Underground HPHT area identification method and device
By combining a multi-dimensional data fusion method with a physical model and an integrated learning model, the problem of low prediction accuracy of high temperature and high pressure in complex geological environments is solved, and HPHT area identification and early warning with higher accuracy is achieved, which improves the safety and efficiency of drilling operations.
Patent Information
- Application Number
- CN202510394806.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-31
- Publication Date
- 2025-06-17
AI Technical Summary
In the prior art, physical models are difficult to apply to complex geological environments, resulting in poor prediction accuracy of high temperature and high pressure.
By extracting timing and static feature data based on the preliminary predicted values of the temperature and pressure physical model and real-time drilling data, the integrated learning model is input to perform multi-dimensional data fusion, and the temperature and pressure values are compensated, and the HPHT region is finally identified.
The prediction accuracy of temperature and pressure under complex geological conditions such as deep wells and ultra-deep wells has been improved, and abnormal HPHT areas have been effectively identified and warned, and the safety and construction efficiency of drilling operations have been improved.
Smart Images

Figure CN120159391A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of petroleum drilling engineering, and in particular to a method and device for identifying an underground HPHT area. Background Art
[0002] Drilling engineering is a general term for engineering that establishes an oil and gas channel between the ground and the bottom of the well by using techniques such as fast and efficient crushing of the rock at the bottom of the well, removing the cuttings at the bottom of the well and protecting the integrity of the well wall. In the field of oil and gas wells, temperature and pressure are important criteria for evaluating drilling work, affecting the drilling cycle and drilling costs. HPHT (High-Pressure High-Temperature) refers to an environment in which the temperature in the formation exceeds 150°C and the pressure exceeds 70MPa. Its definition and countermeasures are one of the core research contents of drilling engineering. With the continuous growth of demand for deep oil and gas resources, especially in deep or ultra-deep well operations, HPHT formations are becoming more and more common. How to accurately and real-time predict HPHT areas is very important in the process of oil drilling.
[0003] Traditional high temperature and high pressure predictions mostly rely on physical models. For example, a gas-water two-phase wellbore temperature distribution model in the directional well production process is established based on the principles of conservation of mass, momentum, and energy. In this model, it is assumed that the heat transfer in the wellbore is steady-state, while the heat transfer in the formation is unsteady-state. The effect of heat generated by friction on the wellbore temperature is also considered. At the same time, the wellbore pressure drop model is combined with the wellbore temperature model to obtain a directional well gas-water two-phase wellbore temperature-pressure coupling model, which is used to predict the wellbore temperature and pressure of the gas-water two-phase directional well.
[0004] However, the accuracy of the physical model depends on empirical parameters, and it is difficult to apply to complex geological environments such as deep wells and ultra-deep wells. There are many factors that affect the prediction of high temperature and high pressure in complex geological environments. It is affected by the combined effects of multiple factors such as formation conditions, well depth, fluid properties, construction parameters, and drill bit type. In addition, the various factors are mutually restricted and related. The traditional high temperature and high pressure prediction model cannot take into account the various influencing factors in complex geological environments, and the prediction accuracy is poor. Summary of the invention
[0005] In view of this, it is necessary to provide a method and device for identifying an underground HPHT area to solve the technical problems existing in the prior art that the physical model is difficult to apply to complex geological environments and has poor prediction accuracy.
[0006] In order to solve the above problems, on the one hand, the present invention provides a method for identifying an underground HPHT area, comprising: Calculate a preliminary temperature prediction value based on a temperature physical model, and calculate a preliminary pressure prediction value based on a pressure physical model; Extract the time-series feature data and static feature data from the preliminary temperature prediction value, the preliminary pressure prediction value, and the real-time obtained drilling data; Input the time-series feature data and the static feature data into two base models of a trained ensemble learning model respectively to obtain a temperature compensation value and a pressure compensation value; Based on the preliminary temperature prediction value and the temperature compensation value, obtain the final temperature prediction value, and based on the preliminary pressure prediction value and the pressure compensation value, obtain the final pressure prediction value; Identify the HPHT area based on the final temperature prediction value and the final pressure prediction value.
[0007] In a possible implementation manner, the calculation formula of the temperature physical model is:
[0008] In the formula, represents the well depth (m), represents the well depth at the preliminary temperature prediction value, represents the surface temperature (°C), represents the regional geothermal gradient (°C / km), represents the lithology correction coefficient (such as sandstone = 1.0, shale = 1.2), represents the bit friction heat generation term, represents the bit friction coefficient (0.2 - 0.4), represents the bit rotation speed (rad / s), represents the bit radius (m), represents the rock thermal conductivity (W / m·K), represents the formation thickness.
[0009] In a possible implementation manner, the method further includes: calculating a first compensation error index of the preliminary temperature prediction value and the measured temperature value in real time, and when the first compensation error index exceeds a preset first threshold, updating the first update parameter of the temperature physical model.
[0010] In a possible implementation manner, the first update parameter includes the regional geothermal gradient and the bit friction coefficient, and the update formula of the first update parameter is:
[0011]
[0012]
[0013] In the formula, represents the preliminary temperature prediction value, Represents the measured temperature value, Represents the regional geothermal gradient before update, Represents the regional geothermal gradient after update, Represents the bit friction coefficient before update, Represents the bit friction coefficient after update, and Represents the preset hyperparameter.
[0014] In a possible implementation, the calculation formula of the pressure physical model is:
[0015] In the formula, Represents the well depth (m), Represents the well depth The preliminary pressure prediction value at the location, Represents the density of the drilling fluid (kg / m 3 ), Represents the acceleration due to gravity (9.81 m / s²), and the formation pressure adopts the Darcy's law correction model, Represents the initial pressure of the reservoir, Represents the formation fluid flow rate (m 3 / s), Represents the fluid viscosity (Pa·s), Represents the reservoir permeability (m 2 ), Represents the reservoir thickness (m), Represents the drainage radius (m), Represents the wellbore radius (m).
[0016] In a possible implementation, the second compensation error index of the preliminary pressure prediction value and the measured pressure value is calculated in real time. When the second compensation error index exceeds the preset second threshold, the second update parameter of the pressure physical model is updated.
[0017] In a possible implementation, the second update parameter includes the reservoir permeability, fluid viscosity, and formation fluid flow rate. The update formula of the second update parameter is:
[0018]
[0019]
[0020]
[0021] In the formula, Represents the preliminary pressure prediction value, Represents the measured pressure value, represents the reservoir permeability before update, represents the reservoir permeability after update, represents the fluid viscosity before update, represents the fluid viscosity after update, represents the formation fluid flow rate before update, represents the formation fluid flow rate after update, , and represent preset hyperparameters.
[0022] In a possible implementation, the ensemble learning model includes two base models and one meta-model. One of the two base models is a long short-term memory model, and the other is a random forest model. The long short-term memory model includes a first temperature compensation network and a first pressure compensation network. The first temperature compensation network is used to input time series feature data and output a first temperature compensation value. The first pressure compensation network is used to input time series feature data and output a first pressure compensation value. The random forest model includes a second temperature compensation network and a second pressure compensation network. The second temperature compensation network is used to input static feature data and output a second temperature compensation value. The second pressure compensation network is used to input static feature data and output a second pressure compensation value. The meta-model is an extreme gradient boosting model, which is used to fuse the first temperature compensation value and the second temperature compensation value to obtain a final temperature compensation value, and is also used to fuse the first pressure compensation value and the second pressure compensation value to obtain a final pressure compensation value.
[0023] In a possible implementation, the method further includes: based on the measured temperature value and the measured pressure value, calculating the compensation error index and the absolute value of the network parameter gradient of the ensemble learning model in real time. When both the compensation error index and the absolute value of the network prediction parameter gradient are greater than the corresponding thresholds, updating the network parameters of the ensemble learning model.
[0024] On the other hand, the present invention also provides a downhole HPHT area identification device, including: A physical model prediction unit, configured to calculate a preliminary temperature prediction value based on a temperature physical model and calculate a preliminary pressure prediction value based on a pressure physical model; A feature data extraction unit, configured to extract time series feature data and static feature data from the preliminary temperature prediction value, the preliminary pressure prediction value, and the drilling data obtained in real time; An integrated model compensation unit is configured to input the timing feature data and the static feature data into two base models of a trained ensemble learning model respectively, to obtain a temperature compensation value and a pressure compensation value, and based on the preliminary temperature prediction value and the temperature compensation value, to obtain a final temperature prediction value, and based on the preliminary pressure prediction value and the pressure compensation value, to obtain a final pressure prediction value; An abnormal area identification unit is configured to identify the HPHT area based on the final temperature prediction value and the final pressure prediction value.
[0025] The beneficial effects of the present invention are as follows: The downhole HPHT area identification method provided by the present invention first makes a preliminary prediction of temperature and pressure through a physical model, and then inputs the preliminary prediction values and multi-factor data under complex geological conditions of drilling into an ensemble learning model for multi-dimensional data fusion to compensate the preliminary prediction values. By combining the prior knowledge of the physical model with the learning ability of the data-driven method, the prediction accuracy of temperature and pressure under complex geological conditions such as deep wells and ultra-deep wells is effectively improved. Furthermore, the abnormal HPHT area is effectively identified and warned, providing high-precision decision support for drilling operations and improving operation safety and construction efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] Figure 1 It is a schematic flowchart of an embodiment of the downhole HPHT area identification method provided by the present invention; Figure 2 It is a schematic structural diagram of an embodiment of the downhole HPHT area identification device provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0027] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present invention.
[0028] It should be understood that the schematic drawings are not drawn to scale. The flowcharts used in the present invention illustrate operations implemented according to some embodiments of the present invention. It should be understood that the operations of the flowchart may not be implemented in sequence, and steps without logical context relationships may be reversed or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart or remove one or more operations from the flowchart under the guidance of the content of the present invention. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor systems and / or microcontroller systems.
[0029] The descriptions such as "first" and "second" involved in the embodiments of the present invention are only for implicit purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Therefore, the technical features defined with "first" and "second" may explicitly or implicitly include at least one such feature. "And / or" describes the association relationship of associated objects and indicates that three relationships can exist. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.
[0030] Referring to "embodiments" herein means that specific features, structures, or characteristics described in connection with the embodiments can be included in at least one embodiment of the present invention. The phrase appears in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.
[0031] Before presenting the embodiments, the following terms are explained first.
[0032] The present invention provides a method and device for identifying downhole HPHT regions, which are described separately below.
[0033] Figure 1 It is a schematic flowchart of an embodiment of the method for identifying downhole HPHT regions provided by the present invention. As Figure 1 shown, the method for identifying downhole HPHT regions, applied to an edge computing device, includes: S101. Calculate a preliminary temperature prediction value based on a temperature physical model and calculate a preliminary pressure prediction value based on a pressure physical model; In order to obtain a more accurate preliminary temperature prediction value, in some embodiments of the present invention, the calculation formula of the temperature physical model is:
[0034] In the formula, represents the well depth (m), represents the well depth at the preliminary temperature prediction value, represents the surface temperature (°C), represents the regional geothermal gradient (°C / km), represents the lithology correction coefficient (e.g., sandstone = 1.0, shale = 1.2), represents the heat generation term due to bit friction, represents the bit friction coefficient (0.2 - 0.4), represents the bit rotation speed (rad / s), represents the bit radius (m), represents the rock thermal conductivity (W / m·K), represents the formation thickness.
[0035] Furthermore, in order to improve the calculation accuracy of the temperature physical model, in some embodiments of the present invention, the temperature physical model is corrected in real time during the drilling process. Specifically, the method further includes: calculating the first compensation error index of the preliminary temperature prediction value and the measured temperature value in real time. When the first compensation error index exceeds a preset first threshold, the first update parameter of the temperature physical model is updated.
[0036] It should be noted that the calculation formula of the first compensation error index is:
[0037] In the formula, represents the first compensation error index, that is, the mean absolute error of temperature, represents the preliminary temperature prediction value, represents the measured temperature value. When the first compensation error index exceeds the first threshold, such as 5%, for 3 consecutive times, the correction of the temperature physical model parameters is triggered.
[0038] The first update parameter includes the regional geothermal gradient and the bit friction coefficient , and the update formula of the first update parameter is:
[0039]
[0040]
[0041] In the formula, represents the preliminary temperature prediction value, represents the measured temperature value, represents the regional geothermal gradient before update, represents the updated regional geothermal gradient, represents the bit friction coefficient before update, represents the bit friction coefficient after update, and represents the preset hyperparameter.
[0042] In order to obtain a more accurate initial pressure prediction value, in some embodiments of the present invention, the calculation formula of the pressure physical model is:
[0043] In the formula, represents the well depth (m), represents the well depth at the initial pressure prediction value, represents the drilling fluid density (kg / m 3 ), represents the acceleration of gravity (9.81 m / s²), and the formation pressure adopts the Darcy's law correction model, represents the initial reservoir pressure, represents the formation fluid flow rate (m 3 / s), represents the fluid viscosity (Pa·s), represents the reservoir permeability (m 2 ), represents the reservoir thickness (m), represents the drainage radius (m), represents the wellbore radius (m).
[0044] Furthermore, in order to improve the calculation accuracy of the pressure physical model, in some embodiments of the present invention, the pressure physical model is corrected in real time during the drilling process. Specifically, the method further includes: calculating the second compensation error index of the initial pressure prediction value and the measured pressure value in real time. When the second compensation error index exceeds the preset second threshold, the second update parameter of the pressure physical model is updated.
[0045] It should be noted that the calculation formula of the second compensation error index is:
[0046] In the formula, represents the second compensation error index, that is, the mean absolute error of the pressure, represents the initial pressure prediction value, represents the measured pressure value. When the second compensation error index exceeds the second threshold such as 5% for 3 consecutive times, the correction of the pressure physical model parameters is triggered.
[0047] The second update parameter includes the reservoir permeability , fluid viscosity and formation fluid flow rate , the update formula for the second update parameter is:
[0048]
[0049]
[0050]
[0051] In the formula, represents the preliminary pressure prediction value, represents the measured pressure value, represents the reservoir permeability before update, represents the reservoir permeability after update, represents the fluid viscosity before update, represents the fluid viscosity after update, represents the formation fluid flow rate before update, represents the formation fluid flow rate after update, , and represent preset hyperparameters.
[0052] S102. Extract time series feature data and static feature data from the preliminary temperature prediction value, the preliminary pressure prediction value, and the drilling data obtained in real time; It should be noted that the drilling data includes geological data and sensor data. The static feature data is extracted from the preliminary prediction value and the geological data. The static feature data includes the preliminary temperature prediction value, the preliminary pressure prediction value, the gradient of the preliminary temperature prediction value, the gradient of the preliminary pressure prediction value, and the lithology code, etc. Among them, the lithology code is encoded by One - Hot, such as sandstone = 0, shale = 1; the time series feature data is extracted from the sensor data.
[0053] It should also be noted that the original sensor data includes well depth, drilling fluid flow rate, measured temperature value, and measured pressure value, etc. The sensor sampling frequency is 10Hz; the original sensor data needs to be pre - processed before use. The pre - processing steps include: first, perform denoising processing, then use the mean imputation method to process missing values, and finally perform standardization processing to obtain the pre - processed sensor data; among them, a fixed - length sliding window (such as data points) is used for standardization processing to eliminate the dimension difference. The standardization formula is:
[0054] Among them, represents the sensor data before standardization, Represents the standardized sensor data, represents the mean value of the data within the window, represents the standard deviation of the data within the window.
[0055] In order to obtain better time-series feature data, data such as the change in the drilling fluid flow rate, the change in the drilling speed, the change in the measured temperature value, and the change in the measured pressure value are further extracted from the preprocessed sensor data to form time-series feature data; among them, the flow rate is specifically the moving window mean, and the change in the drilling speed is specifically the difference between the drilling speed at time and the drilling speed at time The change in the measured temperature value is specifically the difference between the measured temperature value at time and the measured temperature value at time The change in the measured pressure value is specifically the difference between the measured pressure value at time
[0056] S103. Input the time-series feature data and the static feature data into two base models of the trained ensemble learning model respectively to obtain the temperature compensation value and the pressure compensation value; In order to better process the time-series feature data and the static feature data, in some embodiments of the present invention, the ensemble learning model takes the difference between the measured value and the preliminary predicted value, that is, the actual compensation amount, as the target during training. Structurally, the ensemble learning model adopts the stacking algorithm, including two base models and a meta-model. One of the two base models is a Long Short-Term Memory (LSTM) model, and the other is a Random ForestModel (RF) model. The long short-term memory model includes a first temperature compensation network and a first pressure compensation network. The first temperature compensation network is used to input the time-series feature data and output the first temperature compensation value. The first pressure compensation network is used to input the time-series feature data and output the first pressure compensation value. The random forest model includes a second temperature compensation network and a second pressure compensation network. The second temperature compensation network is used to input the static feature data and output the second temperature compensation value. The second pressure compensation network is used to input the static feature data and output the second pressure compensation value. The meta-model is an eXtreme Gradient Boosting (XGboost) model, which is used to fuse the first temperature compensation value and the second temperature compensation value to obtain the final temperature compensation value, and is also used to fuse the first pressure compensation value and the second pressure compensation value to obtain the final pressure compensation value.
[0057] Further, in order to improve the compensation accuracy of the integrated learning model, in some embodiments of the present invention, the integrated learning model is dynamically corrected during the drilling process. The method further includes: based on the measured temperature value and the measured pressure value, calculating the compensation error index and the absolute value of the network parameter gradient of the integrated learning model in real time. When both the compensation error index and the absolute value of the network prediction parameter gradient are greater than the corresponding thresholds, the network parameters of the integrated learning model are updated.
[0058] It should be noted that only the network parameters of the two base models are updated. For the long short-term memory model, since the dynamic correction methods of the first temperature compensation network and the first pressure compensation network are similar, only the dynamic correction method of the first temperature compensation network is described. Specifically, the method further includes: based on the measured temperature value and the first temperature compensation value, calculating the first temperature compensation error index and the absolute value of the first network parameter gradient of the first temperature compensation network in real time. When both the first temperature compensation error index and the absolute value of the first network prediction parameter gradient are greater than the corresponding thresholds, the first network parameters of the first temperature compensation network are updated.
[0059] It should be noted that the first temperature compensation error index is the mean absolute error. For example, when the first temperature compensation error index is greater than 0.05 or the absolute value of the first network prediction parameter gradient is greater than 10 -4 the first network parameters are updated. The first network parameters are the weight matrix, and the update formula for the first network parameters is:
[0060]
[0061] In the formula, represents the first network parameters, that is, the weight matrix, represents the mean square error of the first temperature compensation network before update, represents the first network parameters before update, represents the first network parameters after update, represents the first network parameter gradient before update.
[0062] In order to adapt to the resource limitations of edge devices, in some embodiments of the present invention, when updating the first network parameters of the first temperature compensation network, the weight pruning technology (Pruning) is used to remove redundant connections in the first temperature compensation network, and weight pruning is performed on the updated first network parameters calculated according to the update formula. Among them, the pruning threshold: , and the compression rate: 40%.
[0063] For the random forest model, since the dynamic correction methods of the second temperature compensation network and the second pressure compensation network are similar, only the dynamic correction method of the second temperature compensation network will be described. Specifically, the method further includes: based on the measured temperature value and the second temperature compensation value, calculating in real time the second temperature compensation error index and the absolute value of the second network parameter gradient of the second temperature compensation network. When both the second temperature compensation error index and the absolute value of the second network prediction parameter gradient are greater than the corresponding thresholds, update the second network parameter of the second temperature compensation network.
[0064] It should be noted that the second temperature compensation error index is the mean absolute error. For example, when the second temperature compensation error index is greater than 0.05 or the absolute value of the second network prediction parameter gradient is greater than 10 -4 at this time.
[0065] To adapt to the resource limitations of edge devices, in some embodiments of the present invention, the second network parameter is a weight matrix. During the process of updating the second network parameter, when the importance of a feature is lower than the corresponding threshold, it is regarded as a feature with low contribution. In subsequent newly added trees, this feature is no longer used for node splitting, reducing the computational overhead.
[0066] S104. Based on the preliminary temperature prediction value and the temperature compensation value, obtain the final temperature prediction value. Based on the preliminary pressure prediction value and the pressure compensation value, obtain the final pressure prediction value; S105. Identify the HPHT area based on the final temperature prediction value and the final pressure prediction value.
[0067] It should be noted that an anomaly detection algorithm (such as isolation forest, K-means clustering, etc.) is used to identify the high-temperature and high-pressure abnormal area. Specifically, based on the final prediction value and the isolation forest algorithm, single-point anomaly detection (judging whether the final prediction value exceeds the threshold) is performed to obtain the single-point anomaly detection result. Based on the final prediction value and the K-means clustering algorithm, regional anomaly clustering (judging whether the final prediction value deviates from the centroid) is performed to obtain the regional anomaly clustering result. The single-point anomaly detection result and the regional anomaly clustering result are fused to obtain the final HPHT area identification result.
[0068] Compared with the prior art, the present invention first makes a preliminary prediction of temperature and pressure through a physical model, and then inputs the preliminary prediction values and multi-factor data under complex geological conditions of drilling into an ensemble learning model for multi-dimensional data fusion to compensate the preliminary prediction values. Combining the prior knowledge of the physical model with the learning ability of the data-driven method effectively improves the prediction accuracy of temperature and pressure under complex geological conditions such as deep wells and ultra-deep wells, and further effectively identifies and warns of abnormal HPHT areas, providing high-precision decision support for drilling operations and improving operation safety and construction efficiency.
[0069] Furthermore, the ability of the long short-term memory model to extract temporal features and the random forest model to extract static features is effectively combined by using the ensemble learning model, improving the prediction accuracy of the ensemble learning model.
[0070] To better implement an underground HPHT area identification method in the embodiments of the present invention, correspondingly, on the basis of an underground HPHT area identification method, as Figure 2 shown, the embodiments of the present invention also provide an underground HPHT area identification device 200, including: A physical model prediction unit 201, configured to calculate a preliminary temperature prediction value based on a temperature physical model and calculate a preliminary pressure prediction value based on a pressure physical model; A feature data extraction unit 202, configured to extract temporal feature data and static feature data from the preliminary temperature prediction value, the preliminary pressure prediction value, and the drilling data obtained in real time; An integrated model compensation unit 203, configured to input the temporal feature data and the static feature data into two base models of a trained complete ensemble learning model respectively to obtain a temperature compensation value and a pressure compensation value, and obtain a final temperature prediction value based on the preliminary temperature prediction value and the temperature compensation value, and obtain a final pressure prediction value based on the preliminary pressure prediction value and the pressure compensation value; An abnormal area identification unit 204, configured to identify the HPHT area based on the final temperature prediction value and the final pressure prediction value.
[0071] The underground HPHT area identification device 200 provided in the above embodiments can implement the technical solutions described in the embodiments of the above underground HPHT area identification method. The specific implementation principles of the above units can be referred to the corresponding contents in the embodiments of the above underground HPHT area identification method, and will not be elaborated here.
[0072] The above has introduced in detail an underground HPHT area identification method provided by the present invention. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.
[0073] The above is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.
Claims
1. A method for identifying an underground HPHT region, characterized in that: include: Calculate a preliminary temperature prediction value based on a temperature physical model, and calculate a preliminary pressure prediction value based on a pressure physical model; Extracting time series characteristic data and static characteristic data from the preliminary temperature prediction value, the preliminary pressure prediction value and the drilling data acquired in real time; Inputting the time series feature data and the static feature data into two base models of a fully trained integrated learning model respectively to obtain a temperature compensation value and a pressure compensation value; Based on the preliminary temperature prediction value and the temperature compensation value, a final temperature prediction value is obtained, and based on the preliminary pressure prediction value and the pressure compensation value, a final pressure prediction value is obtained; A HPHT region is identified based on the final temperature prediction value and the final pressure prediction value.
2. The method for identifying downhole HPHT regions according to claim 1, characterized in that: The calculation formula of the temperature physical model is: In the formula, Indicates well depth (m), Indicates well depth The initial temperature prediction value at represents the surface temperature (°C), represents the regional geothermal gradient (℃ / km), Indicates the lithology correction factor (e.g. sandstone = 1.0, shale = 1.2), represents the frictional heat generated by the drill bit, Indicates the friction coefficient of the drill bit (0.2-0.4), Indicates the drill speed (rad / s), Indicates the drill bit radius (m), represents the thermal conductivity of rock (W / m·K), Indicates the thickness of the formation.
3. The method for identifying downhole HPHT regions according to claim 2, characterized in that: The method further includes: calculating in real time a first compensation error index of a preliminary temperature prediction value and a temperature measurement value, and when the first compensation error index exceeds a preset first threshold, updating a first update parameter of the temperature physical model.
4. The method for identifying downhole HPHT regions according to claim 3, characterized in that: The first update parameter includes the regional geothermal gradient and the drill bit friction coefficient, and the update formula of the first update parameter is: In the formula, represents the preliminary temperature prediction value, Indicates the measured temperature value. represents the regional geothermal gradient before updating, represents the updated regional geothermal gradient, Indicates the friction coefficient of the drill bit before updating, represents the updated drill friction coefficient, and Represents the preset hyperparameters.
5. The method for identifying downhole HPHT regions according to claim 1, characterized in that: The calculation formula of the pressure physical model is: In the formula, Indicates well depth (m), Indicates well depth The initial pressure prediction value at Indicates the density of drilling fluid (kg / m 3 ), represents the gravitational acceleration (9.81 m / s²), and the formation pressure adopts the Darcy law correction model. represents the initial reservoir pressure, Indicates the formation fluid flow rate (m 3 / s), represents the fluid viscosity (Pa·s), Reservoir permeability (m 2 ), represents the reservoir thickness (m), represents the oil leakage radius (m), Represents the wellbore radius (m).
6. The method for identifying downhole HPHT regions according to claim 5, characterized in that: A second compensation error index of the preliminary pressure prediction value and the pressure measured value is calculated in real time, and when the second compensation error index exceeds a preset second threshold, a second update parameter of the pressure physics model is updated.
7. The method for identifying downhole HPHT regions according to claim 6, characterized in that: The second update parameters include reservoir permeability, fluid viscosity and formation fluid flow rate, and the update formula of the second update parameters is: In the formula, represents the initial pressure prediction value, Indicates the measured pressure value. represents the reservoir permeability before updating, represents the updated reservoir permeability, Indicates the fluid viscosity before updating, represents the updated fluid viscosity, represents the formation fluid flow rate before updating, represents the updated formation fluid flow rate, , and Represents the preset hyperparameters.
8. The method for identifying downhole HPHT regions according to claim 1, characterized in that: The integrated learning model includes two base models and a meta-model, one of the two base models is a long short-term memory model, and the other is a random forest model. The long short-term memory model includes a first temperature compensation network and a first pressure compensation network. The first temperature compensation network is used to input time series feature data and output a first temperature compensation value. The first pressure compensation network is used to input time series feature data and output a first pressure compensation value. The random forest model includes a second temperature compensation network and a second pressure compensation network. The second temperature compensation network is used to input static feature data and output a second temperature compensation value. The second pressure compensation network is used to input static feature data and output a second pressure compensation value. The meta-model is an extreme gradient boosting model, which is used to fuse the first temperature compensation value and the second temperature compensation value to obtain a final temperature compensation value, and is also used to fuse the first pressure compensation value and the second pressure compensation value to obtain a final pressure compensation value.
9. The method for identifying downhole HPHT regions according to claim 1, characterized in that: The method also includes: based on the actual measured temperature value and the actual measured pressure value, calculating the compensation error index and the absolute value of the network parameter gradient of the integrated learning model in real time, and when the compensation error index and the absolute value of the network prediction parameter gradient are both greater than the corresponding threshold value, updating the network parameters of the integrated learning model.
10. A downhole HPHT area identification device, characterized in that: include: A physical model prediction unit, used to calculate a preliminary temperature prediction value based on a temperature physical model, and to calculate a preliminary pressure prediction value based on a pressure physical model; A feature data extraction unit, used to extract time series feature data and static feature data from the preliminary temperature prediction value, the preliminary pressure prediction value and the drilling data acquired in real time; An integrated model compensation unit, used for inputting the time series feature data and the static feature data into two base models of a fully trained integrated learning model, respectively, to obtain a temperature compensation value and a pressure compensation value, to obtain a final temperature prediction value based on the preliminary temperature prediction value and the temperature compensation value, and to obtain a final pressure prediction value based on the preliminary pressure prediction value and the pressure compensation value; An abnormal region identifying unit is used to identify a HPHT region based on the final temperature prediction value and the final pressure prediction value.
Citation Information
Cited By
Shale parameter prediction method and device based on transfer learning, equipment and medium
CN121071494A
