Deep geothermal reservoir parameter prediction system based on machine learning

Abstract

The invention relates to the technical field of geological data processing, in particular to a deep geothermal reservoir parameter prediction system based on machine learning. The system comprises a data acquisition module used for acquiring gravity data, magnetic field intensity and seismic wave data of an observation point; the predicted temperature acquisition module is used for acquiring the first predicted temperature, the second predicted temperature and the external environment stability of the observation point; the geological interference analysis module is used for judging whether the observation point is an abnormal observation point or not so as to obtain the geological interference of the measurement point; the formation uniformity analysis module is used for calculating formation uniformity based on the travel time curve of the observation point; the fusion prediction temperature acquisition module is used for training a CNN model and acquiring stratum uniformity and geological interference of a target observation point; and obtaining a fusion prediction temperature of the target observation point according to the stratum uniformity, the geological interference, the external environment stability, the first preset value and the second prediction temperature. According to the invention, the accuracy of deep geothermal reservoir temperature prediction can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of geological data processing, and particularly relates to a deep geothermal reservoir parameter prediction system based on machine learning. Background Art

[0002] As a clean and sustainable energy form, deep geothermal energy plays an important role in the global energy transition. China is rich in geothermal resources, and efficient exploration and development of geothermal energy is the key to the development of new alternative energy. Among them, reservoir temperature is the core parameter for calculating the reserves of geothermal resources. Through temperature prediction, the total thermal energy of geothermal fluids can be estimated, thereby judging the development potential of geothermal fields; avoiding equipment damage or insufficient production capacity caused by temperature anomalies; improving development efficiency, optimizing well location selection and production strategies to ensure maximum energy output. Therefore, reservoir temperature prediction is an indispensable key link in the whole life cycle of geothermal exploration and development.

[0003] Traditionally, electromagnetic inversion methods are mostly used to predict the reservoir temperature in undrilled areas. The electromagnetic inversion method mainly uses the characteristics of the conductivity (or resistivity) of rocks changing with temperature to estimate the reservoir temperature. However, this relationship is not a simple linear one, but is affected by various factors (such as rock type, porosity, fluid composition, etc.), and the underground electrical structure obtained by electromagnetic inversion may have multiple solutions, that is, different geological models may produce similar electromagnetic responses. At the same time, due to external interference and the complexity of geological conditions, the electromagnetic response will be complex and uncertain. Therefore, the temperature prediction results obtained only by the electromagnetic inversion method may have errors. Summary of the Invention

[0004] In order to solve the above technical problems, the purpose of the present invention is to provide a deep geothermal reservoir parameter prediction system based on machine learning, and the specific technical solutions adopted are as follows: An embodiment of the present invention provides a deep geothermal reservoir parameter prediction system based on machine learning, and the system includes: A data acquisition module, configured to acquire gravity data, magnetic field intensity, and seismic wave data of observation points; A predicted temperature acquisition module, configured to acquire a first predicted temperature and a second predicted temperature of an observation point; arrange sensors at the observation point, and calculate the stability of the external environment of the observation point according to the data collected by the sensors; A geological interference analysis module, configured to determine whether the observation point is an abnormal observation point according to the gravity data and magnetic field intensity of the observation point. If it is a normal observation point, the geological interference of the observation point is a preset value. If it is an abnormal observation point, calculate the geological interference of the observation point according to the adjacent observation points of the observation point; The formation uniformity analysis module is used to obtain the travel-time curve based on the seismic wave data of the observation points; calculate the formation uniformity based on the change of data points on the travel-time curve; The fusion prediction temperature acquisition module is used to train a CNN model based on the travel-time curves, gravity data, magnetic field intensity, formation uniformity, and geological interference of each observation point; input the travel-time curve, gravity data, and magnetic field intensity of the target observation point into the CNN model to obtain the formation uniformity and geological interference of the target observation point; obtain the fusion prediction temperature of the target observation point according to the formation uniformity, geological interference, external environment stability, first predicted temperature, and second predicted temperature.

[0005] Preferably, obtaining the first predicted temperature and the second predicted temperature of the observation point includes: Obtain the formation temperatures at the same depth of an observation point through the heat conduction model method and the electromagnetic inversion method, and denote them as the first predicted temperature and the second predicted temperature respectively.

[0006] Preferably, calculating the external environment stability of the observation point according to the data collected by the sensor includes: Obtain the absolute value of the difference between the slope of a temperature data in the temperature data collected by a sensor and the slopes of the adjacent temperature data before and after, and denote it as the slope change difference of the temperature data; obtain the reciprocal of the sum of the slope change difference and the first preset value and perform a negative correlation mapping to obtain the mutation rate of the temperature data; denote the temperature data with a mutation rate greater than the mutation rate threshold as mutation data; obtain the proportion of the number of mutation data in the temperature data corresponding to a sensor, and denote it as the mutation data occurrence probability of the sensor, and obtain the difference between the first preset value and the average value of the mutation data occurrence probabilities of all sensors, and denote it as the external environment stability of the observation point.

[0007] Preferably, judging whether the observation point is an abnormal observation point according to the gravity data and magnetic field intensity of the observation point includes: Obtain the average value of the gravity data of an observation point after removing the maximum value and the minimum value, and denote it as the gravity mean value. Arrange the gravity data in ascending order, and obtain the median and the mode among them; calculate the average value of the gravity mean value, the median, and the mode, and denote it as the reference value; use the first preset value to subtract the reciprocal of the difference between a gravity data and the reference value to obtain the gravity data abnormality degree of the gravity data; similarly obtain the magnetic field intensity abnormality degree of each magnetic field intensity; if the gravity data abnormality degree of the gravity data or the magnetic field intensity abnormality degree of the magnetic field intensity at any moment of an observation point is greater than or equal to the abnormality threshold, the observation point is an abnormal observation point.

[0008] Preferably, calculating the geological interference of the observation point according to the adjacent observation points of the observation point includes: Calculate the distance between an observation point and other observation points, and select a preset number of other observation points with the smallest distances as the neighboring observation points of this observation point; calculate the difference between the mean value of the gravity data of each neighboring observation point and the mean value of the gravity data of this observation point respectively, then average and take the absolute value of the differences corresponding to each neighboring observation point to obtain the local difference of the gravity data of this observation point; similarly, obtain the local difference of the magnetic field intensity of this observation point; use a first preset value to perform negative correlation mapping on the reciprocals of the local difference of the gravity data and the local difference of the magnetic field intensity of this observation point respectively to obtain a first mapping result and a second mapping result; the sum of the first mapping result and the second mapping result is the local difference of this observation point; calculate the ratio of the number of types of abnormal data of this observation point to the total number of data types, and subtract this ratio from the first preset value to obtain the abnormal proportion of this observation point; perform weighted summation on the local difference and the abnormal proportion of this observation point to obtain the geological interference of this observation point.

[0009] Preferably, the travel-time curve includes the travel-time curve of the direct wave and the travel-time curve of the reflected wave.

[0010] Preferably, calculating the formation uniformity based on the change of data points on the travel-time curve includes: Obtain the slope between every two adjacent data points on the travel-time curve of the direct wave, and form a slope sequence in the order of the abscissa; use three adjacent data points on the travel-time curve of the reflected wave to construct a triangle, and calculate the curvature of the data point located in the center among the three adjacent data points using the minimum circumscribed circle of the triangle, and form a curvature sequence of each data point in the order of the abscissa; calculate the mean value of the absolute values of the differences between every two adjacent slopes in the slope sequence and add it to the first preset value to obtain an addition result, and take the reciprocal of the addition result to obtain the reflection feature of the direct wave. Similarly, analyze the curvature sequence in the same way to obtain the reflection feature of the reflected wave; average the reflection feature of the direct wave and the reflection feature of the reflected wave to obtain the formation uniformity of the observation point.

[0011] Preferably, training the CNN model includes: Take the travel-time curve, gravity data, and magnetic field intensity of each observation point of an observation point as a sample, and the formation uniformity and geological interference of this observation point are the labels of this sample to obtain the labeled samples corresponding to each observation point, and train the CNN model.

[0012] Preferably, obtaining the fusion prediction temperature of the target observation point according to the formation uniformity, geological interference, external environment stability, first predicted temperature, and second predicted temperature includes: Obtain the weight corresponding to the first predicted temperature and the weight corresponding to the second predicted temperature according to the formation uniformity, geological interference, and external environment stability of the target observation point, and perform weighted summation on the first predicted temperature and the second predicted temperature according to the weight corresponding to the first predicted temperature and the weight corresponding to the second predicted temperature to obtain the fusion predicted temperature of the target observation point.

[0013] Preferably, obtaining the weight corresponding to the first predicted temperature and the weight corresponding to the second predicted temperature includes: Take the formation uniformity of the target observation point as the credibility of the first predicted temperature of the target observation point; multiply the external environment stability of the target observation point by the difference between the first preset value and the geological interference of the target observation point to obtain the credibility of the second predicted temperature of the target observation point; the ratio of the credibility of the first predicted temperature to the sum of the credibility of the first predicted temperature and the credibility of the second predicted temperature is the weight corresponding to the first predicted temperature; the ratio of the credibility of the second predicted temperature to the sum of the credibility of the first predicted temperature and the credibility of the second predicted temperature is the weight corresponding to the second predicted temperature.

[0014] The embodiments of the present invention have at least the following beneficial effects: The present invention predicts the temperature of the deep geothermal reservoir at the observation point by two methods, namely the heat conduction model and the electromagnetic inversion method, to obtain the first predicted temperature and the second predicted temperature. Then, sensors are arranged at the observation point, and the external environment stability of the observation point is calculated according to the data collected by the sensors. The geological interference of the observation point is obtained by analyzing the gravity data and magnetic field intensity of the observation point, thereby obtaining two factors that affect the predicted temperature of the electromagnetic inversion method. Further, the travel time curve is obtained according to the seismic wave data of the observation point; the formation uniformity is calculated based on the change of the data points on the travel time curve, and the factor affecting the predicted temperature of the heat conduction model is obtained; then, the CNN model is trained using the travel time curve, gravity data, magnetic field intensity, formation uniformity, and geological interference of each observation point to reduce the complex calculations in the subsequent actual use process and improve the efficiency of actual use, and then obtain the formation uniformity, geological interference, and external environment stability of the target observation point; finally, the first predicted temperature and the second predicted temperature predicted by the two methods are fused according to the formation uniformity, geological interference, and external environment stability to obtain the fusion predicted temperature of the target observation point, bridging the deficiencies of the temperature predictions of the two methods, reducing the influence of various factors on the temperature prediction of the deep geothermal reservoir, and improving the accuracy of the temperature prediction of the deep geothermal reservoir. Description of the Drawings

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0016] Figure 1 It is a system block diagram of a deep geothermal reservoir parameter prediction system based on machine learning provided by an embodiment of the present invention; Figure 2 It is a schematic diagram of a travel-time curve of a deep geothermal reservoir parameter prediction system based on machine learning provided by an embodiment of the present invention. Detailed Embodiments

[0017] To further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific embodiments, structures, features, and effects of a deep geothermal reservoir parameter prediction system based on machine learning proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0019] The following specifically describes the specific solution of a deep geothermal reservoir parameter prediction system based on machine learning provided by the present invention in conjunction with the accompanying drawings.

[0020] Embodiment: The main application scenario of the present invention is: predicting the temperature of the deep geothermal reservoir during the exploration and development of undeveloped mining areas. For undeveloped mining areas where temperature prediction is to be carried out, since the heat conduction model assumes that the formation is uniform, the more uniform the formation, the greater the weight of the heat conduction model. And the electromagnetic inversion method is affected by the exploration geology and the external environment, that is, the weight of the electromagnetic inversion method is high when the external environment has little influence and the geological interference is small. The gravity and magnetic method and seismic wave velocity can be used to obtain the geological data of the terrain to be explored, analyze the geological data to calculate the formation uniformity, external environment stability, and geological interference of the terrain, obtain the prediction temperature data weights of the two methods, and obtain the final predicted temperature through weighted fusion.

[0021] Please refer to Figure 1 , which shows a system block diagram of a deep geothermal reservoir parameter prediction system based on machine learning provided by an embodiment of the present invention. The system includes the following modules: A data acquisition module, which is used to acquire gravity data, magnetic field intensity, and seismic wave data of the observation points.

[0022] First, acquire the collected data. The geological data used in this application is mainly obtained through the gravity and magnetic method and the seismic wave velocity method. Among them, for the gravity and magnetic method, a gravimeter (such as an absolute gravimeter, a relative gravimeter) is needed to observe in the area to be measured to obtain the gravity values of each measuring point, and a magnetometer (such as a proton precession magnetometer, an optically pumped magnetometer, etc.) is used to observe in the area to be measured to obtain magnetic parameters such as the magnetic field intensity, magnetic dip angle, and magnetic declination of each measuring point; the seismic wave velocity is obtained by exploration equipment (such as a geophone) for the seismic wave signals on the surface of the earth. The method is single-channel excitation, receiving while exciting, that is, after the seismic source is excited, the seismic wave data is received at one position.

[0023] Specifically, during the observation time of this application, the gravity data and magnetic field intensity of each observation point are obtained through a gravimeter and a magnetometer, and the seismic wave data is obtained by using a geophone. The acquisition frequency is set by the implementer according to the actual situation.

[0024] A predicted temperature acquisition module, which is used to acquire the first predicted temperature and the second predicted temperature of the observation points; sensors are arranged at the observation points, and the external environment stability of the observation points is calculated according to the data collected by the sensors.

[0025] The predicted temperature is mainly obtained through the electromagnetic inversion method and the heat conduction model. Among them, for the heat conduction model, the surface temperature data needs to be obtained through methods such as meteorological station observation and remote sensing technology, the geothermal gradient data needs to be obtained through methods such as drilling and temperature measurement wells, and at the same time, other relevant parameters such as the thermal conductivity and density of rocks are obtained through laboratory tests or geological survey data. The electromagnetic inversion method requires using electromagnetic exploration equipment (such as wide-area electromagnetic inversion detection equipment) to arrange an electrode array on the surface of the earth, emit electromagnetic signals and receive the reflected signals to obtain the potential difference or current distribution data between different positions.

[0026] The heat conduction model calculates the formation temperature according to the collected data, and the mathematical model is as follows: ; Wherein, represents the formation temperature at a depth of , represents the surface temperature, represents the geothermal gradient, that is, the change rate of the formation temperature per unit depth, represents the depth.

[0027] Obtaining predicted temperature by electromagnetic inversion method: Through inversion calculation, the observed electromagnetic field data is converted into the resistivity distribution of the underground medium. According to the known resistivity characteristics at different temperatures, the temperature distribution of the geothermal reservoir is inferred. Thus, the formation temperature at the same depth of an observation point can be obtained by the heat conduction model method and the electromagnetic inversion method, which are respectively denoted as the first predicted temperature and the second predicted temperature. The heat conduction model method and the electromagnetic inversion method are well-known technologies and will not be elaborated here.

[0028] The electromagnetic inversion method mainly uses the characteristic that the electrical conductivity (or resistivity) of rocks changes with temperature to estimate the reservoir temperature. Therefore, it is easily affected by external electromagnetic interference, masking or distorting the true electromagnetic response underground, resulting in errors in the predicted temperature. At the same time, due to geological structures (such as rock type, porosity, permeability, etc.), the electromagnetic response is complex and uncertain, so there may be errors in temperature estimation. Therefore, the stability of the external environment and geological interference may affect the error of the predicted temperature by the electromagnetic inversion method.

[0029] Since the electromagnetic inversion method is for geological exploration, underground electromagnetic interference is often related to topography and geology, mainly including underground ore bodies, pipelines, geological structures, etc. Such interference is relatively weak and closely related to the exploration geology, reflecting the underground situation to a certain extent. Therefore, the main electronic interference that needs to be excluded by the electromagnetic inversion method comes from the power system, industrial equipment, and wireless communication system on the ground surface. Since the sensor is vulnerable to electromagnetic interference, therefore, any number of sensors can be placed around each observation point to collect corresponding data, analyze whether the data has mutations, and then analyze the environmental stability. It should be noted that when arranging the sensors, the data collected by the sensors is preferably stable when there is no obvious change in the external environment, that is, the probability of the data itself mutating is low.

[0030] Preferably, the sensors set in this application are temperature sensors, which collect temperature data. The collection duration is the length of the observation time, and the collection frequency is once per second, which is specifically determined according to the actual situation. Moreover, the arrangement of the temperature sensors is in a circular array. With the observation point as the center, a circle is drawn with a preset length as the radius, and the temperature sensors are evenly distributed on the boundary of the circle. The central angle between two adjacent temperature sensors and the center of the observation point is a preset degree. The preset length and the preset degree are set by the implementer according to the actual situation. Preferably, in the embodiments of the present invention, the preset length and the preset degree are 1 meter and 60° respectively.

[0031] Since interference can cause mutations in sensor data, therefore, mutation data judgment is performed on temperature data. Specifically, the absolute value of the difference between the slope of a temperature data in the temperature data collected by a sensor and the slopes of the adjacent temperature data before and after is obtained, which is denoted as the slope change difference of the temperature data; the reciprocal of the sum of the slope change difference and the first preset value is obtained and negatively correlated, and the mutation rate of the temperature data is obtained. The specific calculation model is as follows: ; Among them, A represents the mutation rate of the temperature data at a certain moment of a temperature sensor, represents the slope of the temperature data at this moment and the previous temperature data, represents the slope of the temperature data at this moment and the next temperature data, 1 represents the first preset value, represents the slope change difference. The larger this value is, the larger the value of the negative correlation mapping of the sum of the slope change difference and the first preset value using the first preset value is, indicating that the mutation of the temperature data at this moment is more obvious, that is, the mutation rate A is larger.

[0032] Calculate the mutation rate for the temperature data at each moment collected by all sensors at the observation point. According to experimental statistics, the mutation rates of most data are around 0.3. Therefore, the mutation rate threshold is set to 0.7, and the temperature data with a mutation rate greater than the mutation rate threshold is recorded as mutation data. Further, the external environment stability of the observation point is obtained according to the mutation situation of the temperature data around the observation point. Specifically: obtain the proportion of the number of mutation data in the temperature data corresponding to a sensor, which is denoted as the mutation data occurrence probability of the sensor, and calculate the difference between the first preset value and the average value of the mutation data occurrence probabilities of all sensors, which is denoted as the external environment stability of the observation point. The specific calculation model is as follows: ; Among them, α represents the external environment stability of an observation point, N represents the number of sensors (temperature sensors) placed around the observation point, represents the number of mutation data in the temperature data collected by the i-th sensor, M represents the number of temperature data collected by each sensor, The probability of mutation data appearing in the data collected by the i-th sensor, that is, the mutation data occurrence probability corresponding to the i-th sensor. If the probability of mutation data appearing in the temperature data collected by the sensors placed around an observation point is small, it means that the electromagnetic interference around the observation point is small or not obvious, then the external stability of the observation point is large.

[0033] A geological interference analysis module is used to determine whether an observation point is an abnormal observation point based on the gravity data and magnetic field intensity of the observation point. If it is a normal observation point, the geological interference of the observation point is a preset value. If it is an abnormal observation point, the geological interference of the observation point is calculated based on the adjacent observation points of the observation point.

[0034] The gravity and magnetic method obtains magnetic field data, while the electromagnetic inversion method emits electromagnetic waves and obtains response data. Therefore, the magnetic force data of the gravity and magnetic method is different from the data of the electromagnetic inversion method. This is hereby stated. In the gravity and magnetic method, the gravity data mainly reflects the density distribution of underground substances. According to the gravity distribution, it is possible to judge whether there are geological bodies with high or low density underground, such as ore bodies, rock layers, faults, etc., as well as fracture zones and fold zones of the terrain, and metal ore bodies. The magnetic force is mainly affected by the rock type and metal ore bodies. Therefore, the geological interference can be obtained by combining the differences between the data.

[0035] Because the conditions affecting the gravity data and magnetic force data are slightly different, if the gravity data of a certain observation point is abnormal while the magnetic field intensity may not be abnormal, it may be due to topographical reasons. In places where both types of data are abnormal, if the anomalies around both are consistent, it may be only due to geological characteristics. If there are anomalies that are inconsistent with the surrounding anomalies themselves, there may be interference from influencing factors. That is, if the data are all consistently abnormal and consistent with the surrounding anomalies, the geological interference is small. The geological interference can be obtained by calculating whether the anomaly degree and the anomaly position of the data of each observation point in the data of its own type are consistent.

[0036] Furthermore, obtain the average value of the gravity data of an observation point after removing the maximum and minimum values, denoted as the gravity mean value. Arrange the gravity data in ascending order, and obtain the median and mode among them; calculate the mean value of the gravity mean value, median and mode, denoted as the reference value; use the first preset value minus the reciprocal of the difference between a gravity data and the reference value to obtain the gravity data anomaly degree of the gravity data; similarly obtain the magnetic field intensity anomaly degree of each magnetic field intensity; obtain the anomaly threshold according to experimental statistics, and the anomaly threshold is 0.6. If the gravity data anomaly degree of the gravity data or the magnetic field intensity anomaly degree of the magnetic field intensity of an observation point at any moment is greater than or equal to the anomaly threshold, the observation point is an abnormal observation point.

[0037] Further, analyze the abnormal observation points, and calculate the geological interference of the observation points according to the neighboring observation points of the observation points. Specifically, calculate the distance between an observation point and other observation points, and take the preset number of other observation points with the smallest distance as the neighboring observation points of the observation point; calculate the difference between the mean value of the gravity data of each neighboring observation point and the mean value of the gravity data of the observation point respectively, and then average and take the absolute value of the differences corresponding to each neighboring observation point to obtain the local difference of the gravity data of the observation point; similarly, obtain the local difference of the magnetic field intensity of the observation point; use the first preset value to perform negative correlation mapping on the reciprocals of the local difference of the gravity data and the local difference of the magnetic field intensity of the observation point respectively to obtain the first mapping result and the second mapping result; the mean value of the sum of the first mapping result and the second mapping result is the local difference of the observation point; use the first preset value to subtract the ratio of the number of types of abnormal data of the observation point to the total number of types of data to obtain the abnormal proportion of the observation point; perform weighted summation on the local difference and the abnormal proportion of the observation point to obtain the geological interference of the observation point.

[0038] The specific calculation model is: ; Among them, γ represents the geological interference of an observation point (abnormal observation point), C represents the number of neighboring observation points of the observation point, and in order to prevent a large probability of geological changes due to too far a distance, it is set to 3 in this application; represents the mean value of the gravity data of the i-th neighboring observation point, represents the mean value of the gravity data of the observation point, represents the mean value of the magnetic field intensity of the i-th neighboring observation point, represents the mean value of the magnetic field intensity of the i-th neighboring observation point, h represents the number of types of abnormal data of the observation point. If only one of the abnormal degrees of the gravity data or the abnormal degree of the magnetic field intensity is greater than or equal to the abnormal threshold, h takes 1. If there are two, h takes 2, and H represents the number of types of data, including two types of data: gravity data and magnetic field intensity. represents the weight. Because whether the abnormal manifestations of the data are consistent is more important, so let , .

[0039] is the local difference of the gravity data, which represents the abnormal degree of the gravity data of the observation point compared with the surrounding observation points. The larger the difference, the more abnormal it is compared with the surrounding observation points, and the greater the possibility of geological interference; is the local difference of the magnetic field intensity, which represents the abnormal degree of the magnetic force data of the selected observation point compared with the surrounding observation points. The larger the difference, the more abnormal it is compared with the surrounding observation points, and the greater the possibility of geological interference. The first mapping result and the second mapping result are respectively and They are the first mapping result and the second mapping result respectively.

[0040] For normal observation points, the geological interference cannot be set to 0. This is due to the limitations of experimental conditions, instrument accuracy, environment and other factors, as well as the inevitability of geological interference. The error cannot be completely eliminated. Therefore, according to experimental calculations and statistics, for observation points without anomalies, the geological interference should be set to the error value of 0.1.

[0041] The formation uniformity analysis module is used to obtain the travel-time curve based on the seismic wave data of the observation point; and calculate the formation uniformity based on the changes of the data points on the travel-time curve.

[0042] If the formation is uniform, the propagation speed of seismic waves will not change, and there will be no medium boundaries or interfaces encountered during the propagation process. Therefore, its propagation path is a straight line. At the same time, seismic waves include P-waves and S-waves, and the propagation speed of P-waves is usually faster than that of S-waves. The travel time refers to the propagation time of seismic waves from the seismic source to the observation point for reception. Based on the above characteristics, the main observation and recording objects of the travel-time curve of seismic waves are the direct wave and the reflected wave. The direct wave is the seismic wave that does not encounter a reflecting surface and directly propagates from the excitation point to the reception point. Therefore, on the premise of a uniform formation, it will appear as a straight line passing through the shot point, and the slope is the reciprocal of the speed; the reflected wave assumes that there is a horizontal reflecting interface underground, and there is a difference in wave impedance above and below the interface, and the formation above the interface is a homogeneous isotropic formation. The travel-time curve will then show the non-linear characteristics of the curve and appear as a smooth curve when the formation is uniform. From this, the travel-time curve when the formation is uniform can be obtained. The travel-time curve includes the travel-time curve of the direct wave and the travel-time curve of the reflected wave, as Figure 2 shown.

[0043] Furthermore, calculate the formation uniformity based on the changes of the data points on the travel-time curve. Specifically, obtain the slope between every two adjacent data points on the travel-time curve of the direct wave, and form a slope sequence in the order of the abscissa; use three adjacent data points on the travel-time curve of the reflected wave to construct a triangle, and calculate the curvature of the data point located in the center among the three adjacent data points using the minimum circumcircle of the triangle. Form a curvature sequence with the curvature corresponding to each data point in the order of the abscissa; calculate the mean of the absolute values of the differences between every two adjacent slopes in the slope sequence and add it to the first preset value to obtain an addition result, and take the reciprocal of the addition result to obtain the direct wave reflection feature. Similarly, analyze the curvature sequence in the same way to obtain the reflected wave reflection feature; average the direct wave reflection feature and the reflected wave reflection feature to obtain the formation uniformity of the observation point.

[0044] The specific calculation model is: ; Among them, β represents the formation uniformity of the observation point, n represents the number of slopes in the slope sequence corresponding to the travel-time curve of the direct wave, and m represents the number of curvatures in the curvature sequence corresponding to the travel-time curve of the reflected wave. represents the (i + 1)-th slope in the slope sequence, represents the i-th slope in the slope sequence, represents the average value of the slope differences. If the average value of the slope differences is closer to 0, it indicates that the seismic wave is less affected during propagation and the formation uniformity is stronger. represents the reflection characteristic of the direct wave. The larger this value is, the better the formation uniformity. represents the (j + 1)-th curvature in the curvature sequence, represents the j-th curvature in the curvature sequence, represents the average value of the curvature differences. If the average value of the curvature differences is smaller, it indicates that the curve is smoother, the seismic wave is less affected during underground propagation, and the formation uniformity is stronger. represents the reflection characteristic of the reflected wave.

[0045] The fusion prediction temperature acquisition module is used to train a CNN model based on the travel-time curves, gravity data, magnetic field intensity, formation uniformity, and geological interference of each observation point; input the travel-time curve, gravity data, and magnetic field intensity of the target observation point into the CNN model to obtain the formation uniformity and geological interference of the target observation point; and obtain the fusion prediction temperature of the target observation point according to the formation uniformity, geological interference, external environment stability, first predicted temperature, and second predicted temperature.

[0046] In the above process, the formation uniformity and geological interference of the observation point are obtained. Subsequently, in order to avoid complex calculations during use, a machine learning model is used to learn the relationship between the original data of the observation point and the formation uniformity and geological interference of the observation point. Subsequently, during use, only the original data needs to be input to obtain the formation uniformity and geological interference of the observation point. Preferably, the machine learning model selected in this application is a CNN model. The CNN model is trained based on the travel-time curves, gravity data, magnetic field intensity, formation uniformity, and geological interference of each observation point. Specifically, the travel-time curves, gravity data, and magnetic field intensity of each observation point of an observation point are used as a sample, and the formation uniformity and geological interference of this observation point are used as the label of this sample, obtaining the labeled samples corresponding to each observation point, and training the CNN model to complete the training of the CNN model. It should be noted that when the number of samples is insufficient, the travel-time curves, gravity data, magnetic field intensity, formation uniformity, and geological interference of other detection points in the past can be obtained based on previous detection data to construct samples for training.

[0047] Finally, record the observation point to be analyzed as the target observation point, obtain the stability of the external environment of the target observation point, and input the travel-time curve, gravity data, and magnetic field intensity of the target observation point into the trained CNN model to obtain the formation uniformity and geological interference of the target observation point.

[0048] Finally, obtain the fusion prediction temperature of the target observation point according to the formation uniformity, geological interference, external environment stability, first predicted temperature, and second predicted temperature. Specifically, obtain the weight corresponding to the first predicted temperature and the weight corresponding to the second predicted temperature according to the formation uniformity, geological interference, and external environment stability of the target observation point, and perform weighted summation on the first predicted temperature and the second predicted temperature according to the weight corresponding to the first predicted temperature and the weight corresponding to the second predicted temperature to obtain the fusion prediction temperature of the target observation point.

[0049] The specific calculation model for the fusion prediction temperature of the target observation point is: , where represents the fusion prediction temperature of the target observation point, and respectively represent the weight corresponding to the first predicted temperature and the weight corresponding to the second predicted temperature, and respectively represent the first predicted temperature and the second predicted temperature.

[0050] Among them, obtaining the weight corresponding to the first predicted temperature and the weight corresponding to the second predicted temperature is specifically: taking the formation uniformity of the target observation point as the credibility of the first predicted temperature of the target observation point; multiplying the stability of the external environment of the target observation point by the difference between the first preset value and the geological interference of the target observation point to obtain the credibility of the second predicted temperature of the target observation point; the ratio of the credibility of the first predicted temperature to the sum of the credibility of the first predicted temperature and the credibility of the second predicted temperature is the weight corresponding to the first predicted temperature; the ratio of the credibility of the second predicted temperature to the sum of the credibility of the first predicted temperature and the credibility of the second predicted temperature is the weight corresponding to the second predicted temperature.

[0051] Since the influencing factors of the heat conduction model are mainly formation uniformity, and the influencing factors of the electromagnetic inversion method are mainly external environment stability and geological interference, therefore, based on the above properties, the credibility of the two predicted temperatures of the target observation point can be obtained, and the calculation model is: , ; where represents the credibility of the first predicted temperature, that is, the credibility of selecting the heat conduction model to predict the temperature of the target observation point, represents the credibility of the second predicted temperature, that is, the credibility of selecting the electromagnetic inversion method to predict the temperature of the target observation point, Indicates the stability of the external environment of the target observation point, Indicates the geological interference of the target observation point, Indicates the formation uniformity of the target observation point. Then the weight corresponding to the first predicted temperature is , and the weight corresponding to the second predicted temperature is .

[0052] In summary, by analyzing the formation uniformity, geological interference, and external environment stability of the observation point, this application obtains the weights of the temperatures predicted by the two methods, which can improve the accuracy of the predicted temperature. At the same time, for the calculations of relatively complex formation uniformity and geological interference, in order to improve the boundary line during use, and then train the model, the machine learning model is used to obtain the formation uniformity and geological interference of the observation point, improving the efficiency during actual use.

[0053] It should be noted that the above order of the embodiments of the present invention is only for description and does not represent the advantages or disadvantages of the embodiments. And the above description of specific embodiments of this specification has been made. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0054] Each embodiment in this specification is described in a progressive manner. The same or similar parts between each embodiment can be referred to each other. The key point of each embodiment is to illustrate the differences from other embodiments.

[0055] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A deep geothermal reservoir parameter prediction system based on machine learning, characterized in that: The system includes: Data acquisition module, used to collect gravity data, magnetic field intensity and seismic wave data of observation points; A predicted temperature acquisition module is used to acquire a first predicted temperature and a second predicted temperature of an observation point; a sensor is arranged at the observation point, and the stability of the external environment of the observation point is calculated based on the data collected by the sensor; The geological interference analysis module is used to determine whether the observation point is an abnormal observation point based on the gravity data and magnetic field strength of the observation point. If it is a normal observation point, the geological interference of the observation point is a preset value. If it is an abnormal observation point, the geological interference of the observation point is calculated based on the neighboring observation points of the observation point. The formation uniformity analysis module is used to obtain the travel time curve according to the seismic wave data of the observation point; and calculate the formation uniformity based on the change of the data points on the travel time curve; The fused predicted temperature acquisition module is used to train the CNN model based on the travel time curve, gravity data, magnetic field strength, stratigraphic uniformity and geological interference of each observation point; the travel time curve, gravity data and magnetic field strength of the target observation point are input into the CNN model to obtain the stratigraphic uniformity and geological interference of the target observation point; the fused predicted temperature of the target observation point is obtained according to the stratigraphic uniformity, geological interference, external environmental stability, the first predicted temperature and the second predicted temperature.

2. A deep geothermal reservoir parameter prediction system based on machine learning according to claim 1, characterized in that: The obtaining of the first predicted temperature and the second predicted temperature of the observation point comprises: The formation temperature at the same depth of an observation point is obtained by the heat conduction model method and the electromagnetic inversion method, and is recorded as the first predicted temperature and the second predicted temperature respectively.

3. The deep geothermal reservoir parameter prediction system based on machine learning according to claim 1, characterized in that: The step of calculating the external environment stability of the observation point according to the data collected by the sensor includes: The absolute value of the difference between the slope of a temperature data and the adjacent temperature data in the temperature data collected by a sensor is obtained, and recorded as the slope change difference of the temperature data; the inverse of the slope change difference plus the first preset value is obtained and negative correlation mapping is performed to obtain the mutation rate of the temperature data; the temperature data with a mutation rate greater than the mutation rate threshold is recorded as mutation data; the proportion of the number of mutation data in the temperature data corresponding to a sensor is obtained, and recorded as the probability of occurrence of mutation data of the sensor; the difference between the first preset value and the mean value of the probability of occurrence of mutation data of all sensors is calculated, and recorded as the external environment stability of the observation point.

4. The deep geothermal reservoir parameter prediction system based on machine learning according to claim 1, characterized in that: The step of judging whether an observation point is an abnormal observation point according to the gravity data and magnetic field strength of the observation point includes: The average value of the gravity data of an observation point after removing the maximum and minimum values ​​is obtained, recorded as the gravity mean, and the gravity data is arranged in order from small to large to obtain the median and mode therein; the average of the gravity mean, median and mode is calculated, recorded as the reference value; the gravity data anomaly degree of the gravity data is obtained by subtracting the reciprocal of the difference between a gravity data and the reference value from the first preset value; the magnetic field intensity anomaly degree of each magnetic field intensity is obtained in the same way; if the gravity data anomaly degree of the gravity data or the magnetic field intensity anomaly degree of the magnetic field intensity of an observation point at any time is greater than or equal to the anomaly threshold, the observation point is an abnormal observation point.

5. The deep geothermal reservoir parameter prediction system based on machine learning according to claim 1, characterized in that: The step of calculating the geological interference of an observation point according to the neighboring observation points of the observation point comprises: Calculate the distance between an observation point and other observation points, and take a preset number of other observation points with the smallest distance as the neighboring observation points of the observation point; calculate the difference between the mean value of the gravity data of each neighboring observation point and the mean value of the gravity data of the observation point, and then average the difference corresponding to each neighboring observation point and take the absolute value to obtain the local difference of the gravity data of the observation point; similarly obtain the local difference of the magnetic field intensity of the observation point; use the first preset value to negatively correlate the inverse of the local difference of the gravity data of the observation point and the inverse of the local difference of the magnetic field intensity to obtain the first mapping result and the second mapping result; the sum of the first mapping result and the second mapping result is the local difference of the observation point; calculate the ratio of the number of types of abnormal data to the total number of data types at the observation point, and use the first preset value to subtract the ratio to obtain the abnormal proportion of the observation point; weighted sum the local difference and the abnormal proportion of the observation point to obtain the geological interference of the observation point.

6. The deep geothermal reservoir parameter prediction system based on machine learning according to claim 1, characterized in that: The travel time curve includes a travel time curve of a direct wave and a travel time curve of a reflected wave.

7. The deep geothermal reservoir parameter prediction system based on machine learning according to claim 1, characterized in that: The calculation of formation uniformity based on the change of data points on the travel time curve includes: The slope between every two adjacent data points on the travel time curve of the direct wave is obtained, and a slope sequence is formed in the order of the horizontal coordinates; a triangle is constructed using three adjacent data points on the travel time curve of the reflected wave, and the curvature of the data point located at the center of the three adjacent data points is calculated using the minimum circumscribed circle of the triangle, and the curvature corresponding to each data point is formed into a curvature sequence in the order of the horizontal coordinates; the average of the absolute values ​​of the differences between every two adjacent slopes in the slope sequence is calculated and added to a first preset value to obtain an addition result, and the addition result is inverted to obtain the direct wave reflection feature. Similarly, the curvature sequence is analyzed in the same way to obtain the reflection feature of the reflected wave; the direct wave reflection feature and the reflection wave reflection feature are averaged to obtain the formation uniformity of the observation point.

8. The deep geothermal reservoir parameter prediction system based on machine learning according to claim 1, characterized in that: The training CNN model comprises: The travel time curves, gravity data and magnetic field intensity of each observation point of an observation point are taken as a sample, and the stratigraphic uniformity and geological interference of the observation point are taken as the label of the sample. The labeled samples corresponding to each observation point are obtained to train the CNN model.

9. The deep geothermal reservoir parameter prediction system based on machine learning according to claim 1, characterized in that: The method of obtaining the fused predicted temperature of the target observation point according to the formation uniformity, geological interference, external environment stability, the first predicted temperature and the second predicted temperature includes: According to the stratigraphic uniformity, geological interference, and external environmental stability of the target observation point, the weight corresponding to the first predicted temperature and the weight corresponding to the second predicted temperature are respectively obtained, and the first predicted temperature and the second predicted temperature are weightedly summed according to the weight corresponding to the first predicted temperature and the weight corresponding to the second predicted temperature to obtain the fused predicted temperature of the target observation point.

10. A deep geothermal reservoir parameter prediction system based on machine learning according to claim 9, characterized in that: The obtaining of the weight corresponding to the first predicted temperature and the weight corresponding to the second predicted temperature includes: The stratigraphic uniformity of the target observation point is used as the credibility of the first predicted temperature of the target observation point; the credibility of the second predicted temperature of the target observation point is obtained by multiplying the external environmental stability of the target observation point by the difference between the first preset value and the geological interference of the target observation point; the ratio of the credibility of the first predicted temperature to the sum of the credibility of the first predicted temperature and the credibility of the second predicted temperature is the weight corresponding to the first predicted temperature; the ratio of the credibility of the second predicted temperature to the sum of the credibility of the first predicted temperature and the credibility of the second predicted temperature is the weight corresponding to the second predicted temperature.

Images