0043] 2. Dynamic trend feature extraction: Set trend calculation sliding window Minutes; the system uses the least squares method to calculate the regression slope of each parameter in real time. The slope of the mud pit volume during the current time period. The threshold value is 0.3; according to the system's preset configuration, the threshold for rapid increase in mud pit volume is... ;because The system discretizes its trend feature code C and maps it to 2 (rapid increase); similarly, the current outlet flow rate trend is calculated to be 2 (rapid increase), the riser pressure trend is 2 (rapid increase), and the total hydrocarbon value trend is 2 (rapid increase).

[0044] 3. Rule base matching and matching degree calculation: The system will input the extracted real-time multi-parameter trend features into Table 1 for traversal and matching; Table 1. Expert Knowledge Rule Base Based on Multi-Parameter Dynamic Trends

[0045] The preset core rules for detecting overflow include: rapid increase in mud pit volume, rapid increase in outlet flow rate, rapid increase in riser pressure, and rapid increase in total hydrocarbon value; compared with the currently extracted characteristics, the indicator functions of the above four key parameters... All conditions are met; the system calculates the current risk matching score for the overflow by summing the results according to preset weights. .

[0046] 4. Dynamic suppression and early warning generation: Set trigger threshold ,current The triggering condition is met; the system obtains the current drill bit depth. A database search revealed the depth of the previous similar drill string leakage warning. ;calculate Larger than the set suppression window Therefore, the system does not suppress the request and allows it to pass.

[0047] like Figure 2 This is a schematic diagram illustrating the implementation effect of the spatial risk probability prediction module provided by the present invention. The red dots in the diagram indicate high-risk trajectory points predicted based on the three-dimensional Kriging model.

[0048] The successfully matched sub-rules are semantically merged to generate, such as Figure 3 The structured early warning record shown is as follows: Risk type: Overflow; Matching probability: 0.85; Detailed description: Rapid increase in mud pit volume; Rapid increase in outlet flow rate; Rapid increase in riser pressure; Rapid increase in total hydrocarbon value; This record is pushed to the front-end monitoring interface in real time and entered into the database, completing a complete closed-loop diagnostic operation.

[0049] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for early warning and perception of complex drilling conditions based on spatiotemporal data fusion, characterized in that, Includes the following steps: Set up a set of operating conditions that cover different types of complex drilling conditions; A spatial risk probability prediction module is constructed, including: constructing a spatial risk indicator sample set based on the coordinates of observation points and indicator variables of known wells; and calculating the prior risk probability of the well to be logged using a three-dimensional kriging model based on the spatial risk indicator sample set and the coordinates of the well to be logged. and / or Set up a set of working conditions covering different types of complex drilling conditions, build a real-time working condition perception and early warning module, and output a trend description of the working conditions.

2. The method for early warning and perception of complex drilling conditions based on spatiotemporal data fusion according to claim 1, characterized in that, The three-dimensional kriging model includes: Calculate the risk probability of the sample points of the well to be logged ; As a working condition indicator variable, These are the weighting coefficients; The system of Kriging equations that satisfy the unbiased constraint and minimize the estimated variance is as follows. , It is the spatial variability function. For Lagrange multipliers, Given the spatial coordinates of the well sample points; These are the spatial coordinates of the well sample points to be logged.

3. The method for early warning and perception of complex drilling conditions based on spatiotemporal data fusion according to claim 2, characterized in that, When the prior probability of risk exceeds the risk threshold, output the risk trajectory point.

4. The method for early warning and perception of complex drilling conditions based on spatiotemporal data fusion according to claim 2, characterized in that, A continuous risk probability profile of the well to be logged is constructed using prior risk probability and well depth.

5. The method for early warning and perception of complex drilling conditions based on spatiotemporal data fusion according to claim 1, characterized in that, The real-time operating condition perception and early warning module includes: Real-time acquisition of engineering parameter data for the well to be logged; Calculate the slope using the change sequence of engineering parameters and time index ; Based on the range of slope values, trends are defined, and an expert knowledge rule base is constructed based on engineering parameters, trends, and trend weights. At any point in time t， The engineering parameters are compared with the rule set of the expert knowledge rule base to obtain the matching indicator function. ; Calculation of working condition types in t Risk matching score ; Trend weight.

6. The method for early warning and perception of complex drilling conditions based on spatiotemporal data fusion according to claim 5, characterized in that, When the risk matching value exceeds the trigger threshold, a trend description is output.

7. The method for early warning and perception of complex drilling conditions based on spatiotemporal data fusion according to claim 5, characterized in that, The engineering parameters include: mud pit volume, outlet flow rate, riser pressure, total hydrocarbon value, rotary table speed, rotary table torque, hook load, pump flow rate, and outlet temperature.

8. The method for early warning and perception of complex drilling conditions based on spatiotemporal data fusion according to claim 1, characterized in that, Operating conditions include: well leakage, overflow, stuck drill bit, and drill string leakage.

9. A drilling complex condition early warning and perception system based on spatiotemporal data fusion, characterized in that, include: Memory is used to store instructions that can be executed by the processor; A processor for executing instructions to implement the method for early warning and perception of complex drilling conditions based on spatiotemporal data fusion as described in any one of claims 1-8.

10. A computer-readable medium storing computer program code, characterized in that, The computer program code, when executed by a processor, implements the method for early warning and perception of complex drilling conditions based on spatiotemporal data fusion as described in any one of claims 1-8.

Images