An organic matter hydrocarbon generation kinetics parameter calibration method for eliminating overfitting and multi-solution problems

Abstract

The present application relates to a kind of organic matter hydrocarbon generation kinetics parameter calibration method for eliminating overfitting and multi-solution problem, belong to oil and gas resource exploration technical field;Including: step 1: obtaining temperature-hydrocarbon generation conversion rate experimental data;Step 2: EasyRo-hydrocarbon generation conversion rate data set is constructed;Step 3: EasyRo-hydrocarbon generation conversion rate benchmark relationship is constructed;Step 4: temperature-hydrocarbon generation conversion rate data is obtained by inverse calculation;Step 5: parallel first-order reaction organic matter hydrocarbon generation kinetics model is constructed;Step 6: inner layer establishes organic matter hydrocarbon generation kinetics parameter calibration;Step 7: outer layer is established based on iterative verification of geological application, and global optimization is carried out to dynamic parameter;Step 8: whether the dynamic parameter meets experimental fitting precision and geological constraint condition is judged;Step 9: output optimal hydrocarbon generation kinetics parameter of hydrocarbon source rock sample.The present application provides more reliable geochemical parameter support.

Description

Technical Field

[0001] This invention relates to a method for calibrating the kinetic parameters of organic hydrocarbon generation to eliminate overfitting and multiple solutions, belonging to the field of oil and gas resource exploration technology. Background Technology

[0002] With long-term high-intensity development, conventional and easily exploitable shallow and medium-depth oil and gas resources are becoming increasingly depleted. Globally and in my country, the focus of oil and gas exploration and development is rapidly shifting strategically towards unconventional oil and gas resources (such as tight oil and gas, shale oil and gas) and deep-to-ultra-deep oil and gas fields (burial depths exceeding 6,000 meters or even 10,000 meters). However, these fields generally feature complex geological conditions, strong reservoir heterogeneity, and high drilling and testing costs. Accurately identifying recoverable underground resources and significantly reducing the trial-and-error risks in exploration deployment has become a key technical issue in current oil and gas exploration. Quantitative evaluation and sweet spot prediction of oil and gas resources are core tasks in modern oil and gas basin exploration. Especially in unconventional shale oil and gas and deep-to-ultra-deep oil and gas exploration, higher demands are placed on the accurate prediction of organic matter hydrocarbon generation thresholds, hydrocarbon production rates, and fluid phase states (oil and gas ratios). Organic matter hydrocarbon generation kinetics, as an important bridge connecting laboratory thermal simulation and geological historical thermal evolution processes, is the theoretical basis for achieving the above-mentioned quantitative evaluation and prediction.

[0003] Since the 1980s, Tissot and other Chinese and foreign geologists and geochemists have established the theoretical foundation of hydrocarbon generation kinetics based on the classical Arrhenius equation. This theoretical breakthrough completely changed the traditional petroleum geology's reliance on empirical formulas and qualitative observations, enabling the process of organic matter transforming into hydrocarbons to move from qualitative description to rigorous mathematical and physical quantitative calculations. Over decades of technological development, the hydrocarbon generation kinetics theory has been continuously deepened and expanded: from early bulk kinetics models that could only predict the total evolution of hydrocarbons generated from organic matter, it has gradually evolved into compositional kinetics capable of predicting the specific phases of hydrocarbons (such as the proportions of each component in crude oil, methane, ethane, etc.), as well as isotope kinetics for tracking gas source characteristics. The core idea is to use laboratory instruments (such as Rock-Eval rock pyrolysis apparatus or high-pressure closed gold tube pyrolysis system) to artificially accelerate the evolution process of organic matter at high temperatures of 300℃ to 600℃ for several hours to several days and at different fixed heating rates, to obtain high-precision hydrocarbon generation conversion curves, and to inversely calculate the kinetic parameters of kerogen pyrolysis (activation energy, pre-exponential factor and reaction fraction), and then extrapolate them to geological historical conditions (millions of years, low temperatures of 100℃ to 200℃) to reconstruct the actual hydrocarbon generation process in geological periods.

[0004] Currently, the most widely adopted hydrocarbon generation kinetics solution in industry and academia is the parallel first-order reaction model, and the relevant technical solution has been disclosed in Chinese patent document CN104156593A (Establishment of Shale Oil and Gas Yield Evaluation Model and Parameter Calibration Method under Closed System). Although the parallel first-order reaction model has been widely used in basin simulation, there are still shortcomings in the actual parameter calibration process: In the Arrhenius equation, the pre-exponential factor (A) and activation energy (E) have a strong mathematical coupling relationship, and the parallel first-order reaction model has a large number of parameters to be optimized, which easily leads to multiple solutions in the parameter calibration process; at the same time, under the constraint of limited experimental data, due to the small number of experimental measurement points and the unavoidable experimental errors and dispersion, traditional parameter calibration methods often take minimizing experimental fitting errors as the main goal, which easily leads to over-approximation of the limited measurement points, resulting in numerical overfitting problems.

[0005] Therefore, there is an urgent need for a method to calibrate the kinetic parameters of organic matter hydrocarbon generation, in order to effectively reduce the effects of overfitting under limited experimental data conditions, alleviate the ambiguity caused by the coupling of pre-exponential factors and activation energy, and combine experimental fitting with the long-term underground thermal evolution law, thereby improving the stability, uniqueness and geological applicability of the kinetic parameter calibration results.

[0006] In summary, the main technical problems existing in the prior art include: (1) How to reduce the overfitting of traditional dynamic parameter calibration methods to discrete experimental data and improve the continuity and stability of input data under the conditions of limited experimental measurement points and experimental noise.

[0007] (2) How to alleviate the parameter polysemy caused by the coupling between the pre-exponential factor and the activation energy in a parallel first-order reaction kinetic model, and improve the uniqueness and identifiability of the kinetic parameter calibration results.

[0008] (3) How to combine experimental fitting constraints with the geochemical laws of organic matter thermal evolution so that the obtained kinetic parameters can not only match the results of thermal simulation experiments, but also reasonably characterize the hydrocarbon generation and transformation process under long-term underground thermal evolution conditions.

[0009] The minor technical problems existing in the prior art include: (1) How to uniformly convert temperature-hydrogenation conversion rate data under multiple heating rate conditions to maturity-hydrogenation conversion rate dimension, so as to eliminate the thermal maturity scale deviation caused by the difference in heating rate, construct a unified scale for indoor thermal simulation experiment and underground geological thermal evolution process, and lay the foundation for subsequent geological constraints and accurate parameter calibration of dynamic parameters.

[0010] (2) How to construct a maturity-hydrogenation conversion rate relationship that satisfies the range of hydrocarbon generation conversion rate from 0 to 1 and increases monotonically with maturity, so as to ensure that the data processing results conform to the basic geochemical laws of organic matter thermal evolution. Summary of the Invention

[0011] To address the shortcomings of existing technologies, this invention provides a method for calibrating organic hydrocarbon generation kinetic parameters to eliminate overfitting and multiple solutions. The technical solution of this invention is as follows: A method for calibrating organic hydrocarbon generation kinetic parameters to eliminate overfitting and multiple solutions includes: Step 1: Select low-maturity source rock samples for thermal simulation experiments to obtain temperature-hydrogenation conversion rate experimental data; Step 2: Based on the Sweeney & Burnham vitrinite reflectance equivalent model, establish the temperature-maturity correspondence and construct the EasyRo-hydrocarbon conversion rate dataset; Step 3: Establish a bounded monotonic constrained Gaussian process regression model and construct a continuous and stable EasyRo-hydrogenation conversion rate benchmark relationship; Step 4: Based on the EasyRo-hydrogenation conversion rate benchmark relationship, perform noise reduction, smoothing, and interpolation on the EasyRo-hydrogenation conversion rate data, and back-calculate the temperature-hydrogenation conversion rate data as input data for the kinetic model; Step 5: Construct a parallel first-order reaction kinetic model for hydrocarbon generation from organic matter; Step 6: Based on temperature-hydrogenation conversion data, establish calibration of organic matter hydrocarbon generation kinetic parameters in the inner layer; Step 7: Establish iterative verification based on geological applications in the outer layer and perform global optimization of dynamic parameters; Step 8: Determine whether the dynamic parameters meet the experimental fitting accuracy and geological constraints. If yes, proceed to step 9; otherwise, if it is not the optimal solution, adjust the pre-exponential factor and return to step 6. Step 9: Output the optimal hydrocarbon generation kinetic parameters for the source rock sample.

[0012] According to a preferred embodiment of the present invention, the specific implementation process of step 1 includes: Low-maturity source rock samples were selected, and thermal simulation experiments were carried out. The changes in hydrocarbon generation products of low-maturity source rock samples during the heating process were continuously recorded, and experimental data on temperature-hydrogenation conversion rate under different heating rates were obtained. Based on the cumulative hydrocarbon generation measured experimentally, temperature-hydrocarbon generation conversion rate experimental data were obtained at different heating rates; the hydrocarbon generation conversion rate is defined as the proportion of the cumulative hydrocarbon generation at a certain temperature to the total hydrocarbon generation potential, i.e.: (1); Where X(T) is the hydrocarbon conversion rate at temperature T; Y(T) is the cumulative hydrocarbon generation, in mg HC / g TOC; Y max The total hydrocarbon generation amount or the corrected total hydrocarbon generation potential corresponding to the experimental endpoint is expressed in mg HC / g TOC.

[0013] According to a preferred embodiment of the present invention, step 2 includes the following specific implementation process: In the data conversion process, the Sweeney & Burnham vitrinite reflectance equivalent model was used to characterize the degree of organic matter thermal evolution. The vitrinite thermal maturation process was regarded as a multi-component parallel first-order Arrhenius reaction, with each reaction having the same pre-exponential factor A and different activation energies E. i The reaction rate equation for a single component is: (2); in, Let A be the mass of unreacted organic matter of the i-th component at any time t; t is time, in seconds; A is the pre-exponential factor; E is the mass of unreacted organic matter of the i-th component at any time t ... A is the mass of unreacted organic matter of the i-th component at any time t; E is the mass of unreacted organic matter of the i-th component at any time t; E is the mass of unreacted organic matter of the i-th component at any time t; E is the mass of unreacted organic matter of the i-th component at any time t; E is the mass of unreacted organic matter i Let R be the activation energy of the i-th reaction group; R be the gas constant; and T(t) be the absolute temperature as a function of time. The overall reaction rate is the sum of the reaction rates of each component: (3); in, The number of components for parallel first-order reactions; The total unreacted organic mass of all components at time t; Introducing the stoichiometric factor f i The overall degree of reaction is obtained by weighted summation of the conversion ratios of each component. : (4); in, This represents the initial mass of the organic matter. Let be the initial mass of organic matter in the i-th component; The empirical formula for the relationship between total reaction degree F and maturity level EasyRo is as follows: (5); Through data conversion, the temperature-hydrogenation conversion rate data obtained under different heating rate experimental conditions were uniformly converted into EasyRo-hydrogenation conversion rate data.

[0014] According to a preferred embodiment of the present invention, step 3 specifically includes the following steps: Let EasyRo be the input variable x, and the hydrocarbon generation conversion rate be the output variable y; introduce the latent function f(x), and establish the relationship between the hydrocarbon generation conversion rate and the latent function through bounded monotonic mapping: (6); In the formula, f(x) represents the potential functional relationship between EasyRo and hydrocarbon generation conversion; the predicted hydrocarbon generation conversion rate is guaranteed to satisfy: (7); Assume that the function f(x) follows a Gaussian process distribution: (8); in, Let f(x) represent a Gaussian process; x and x′ represent two different samples; m(x) is the mean function, representing the prior expectation of f(x); k(x,x′) is the covariance function, which is used to characterize the correlation of hydrocarbon conversion rate changes between different maturity points x and x′. The covariance function uses a squared exponential kernel function: (9); Where, σ f 2 The variable represents the variance of the function signal; l represents the characteristic length scale. For the training sample input vector: , and their corresponding predicted values: ; This represents the nth EasyRo input value. The output value represents the conversion rate of the nth hydrocarbon generation. The latent variable observation vector is obtained by performing an inverse function transformation on the observed values ​​using equation (6). : , (10); Constructing the covariance matrix of training samples : (11); The joint distribution of the training samples after adding the noise term is: (12); Where N is a multivariate normal distribution; Indicates the variance of observation noise; It is an n-dimensional identity matrix; Observed value z and predicted value The joint prior distribution is: (13); In the formula, , for test points The n×1 covariance matrix between the input X of the training set and the input X; For test points Its own covariance; The test point is obtained from equation (13). latent function value The posterior distribution is: (14); Among them, the mean of the posterior distribution for: (15); Posterior prediction variance for: (16); The predicted mean of the latent function is converted into the predicted value of the hydrocarbon generation conversion rate through the bounded mapping in equation (5), and finally a continuous and stable maturity-hydrogenation conversion rate benchmark relationship is generated, namely EasyRo-hydrogenation conversion rate benchmark relationship: (17); At the preset maturity nodes, a non-negativity constraint is set on the first derivative of the latent function, namely: (18); The obtained hydrocarbon generation conversion rate prediction curve also satisfies: (19); This ensures that the baseline relationship changes incrementally overall.

[0015] According to a preferred embodiment of the present invention, in step 4, the continuous and stable EasyRo-hydrogenation conversion rate benchmark relationship obtained in step 3 is used for experimental data interpolation reconstruction and geological application iterative verification, respectively. On the one hand, under the constraints of experimental heating conditions, based on the EasyRo-hydrogenation conversion rate benchmark relationship, the temperature-hydrogenation conversion rate experimental data are interpolated and reconstructed; specifically, the experimental heating rate is used as the preset heating rate, the temperature range defined by the experimental start temperature and end temperature is used as the interpolation range, and multiple interpolation temperature nodes are set within the interpolation range according to the preset temperature step size. For each interpolated temperature node, the Sweeney & Burnham vitrinite reflectance equivalent model in step 2 is used to convert the corresponding time-temperature history into the maturity parameter EasyRo, and the obtained EasyRo value is substituted into the equation (17) obtained in step 3 to obtain the hydrocarbon generation conversion rate corresponding to each interpolated temperature node, and then the interpolated temperature-hydrocarbon generation conversion rate data sequence is constructed. On the other hand, a parallel first-order reaction organic hydrocarbon generation kinetic model suitable for the thermal evolution process of organic matter is established, and the kinetic parameters of the parallel first-order reaction organic hydrocarbon generation kinetic model are calibrated. The parallel first-order reaction organic hydrocarbon generation kinetic model is used to describe the organic hydrocarbon generation process, which includes the stages of oil generation, gas generation, and oil-to-gas conversion.

[0016] More preferably, the temperature step size is 5°C.

[0017] According to a preferred embodiment of the present invention, step 5 involves constructing a parallel first-order reaction kinetic model for hydrocarbon generation from organic matter; including: Suppose there is an organic hydrocarbon generation process consisting of NH4+ and NH4+ parallel first-order reactions, with the activation energy of each parallel first-order reaction being EH2. i Pre-exponential factor AH i Let XH be the initial potential amount of organic matter corresponding to each reaction. i0 If i = 1, 2, ..., NH, then the total hydrocarbon generation of NH parallel reactions is... for: (20); in, Let XH be the amount of hydrocarbon generated in the i-th reaction, D be the heating rate, R be the gas constant, and T be the absolute temperature. When XH is XO, it is oil generation; when XH is XG, it is gas generation; and when XH is XOG, it is oil-to-gas conversion.

[0018] According to a preferred embodiment of the present invention, in step 6, based on temperature-hydrogenation conversion data, the inner layer establishes calibration of organic matter hydrocarbon generation kinetic parameters; including: Step 6.1: Construct the objective function; Suppose that at a certain heating rate l, when a certain temperature j is reached, the experimentally measured hydrocarbon conversion rate is XH1. lj Under the same conditions, assuming EH i AH i XH i Then, the hydrocarbon generation conversion rate calculated by equation (20) is XH. lj Construct the objective function : (twenty one); Where L0 is the number of experiments with different heating rates, and J0 is the number of sampling points from an experimental curve; Then equation (21) is transformed into equation (22): (twenty two); In equation (20), AH i XH i0 satisfy: (twenty three); in, It is a small integer; Step 6.2: Construct the penalty function; The penalty function method is used to transform the process of finding the minimum objective function under constraints into finding it under unconstrained conditions; the process is as follows: For any given constraint, construct a penalty term G. When the minimum value of the obtained function satisfies the constraint, the function value is 0; otherwise, it is a positive number. For AH i Given the constraint >0, we have: (twenty four); Right now: (25); Similarly, we have: (26); Right now: (27); Similarly, we have: (28); Right now: (29); Based on the established function, the penalty term is: (30); Take a sufficiently large positive integer R1, and construct the penalty function using equations (22) and (30): (31); If the minimum point obtained exceeds the constraint conditions, then gradually increase R1. When R1 is sufficiently large, the minimum solution of equation (31) is the minimum solution of the objective function equation (22). Step 6.3: Solve for the first-order partial derivatives of the objective function and the penalty function; A necessary condition for the existence of a minimum is that the first-order partial derivative of the penalty function is 0; first, take the partial derivative with respect to the objective function: (32); In the formula: (33); In the formula, m = 1, 2, 3, ... NH; Similarly, the partial derivative of the penalty term is: (34); (35); In the formula, FN is the expression symbol inside the parentheses, that is: (36); At the minimum point, we have: (37); Step 6.4: Finding the approximate minimum point; The objective function corresponding to equation (31) is optimized and solved using the variable metric method of the second derivative matrix and its inverse matrix, in order to calibrate the parameters of the organic matter hydrocarbon generation kinetic model, including: Given an initial point The first derivative of function (31) at the initial point is calculated using equation (37). Simultaneously, the approximate inverse of the second derivative matrix of the function at that initial point is calculated. According to the variable metric method, To ensure that the function value of equation (31) decreases in a certain direction, a one-dimensional search is performed in that direction to determine the step size that optimizes the decrease in the objective function value. That is, to find an approximate solution that is closer to the minimum point. Calculate the gradient at that point. If the gradient is less than a given small positive number That is, we consider the point to be an approximate solution to the minimum; otherwise, when If the convergence condition has not yet been met, the approximate value of the inverse of the second derivative matrix is ​​updated based on the parameter changes and gradient changes between two adjacent iteration points to obtain the approximate value of the inverse matrix at the new point. Find the direction in which the function value decreases at the new point. The process involves one-dimensional search, parameter update, and convergence determination steps to obtain subsequent iteration points. ; After the above iterative process, at a certain iteration point The magnitude of the gradient of the objective function is less than When, the iteration point The approximate minimum point of the objective function is used as the parameter value corresponding to the approximate minimum point, and the calibration result of the organic matter hydrocarbon generation kinetic model, i.e., equation (20), is used.

[0019] According to a preferred embodiment of the present invention, in step 7, the outer layer establishes an iterative verification based on geological applications and performs global optimization of dynamic parameters; including: Suppose the strata under investigation are divided into M layers, denoted from top to bottom as follows: The time when the i-th layer begins deposition is denoted as t, and the time when it ends is denoted as t. Let the l-th sublayer of the i-th layer be the source rock. The burial depth during the layer deposition was , The geothermal gradient and surface temperature during the deposition of the i-th layer are: and The hydrocarbon generation conversion rate of the i-th layer source rock during the deposition of the j-th layer was... The hydrocarbon generation conversion rate at the end of sedimentation was... The amount of oil generated during the deposition of layer j alone was Then, the hydrocarbon generation conversion rate at any time t can be obtained from equation (20). for: (38); (39); The conversion rate of organic matter to hydrocarbons can be calculated from equation (38). When XH is XO, it is oil production; when XH is XG, it is gas production; and when XH is XOG, it is oil to gas. After the first calibration of the hydrocarbon generation kinetic parameters of the inner organic matter, the combination of kinetic parameters was obtained. Using equation (38) for geological applications, the current hydrocarbon generation conversion rate is obtained. The EasyRo-hydrogenation conversion benchmark relationship obtained after processing by a bounded monotonic constrained Gaussian process regression model is as follows: Calculate the average difference between the hydrocarbon generation conversion rate after the geological application of the current kinetic parameters and the baseline hydrocarbon generation conversion rate. The sum of squared residuals is : (40); (41); in, Used for judgment Adjustment direction, Used to characterize the degree of matching between the geological application results under the nth round of outer layer iteration and the EasyRo-hydrogenation conversion rate benchmark relationship; In adjacent outer iterations, the pre-exponential factor is adjusted using an order-of-magnitude step search method, and the update method is expressed as follows: (42); In the formula, Preset step size; After each completion After adjustment, the updated version will be available. The inner layer parameters were recalibrated using the new initial values ​​to simultaneously obtain the corresponding activation energy distribution. and reaction fraction Furthermore, geological application calculations were conducted to obtain a new round of residual sum of squares. ; First adjustment At that time, with As a criterion, when Then decrease Conversely, when Then increase After the initial adjustment Using these as initial values, internal parameter calibration and external geological verification were performed to obtain... For the results after the initial adjustment, if If the adjustment indicates a worse matching degree, then stop the outer layer iterative verification and output the current dynamic parameters; if If the adjustment is effective, then continue to adjust according to formula (42). Perform iterative corrections; During continuous iteration, record the corresponding time intervals. Change; when the following conditions are met: and When determining the iteration corresponding to the (n-1)th iteration To approximate the minimum value, the pre-exponential factor, activation energy distribution, and reaction fraction corresponding to the current cycle are determined as the optimal organic matter hydrocarbon generation kinetic parameters for the source rock sample, namely: (43)

[0020] The beneficial effects of this invention are as follows: This invention addresses the overfitting and multiple-solution problems commonly found in existing organic hydrocarbon generation kinetic parameter calibration processes by proposing a method to eliminate these issues. First, the temperature-hydrocarbon generation conversion rate data obtained from thermal simulation experiments are uniformly converted into an EasyRo-hydrocarbon generation conversion rate relationship using the Sweeney & Burnham vitrinite reflectance equivalent model. A continuous and stable EasyRo-hydrocarbon generation conversion rate benchmark relationship is constructed using a bounded monotonically constrained Gaussian process regression model. Based on this benchmark relationship, the EasyRo-hydrocarbon generation conversion rate data is denoised, smoothed, and interpolated. Without altering the original experimental heating series and temperature range, the discrete experimental measurement points are expanded into a high-resolution temperature-hydrocarbon generation conversion rate data sequence, reducing the influence of experimental noise and local outliers. This avoids overfitting of traditional kinetic parameter calibration methods to a limited number of experimental points, improving the robustness and accuracy of the kinetic parameter calibration. Based on this, this invention establishes a two-layer iterative optimization mechanism that combines the calibration of inner-layer organic matter hydrocarbon generation kinetic parameters with outer-layer geological application verification. Using EasyRo as a unified maturity characterization parameter, it establishes an equivalent constraint relationship between experimental thermal simulation conditions and geological basin thermal evolution conditions. The laboratory calibration results are then placed under the actual geological thermal evolution background for iterative geological application verification, thereby resolving the ambiguity caused by the coupling of pre-exponential factors and activation energy in parallel first-order reaction models. The kinetic parameters obtained by this invention are more consistent with actual subsurface thermal evolution processes, providing more reliable geochemical parameter support for hydrocarbon generation history reconstruction from source rocks, resource potential assessment, and prediction of unconventional and deep-ultra-deep oil and gas sweet spots. This has significant theoretical and practical value. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating a method for calibrating kinetic parameters of organic hydrocarbon generation to eliminate overfitting and multiple solutions according to the present invention. Figure 2 A comparison chart of the experimental value of organic matter gas conversion rate in the thermal simulation experiment of mudstone gold tube in Well X1 of Bohai Bay Basin and the calculated value obtained by calibration of this invention. Figure 3 The distribution of organic matter gas generation activation energy corresponding to the thermal simulation experiment of mudstone gold tube in Well X1 in Bohai Bay Basin. Figure 4 A comparison chart of the experimental values ​​of EasyRo-gas conversion rate and the calculated values ​​after iterative verification and optimization of the outer geological application in the thermal simulation experiment of mudstone gold pipe in Well X1 of Bohai Bay Basin. Figure 5 A comparison chart of the experimental value of organic matter gas conversion rate in the Rock-Eval thermal simulation experiment of the X2 shale in the Ordos Basin and the calculated value obtained by calibration according to the present invention. Figure 6The diagram shows the distribution of organic matter gas generation activation energy corresponding to the Rock-Eval thermal simulation experiment of the X2 shale in the Ordos Basin. Figure 7 This is a comparison chart of the experimental values ​​of EasyRo—gas conversion rate in the Rock-Eval thermal simulation experiment of the X2 shale in the Ordos Basin and the calculated values ​​after iterative verification and optimization using outer geological applications. Detailed Implementation

[0022] To make the technical solution, beneficial effects and technical advantages of the present invention clearer and more complete, the technical solution of the present invention will be described in detail and completely below with reference to the accompanying drawings and specific examples.

[0023] Terminology Explanation: The Sweeney & Burnham vitrinite reflectance equivalent model, proposed by Sweeney and Burnham in 1990, is based on the theory of multicomponent parallel first-order Arrhenius reactions. It decomposes the thermal maturation process of vitrinite into a series of parallel hydrocarbon generation reactions with different activation energies and unified pre-exponential factors, comprehensively characterizing the cumulative thermal effect of temperature and time. This model can convert any thermal history (temperature-time series) into an equivalent vitrinite reflectance (EasyRo), uniformly correcting the maturity differences in thermal simulation experiments at different heating rates and long-term thermal evolution processes at geological scales. It achieves normalized matching between indoor experiments and underground geological thermal evolution scales, providing a unified maturity scale for experimental fitting and geological applications of hydrocarbon generation kinetic parameters.

[0024] Example 1 A method for calibrating organic hydrocarbon generation kinetic parameters to eliminate overfitting and multiple solutions, such as... Figure 1 As shown, it includes: Step 1: Select low-maturity source rock samples for thermal simulation experiments to obtain temperature-hydrogenation conversion rate experimental data; Step 2: Based on the Sweeney & Burnham vitrinite reflectance equivalent model, establish the temperature-maturity correspondence and construct the EasyRo-hydrocarbon conversion rate dataset; Step 3: Establish a bounded monotonic constrained Gaussian process regression model and construct a continuous and stable EasyRo-hydrogenation conversion rate benchmark relationship; Step 4: Based on the EasyRo-hydrogenation conversion rate benchmark relationship, perform noise reduction, smoothing, and interpolation on the EasyRo-hydrogenation conversion rate data, and back-calculate the temperature-hydrogenation conversion rate data as input data for the kinetic model; Step 5: Construct a parallel first-order reaction kinetic model for hydrocarbon generation from organic matter; Step 6: Based on temperature-hydrogenation conversion data, establish calibration of organic matter hydrocarbon generation kinetic parameters in the inner layer; Step 7: Establish iterative verification based on geological applications in the outer layer and perform global optimization of dynamic parameters; Step 8: Determine whether the dynamic parameters meet the experimental fitting accuracy and geological constraints. If yes, proceed to step 9; otherwise, if it is not the optimal solution, adjust the pre-exponential factor and return to step 6. Step 9: Output the optimal hydrocarbon generation kinetic parameters for the source rock sample.

[0025] Example 2 The method for calibrating organic hydrocarbon generation kinetic parameters according to Example 1, which eliminates overfitting and multiple solutions, differs in that: This embodiment uses the mudstone from well X1 in the Bohai Bay Basin as an example to provide a more detailed explanation of the organic matter hydrocarbon generation kinetic parameter calibration method established in this invention to eliminate overfitting and multiple solutions.

[0026] This experiment employed a closed-system gold tube thermal simulation, using a gold tube with an inner diameter of 4 mm and a length of 40 mm. The gold tube, containing mudstone samples from well X1 in the Bohai Bay Basin, was placed inside an autoclave. Water was introduced into the high-pressure pump to apply pressure to the sample within the gold tube. A pressure and temperature control system maintained the set temperature and pressure in the autoclave. The temperature was initially rapidly increased from room temperature to 200℃, with subsequent increases at rates of 2℃ / h and 20℃ / h, reaching a final temperature of 600℃. Upon reaching the target temperature, the autoclave was immediately removed and rapidly cooled to terminate further pyrolysis of the sample as much as possible. After the thermal simulation, the hydrocarbon products generated within the gold tube were separated and analyzed. These included gaseous hydrocarbons (C1–C5) and light hydrocarbons (C6–C5). 14 Gas chromatography was used for detection, combined with internal standard method for quantification; heavy hydrocarbons and pyrolytic asphalt were determined by gravimetric method. The yield of each product was calculated based on the initial sample mass.

[0027] The specific implementation process of step 1 includes: Based on the experimentally measured gaseous hydrocarbon yield, temperature-gas conversion rate experimental data were obtained at different heating rates; the gas conversion rate is defined as the proportion of cumulative gas production at a certain temperature to the total gas production potential, i.e.: (1); Where X(T) is the gas conversion rate at temperature T; Y(T) is the cumulative gas production, in mg HC / g TOC; Y max The total gas production or corrected total gas production potential corresponds to the experimental endpoint, expressed in mg HC / g TOC.

[0028] The specific implementation process of step 2 includes: The temperature and gas conversion rate data obtained in step 1 under multiple heating rates were corrected for temperature-time cumulative effect based on the Sweeney & Burnham classic organic matter thermal evolution model. The experimental data under different heating rates were uniformly converted into EasyRo maturity parameters, and a dataset of correspondence between EasyRo and gas conversion rate was established.

[0029] In the data conversion process, the Sweeney & Burnham vitrinite reflectance equivalent model was used to characterize the degree of organic matter thermal evolution. The vitrinite thermal maturation process was regarded as a multi-component parallel first-order Arrhenius reaction, with each reaction having the same pre-exponential factor A and different activation energies E. i The reaction rate equation for a single component is: (2); in, Let A be the mass of unreacted organic matter of the i-th component at any time t; t is time, in seconds; A is the pre-exponential factor; E is the mass of unreacted organic matter of the i-th component at any time t ... A is the mass of unreacted organic matter of the i-th component at any time t; E is the mass of unreacted organic matter of the i-th component at any time t; E is the mass of unreacted organic matter of the i-th component at any time t; E is the mass of unreacted organic matter of the i-th component at any time t; E is the mass of unreacted organic matter i Let be the activation energy of the i-th reaction group; R is the gas constant, 8.31447 kJ·mol⁻¹. -1 ·K -1 T(t) is the absolute temperature (K) that varies with time. The overall reaction rate is the sum of the reaction rates of each component: (3); in, The number of components for parallel first-order reactions; The total unreacted organic mass of all components at time t; Introducing stoichiometric factors (weighting coefficients) f i The overall degree of reaction is obtained by weighted summation of the conversion ratios of each component. : (4); in, This represents the initial mass of the organic matter. Let be the initial mass of organic matter in the i-th component; The empirical formula for the relationship between total reaction degree F and maturity level EasyRo is as follows: (5); Through data conversion, the temperature-gas conversion rate data obtained under different heating rate experimental conditions were uniformly converted into EasyRo-gas conversion rate data. The time-temperature effects corresponding to different heating processes were normalized to the same maturity scale, thereby reducing the impact of heating path differences on experimental results and improving the comparability and consistency between different experimental data.

[0030] The specific implementation process of step 3 includes: A bounded monotonic constrained Gaussian process regression model is established, and the correlation between conversion rate and EasyRo is characterized by the covariance function. A continuous and stable EasyRo-gas conversion rate benchmark relationship is constructed, specifically including: To ensure that the EasyRo-gas conversion rate relationship conforms to the geochemical laws governing the thermal evolution of source rock organic matter, a range constraint and a monotonicity constraint are imposed on the gas conversion rate during the Gaussian process regression modeling. Specifically, EasyRo is denoted as the input variable x, and the gas conversion rate as the output variable y. Considering that the physical meaning of the gas conversion rate requires a value between 0 and 1, and that it generally increases with increasing thermal maturity, an unconstrained Gaussian process model is not directly established for y. Instead, a latent function f(x) is introduced, and the relationship between the gas conversion rate and the latent function is established through a bounded monotonic mapping.

[0031] (6); In the formula, f(x) represents the latent functional relationship between EasyRo and the life conversion; since the Sigmoid mapping function used in formula (6) has a range of (0,1), it can be guaranteed that the predicted life conversion rate satisfies: (7); This aligns with the physical definition of gas conversion rate and the requirements for geological applications.

[0032] Considering that the gas conversion rate changes continuously with increasing maturity during the thermal evolution of organic matter, and that there is a strong correlation between adjacent maturity intervals, the function f(x) is assumed to follow a Gaussian process distribution: (8); in, Let f(x) represent a Gaussian process; x and x′ represent two different samples; m(x) is the mean function, representing the prior expectation of f(x); this invention adopts the zero-mean prior hypothesis (m(x)=0); k(x,x′) is the covariance function, which is used to characterize the correlation between the changes in the life conversion rate between different maturity points x and x′; for sample points with similar maturity, a larger covariance value indicates that the trends of the life conversion rate changes between the two are more similar; for sample points with large differences in maturity, a smaller covariance value indicates that the correlation is weakened.

[0033] The covariance function uses a squared exponential kernel function: (9); Where, σ f 2The function represents the variance of the signal; l represents the characteristic length scale; the parameter l is used to control the smoothness of the hydrocarbon conversion rate as it changes with maturity; when l is larger, the prediction curve is smoother; when l is smaller, the curve is more sensitive to local changes.

[0034] For the training sample input vector: , and their corresponding predicted values: ; This represents the nth EasyRo input value. This represents the output value of the nth anger conversion rate; The latent variable observation vector is obtained by performing an inverse function transformation on the observed values ​​using equation (6). : , (10); Constructing the covariance matrix of training samples : (11); The joint distribution of the training samples after adding the noise term is: (12); Where N is a multivariate normal distribution; Indicates the variance of observation noise; It is an n-dimensional identity matrix; Observed value z and predicted value The joint prior distribution is: (13); In the formula, , for test points The n×1 covariance matrix between the input X of the training set and the input X; For test points Its own covariance; The test point is obtained from equation (13). latent function value The posterior distribution is: (14); Among them, the mean of the posterior distribution for: (15); Posterior prediction variance for: (16); The latent function prediction mean is converted into the gaseous energy conversion rate prediction value through the bounded mapping in equation (5), and finally a continuous and stable maturity-gaseous energy conversion rate benchmark relationship, namely EasyRo-gaseous energy conversion rate benchmark relationship, is generated: (17); To ensure that the baseline relationship conforms to the geochemical characteristic that the gas conversion rate monotonically increases with maturity, a non-negative constraint is set on the first derivative of the potential function at a preset maturity node, namely: (18); Since the mapping function in equation (6) is a monotonically increasing function, when the latent function satisfies equation (18), the resulting gas conversion rate prediction curve also satisfies: (19); This ensures that the baseline relationship changes incrementally overall.

[0035] In step 4, the continuous and stable EasyRo-gas conversion rate benchmark relationship obtained in step 3 is used for experimental data interpolation reconstruction and geological application iterative verification. On the one hand, under the constraints of experimental heating conditions, the temperature-gas conversion rate experimental data are interpolated and reconstructed based on the EasyRo-gas conversion rate benchmark relationship; specifically, the experimental heating rate is used as the preset heating rate, the temperature range defined by the experimental start temperature and end temperature is used as the interpolation range, and multiple interpolation temperature nodes are set within the interpolation range according to the preset temperature step size. For each interpolated temperature node, the Sweeney & Burnham vitrinite reflectance equivalent model in step 2 is used to convert the corresponding time-temperature history into the maturity parameter EasyRo, and the obtained EasyRo value is substituted into equation (17) obtained in step 3 to obtain the gas conversion rate corresponding to each interpolated temperature node, and then the interpolated temperature-gas conversion rate data sequence is constructed; as the input data for the subsequent gas dynamics model calibration.

[0036] On the other hand, the continuous and stable EasyRo-gas conversion rate benchmark relationship is used for iterative verification under subsequent geological application conditions to achieve maturity-conversion rate comparison and kinetic parameter correction between experimental conditions and geological conditions. A parallel first-order reaction organic matter gas generation kinetic model suitable for the thermal evolution process of organic matter was established, and the kinetic parameters of the parallel first-order reaction organic matter gas generation kinetic model were calibrated. The parallel first-order reaction organic matter gas generation kinetic model is used to describe the organic matter gas generation process, which includes the stages of oil generation, gas generation, and oil-to-gas conversion.

[0037] The temperature step size is 5°C.

[0038] In step 5, a parallel first-order reaction kinetic model for the generation of organic matter is constructed, including: Suppose that an organic gas generation process consists of NG parallel first-order reactions, and the activation energy for each parallel first-order reaction is EG. i Pre-exponential factor AG i Let XG be the initial potential amount of organic matter corresponding to each reaction. i0 If i = 1, 2, ..., NG, then the total amount of gas produced by NG parallel reactions is... for: (20); in, Let be the amount of gas produced in the i-th reaction, D be the heating rate, and R be the gas constant (8.31447 kJ·mol⁻¹). -1 ·K -1 T represents absolute temperature (K).

[0039] In step 6, based on the temperature-gas conversion rate data, the kinetic parameters of organic matter gas generation in the inner layer are calibrated; including: Taking the parameter calibration of the kerogen oil extraction model as an example, the following steps are performed: Step 6.1: Construct the objective function; Suppose that at a certain heating rate l, when a certain temperature j is reached, the gas conversion rate measured by the experiment is XG1. lj Under the same conditions, assuming EG i AG i XG i Then, the life force conversion rate calculated by equation (20) is XG. lj If there exists a set EG i AG i XG i0 The value of is such that for all 1 and j, XG1 lj — XG lj =0, then the group EG i AG i XG i0 This is what we are looking for. However, due to experimental errors and other reasons, this is practically impossible. Therefore, we can only find the value that makes XG1... lj —XG lj Smallest possible EG i AG i XG i0 The value of is determined. Therefore, an objective function is constructed. :

[0040] (twenty one); Where L0 is the number of experiments with different heating rates, and J0 is the number of sampling points from an experimental curve; In principle, a larger number of parallel reactions would be more likely to include all types of reactions involved in the generation of organic matter, and therefore should be more accurate. However, this approach is impractical because the computational burden during model calibration and subsequent application becomes too great. Furthermore, in actual calibration, it was found that while the model's fit to experimental data generally improves with a greater number of parallel reactions, this improvement becomes insignificant once the number of parallel reactions reaches a certain threshold. Therefore, we only need a finite number of parallel reactions with a certain interval.

[0041] Due to EH i The solution can be obtained by determining the distribution range of activation energies of parallel reactions and the interval between activation energies of adjacent parallel reactions, thus transforming equation (21) into equation (22): (twenty two); In equation (20), AG i XG i0 (Expressed as a percentage of total reactive matter) satisfies: (twenty three); in, It is a small integer; Therefore, the problem of obtaining the dynamic parameters of the model (Equation (20)) is transformed into the problem of finding the minimum point of the non-negative objective function (Equation (22)) when the constraint condition (Equation (23)) is satisfied.

[0042] Step 6.2: Construct the penalty function; The penalty function method is used to transform the process of finding the minimum objective function under constraints into finding it under unconstrained conditions; the process is as follows: For any given constraint, construct a penalty term G. When the minimum value of the obtained function satisfies the constraint, the function value is 0; otherwise, it is a positive number. For AG i Given the constraint >0, we have: (twenty four); Right now: (25); Similarly, we have: (26); Right now: (27); Similarly, we have: (28); Right now: (29); Based on the established function, the penalty term is: (30); Take a sufficiently large positive integer R1, and construct the penalty function using equations (22) and (30): (31); If the minimum point obtained exceeds the constraint conditions, then gradually increase R1. When R1 is sufficiently large, the minimum solution of equation (31) is the minimum solution of the objective function equation (22). Step 6.3: Solve for the first-order partial derivatives of the objective function and the penalty function; A necessary condition for the existence of a minimum is that the first-order partial derivative of the penalty function is 0; first, take the partial derivative with respect to the objective function: (32); In the formula: (33); In the formula, m = 1, 2, 3, ... NG; Similarly, the partial derivative of the penalty term is: (34); (35); In the formula, FN is the expression symbol inside the parentheses, that is: (36); At the minimum point, we have: (37); Therefore, if the 2×NG equations and 2×NG undetermined variables of the system of equations (equation (37)) can be solved precisely, then several possible minimum points can be obtained, thus achieving the solution of the 2×NG undetermined dynamic parameters AG. i XG i0 The purpose is to obtain an approximate solution for complex non-polynomial functions such as equations (37).

[0043] Step 6.4: Finding the approximate minimum point; The objective function corresponding to equation (31) is optimized using the variable metric method of the second derivative matrix and its inverse matrix (Li Weizheng et al., 1982) to calibrate the parameters of the organic matter gas generation kinetic model, including: Given an initial point The first derivative of function (31) at the initial point is calculated using equation (37). At the same time, an approximate inverse of the second derivative matrix of the function at that initial point is calculated using an appropriate method. According to the variable metric method, To ensure that the function value of equation (31) decreases in a certain direction, a one-dimensional search is performed in that direction to determine the step size that optimizes the decrease in the objective function value. That is, to find an approximate solution that is closer to the minimum point. Calculate the gradient at that point. (The magnitude of the first derivative vector), if the gradient is less than a given small positive number. (Used to characterize the required optimization accuracy), that is, considering this point as an approximate solution to the minimum point; otherwise, when If the convergence condition has not yet been met, the approximate value of the inverse of the second derivative matrix is ​​updated based on the parameter changes and gradient changes between two adjacent iteration points to obtain the approximate value of the inverse matrix at the new point. Find the direction in which the function value decreases at the new point. The process involves one-dimensional search, parameter update, and convergence determination steps to obtain subsequent iteration points. ; After the above iterative process, at a certain iteration point The magnitude of the gradient of the objective function is less than When, the iteration point The approximate minimum point of the objective function is used as the parameter value corresponding to the approximate minimum point, and the calibration result of the organic matter gas generation dynamics model, i.e., equation (20), is used as the calibration result.

[0044] In step 7, the outer layer establishes iterative verification based on geological applications and performs global optimization of dynamic parameters; including: After completing the calibration of the inner layer gas dynamics parameters, the dynamic parameters are obtained. Based on Equation (20), the conversion rate of gas generated from organic matter in the source rock at any given time is calculated by combining this parameter with the sedimentary and thermal history of the source rock. See Equation (38).

[0045] Suppose the strata under investigation are divided into M layers, denoted from top to bottom as follows: The time when the i-th layer begins deposition is denoted as t, and the time when it ends is denoted as t. Let the l-th sublayer of the i-th layer be the source rock. The burial depth during the layer deposition was (Can be obtained from sedimentary burial history) The geothermal gradient and surface temperature during the deposition of the i-th layer are: and The gas conversion rate of the source rock in layer i during the deposition of layer j was... The gas conversion rate at the end of sedimentation was... The amount of oil generated during the deposition of layer j alone was Then, the life conversion rate at any time t can be obtained from equation (20). for: (38); (39); After the first calibration of the gas generation kinetic parameters of the inner organic matter, the combination of kinetic parameters was obtained. Using equation (38) for geological applications, the current gas conversion rate is obtained. The EasyRo-vitamin conversion rate benchmark relationship obtained after processing by a bounded monotonic constrained Gaussian process regression model is as follows: Calculate the average difference between the gas conversion rate after the geological application of the current dynamic parameters and the baseline gas conversion rate. The sum of squared residuals is : (40); (41); in, Used for judgment Adjustment direction, Used to characterize the degree of matching between geological application results and the EasyRo-gas conversion rate benchmark relationship under the nth round of outer layer iteration; In adjacent outer iterations, the pre-exponential factor is adjusted using an order-of-magnitude step search method, and the update method is expressed as follows: (42); In the formula, This is the preset step size; the default value is 0.3. After each completion After adjustment, the updated version will be available. The inner layer parameters were recalibrated using the new initial values ​​to simultaneously obtain the corresponding activation energy distribution. and reaction fraction Furthermore, geological application calculations were conducted to obtain a new round of residual sum of squares. ; First adjustment At that time, with As a criterion, when Then decrease Conversely, when Then increase After the initial adjustment Using these as initial values, internal parameter calibration and external geological verification were performed to obtain... For the results after the initial adjustment, if If the adjustment indicates a worse matching degree, then stop the outer layer iterative verification and output the current dynamic parameters; if If the adjustment is effective, then continue to adjust according to formula (42). Perform iterative corrections; During continuous iteration, record the corresponding time intervals. Change; when the following conditions are met: and When determining the iteration corresponding to the (n-1)th iteration To approximate the minimum value, the pre-exponential factor, activation energy distribution, and reaction fraction corresponding to the current cycle are determined as the optimal organic matter gas generation kinetic parameters for the source rock sample, namely: (43)

[0046] In a preferred embodiment, the approximate minimum sum of squared residuals is used. The order of magnitude can reach 10 −3 The maximum number of iterations n is set to 10.

[0047] Based on the method and principle in the above technical solution, the kinetic parameters of the organic gas generation reaction of the mudstone calibration in the Bohai Bay Basin X1 well were obtained, as shown in Table 1; Figure 2 A comparison chart of the experimental value of organic matter gas conversion rate in the thermal simulation experiment of mudstone gold tube in Well X1 of Bohai Bay Basin and the calculated value obtained by calibration of this invention. Figure 3 The distribution of organic matter gas generation activation energy corresponding to the thermal simulation experiment of mudstone gold tube in Well X1 in Bohai Bay Basin. Figure 4 This is a comparison chart of the experimental value of EasyRo-gas conversion rate and the calculated value after iterative verification and optimization of the outer geological application in the thermal simulation experiment of mudstone gold pipe in Well X1 of the Bohai Bay Basin.

[0048] Table 1. Kinetic parameters of organic matter gas generation reaction in mudstone of Well X1 in Bohai Bay Basin; Example 3 The method for calibrating organic hydrocarbon generation kinetic parameters according to Example 2, which eliminates overfitting and multiple solutions, differs in that: Taking the X2 well shale in the Ordos Basin as an example, the following steps are carried out using the organic matter hydrocarbon generation kinetic parameter calibration method established in this invention to eliminate overfitting and multiple solutions: This experiment used the Rock-Eval thermal simulation system to test the shale from well X2 in the Ordos Basin. The samples were heated from 200℃ to 650℃ at heating rates of 15℃ / min and 25℃ / min, respectively, to ensure complete pyrolysis of organic matter. Pyrolysis temperature and chromatographic signals were collected in real time throughout the experiment. The pyrolysis products were enriched by liquid nitrogen cooling and then rapidly desorbed into a gas chromatography system for component separation and quantitative detection. Based on the detected gaseous hydrocarbons (C1-C5) and liquid hydrocarbons (C6-C5), the results were analyzed. 6+ Based on the yield data, the corresponding organic matter-to-oil conversion rate was calculated. Kinetic parameters of the organic matter-to-oil reaction in the X2 well shale of the Ordos Basin were obtained according to the method and principle described in the above technical scheme, as shown in Table 2. Figure 5 A comparison chart of the experimental value of organic matter gas conversion rate in the Rock-Eval thermal simulation experiment of the X2 shale in the Ordos Basin and the calculated value obtained by calibration according to the present invention. Figure 6 The diagram shows the distribution of organic matter gas generation activation energy corresponding to the Rock-Eval thermal simulation experiment of the X2 shale in the Ordos Basin. Figure 7 This is a comparison chart of the experimental values ​​of EasyRo—gas conversion rate in the Rock-Eval thermal simulation experiment of the X2 shale in the Ordos Basin and the calculated values ​​after iterative verification and optimization using outer geological applications.

[0049] Table 2. Kinetic parameters of organic matter generation reaction in the X2 shale of the Ordos Basin.

Claims

1. A method for calibrating kinetic parameters of organic hydrocarbon generation to eliminate overfitting and multiple solutions, characterized in that, include: Step 1: Select low-maturity source rock samples for thermal simulation experiments to obtain temperature-hydrogenation conversion rate experimental data; Step 2: Based on the Sweeney & Burnham vitrinite reflectance equivalent model, establish the temperature-maturity correspondence and construct the EasyRo-hydrocarbon conversion rate dataset; Step 3: Establish a bounded monotonic constrained Gaussian process regression model and construct a continuous and stable EasyRo-hydrogenation conversion rate benchmark relationship; Step 4: Based on the EasyRo-hydrogenation conversion rate benchmark relationship, perform noise reduction, smoothing, and interpolation on the EasyRo-hydrogenation conversion rate data, and back-calculate the temperature-hydrogenation conversion rate data as input data for the kinetic model; Step 5: Construct a parallel first-order reaction kinetic model for hydrocarbon generation from organic matter; Step 6: Based on temperature-hydrogenation conversion data, establish calibration of organic matter hydrocarbon generation kinetic parameters in the inner layer; Step 7: Establish iterative verification based on geological applications in the outer layer and perform global optimization of dynamic parameters; Step 8: Determine whether the dynamic parameters meet the experimental fitting accuracy and geological constraints. If yes, proceed to step 9; otherwise, if it is not the optimal solution, adjust the pre-exponential factor and return to step 6. Step 9: Output the optimal hydrocarbon generation kinetic parameters for the source rock sample.

2. The method for calibrating organic hydrocarbon generation kinetic parameters to eliminate overfitting and multiple solutions as described in claim 1, characterized in that, The specific implementation process of step 1 includes: Low-maturity source rock samples were selected, and thermal simulation experiments were carried out. The changes in hydrocarbon generation products of low-maturity source rock samples during the heating process were continuously recorded, and experimental data on temperature-hydrogenation conversion rate under different heating rates were obtained. Based on the cumulative hydrocarbon generation measured experimentally, temperature-hydrocarbon generation conversion rate experimental data were obtained at different heating rates; the hydrocarbon generation conversion rate is defined as the proportion of the cumulative hydrocarbon generation at a certain temperature to the total hydrocarbon generation potential, i.e.: (1)； Where X(T) is the hydrocarbon conversion rate at temperature T; Y(T) is the cumulative hydrocarbon generation, in mg HC / g TOC; Y max The total hydrocarbon generation amount or the corrected total hydrocarbon generation potential corresponding to the experimental endpoint is expressed in mg HC / g TOC.

3. The method for calibrating organic hydrocarbon generation kinetic parameters to eliminate overfitting and multiple solutions as described in claim 1, characterized in that, The specific implementation process of step 2 includes: In the data conversion process, the Sweeney & Burnham vitrinite reflectance equivalent model was used to characterize the degree of organic matter thermal evolution. The thermal maturation process of vitrinite was regarded as a multi-component parallel first-order Arrhenius reaction, with each reaction having the same pre-exponential factor A and different activation energies E. i The reaction rate equation for a single component is: (2)； in, Let A be the mass of unreacted organic matter of the i-th component at any time t; t is time in seconds; A is the pre-exponential factor; E is ... i Let R be the activation energy of the i-th reaction group; R be the gas constant; and T(t) be the absolute temperature as a function of time. The overall reaction rate is the sum of the reaction rates of each component: (3)； in, The number of components for parallel first-order reactions; The total unreacted organic mass of all components at time t; Introducing the stoichiometric factor f i The overall degree of reaction is obtained by weighted summation of the conversion ratios of each component. : (4)； in, This represents the initial mass of the organic matter. Let be the initial mass of organic matter in the i-th component; The empirical formula for the relationship between total reaction degree F and maturity level EasyRo is as follows: (5)； Through data conversion, the temperature-hydrogenation conversion rate data obtained under different heating rate experimental conditions were uniformly converted into EasyRo-hydrogenation conversion rate data.

4. The method for calibrating organic hydrocarbon generation kinetic parameters to eliminate overfitting and multiple solutions as described in claim 1, characterized in that, The specific implementation process of step 3 includes: Let EasyRo be the input variable x, and the hydrocarbon generation conversion rate be the output variable y; introduce the latent function f(x), and establish the relationship between the hydrocarbon generation conversion rate and the latent function through bounded monotonic mapping: (6)； In the formula, f(x) represents the potential functional relationship between EasyRo and hydrocarbon generation conversion; the predicted hydrocarbon generation conversion rate is guaranteed to satisfy: (7)； Assume that the function f(x) follows a Gaussian process distribution: (8)； in, Let f(x) represent a Gaussian process; x and x′ represent two different samples; m(x) is the mean function, representing the prior expectation of f(x); k(x,x′) is the covariance function, which is used to characterize the correlation of hydrocarbon conversion rate changes between different maturity points x and x′. The covariance function uses a squared exponential kernel function: (9)； Where, σ f 2 The variable represents the variance of the function signal; l represents the characteristic length scale. For the training sample input vector: , and their corresponding predicted values: ; This represents the nth EasyRo input value. The output value represents the conversion rate of the nth hydrocarbon generation. The latent variable observation vector is obtained by performing an inverse function transformation on the observed values ​​using equation (6). : , (10)； Constructing the covariance matrix of training samples : (11)； The joint distribution of the training samples after adding the noise term is: (12)； Where N is a multivariate normal distribution; Indicates the variance of observation noise; It is an n-dimensional identity matrix; Observed value z and predicted value The joint prior distribution is: (13)； In the formula, , for test points The n×1 covariance matrix between the input X of the training set and the input X; For test points Its own covariance; The test point is obtained from equation (13). latent function value The posterior distribution is: (14)； Among them, the posterior distribution mean for: (15)； Posterior prediction variance for: (16)； The predicted mean of the latent function is converted into the predicted value of the hydrocarbon generation conversion rate through the bounded mapping in equation (5), and finally a continuous and stable maturity-hydrogenation conversion rate benchmark relationship is generated, namely EasyRo-hydrogenation conversion rate benchmark relationship: (17)； At the preset maturity nodes, a non-negativity constraint is set on the first derivative of the latent function, namely: (18)； The obtained hydrocarbon generation conversion rate prediction curve also satisfies: (19)； This ensures that the baseline relationship changes incrementally overall.

5. The method for calibrating organic hydrocarbon generation kinetic parameters to eliminate overfitting and multiple solutions as described in claim 4, characterized in that, In step 4, the continuous and stable EasyRo-hydrogenation conversion rate benchmark relationship obtained in step 3 is used for experimental data interpolation reconstruction and geological application iterative verification. On the one hand, under the constraints of experimental heating conditions, based on the EasyRo-hydrogenation conversion rate benchmark relationship, the temperature-hydrogenation conversion rate experimental data are interpolated and reconstructed; specifically, the experimental heating rate is used as the preset heating rate, the temperature range defined by the experimental start temperature and end temperature is used as the interpolation range, and multiple interpolation temperature nodes are set within the interpolation range according to the preset temperature step size. For each interpolated temperature node, the Sweeney & Burnham vitrinite reflectance equivalent model in step 2 is used to convert the corresponding time-temperature history into the maturity parameter EasyRo, and the obtained EasyRo value is substituted into equation (17) obtained in step 3 to obtain the hydrocarbon generation conversion rate corresponding to each interpolated temperature node, and then the interpolated temperature-hydrocarbon generation conversion rate data sequence is constructed. On the other hand, a parallel first-order reaction organic hydrocarbon generation kinetic model suitable for the thermal evolution process of organic matter is established, and the kinetic parameters of the parallel first-order reaction organic hydrocarbon generation kinetic model are calibrated. The parallel first-order reaction organic hydrocarbon generation kinetic model is used to describe the organic hydrocarbon generation process, which includes the stages of oil generation, gas generation, and oil-to-gas conversion.

6. The method for calibrating organic hydrocarbon generation kinetic parameters to eliminate overfitting and multiple solutions as described in claim 5, characterized in that, The temperature step size is 5°C.

7. The method for calibrating organic hydrocarbon generation kinetic parameters to eliminate overfitting and multiple solutions as described in claim 1, characterized in that, In step 5, a parallel first-order reaction kinetic model for hydrocarbon generation from organic matter is constructed, including: Suppose there is an organic hydrocarbon generation process consisting of NH4+ and NH4+ parallel first-order reactions, with the activation energy of each parallel first-order reaction being EH2. i Pre-exponential factor AH i Let XH be the initial potential amount of organic matter corresponding to each reaction. i0 If i = 1, 2, ..., NH, then the total hydrocarbon generation of NH parallel reactions is... for: (20)； in, Let XH be the amount of hydrocarbon generated in the i-th reaction, D be the heating rate, R be the gas constant, and T be the absolute temperature. When XH is XO, it is oil generation; when XH is XG, it is gas generation; and when XH is XOG, it is oil-to-gas conversion.

8. The method for calibrating organic hydrocarbon generation kinetic parameters to eliminate overfitting and multiple solutions as described in claim 7, characterized in that, In step 6, based on temperature-hydrogenation conversion data, the inner layer establishes calibration of organic matter hydrocarbon generation kinetic parameters; including: Step 6.1: Construct the objective function; Suppose that at a certain heating rate l, when a certain temperature j is reached, the experimentally measured hydrocarbon conversion rate is XH1. lj Under the same conditions, assuming EH i AH i XH i Then, the hydrocarbon generation conversion rate calculated by equation (20) is XH. lj Construct the objective function : (21)； Where L0 is the number of experiments with different heating rates, and J0 is the number of sampling points from an experimental curve; Then equation (21) is transformed into equation (22): (22)； In equation (20), AH i XH i0 satisfy: (23)； in, It is a small integer; Step 6.2: Construct the penalty function; The penalty function method is used to transform the process of finding the minimum objective function under constraints into finding it under unconstrained conditions; the process is as follows: For any given constraint, construct a penalty term G. When the minimum value of the obtained function satisfies the constraint, the function value is 0; otherwise, it is a positive number. For AH i Given the constraint >0, we have: (24)； Right now: (25)； Similarly, we have: (26)； Right now: (27)； Similarly, we have: (28)； Right now: (29)； Based on the established function, the penalty term is: (30)； Take a sufficiently large positive integer R1, and construct the penalty function using equations (22) and (30): (31)； If the minimum point obtained exceeds the constraint conditions, then gradually increase R1. When R1 is sufficiently large, the minimum solution of equation (31) is the minimum solution of the objective function equation (22). Step 6.3: Solve for the first-order partial derivatives of the objective function and the penalty function; A necessary condition for the existence of a minimum is that the first-order partial derivative of the penalty function is 0; first, take the partial derivative with respect to the objective function: (32)； In the formula: (33)； In the formula, m = 1, 2, 3, ... NH; Similarly, the partial derivative of the penalty term is: (34)； (35)； In the formula, FN is the expression symbol inside the parentheses, that is: (36)； At the minimum point, we have: (37)； Step 6.4: Finding the approximate minimum point; The objective function corresponding to equation (31) is optimized and solved using the variable metric method of the second derivative matrix and its inverse matrix, in order to calibrate the parameters of the organic matter hydrocarbon generation kinetic model, including: Given an initial point The first derivative of function (31) at the initial point is calculated using equation (37). Simultaneously, the approximate inverse of the second derivative matrix of the function at that initial point is calculated. According to the variable metric method, To ensure that the function value of equation (31) decreases in a certain direction, a one-dimensional search is performed in that direction to determine the step size that optimizes the decrease in the objective function value. That is, to find an approximate solution that is closer to the minimum point. Calculate the gradient at that point. If the gradient is less than a given small positive number That is, we consider the point to be an approximate solution to the minimum; otherwise, when If the convergence condition has not yet been met, the approximate value of the inverse of the second derivative matrix is ​​updated based on the parameter changes and gradient changes between two adjacent iteration points to obtain the approximate value of the inverse matrix at the new point. Find the direction in which the function value decreases at the new point. The process involves one-dimensional search, parameter update, and convergence determination steps to obtain subsequent iteration points. ; After the above iterative process, at a certain iteration point The magnitude of the gradient of the objective function is less than When, the iteration point The approximate minimum point of the objective function is used as the parameter value corresponding to the approximate minimum point, and the calibration result of the organic matter hydrocarbon generation kinetic model, i.e., equation (20), is used.

9. A method for calibrating organic hydrocarbon generation kinetic parameters to eliminate overfitting and multiple solutions as described in any one of claims 1-8, characterized in that, In step 7, the outer layer establishes iterative verification based on geological applications and performs global optimization of dynamic parameters; including: Suppose the strata under investigation are divided into M layers, denoted from top to bottom as follows: The time when the i-th layer begins deposition is denoted as t, and the time when it ends is denoted as t. Let the l-th sublayer of the i-th layer be the source rock. The burial depth during the layer deposition was , The geothermal gradient and surface temperature during the deposition of the i-th layer are: and The hydrocarbon generation conversion rate of the i-th layer source rock during the deposition of the j-th layer was... The hydrocarbon generation conversion rate at the end of sedimentation was... The amount of oil generated during the deposition of layer j alone was Then, the hydrocarbon generation conversion rate at any time t can be obtained from equation (20). for: (38)； (39)； The conversion rate of organic matter to hydrocarbons can be calculated from equation (38). When XH is XO, it is oil production; when XH is XG, it is gas production; and when XH is XOG, it is oil to gas. After the first calibration of the hydrocarbon generation kinetic parameters of the inner organic matter, the combination of kinetic parameters was obtained. Using equation (38) for geological applications, the current hydrocarbon generation conversion rate is obtained. The EasyRo-hydrogenation conversion benchmark relationship obtained after processing by a bounded monotonic constrained Gaussian process regression model is as follows: Calculate the average difference between the hydrocarbon generation conversion rate after the geological application of the current kinetic parameters and the baseline hydrocarbon generation conversion rate. The sum of squared residuals is : (40)； (41)； in, Used for judgment Adjustment direction, Used to characterize the degree of matching between the geological application results under the nth round of outer layer iteration and the EasyRo-hydrogenation conversion rate benchmark relationship; In adjacent outer iterations, the pre-exponential factor is adjusted using an order-of-magnitude step search method, and the update method is expressed as follows: (42)； In the formula, Preset step size; After each completion After adjustment, the updated version will be available. The inner layer parameters were recalibrated using the new initial values ​​to simultaneously obtain the corresponding activation energy distribution. and reaction fraction Furthermore, geological application calculations were conducted to obtain a new round of residual sum of squares. ; First adjustment At that time, with As a criterion, when Then decrease Conversely, when Then increase After the initial adjustment Using these as initial values, internal parameter calibration and external geological verification were performed to obtain... For the results after the initial adjustment, if If the adjustment indicates a worse matching degree, then stop the outer layer iterative verification and output the current dynamic parameters; if If the adjustment is effective, then continue to adjust according to formula (42). Perform iterative corrections; During continuous iteration, record the corresponding time intervals. Change; when the following conditions are met: and At that time, determine the corresponding iteration of the (n-1)th iteration. To approximate the minimum value, the pre-exponential factor, activation energy distribution, and reaction fraction corresponding to the current cycle are determined as the optimal organic matter hydrocarbon generation kinetic parameters for the source rock sample, namely: (43)。

Images