Low-permeability thin-difference oil reservoir production degree fine evaluation method based on thermal ultrasonic logging

Abstract

The invention relates to the technical field of feature recognition, in particular to a low-permeability thin-difference oil reservoir production degree fine evaluation method based on thermal ultrasonic logging, and the method comprises the steps: obtaining the temperature and flow data of each liquid production well section in a production well in real time, and screening out a detection well section; dividing the flow data of each detection well section into flow subsequences; constructing an oil reservoir silence index based on the mean value and the change trend of each flow sub-sequence; based on the temperature change trend and the flow dispersion degree of each detection well section, an oil layer liquid production index is constructed; and training a logistic regression model based on the oil layer liquid production index, the oil layer silence index and the flow dispersion degree of each low-liquid-production layer so as to judge whether each detection well section is a real low-liquid-production layer or not, and correcting an injection profile in an injection well so as to calculate the oil layer production degree. By analyzing the characteristics of the low-yield liquid layer, the leakage of the low-yield liquid layer can be avoided, and the evaluation precision of the oil layer production degree is improved.

Description

Technical Field

[0001] This application relates to the field of feature recognition technology, specifically to a method for fine evaluation of the utilization degree of low-permeability, thin-layer oil reservoirs based on thermal ultrasonic logging. Background Technology

[0002] In the mid-to-late stages of oilfield development, accurate assessment of reservoir utilization is crucial for formulating adjustment and potential-tapping measures. Currently, conventional flow logging methods, such as electromagnetic flowmeters and turbine flowmeters, are commonly used for injection profiles. However, these technologies have limited resolution, making it difficult to accurately identify oil-bearing layers with low single-layer fluid uptake. While tracer monitoring can qualitatively display fluid flow direction, it is limited by interpretation models, making precise quantitative stratification interpretation difficult. Therefore, utilizing machine learning technology to construct a method that can overcome the current testing resolution limits and integrate multiple information sources for refined and quantitative evaluation of low-permeability, thin, and poor-quality oil layers is of great significance for solving the technical bottleneck of "inability to measure and inaccurate evaluation."

[0003] While existing technologies have taken into account the shortcomings of flow-collecting testing methods and have utilized non-flow-collecting detection techniques to analyze produced profile data, they have not fully considered that the collected data from low-yield fluid layers are relatively weak compared to normal data. Furthermore, due to the complex downhole environment, low-yield fluid layers are often misclassified as unusable layers during analysis, thus missing low-yield fluid layers and causing errors in the analysis results. This results in poor correction of the injection profile and affects the accuracy of the final calculation of the oil layer utilization. Summary of the Invention

[0004] To address the aforementioned technical problems, this application provides a method for finely evaluating the utilization degree of low-permeability, thin-layer oil reservoirs based on thermal ultrasonic logging, thereby resolving existing issues.

[0005] The method for fine evaluation of the activation degree of low-permeability, thin-diameter oil reservoirs based on thermal ultrasonic logging in this application adopts the following technical solution: One embodiment of this application provides a method for fine evaluation of the utilization degree of low-permeability, thin-diameter oil reservoirs based on thermal ultrasonic logging. The method includes the following steps: Real-time acquisition of temperature and flow data of each producing section within the well within a preset time period; screening of each detection section from all producing sections according to flow rate to achieve initial location of low-producing layers; All flow data of each monitoring well section are divided into several flow subsequences. Based on the mean and trend of each flow subsequence, the oil layer quiescent index of each flow subsequence is constructed to characterize the probability of no fluid production in the monitoring well section within the time period corresponding to each flow subsequence. In this way, all flow subsequences are divided into quiescent subsequences and seepage subsequences. Based on the existence of silent subsequences in each detection well section, the changing trends of temperature data corresponding to each silent subsequence and each seepage subsequence in each detection well section, and combined with the flow dispersion of all seepage subsequences, an oil layer production index is constructed to characterize the possibility that each detection well section is a low-production layer. Several real and false low-production fluid layers are obtained, and a logistic regression model is trained based on the fluid production index of each low-production fluid layer, the quiescent index of all oil layers, and the flow dispersion. Then, the trained logistic regression model is used to determine whether each detection well section is a real low-production fluid layer. Based on the location and production volume of the real low-production fluid layers in the production well, the injection profile in the injection well is corrected, and the degree of oil layer utilization is calculated.

[0006] Preferably, each detection well segment refers to a producing well segment where the mean of all flow data within a preset time period is less than or equal to a preset segmentation threshold and is not equal to 0.

[0007] Preferably, the reservoir quiescence index of each flow subsequence is negatively correlated with the mean of each flow subsequence and the absolute value of the slope of the fitted line of each flow subsequence.

[0008] Preferably, the quiescent subsequence refers to the flow rate subsequence where the reservoir quiescent index is greater than or equal to a preset quiescent threshold; the seepage subsequence refers to the flow rate subsequence where the reservoir quiescent index is less than the preset quiescent threshold.

[0009] Preferably, the process for constructing the reservoir production index is as follows: The mean slope value of the temperature data corresponding to all seepage subsequences in each detection well section is calculated and recorded as the seepage temperature slope of each detection well section. When the flow rate subsequence of each monitoring well section contains a silent subsequence: the mean slope value of the temperature data corresponding to all silent subsequences of each monitoring well section is calculated and recorded as the silent temperature slope of each monitoring well section; the oil production index of each monitoring well section is positively correlated with the seepage temperature slope and the mean value of the oil layer silent index of all silent subsequences, and negatively correlated with the silent temperature slope and the flow rate dispersion of all seepage subsequences; When the flow rate subsequence of each monitoring well section does not contain a silent subsequence: the oil production index of each monitoring well section is positively correlated with the seepage temperature slope and negatively correlated with the flow rate dispersion of all seepage subsequences.

[0010] Preferably, the flow dispersion of all seepage subsequences refers to the mean of the variances of all seepage subsequences within each detection well section.

[0011] Preferably, the specific process of training the logistic regression model is as follows: Based on the oil layer production index of each low-yield fluid layer in each data acquisition process, the mean of the oil layer quiescence index of all flow subsequences, and the flow dispersion of all seepage subsequences, each low-yield fluid layer's production vector is constructed. Set the label of the production vector of the true low-production layer to 1 and the label of the production vector of the false low-production layer to 0. The production vectors and labels of all real low-production layers and all false low-production layers are used as inputs to the logistic regression model to train the model and obtain the trained logistic regression model.

[0012] Preferably, the specific process for determining whether each detection well section is a true low-production fluid layer is as follows: the production vector of each detection well section is used as the input of the trained logistic regression model, and the output is the label corresponding to each detection well section; if the label is 1, then the corresponding detection well section is determined to be a true low-production fluid layer.

[0013] Preferably, the specific process of correcting the injection profile in the injection well is as follows: Oil-bearing segments that are identified as liquid-absorbing layers by tracers injected into the well but are not assessed as liquid-absorbing layers by electromagnetic flow logging are all recorded as potential low-liquid-absorbing layers. Based on the distance between each potential low-supply layer and each actual low-yield layer in the production well, a one-to-one correspondence is established between the actual low-yield layers in the production well and the potential low-supply layers in the injection well. The production volume of all real low-yield fluid layers in the production well is counted within a preset time period and summed to obtain the total production volume; and the proportion of the production volume of each real low-yield fluid layer in the total production volume is calculated. Obtain the total amount of liquid absorbed by all potential low-suction layers in the injection well; based on the proportion of the actual low-production layer corresponding to each potential low-suction layer, allocate the total amount of liquid absorbed to each potential low-suction layer according to the corresponding proportion, and obtain the amount of liquid absorbed by each potential low-suction layer, thereby generating the corrected injection profile.

[0014] Preferably, the process of establishing the one-to-one correspondence is as follows: The absolute difference between the depth range of each potential low-suction layer and the median value of all actual low-production layers in the produced well is calculated. The actual low-yield liquid layer corresponding to the minimum absolute difference is denoted as the actual low-yield liquid layer corresponding to each potential low-absorption liquid layer.

[0015] This application has at least the following beneficial effects: To address the problem that existing technologies fail to adequately consider the weak and easily interfered signals of low-production fluid layers in downhole wells, leading to misjudgments of these layers and consequently errors in the evaluation of their utilization, this application overcomes these shortcomings by providing a refined evaluation method for the utilization of low-permeability, thin-yield oil layers based on thermal ultrasonic logging using machine learning technology. Specifically, by integrating high-resolution logging technology with dynamic monitoring data, a multi-source information fusion interpretation model is established to achieve a refined and quantitative characterization of the fluid absorption and production status of low-permeability, thin-yield oil layers. Furthermore, by analyzing the flow characteristics of each monitored well section, an oil layer quiescence index is constructed, thereby enabling... Accurately identify the quiescent characteristics of low-yield reservoirs; by analyzing the flow and temperature variation characteristics of the monitored well sections when they are low-yield reservoirs, an oil layer production index is constructed, which can integrate temperature response and production characteristics, breaking through the resolution limitations of traditional testing, accurately identifying the production characteristics of thin and poor-yield reservoirs, and accurately determining the probability that each monitored well section belongs to a low-yield reservoir; by using a trained logistic regression model to judge whether each monitored well section is a low-yield reservoir, the corrected injection profile has a higher resolution, thereby improving the accuracy of reservoir utilization evaluation, which in turn can improve oil recovery rate and save development costs. Attached Figure Description

[0016] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating the steps of the method for fine evaluation of the activation degree of low-permeability, thin-differential oil reservoirs based on thermal ultrasonic logging provided in this application; Figure 2 A flowchart for obtaining the reservoir production index provided in this application. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the method for fine evaluation of the utilization degree of low-permeability, thin-layer oil reservoirs based on thermal ultrasonic logging proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0020] The following description, in conjunction with the accompanying drawings, details the specific scheme of the method for fine evaluation of the activation degree of low-permeability, thin-layer oil reservoirs based on thermal ultrasonic logging provided in this application.

[0021] This application provides an embodiment of a method for finely evaluating the utilization degree of low-permeability, thin-layer, poor-permeability oil reservoirs based on thermal ultrasonic logging. Specifically, the method is described below. Please refer to [link to relevant documentation]. Figure 1 The method includes the following steps: Step 1: Acquire temperature and flow data of each producing well section within the production well in real time within a preset time period; select each detection well section from all producing well sections according to the flow rate to achieve initial positioning of low-producing layers.

[0022] This application takes any group of production wells and injection wells in the development well group of the three types of oil layers in the oilfield as an example for analysis.

[0023] Within the wellbore of a produced well, there are multiple rock formations and oil layers. In this embodiment, the wellbore is first divided into multiple segments based on the depth range corresponding to each oil layer segment, and the segments corresponding to the oil layer segments are designated as production segments. The number of segments can be set by the implementer according to the actual conditions of the reservoir; this embodiment does not impose any restrictions.

[0024] Subsequently, this application sends an optical fiber to the bottom of the production well via coiled tubing to collect temperature data of each producing well section; since the distributed optical fiber temperature measurement technology will collect temperature values ​​at multiple locations in each producing well section at the same time, the average of all temperature values ​​collected in the same producing well section at the same time is recorded as the temperature data of the producing well section at that time.

[0025] This application uses a non-collector type ultrasonic flow meter to collect the flow rate data of the produced fluid in each producing section of the production well.

[0026] In this embodiment, temperature and flow rate data are collected synchronously and in real time within each producing well section, with a data collection interval of 1 second. The units for temperature and flow rate data are ℃ and m³ / s, respectively. All temperature and flow rate data for each producing well section within a preset time period are arranged in chronological order to obtain the temperature sequence and flow rate sequence for each producing well section. In this embodiment, the preset time period is 2 hours.

[0027] It should be noted that data acquisition for all well sections is not synchronous; however, temperature and flow rate data for a single well section are acquired synchronously.

[0028] Next, the mean flow rate sequence of each producing well section is calculated and denoted as the mean flow rate of each producing well section, reflecting the fluid production flow rate of the corresponding oil layer. Then, producing well sections with a mean flow rate of 0 are removed from all producing well sections, i.e., oil layer sections with no fluid production, which improves calculation efficiency and avoids misjudgments in subsequent calculations.

[0029] Finally, the average flow rate of all remaining producing well sections within the produced well is used as the input to the Otsu threshold method, and a segmentation threshold is output, which is recorded as the preset segmentation threshold. Producing well sections with an average flow rate less than or equal to the preset segmentation threshold are recorded as detection well sections, thereby achieving the initial location of low-producing layers within the produced well. The Otsu threshold method is a well-known technique, and its specific process will not be elaborated further.

[0030] Step 2: Divide all flow data of each monitoring well section into several flow subsequences; based on the mean and trend of each flow subsequence, construct the reservoir quiescent index for each flow subsequence to characterize the probability of no fluid production in the monitoring well section within the time period corresponding to each flow subsequence, thereby dividing all flow subsequences into quiescent subsequences and seepage subsequences.

[0031] Compared to fiber optic temperature measurement technology, which has strong anti-interference capabilities and high measurement stability, non-collector ultrasonic flowmeters are susceptible to multiple factors in produced fluid measurement: on the one hand, the components of produced fluid, such as oil, water, and impurities, are dynamically changing, and produced fluid containing a large number of pollutants will cause ultrasonic signal scattering, thus affecting ultrasonic accuracy; on the other hand, due to the harsh working environment downhole, ultrasonic probes are prone to scaling, aging, or drift, which will also affect measurement accuracy and the reliability of the acquired data.

[0032] Therefore, in order to ensure the accuracy of the evaluation of the oil reservoir utilization level, it is necessary to accurately determine the authenticity of the flow data of each detection well section based on machine learning technology, and effectively distinguish between real production data and false data caused by instrument errors or interference. This is necessary to avoid misjudging low-production layers and non-utilized layers, thereby improving the reliability of the utilization status identification of thin and poor layers.

[0033] Taking the i-th detection well section as an example for analysis, we use machine learning technology to evaluate whether it is a true low-yield fluid layer.

[0034] For genuine low-permeability producing fluid layers, due to their low reservoir saturation, weak permeability, and limited displacement efficiency, the produced fluid typically exhibits weak flow and intermittent seepage, without strong pulsed or turbulent flow. Furthermore, because the produced fluid exhibits intermittent overflow characteristics, the produced fluid flow rate measured by the ultrasonic flowmeter is extremely small and stable during the quiescent phase before fluid overflow.

[0035] Since the flow rate data measured during oil spill and quiescent periods differ significantly, it is necessary to first extract the quiescent period data from the overall data to avoid errors in the analysis of the overall data, which could lead to misjudgment.

[0036] The flow rate sequence of the i-th detection well section is divided into multiple flow rate subsequences using a sequence segmentation algorithm. Taking the u-th flow rate subsequence as an example, the analysis is conducted to assess whether it might be flow rate data collected during the silent phase. Sequence segmentation algorithms include, but are not limited to, the BG segmentation algorithm and the moving average method. In this embodiment, the BG segmentation algorithm is used for segmentation. The BG segmentation algorithm is a well-known technology, and its specific process will not be described in detail.

[0037] The mean of the u-th flow rate subsequence is obtained. The mean value reflects the probability that the production of the oil layer corresponding to the i-th monitoring well section is extremely weak during the collection period of the u-th flow rate subsequence. The smaller the value, the weaker the data collected by the i-th monitoring well section during the collection period of the u-th flow rate subsequence and the smaller the data change, thus reflecting the greater the probability that the corresponding oil layer did not produce fluid during the corresponding collection period.

[0038] The u-th flow rate subsequence is used as input to a linear fitting algorithm for fitting, and the absolute value of the slope of the fitted line is denoted as the stationarity factor of the u-th flow rate subsequence. The stationarity factor reflects the overall data change trend of the u-th flow rate subsequence; the smaller the value, the closer the produced fluid volume of the oil reservoir is to a trendless steady-state process during the collection period corresponding to the u-th flow rate subsequence, and the more likely it corresponds to the quiescent state of the oil reservoir. The linear fitting algorithm is a well-known technique, and its specific process will not be described in detail.

[0039] To eliminate the influence of dimensions, the mean and stationarity factor of each flow subsequence of all monitored well sections were calculated, and the maximum value normalization was performed on all the mean and stationarity factors. Maximum value normalization is a well-known technique, and the specific process will not be described in detail.

[0040] In a preferred embodiment, an oil reservoir quiescence index is constructed based on the mean and trend of each flow rate subsequence. This index characterizes the probability of no fluid production in the corresponding monitoring well section within the time period corresponding to each flow rate subsequence. The oil reservoir quiescence index of each flow rate subsequence is negatively correlated with the mean of each flow rate subsequence and the absolute value of the slope of the fitted straight line of each flow rate subsequence. This negative correlation means that the dependent variable decreases (increases) as the independent variable increases (decreases).

[0041] Preferably, in this embodiment, the reservoir quiescence index of the u-th flow subsequence is denoted as... Its specific expression is: In the formula, Let be the reservoir quiescence index for the u-th flow subsequence; This is the normalized value of the mean of the u-th flow subsequence; Let be the normalized value of the stationarity factor of the u-th flow subsequence; This is a preset constant used to avoid the denominator being 0. Its value range is [0.005, 0.01]. The value has a very small impact on the calculation and can be ignored. In this embodiment, it is taken as 0.008.

[0042] The reservoir quiescence index reflects the fluid production intensity of the reservoir through the mean of the flow rate subsequence; it reflects the presence of a trend in the flow rate data through the first slope; and finally, it comprehensively reflects the probability that the reservoir did not produce fluid during the corresponding data collection period through the combination of these two factors. It can reflect the possibility that the corresponding detection well section did not produce fluid during the collection period corresponding to the u-th flow rate subsequence; the larger the value, the more it indicates that the fluid production of the oil layer did not show a significant trend change during the collection period corresponding to the u-th flow rate subsequence, and the fluid production was very weak, which is more consistent with the characteristics of the oil layer being quiescent and not producing fluid.

[0043] Step 3: Based on the existence of silent subsequences in each detection well section, the changing trends of temperature data corresponding to each silent subsequence and each seepage subsequence in each detection well section, and combined with the flow dispersion of all seepage subsequences, an oil layer production index is constructed to characterize the possibility that each detection well section is a low-production layer.

[0044] Furthermore, when intermittent overflow occurs in the oil layer, it indicates that the oil layer has indeed produced fluid intermittently, thus increasing the likelihood that the oil layer is a true producing layer. However, considering that some true low-producing layers may not have intermittent overflow during data acquisition, further processing is required.

[0045] First, the flow rate subsequences corresponding to the quiescent periods of the oil reservoir are located. Specifically, due to the numerous segments of the oil reservoir and the consistency of flow rate data collected during quiescence, there are significant differences compared to flow rate data during normal production. Therefore, the reservoir quiescence index of each flow rate subsequence within each monitoring well segment is calculated. The reservoir quiescence index of all flow rate subsequences within all monitoring well segments is used as the input to the Otsu threshold method, and the output segmentation threshold is recorded as the preset quiescence threshold. Flow rate subsequences with a reservoir quiescence index greater than or equal to the preset quiescence threshold are recorded as quiescent subsequences; otherwise, they are recorded as seepage subsequences. This allows for the differentiation of each flow rate subsequence.

[0046] Because the permeability and liquid content of the low-yield liquid layer are low, the liquid production is usually relatively stable. Conversely, instrument or environmental errors such as probe scaling and shedding, noise, and contaminant scattering usually manifest as sudden and drastic changes in data characteristics.

[0047] Therefore, the dispersion degree between elements within each seepage subsequence of the i-th monitoring well section is calculated, and the mean of the dispersion degree of each seepage subsequence is denoted as the flow dispersion factor of the i-th monitoring well section. The flow dispersion factor can reflect the unsteady-state characteristics of the fluid produced by the oil-bearing layer corresponding to the i-th monitoring well section; the smaller the value, the more stable the overall fluctuation within the seepage subsequence, and the more consistent it is with the physical characteristics of weak seepage in low-permeability thin and poor-permeability layers. The methods for calculating the dispersion degree include, but are not limited to, variance, root mean square error, and coefficient of variation. In this embodiment, variance is used for calculation.

[0048] In the same manner, the flow dispersion factor of all monitoring well sections is calculated, and the maximum value of the flow dispersion factor of the i-th monitoring well section is normalized to eliminate the influence of dimensions.

[0049] Furthermore, due to the Joule-Thomson effect, the temperature of the oil layer will rise when producing fluid. Therefore, temperature data can be combined to further analyze the possibility that the i-th detection well section is a true low-permeability producing layer.

[0050] Based on the range of each flow rate subsequence in the flow rate sequence of the i-th detection well section, the temperature sequence of the i-th detection well section is also divided into corresponding ranges to obtain multiple temperature subsequences; and the temperature subsequence corresponding to the seepage subsequence is denoted as the seepage temperature subsequence, and the temperature subsequence corresponding to the quiescent subsequence is denoted as the quiescent temperature subsequence.

[0051] Each seepage temperature subsequence of the i-th monitoring well section is used as input to the Sen slope estimation algorithm to obtain the slope value of each seepage temperature subsequence, and the average value is calculated, which is denoted as the seepage temperature slope of the i-th monitoring well section. The seepage temperature slope can reflect the temperature evolution trend of the oil layer during fluid production; the smaller the negative value, the less it conforms to the characteristics of oil layer temperature rise during fluid production, thus indicating that there is a large error in the collected data and the oil layer is more likely not actually seeping; conversely, the larger the positive value, the more it conforms to the temperature change characteristics of oil layer during fluid production.

[0052] In addition, when the oil layer enters a quiescent state and production stops, the temperature of the monitoring section will also begin to drop because there is no continuous heat-carrying fluid injection.

[0053] Construct the intermittent factor for the i-th detection well section If the flow rate subsequence after segmentation of the i-th detection well section contains a silent subsequence, then... Setting it to 1 indicates that an overflow has occurred; otherwise, Setting it to 0 indicates that no intermittent overflow has occurred. It should be noted that since the classification is performed on all flow subsequences within all well segments, it is possible to analyze whether the flow subsequence after segmenting the i-th detection well segment contains a silent subsequence.

[0054] When the intermittent factor of the i-th monitoring interval is 1, the following analysis is performed: Each silent temperature subsequence of the i-th monitoring interval is used as input to the Sen slope estimation algorithm to obtain the slope value of each silent temperature subsequence, and the average value is calculated, denoted as the silent temperature slope of the i-th monitoring interval. The silent temperature slope reflects the temperature evolution trend of the oil reservoir during the silent process; a negative value and a smaller value indicate a decreasing trend in oil reservoir temperature during the silent period, and the more pronounced the decreasing trend, which is consistent with the physical process of thermal equilibrium recovery after production stops.

[0055] To avoid the influence of dimensions on subsequent calculation results, the reservoir quiescence index of all flow subsequences in all monitored well sections was calculated, and the maximum value normalization method was used to normalize all reservoir quiescence indices.

[0056] As a preferred implementation, an oil production index is constructed based on the presence of silent subsequences in each monitoring well section, the changing trends of temperature data corresponding to each silent subsequence and each seepage subsequence in each monitoring well section, and the flow dispersion of all seepage subsequences. This index is used to characterize the probability that each monitoring well section is a low-production layer. The construction process of the oil production index is as follows: when the flow subsequences of each monitoring well section contain silent subsequences, the oil production index of each monitoring well section is positively correlated with the seepage temperature slope and the mean of the oil production index of all silent subsequences, and negatively correlated with the silent temperature slope and the flow dispersion factor. When the flow subsequences of each monitoring well section do not contain silent subsequences, the oil production index of each monitoring well section is positively correlated with the seepage temperature slope and negatively correlated with the flow dispersion factor. The positive correlation means that the dependent variable increases (decreases) as the independent variable increases (decreases), and the negative correlation means that the dependent variable decreases (increases) as the independent variable increases (decreases).

[0057] The flowchart for obtaining the reservoir production index in this embodiment is as follows: Figure 2 As shown.

[0058] Preferably, in this embodiment, the oil production index of the i-th detection well section is denoted as... Its expression is: In the formula, Let be the oil production index of the i-th monitoring well section; It is the normalized mean of the reservoir silencing index of all silencing subsequences within the i-th detection well section; The normalized flow dispersion factor for the i-th monitoring well section; , These are the seepage temperature slope and the quiescent temperature slope of the i-th monitoring well section, respectively. This is the sigmoid function, used to normalize input data.

[0059] The reservoir production index reflects the presence of intermittent overflow characteristics in the corresponding reservoir by using the mean of the reservoir quiescent index; it reflects the stability of seepage in low-production reservoirs by using the flow dispersion factor; and it reflects whether the temperature response of the monitored well section conforms to the production pattern by using the seepage temperature slope and the quiescent temperature slope. The obtained reservoir production index can comprehensively reflect the probability that the reservoir corresponding to the i-th monitored well section is a true low-production reservoir by combining flow and temperature data; the larger the value, the more the reservoir corresponding to the i-th monitored well section conforms to the production characteristics of a true low-permeability production reservoir in terms of seepage pattern and thermal response, and the greater the probability that it is a true low-production reservoir.

[0060] Step 4: Obtain several real low-production fluid layers and false low-production fluid layers, and train a logistic regression model based on the oil layer production index of each low-production fluid layer, the quiescent index of all oil layers, and the flow dispersion. Then, determine whether each detection well section is a real low-production fluid layer based on the trained logistic regression model. Based on the location and production volume of the real low-production fluid layers in the production well, correct the injection profile in the injection well, and then calculate the degree of oil layer utilization.

[0061] Logistic regression is a generalized linear regression analysis model, belonging to supervised learning in machine learning. By training the logistic regression model with the calculated oil production index, the model's ability to distinguish real producing layers can be effectively improved, thereby enhancing the accuracy and reliability of identifying the activation state of low-permeability, thin, and poor-permeability layers.

[0062] According to the calculation method of reservoir production index, T (10 in this embodiment) low-production layers with actual production and T false low-production layers without actual production are selected from the existing reservoir production well group. Data acquisition and reservoir production index calculation are performed f (50 in this embodiment) times for each of the actual and false low-production layers. Then, the mean value of the reservoir quiescent index of all flow subsequences of each low-production layer in each data acquisition is calculated. Based on the reservoir production index, the mean value, and the flow dispersion factor of each low-production layer in each data acquisition process, a production vector of each low-production layer in each data acquisition process is constructed, and the label of the production vector of the actual low-production layer is set to 1, while the label of the production vector of the false low-production layer is set to 0.

[0063] The calculated production vectors and labels of all true low-production layers and all false low-production layers are used as input to the logistic regression model, which is then trained to obtain the trained logistic regression model. The loss function for training the logistic regression model is not limited to binary cross-entropy loss or weighted cross-entropy loss; the optimization algorithm is not limited to stochastic gradient descent or Adam. The training process of the logistic regression model is a well-known technique and will not be elaborated upon here.

[0064] The produced fluid vector integrates physical constraints such as seepage stability, intermittent characteristics, and thermal response consistency, transforming raw data into discriminative features with clear physical interpretations. This significantly improves the logistic regression model's ability to identify weak signals. Furthermore, using the produced fluid vector as input to the logistic regression model allows it to maintain structural simplicity while adaptively learning the weights of each feature, dynamically balancing the contribution of different physical processes to the actual produced fluid.

[0065] If the raw data is used directly to train and analyze the logistic regression model, the model is easily affected by sudden interference and dimensional differences, making it difficult to accurately identify the low permeability characteristics of the oil layer, which will lead to a high failure rate and poor generalization performance of low-yield liquid layers.

[0066] The production vectors of each monitoring section within the produced well are used as input to a trained logistic regression model. The output of the logistic regression model is the label corresponding to each monitoring section. If the label is 1, the corresponding monitoring section is determined to be a true low-production layer. This enables accurate assessment of the oil layer status corresponding to each monitoring section in the current produced well and precise location of the true low-production layer in the current produced well.

[0067] Furthermore, for the injection well, the injection flow rate into the well is measured using existing electromagnetic flow logging methods, and the location of the fluid-absorbing layers is analyzed. Simultaneously, a tracer is injected into the injection well; in this embodiment, fluorobenzoic acid is selected as the tracer, and the tracer is monitored and analyzed to locate each fluid-absorbing layer. The methods of locating the fluid-absorbing layers by injection flow rate, and the monitoring and analysis of the tracer and fluid-absorbing layers, are well-known techniques in the art and will not be elaborated upon here.

[0068] Oil layers that are identified as liquid-absorbing layers by tracers but are not assessed as liquid-absorbing layers by electromagnetic flow logging are all recorded as potential low-liquid-absorbing layers.

[0069] In this embodiment, the correspondence is defined by the depth range of the oil layer: the median value of the depth range of each potential low-suction layer is recorded as the center depth of each potential low-suction layer, and the median value of the depth range of each actual low-production layer in the produced well is recorded as the center depth of each actual low-production layer. Taking the c-th potential low-suction layer as an example, the absolute difference between it and the center depth of all actual low-production layers in the produced well is calculated successively, and the actual low-production layer corresponding to the minimum absolute difference is recorded as the actual low-production layer corresponding to the c-th potential low-suction layer.

[0070] The actual low-yield fluid layers corresponding to each potential low-supply layer are calculated sequentially to establish a one-to-one correspondence between the actual low-yield fluid layers in the production wells and the potential low-supply fluid layers in the injection wells. It should be noted that if the number of potential low-supply fluid layers and actual low-yield fluid layers is inconsistent, oil layers without a corresponding relationship are removed; and each fluid-supply layer can only correspond to one producing layer.

[0071] Next, based on the flow rate data of each actual low-yield fluid layer in the production well, the production volume of all actual low-yield fluid layers within a preset time period is calculated and summed to obtain the total production volume; then, the proportion of the production volume of each actual low-yield fluid layer in the total production volume is calculated. The calculation of the production volume is a well-known technique, and the specific process will not be elaborated further.

[0072] The liquid absorption volume of each non-potentially low-suction layer in the injection well is obtained within a preset time period. The total liquid absorption volume of all non-potentially low-suction layers in the injection well within the preset time period is calculated. The total injection volume of the injection well within the preset time period is subtracted from the total liquid absorption volume to obtain the total liquid absorption volume of all potentially low-suction layers in the injection well, which is recorded as the total liquid absorption volume. Then, according to the proportion of the actual low-production layer corresponding to each potentially low-suction layer, the total liquid absorption volume is allocated to each potentially low-suction layer according to the corresponding proportion to obtain the liquid absorption volume of each potentially low-suction layer, thereby generating a corrected injection profile.

[0073] Based on the corrected injection profile, the reservoir utilization level of the current well group is calculated, thereby accurately evaluating the reservoir utilization status and providing a reliable basis for subsequent adjustment measures. The calculation of reservoir utilization level is a well-known technique and will not be elaborated here.

[0074] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0075] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0076] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them; modifications to the technical solutions described in the foregoing embodiments, or equivalent substitutions of some of the technical features, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A refined evaluation method for the utilization degree of low-permeability, thin-layer oil reservoirs based on thermal ultrasonic logging, characterized in that, The method includes the following steps: Real-time acquisition of temperature and flow data of each producing section within the well within a preset time period; screening of each detection section from all producing sections according to flow rate to achieve initial location of low-producing layers; All flow data of each monitoring well section are divided into several flow subsequences. Based on the mean and trend of each flow subsequence, the oil layer quiescent index of each flow subsequence is constructed to characterize the probability of no fluid production in the monitoring well section within the time period corresponding to each flow subsequence. In this way, all flow subsequences are divided into quiescent subsequences and seepage subsequences. Based on the existence of silent subsequences in each detection well section, the changing trends of temperature data corresponding to each silent subsequence and each seepage subsequence in each detection well section, and combined with the flow dispersion of all seepage subsequences, an oil layer production index is constructed to characterize the possibility that each detection well section is a low-production layer. Several real and false low-production fluid layers are obtained, and a logistic regression model is trained based on the fluid production index of each low-production fluid layer, the quiescent index of all oil layers, and the flow dispersion. Then, the trained logistic regression model is used to determine whether each detection well section is a real low-production fluid layer. Based on the location and production volume of the real low-production fluid layers in the production well, the injection profile in the injection well is corrected, and the degree of oil layer utilization is calculated.

2. The method for fine evaluation of the utilization degree of low-permeability, thin-layer oil reservoirs based on thermal ultrasonic logging as described in claim 1, characterized in that, Each detection well segment refers to a producing well segment where the mean of all flow data within a preset time period is less than or equal to a preset segmentation threshold and is not equal to 0.

3. The method for fine evaluation of the utilization degree of low-permeability, thin-layer oil reservoirs based on thermal ultrasonic logging as described in claim 1, characterized in that, The reservoir quiescence index of each flow subsequence is negatively correlated with the mean of each flow subsequence and the absolute value of the slope of the fitted line of each flow subsequence.

4. The method for fine evaluation of the utilization degree of low-permeability, thin-layer oil reservoirs based on thermal ultrasonic logging as described in claim 1, characterized in that, The quiescent subsequence refers to the flow rate subsequence where the reservoir quiescent index is greater than or equal to a preset quiescent threshold; the seepage subsequence refers to the flow rate subsequence where the reservoir quiescent index is less than a preset quiescent threshold.

5. The method for fine evaluation of the utilization degree of low-permeability, thin-layer oil reservoirs based on thermal ultrasonic logging as described in claim 1, characterized in that, The process for constructing the reservoir production index is as follows: The mean slope value of the temperature data corresponding to all seepage subsequences in each detection well section is calculated and recorded as the seepage temperature slope of each detection well section. When the flow rate subsequence of each monitoring well section contains a silent subsequence: the mean slope value of the temperature data corresponding to all silent subsequences of each monitoring well section is calculated and recorded as the silent temperature slope of each monitoring well section; the oil production index of each monitoring well section is positively correlated with the seepage temperature slope and the mean value of the oil layer silent index of all silent subsequences, and negatively correlated with the silent temperature slope and the flow rate dispersion of all seepage subsequences; When the flow rate subsequence of each monitoring well section does not contain a silent subsequence: the oil production index of each monitoring well section is positively correlated with the seepage temperature slope and negatively correlated with the flow rate dispersion of all seepage subsequences.

6. The method for fine evaluation of the utilization degree of low-permeability, thin-layer oil reservoirs based on thermal ultrasonic logging as described in claim 5, characterized in that, The dispersion of flow rate of all seepage subsequences refers to the mean of the variances of all seepage subsequences within each monitoring well section.

7. The method for fine evaluation of the utilization degree of low-permeability, thin-layer oil reservoirs based on thermal ultrasonic logging as described in claim 1, characterized in that, The specific process for training the logistic regression model is as follows: Based on the oil layer production index of each low-yield fluid layer in each data acquisition process, the mean of the oil layer quiescence index of all flow subsequences, and the flow dispersion of all seepage subsequences, each low-yield fluid layer's production vector is constructed. Set the label of the production vector of the true low-production layer to 1 and the label of the production vector of the false low-production layer to 0. The production vectors and labels of all real low-production layers and all false low-production layers are used as inputs to the logistic regression model to train the model and obtain the trained logistic regression model.

8. The method for fine evaluation of the utilization degree of low-permeability, thin-layer oil reservoirs based on thermal ultrasonic logging as described in claim 7, characterized in that, The specific process for determining whether each detection well section is a true low-production fluid layer is as follows: the production vector of each detection well section is used as the input of the trained logistic regression model, and the output is the label corresponding to each detection well section. If its label is 1, then the corresponding detection well section is determined to be a real low-yield fluid layer.

9. The method for fine evaluation of the utilization degree of low-permeability, thin-layer oil reservoirs based on thermal ultrasonic logging as described in claim 1, characterized in that, The specific process for correcting the injection profile in the injection well is as follows: Oil-bearing segments that are identified as liquid-absorbing layers by tracers injected into the well but are not assessed as liquid-absorbing layers by electromagnetic flow logging are all recorded as potential low-liquid-absorbing layers. Based on the distance between each potential low-supply layer and each actual low-yield layer in the production well, a one-to-one correspondence is established between the actual low-yield layers in the production well and the potential low-supply layers in the injection well. The production volume of all real low-yield fluid layers in the well is counted within a preset time period and summed to obtain the total production volume. And calculate the proportion of the liquid production of each real low-yield liquid layer in the total liquid production; Obtain the total amount of liquid absorbed by all potential low-suction layers in the injection well; based on the proportion of the actual low-production layer corresponding to each potential low-suction layer, allocate the total amount of liquid absorbed to each potential low-suction layer according to the corresponding proportion, and obtain the amount of liquid absorbed by each potential low-suction layer, thereby generating the corrected injection profile.

10. The method for fine evaluation of the utilization degree of low-permeability, thin-layer oil reservoirs based on thermal ultrasonic logging as described in claim 9, characterized in that, The process of establishing the one-to-one correspondence is as follows: The absolute difference between the depth range of each potential low-suction layer and the median value of all actual low-production layers in the produced well is calculated. The actual low-yield liquid layer corresponding to the minimum absolute difference is denoted as the actual low-yield liquid layer corresponding to each potential low-absorption liquid layer.

Images