Construction method of pyrolysis kinetic model for in-situ conversion and production of low-maturity shale oil

Abstract

The present invention provides a method for constructing a pyrolysis kinetic model for in-situ conversion and exploitation of medium-low maturity shale oil, which belongs to the technical field of in-situ development of underground resources, and comprises: first, preparing medium-low maturity shale samples with different organic matter compositions, carrying out pyrolysis experiments, and collecting and quantitatively analyzing pyrolysis products. Based on the original experimental data, a data augmentation model is constructed, and an adversarial generation network model is established to optimize the generation effect of pyrolysis material mass change data. The generated material mass change data is merged with the experimental data to obtain merged data. Finally, based on the merged data, a genetic algorithm is used to optimize key parameters such as reaction kinetic equations, reaction activation energy, and frequency factors, so as to provide a reliable pyrolysis kinetic model for in-situ conversion and exploitation of medium-low maturity shale oil, and solve the problem that the existing models are difficult to accurately describe the real kinetic behavior of organic matter pyrolysis conversion.

Description

Technical Field

[0001] The present invention belongs to the technical field of in-situ development of underground resources, and specifically relates to a method for constructing a pyrolysis kinetic model for in-situ conversion and exploitation of medium- and low-maturity shale oil. Background Art

[0002] In recent years, with the increasing depletion of conventional oil and gas resources, shale oil has attracted widespread attention as an important unconventional oil and gas resource. The development of shale oil reservoirs mainly adopts horizontal well fracturing technology, but due to the extremely low permeability of medium and low maturity shale reservoirs, conventional mining methods are difficult to achieve economically feasible development. As a new mining method, in-situ conversion technology injects heat into the reservoir to induce pyrolysis of organic matter in shale to produce oil, which has great application prospects. However, the pyrolysis and conversion process of shale organic matter is very complex, involving multiple reaction mechanisms and kinetic behaviors, which brings great challenges to the optimization design of in-situ conversion technology. Most existing studies use a single kinetic model construction method, such as the equivalent first-order reaction model or the distributed activation energy model, which often oversimplifies the actual reaction process. Especially when dealing with complex systems where kerogen and heavy hydrocarbons coexist, it is difficult to accurately describe the interaction and conversion relationship between the components. In addition, the traditional method of obtaining kinetic parameters mainly relies on thermogravimetric analysis data, lacks in-depth analysis of the composition and yield of pyrolysis products, resulting in a large deviation between the established kinetic model and the actual production process. With the development of artificial intelligence technology, deep learning methods have shown unique advantages in complex system modeling, but their application in the construction of shale oil pyrolysis kinetic model is still in its infancy, and the specific mechanism functions of each reaction have not been explored. The kinetic model determined by it is difficult to accurately describe the real kinetic behavior of organic matter pyrolysis conversion. Therefore, how to establish a pyrolysis kinetic model that can accurately describe the in-situ conversion process of medium and low maturity shale oil is a technical problem that needs to be solved urgently. Summary of the invention

[0003] In view of this, the present invention provides a method for constructing a pyrolysis kinetic model for in-situ conversion and exploitation of medium- and low-maturity shale oil, which can solve the technical problem that the existing kinetic models are difficult to accurately describe the real kinetic behavior of pyrolysis conversion of organic matter.

[0004] The present invention is achieved in that:

[0005] The present invention provides a method for constructing a pyrolysis kinetic model for in-situ conversion and exploitation of medium- and low-maturity shale oil, which comprises the following steps:

[0006] S10, preparing low-mature shale samples with different organic matter compositions;

[0007] S20. Carry out pyrolysis weight loss experiments on low-mature shale at different heating rates;

[0008] S30. Conduct high-temperature pyrolysis experiments on low-mature shale with different pyrolysis final temperatures, collect pyrolysis oil and gas products, and quantitatively characterize the mass changes of kerogen, heavy hydrocarbons, pyrolysis oil, pyrolysis gas and coke in the pyrolysis reaction, which are recorded as the mass change data of the experimental materials;

[0009] S40, constructing a pyrolysis kinetics data augmentation model to generate an expanded data set based on the experimental substance mass change data;

[0010] S50, establishing a generative adversarial network model, wherein the generator of the generative adversarial network model is responsible for generating theoretical material mass change data that conforms to physical laws, and the discriminator of the generative adversarial network model is responsible for distinguishing the difference between the generated data and the real experimental data;

[0011] S60, using the expanded data set to train a generative adversarial network, and optimizing the generation effect of theoretical material mass change data by minimizing the loss function between the generator and the discriminator;

[0012] S70, merging the theoretical material mass change data generated by the generative adversarial network with the experimental material mass change data to form merged data;

[0013] S80. Based on the merged data, combined with the Arrhenius formula and the reaction mechanism function, a genetic algorithm is used to iteratively optimize the fitting accuracy of the prediction results and the experimental data to be no less than 90%, and the final reaction kinetics equation, reaction activation energy and frequency factor are output as the pyrolysis kinetics model for in-situ conversion and exploitation of medium- and low-maturity shale oil.

[0014] The heating rates are 5 to 40 degrees Celsius per minute.

[0015] On the basis of the above technical solution, the method for constructing a pyrolysis kinetic model for in-situ conversion and exploitation of medium- and low-maturity shale oil of the present invention can also be improved as follows:

[0016] The process of preparing low-middle-mature shale samples with different organic matter compositions includes extracting low-middle-mature shale powder samples with dichloromethane, rotary evaporating the extract to obtain heavy hydrocarbon samples, and acid washing the extracted solid powder to obtain kerogen samples.

[0017] Specifically, the step S10 specifically includes:

[0018] Step 101, obtaining low-mature shale samples and crushing them to control the particle size to be less than 0.2 mm;

[0019] Step 102, using dichloromethane to perform Soxhlet extraction on the low-maturity shale powder sample, and the extraction time is not less than 48 hours;

[0020] Step 103, performing a rotary evaporation process on the extract, with the temperature controlled at 40 degrees Celsius, to obtain a heavy hydrocarbon sample;

[0021] Step 104, the solid powder after extraction is pickled with hydrochloric acid and hydrofluoric acid in sequence, and the pickling temperature is controlled at 60 degrees Celsius;

[0022] Step 105, filtering, washing and drying the acid-washed solid powder to obtain a kerogen sample;

[0023] Step 106, determining the contents of kerogen and heavy hydrocarbons in the obtained low-maturity shale samples;

[0024] Step 107: design different proportions of kerogen and heavy hydrocarbon according to experimental requirements, and mix the original low-mature shale powder, heavy hydrocarbon sample and kerogen sample according to a preset proportion;

[0025] Step 108: Use a ball mill to fully mix the mixed sample to obtain medium- and low-maturity shale samples with different organic matter compositions.

[0026] Specifically, the step S20 includes:

[0027] Step 201, using a thermogravimetric analyzer to perform a pyrolysis weight loss experiment on medium- and low-maturity shale samples with different organic matter compositions;

[0028] Step 202, placing the low-maturity shale samples with different organic matter compositions in a crucible of a thermogravimetric analyzer, and controlling the sample mass to be 10 to 15 mg;

[0029] Step 203, using nitrogen as carrier gas, and controlling the gas flow rate to 50 ml per minute;

[0030] Step 204, conduct a first set of experiments at a heating rate of 5 degrees Celsius per minute, and record temperature and mass change data;

[0031] Step 205, conduct a second set of experiments at a heating rate of 10 degrees Celsius per minute, and record temperature and mass change data;

[0032] Step 206, conduct a third set of experiments at a heating rate of 15 degrees Celsius per minute, and record temperature and mass change data;

[0033] Step 207, conduct a fourth set of experiments at a heating rate of 20 degrees Celsius per minute, and record temperature and mass change data;

[0034] Step 208, conduct a fifth set of experiments at a heating rate of 30 degrees Celsius per minute, and record temperature and mass change data;

[0035] Step 209, conduct a sixth set of experiments at a heating rate of 40 degrees Celsius per minute, and record temperature and mass change data;

[0036] Step 210: Arrange and store the obtained temperature and mass change data.

[0037] Specifically, the step S30 includes:

[0038] Step 301: Determine the temperature range of the second stage of shale pyrolysis based on pyrolysis weight loss experimental data;

[0039] Step 302, uniformly selecting 8 to 12 temperature points within the temperature range as the pyrolysis final temperature;

[0040] Step 303, placing low-mature shale samples with different organic matter compositions in a high-temperature and high-pressure reactor;

[0041] Step 304, after evacuating the vacuum, introduce nitrogen gas to control the initial pressure to be 0.5 MPa;

[0042] Step 305, heating to a target pyrolysis final temperature at a heating rate of 10 degrees Celsius per minute;

[0043] Step 306, collecting and measuring pyrolysis gas products;

[0044] Step 307, collecting and weighing the pyrolysis oil product;

[0045] Step 308, extracting the pyrolysis residue and testing the organic carbon content;

[0046] Step 309: record and organize the mass change data of each substance, and record it as the experimental substance mass change data.

[0047] Specifically, the step S40 includes:

[0048] Step 401, constructing a data augmentation basic model based on the basic equation of pyrolysis kinetics;

[0049] Step 402: setting the diffusion coefficient in the basic model as a function of temperature;

[0050] Step 403, introducing a time-varying random disturbance term into the basic model to construct a pyrolysis kinetic disturbance equation;

[0051] Step 404, establishing a temperature field distribution equation taking into account heat conduction and reaction heat;

[0052] Step 405, constructing a component diffusion equation including a chemical potential gradient;

[0053] Step 406: design a noise disturbance model based on Gaussian distribution and introduce a time attenuation term;

[0054] Step 407: Establish a temperature gradient transformation model, and introduce periodic fluctuation and spatial attenuation terms;

[0055] Step 408: construct a component ratio adjustment model to ensure the conservation of mass of each component;

[0056] Step 409: Combine the above models to generate an expanded data set.

[0057] Specifically, the step S50 includes:

[0058] Step 501: construct a generator network, including a multi-layer convolutional neural network and a fully connected layer;

[0059] Step 502: adding a residual connection to the generator to improve network performance;

[0060] Step 503: design a discriminator network, using a deep convolutional discriminator structure;

[0061] Step 504: Add a spectral normalization layer to the discriminator to improve training stability;

[0062] Step 505: designing a loss function based on physical constraints;

[0063] Step 506: adding mass conservation and energy conservation constraints;

[0064] Step 507, designing rationality constraints of reaction kinetic parameters;

[0065] Step 508: Integrate all constraints into the adversarial loss function.

[0066] Specifically, the step S60 includes:

[0067] Step 601, performing normalization preprocessing on the expanded data set;

[0068] Step 602: Divide the data set into a training set and a validation set;

[0069] Step 603, initializing the generator and discriminator network parameters;

[0070] Step 604: design a pre-training strategy, using a progressive training method;

[0071] Step 605: perform adversarial training and update generator parameters;

[0072] Step 606: Calculate the discriminator loss and update the discriminator parameters;

[0073] Step 607: Calculate physical constraint loss and adjust network parameters;

[0074] Step 608: Evaluate model performance and save the optimal model.

[0075] Specifically, the step S80 includes:

[0076] Step 801, determine that the frequency factor in the Arrhenius equation is 10-16 per hour by default, the activation energy range is 100 to 300 kilojoules per mole, and the stoichiometric number in the reaction is searched in the range of 0-1;

[0077] Step 802, setting genetic algorithm parameters, wherein the population size is 200 to 500, the maximum evolutionary generation is 1000, the crossover probability is 0.7 to 0.9, and the mutation probability is 0.01 to 0.05;

[0078] Step 803: Encode the merged data using a real number encoding method to generate an initialization population;

[0079] Step 804, using the Arrhenius formula to calculate the reaction rate constant, and solving the reaction kinetic equations in combination with the reaction mechanism function;

[0080] Step 805: Calculate the root mean square error between the prediction result and the experimental data for each individual, and use the fitting accuracy as the fitness function;

[0081] Step 806: Use roulette wheel selection, uniform crossover and Gaussian mutation operators to evolve the population until the fitting accuracy reaches more than 90% or the maximum evolutionary generation is reached;

[0082] Step 807, extracting the reaction kinetic parameters corresponding to the optimal individual, including reaction activation energy and frequency factor, and the stoichiometric coefficients of each substance in the reaction;

[0083] Step 808: Use the obtained reaction kinetic equation, reaction activation energy and frequency factor as the final output result of the pyrolysis kinetic model for in-situ conversion and production of medium- and low-maturity shale oil.

[0084] Furthermore, the kerogen mass calculation is specifically expressed as follows:

[0085] ;

[0086] In the formula, is the mass of remaining kerogen in the residue after the pyrolysis experiment, in g; is the mass of the residue after removing heavy hydrocarbons after pyrolysis experiments at different temperatures, in g; is the organic carbon content in shale residue, unit: % is the organic carbon content of the original kerogen, in %.

[0087] The coke quality calculation is specifically expressed as follows:

[0088] ;

[0089] In the formula, is the mass of coke, in g; is the total mass of kerogen and heavy hydrocarbons in the initial sample, in g; is the remaining kerogen mass, in g; is the mass of remaining heavy hydrocarbons, in g; is the mass of pyrolysis oil, in g; is the mass of pyrolysis gas, in g.

[0090] The conversion rate calculation is specifically expressed as follows:

[0091] ;

[0092] In the formula, is the conversion rate; is the initial mass, in g; is the mass at time t, in g; is the mass at the reaction endpoint, in g.

[0093] The conversion rate calculation is specifically expressed as follows:

[0094] ;

[0095] In the formula, is the conversion rate, in min -1 ; is the heating rate, in K / min; is the weight loss rate, in g / K; is the initial mass in g.

[0096] The Gaussian activation energy distribution calculation double parallel Gaussian distribution function is specifically expressed as follows:

[0097] ;

[0098] In the formula, is the distribution peak amplitude, ranging from 0 to 1; is the center temperature of the distribution peak, in K; is the standard deviation, reflecting the ratio of peak width to reaction temperature range; is the temperature variable, the unit is K.

[0099] The mechanism function is specifically expressed as follows:

[0100] Reaction level F2 model: ;

[0101] Geometry contraction R3 model: ;

[0102] Reaction rate characterization: ;

[0103] Arrhenius formula: ;

[0104] Where α is the conversion rate; t is the reaction time, in h; k is the reaction rate constant, in h -1 ; E is the activation energy, unit is kJ / mol; A is the frequency factor, unit is h -1 ; T is the reaction temperature, unit is K; R is the universal gas constant, which is 8.314 J / (mol·K).

[0105] The determination of the integral form of the mechanism function is specifically expressed as follows:

[0106] ;

[0107] In the formula, is the integral form of the mechanism function; is the differential form of the mechanism function; is the conversion rate.

[0108] The fitting accuracy calculation formula is specifically expressed as follows:

[0109] ;

[0110] In the formula, is the fitting accuracy; is the experimental value; is the calculated value; is the average value; is the number of data points.

[0111] The main graph method mechanism function determination formula is specifically expressed as follows:

[0112] ;

[0113] In the formula, is the current temperature in K; is the temperature when the conversion rate is 0.5, in K; is the conversion rate; is the experimental conversion rate when the conversion rate is 0.5; is the differential form of the reaction mechanism function; is the integral form of the reaction mechanism function.

[0114] Among them, the pyrolysis kinetics data augmentation model includes the pyrolysis kinetics basic equation, the pyrolysis kinetics perturbation equation, the temperature field distribution equation, the component diffusion equation, as well as the Gaussian noise perturbation model, the temperature gradient transformation model, and the component ratio adjustment model.

[0115] The basic equation of pyrolysis kinetics is specifically expressed as follows:

[0116] ;

[0117] In the formula, is the concentration of component i, in mol / m3; is the diffusion coefficient of component i, in m2 / s; is the reaction rate of component i in reaction j, in mol / (m^3·s); is the number of reactions; is time, unit is s; is the spatial coordinate, the unit is m.

[0118] The pyrolysis kinetics perturbation equation is specifically expressed as follows:

[0119] ;

[0120] In the formula, is the concentration of component i after disturbance; is a time-varying random disturbance term with a mean of 0 and a variance of Gaussian distribution.

[0121] The temperature field distribution equation is specifically expressed as follows:

[0122] ;

[0123] In the formula, is the density, in kg / m3; is the specific heat capacity, in J / (kg·K); is the thermal conductivity, unit is W / (m·K); is the reaction heat, in W / m3; It is an external heat source, and its unit is W / m^3.

[0124] The component diffusion equation is specifically expressed as follows:

[0125] ;

[0126] In the formula, is the diffusion flux of component i, in mol / (m^2·s); is the chemical potential of component i, in J / mol.

[0127] The Gaussian noise disturbance model is specifically expressed as follows:

[0128] ;

[0129] In the formula, is the data after adding noise; is the original data; is the noise intensity coefficient, ranging from 0.01 to 0.1; is the noise decay time constant, in s.

[0130] The temperature gradient transformation model is specifically expressed as follows:

[0131] ;

[0132] In the formula, is the temperature field after transformation; is the temperature fluctuation amplitude, ranging from 1 to 10K; is the angular frequency; is the phase; is the characteristic length in m.

[0133] The component ratio adjustment model is specifically expressed as follows:

[0134] ;

[0135] In the formula, is the concentration of component i after adjustment; is the adjustment coefficient of component i, ranging from -0.1 to 0.1, and satisfies .

[0136] The adversarial generation network loss function is specifically expressed as follows:

[0137] ;

[0138] In the formula, To combat losses; is the discriminator output; Output for the generator; is the real data distribution; is the noise distribution; is the expectation operator.

[0139] The fitness function of the genetic algorithm is specifically expressed as follows:

[0140] ;

[0141] In the formula, is the fitness function value; is the experimental value; is the calculated value; is the number of data points; To prevent a small constant with a denominator of zero, the value is 10^{-6}.

[0142] The final kinetic model is specifically expressed as follows:

[0143] ;

[0144] In the formula, is the pre-exponential factor; is the activation energy, in kJ / mol; is the gas constant, which is taken as 8.314 J / (mol·K); is the temperature, in K; is the diffusion coefficient; is the space coordinate.

[0145] Compared with the prior art, the method for constructing a pyrolysis kinetic model for in-situ conversion and exploitation of medium- and low-maturity shale oil provided by the present invention has the following beneficial effects:

[0146] First, the present invention adopts a systematic sample preparation method, and obtains standard samples with different organic matter compositions by controlling the ratio of kerogen and heavy hydrocarbons, providing a reliable experimental basis for subsequent model construction. Secondly, through the organic combination of pyrolysis weight loss experiment and high-temperature pyrolysis experiment, not only the overall pyrolysis behavior characteristics of the sample are obtained, but also the accurate quantitative analysis of each reaction product is realized, which significantly improves the accuracy of the kinetic model. In particular, in terms of data processing, the data augmentation model developed by the present invention can effectively expand the experimental data set, overcoming the problem of limited experimental data volume. By introducing the adversarial generative network, the present invention realizes the generation of material mass change data in accordance with physical laws, avoiding the problem that parameter optimization in traditional methods is prone to fall into local optimality. In addition, on the basis of preparing medium and low-mature shale samples with different kerogen and heavy hydrocarbon contents, the present invention carries out pyrolysis weight loss experiments and high-temperature pyrolysis experiments, uses the Gaussian activation energy distribution method to separate kerogen reaction and heavy hydrocarbon reaction, uses the main graph method to determine the mechanism functions of kerogen reaction and heavy hydrocarbon reaction respectively, combines genetic algorithm fitting to determine the reaction kinetic equation and kinetic parameters, and solves the technical problem that the existing kinetic model is difficult to accurately describe the real kinetic behavior of organic matter pyrolysis conversion. BRIEF DESCRIPTION OF THE DRAWINGS

[0147] Figure 1 A flow chart of the method provided by the present invention;

[0148] Figure 2 The pyrolysis weight loss curve of low-mature shale with a kerogen to heavy hydrocarbon mass ratio of 20 in the embodiment;

[0149] Figure 3 The pyrolysis weight loss curve of low-maturity shale with a mass ratio of kerogen to heavy hydrocarbons of 5 in the embodiment;

[0150] Figure 4 The pyrolysis conversion rate fitting curve of low-mature shale with a mass ratio of kerogen to heavy hydrocarbon of 20 in the embodiment;

[0151] Figure 5 The pyrolysis conversion rate fitting curve of low-mature shale with a mass ratio of kerogen to heavy hydrocarbon of 5 in the embodiment;

[0152] Figure 6 The reaction mechanism function fitting curve of low-mature shale with a mass ratio of kerogen to heavy hydrocarbon of 20 in the embodiment;

[0153] Figure 7 The reaction mechanism function fitting curve of the low-mature shale with a mass ratio of kerogen to heavy hydrocarbon of 5 in the embodiment;

[0154] Figure 8 The fitting results of the mass changes of various substances in the reaction of low-mature shale with a mass ratio of kerogen to heavy hydrocarbon of 20 in the embodiment;

[0155] Fig. 9 This is the fitting result of the mass change of each substance in the reaction of low-mature shale with a mass ratio of kerogen to heavy hydrocarbons of 5 in the example. DETAILED DESCRIPTION

[0156] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0157] like Figure 1 As shown, it is a flow chart of a method for constructing a pyrolysis kinetic model for in-situ conversion and exploitation of medium- and low-maturity shale oil provided by the present invention. The method comprises the following steps:

[0158] S10, preparing low-mature shale samples with different organic matter compositions;

[0159] S20. Carry out pyrolysis weight loss experiments on low-mature shale at different heating rates;

[0160] S30. Conduct high-temperature pyrolysis experiments on low-mature shale with different pyrolysis final temperatures, collect pyrolysis oil and gas products, and quantitatively characterize the mass changes of kerogen, heavy hydrocarbons, pyrolysis oil, pyrolysis gas and coke in the pyrolysis reaction, which are recorded as the mass change data of the experimental materials;

[0161] S40, constructing a pyrolysis kinetics data augmentation model to generate an expanded data set based on the experimental substance mass change data;

[0162] S50, establishing a generative adversarial network model, wherein the generator of the generative adversarial network model is responsible for generating theoretical material mass change data that conforms to physical laws, and the discriminator of the generative adversarial network model is responsible for distinguishing the difference between the generated data and the real experimental data;

[0163] S60, using the expanded data set to train a generative adversarial network, and optimizing the generation effect of theoretical material mass change data by minimizing the loss function between the generator and the discriminator;

[0164] S70, merging the theoretical material mass change data generated by the generative adversarial network with the experimental material mass change data to form merged data;

[0165] S80. Based on the merged data, combined with the Arrhenius formula and the reaction mechanism function, a genetic algorithm is used to iteratively optimize the fitting accuracy of the prediction results and the experimental data to be no less than 90%, and the final reaction kinetics equation, reaction activation energy and frequency factor are output as the pyrolysis kinetics model for in-situ conversion and exploitation of medium- and low-maturity shale oil.

[0166] The specific implementation methods of the above steps are described in detail below:

[0167] The specific implementation method of step S10 is to first pretreat and separate the components of the medium and low-maturity shale samples. The specific implementation process includes crushing the original shale sample to less than 200 mesh, extracting it with dichloromethane as solvent at 75°C for 24 hours by Soxhlet extraction method, and concentrating the extract by rotary evaporator to obtain a heavy hydrocarbon sample. The extracted solid sample is pickled with hydrochloric acid and hydrofluoric acid in turn, and the kerogen sample is obtained after removing inorganic minerals. The content of carbon, hydrogen, oxygen, nitrogen and sulfur in the original shale sample, heavy hydrocarbon sample and kerogen sample is determined by elemental analyzer, and the thermogravimetric properties of each sample are determined by thermogravimetric analyzer. After obtaining the basic sample, different kerogen and heavy hydrocarbon ratios are designed, and the ratioed mixture is mixed in a ball mill at a speed of 300 rpm for 2 hours to prepare experimental samples with different organic matter compositions. The purpose of this step is to obtain standard samples with different organic matter composition ratios to provide basic materials for subsequent pyrolysis kinetic experiments.

[0168] The specific implementation method of step S20 is to conduct a thermogravimetric experiment to obtain the weight loss data of the sample at different heating rates. The specific implementation process includes placing the prepared sample in the sample cell of the thermogravimetric analyzer, controlling the heating rates to 5, 10, 15, 20, 30 and 40 degrees Celsius per minute, and the heating range is from room temperature to 900°C. High-purity nitrogen is used as the carrier gas with a flow rate of 50 ml per minute. During the experiment, the change in sample mass with temperature is continuously recorded. The experiment is repeated 3 times for each heating rate condition, and the average value is taken as the final result. The weight loss rate and weight loss rate of the sample are calculated according to the thermogravimetric curve, and the formula is used. Calculate the conversion rate, where is the initial mass, is the mass at time t, The purpose of this step is to obtain basic experimental data reflecting the kinetic characteristics of organic matter pyrolysis.

[0169] The specific implementation method of step S30 is to carry out a high-temperature pyrolysis experiment and perform quantitative analysis on the product. The specific implementation process includes selecting a pyrolysis end temperature point every 25°C in the range of 300°C to 500°C, loading the sample into a high-temperature and high-pressure reactor, and heating it to the target temperature at a heating rate of 10°C per minute. A gas chromatograph is used to detect the pyrolysis gas components online, and the pyrolysis oil is collected and weighed with a cold trap. The pyrolysis residue is extracted with dichloromethane to determine the residual heavy hydrocarbon content, and an elemental analyzer is used to determine the total organic carbon content in the residue. The coke yield is calculated using the mass balance principle, and a database of the mass changes of each component with temperature is established, which is recorded as the mass change data of the experimental substance. The purpose of this step is to obtain the conversion relationship and mass change law of each component during the pyrolysis process.

[0170] The specific implementation of step S40 is to construct a data augmentation model to expand the experimental data set. The model includes the following components: the basic equation of thermal decomposition kinetics based on the Arrhenius equation ,in is the pre-exponential factor, is the activation energy, is the gas constant, is the temperature, is the reaction mechanism function; the thermal decomposition kinetic perturbation equation , used to simulate experimental condition fluctuations; temperature field distribution equation , describing the temperature gradient inside the sample; component diffusion equation ,in is the component concentration, is the diffusion coefficient. By adding Gaussian noise to the original data , perform temperature gradient transformation and component ratio adjustment to generate an expanded data set. The purpose of this step is to increase the diversity and representativeness of the training data.

[0171] The specific implementation of step S50 is to construct a generative adversarial network model. The generator adopts a multi-layer perceptron structure, the input layer contains random noise vectors and conditional variables (temperature, pressure and other working parameters), the hidden layer uses a drain rectified linear unit activation function, and the output layer generates theoretical material mass change data. The discriminator also adopts a multi-layer perceptron structure, the input layer receives the generator output or real experimental data, and outputs the discrimination probability after passing through multiple fully connected layers. The generator loss function is , the discriminator loss function is ,in and denote the discriminator and the generator respectively, is the noise distribution, is the real data distribution. The purpose of this step is to establish a deep learning model that can generate material mass change data that conforms to physical laws.

[0172] The specific implementation of step S60 is to train the adversarial generative network model. The small batch stochastic gradient descent algorithm is used for optimization, the batch size is set to 32, the initial value of the learning rate is 0.001, and the cosine annealing strategy is used for dynamic adjustment. In each training cycle, the generator parameters are first fixed, the discriminator parameters are updated several times, and then the discriminator parameters are fixed and the generator parameters are updated. The gradient penalty term is introduced To improve training stability, is the weight coefficient, with a value of 0.1. An early stopping strategy is used to prevent overfitting, and training is stopped when the validation set loss has not improved for 5 consecutive cycles. The purpose of this step is to optimize the model parameters so that the generated theoretical material mass change data is consistent with both physical laws and experimental data.

[0173] The specific implementation of step S70 is to directly merge the theoretical material mass change data generated by the generative adversarial network with the experimental material mass change data to form merged data.

[0174] The specific implementation method of step S80 is to combine the merged data with the genetic algorithm, set the population size to 100, the maximum evolutionary generation to 200, the crossover probability to 0.8, and the mutation probability to 0.1. The chromosome encoding adopts real number encoding to represent the kinetic parameters such as reaction activation energy and frequency factor. The selection operator adopts tournament selection, the crossover operator adopts simulated binary crossover, and the mutation operator adopts polynomial mutation. The fitness function is defined as the determination coefficient of the predicted value and the experimental value, and the fitting accuracy is required to be not less than 90%. During the iterative optimization process, the kinetic parameters of each generation of individuals must satisfy the Arrhenius equation and the reaction mechanism function determined experimentally. The final output kinetic equation includes the kinetic parameter set of three sub-reactions: kerogen pyrolysis reaction, heavy hydrocarbon cracking reaction, and pyrolysis oil cracking reaction. The purpose of this step is to obtain a pyrolysis kinetic model that can accurately describe the in-situ conversion process of medium- and low-maturity shale oil.

[0175] Specifically, the principle of the present invention is:

[0176] Medium-low maturity shale oil is an unconventional fossil resource with great oil generation potential and abundant reserves. Due to the depletion of conventional energy and the huge growth in energy demand, the exploitation of medium-low maturity shale oil has attracted widespread attention. Medium-low maturity shale oil refers to shale oil resources with a vitrinite reflectance (Ro) in the range of 0.5~1.0. In-situ conversion technology is considered to be a proven and effective technology for the exploitation of medium-low maturity shale. Unlike oil shale reservoirs, kerogen and heavy hydrocarbon organic matter coexist in medium-low maturity shales, which makes the pyrolysis conversion of organic matter more complicated during the in-situ conversion process. It is difficult and important to characterize the multiphase and multi-component chemical process and accurately predict the pyrolysis products.

[0177] The pyrolysis conversion of shale organic matter is a complex chemical process involving multiple stages and multiple components. Kerogen, as the main hydrocarbon-generating parent material, first undergoes pyrolysis reaction during heating to generate heavy hydrocarbons, light oil and gas products; at the same time, primary or secondary heavy hydrocarbons will also be further cracked to form lighter hydrocarbon products. These reaction processes are affected by multiple factors such as temperature, pressure, and mineral composition, and show obvious nonlinear characteristics. The present invention establishes a complete experimental data system by precisely controlling the sample composition and experimental conditions, laying the foundation for subsequent model construction. In terms of data processing, the data augmentation model introduced in the present invention is based on the basic principles of pyrolysis dynamics. By adding disturbance terms with clear physical meanings, it simulates various operating conditions that may occur in the actual production process, effectively expanding the coverage of training data. The introduction of adversarial generative networks is an important innovation of the present invention, in which the generator can generate kinetic parameters that conform to physical laws by continuously learning the distribution characteristics of real data, and the discriminator prompts the generator to continuously improve its output quality by evaluating the generated data. This adversarial learning mechanism effectively avoids the problem that traditional optimization methods are prone to fall into local optimality. In order to further improve the reliability of the model, the present invention adopts an integrated learning strategy, integrates the prediction results of multiple basic models, fully utilizes the advantages of different models, and improves the accuracy and stability of the prediction. The application of genetic algorithms ensures that the final output kinetic parameters meet the constraints of basic physical and chemical laws such as the Arrhenius equation.

[0178] A specific embodiment of the present invention is provided below: This embodiment mainly comprises the following steps:

[0179] Step 101, based on the medium-low maturity shale samples, prepare medium-low maturity shale samples with different organic matter compositions.

[0180] The low-mature shale samples with different organic matter compositions refer to the low-mature shale samples with different kerogen and heavy hydrocarbon ratios. According to the experimental needs, various kerogen and heavy hydrocarbon ratios are designed, and the low-mature shale samples with different organic matter compositions are prepared by fully mixing a certain amount of original low-mature shale powder, heavy hydrocarbon samples and kerogen samples. Specifically include:

[0181] (1) Determine the composition of organic matter in low-middle-mature shale samples. One embodiment of the present invention is based on low-middle-mature shale samples of Chang 73 section in the Ordos Basin. The low-middle-mature shale powder samples are extracted with dichloromethane, and the extract is subjected to rotary evaporation to obtain a heavy hydrocarbon sample. The extracted solid powder is acid-washed to remove inorganic minerals, thereby obtaining a kerogen sample. This process is strictly carried out in accordance with the kerogen separation standard in sedimentary rocks "GB / T19144-2010". At the same time, the content of kerogen and heavy hydrocarbons in the existing low-middle-mature shale samples can be clarified. In one embodiment of the present invention, it is determined that the existing low-middle-mature shale samples contain 40wt.% kerogen and 2wt.% heavy hydrocarbons, and the mass ratio of kerogen to heavy hydrocarbons is 20.

[0182] (2) Designing the organic matter composition according to research needs. In one embodiment of the present invention, the mass ratio of kerogen to heavy hydrocarbons is designed to be 5:1 to study the effect of the relative content of kerogen and heavy hydrocarbons on the pyrolysis kinetics of low-mature shale oil.

[0183] (3) Prepare low-middle-mature shale samples with different organic matter compositions. The content of the original low-middle-mature shale powder and heavy hydrocarbon samples that need to be mixed to prepare low-middle-mature shale samples with different organic matter compositions is calculated by the following formula. In one embodiment of the present invention, each group of pyrolysis experiments requires 30g of sample powder. According to calculations, it is determined that 28.3g of original low-middle-mature shale powder and 1.7g of heavy hydrocarbons need to be fully mixed to prepare the low-middle-mature shale samples. At this time, the mass ratio of kerogen to heavy hydrocarbons in the low-middle-mature shale samples is 5.

[0184] (1);

[0185] in, is the mass of the original medium- and low-maturity shale samples that need to be mixed, g; is the mass of heavy hydrocarbons to be mixed, g; W is the mass of low-mature shale samples composed of new organic matter, g; and are the masses of heavy hydrocarbons and kerogen contained in the medium- and low-maturity shale samples composed of new organic matter, g; k is the designed mass ratio of kerogen to heavy hydrocarbons.

[0186] In addition to the designed kerogen and heavy hydrocarbon ratio in one embodiment of the present invention being lower than that of the original low-middle-mature shale sample, if the designed kerogen and heavy hydrocarbon ratio is higher than that of the original low-middle-mature shale sample, this means less heavy hydrocarbon content, and it is necessary to mix the shale powder from which the heavy hydrocarbons are removed and the heavy hydrocarbon sample for preparation. At this time, the mass of the shale powder and heavy hydrocarbon sample to be mixed is calculated by formula (2).

[0187] (2);

[0188] in, is the mass of the original low-maturity shale sample with heavy hydrocarbons removed that needs to be mixed, g. This sample only contains kerogen and inorganic minerals; is the mass of heavy hydrocarbons to be mixed, g; W is the mass of low-mature shale samples composed of new organic matter, g; and are the masses of heavy hydrocarbons and kerogen contained in the medium- and low-maturity shale samples composed of new organic matter, g; k is the designed mass ratio of kerogen to heavy hydrocarbons.

[0189] Step 102, based on the low-mature shale samples with different organic matter compositions, conduct pyrolysis weight loss experiments of low-mature shale at different heating rates.

[0190] Based on the thermogravimetric analyzer, a pyrolysis weight loss experiment of low-mature shale with a heating rate of 5K / min was carried out. The pyrolysis weight loss curves of low-mature shale with different organic matter compositions were obtained as follows: Figure 2~Figure 3 shown.

[0191] Step 103, based on the low-mature shale samples with different organic matter compositions, high-temperature pyrolysis experiments of low-mature shale with different pyrolysis final temperatures are carried out, pyrolysis oil and gas products are collected, and the mass changes of substances involved in the pyrolysis reaction are quantitatively characterized.

[0192] High-temperature pyrolysis experiments of medium- and low-maturity shales were carried out at target temperatures of 320°C, 360°C, 400°C, 440°C, 480°C, and 520°C. The pyrolysis oil and pyrolysis gas produced in the experiments at different pyrolysis final temperatures were collected and weighed. The TOC test was performed on the medium- and low-maturity shale and pure kerogen samples obtained in step 101 to obtain the organic carbon content of the original sample and the kerogen. The residues of medium- and low-maturity shales with different organic matter compositions after pyrolysis experiments at different pyrolysis final temperatures were extracted with dichloromethane to determine the mass of heavy hydrocarbons remaining in the residues. The TOC test was performed on the residues after extraction to determine the organic carbon content in the residues, and the residual kerogen content after the pyrolysis experiment was calculated using formula (3). Finally, combined with the principle of mass conservation, the mass of coke generated by the reaction was determined using the difference subtraction method. In this way, the mass changes of various substances in medium- and low-maturity shales with different organic matter compositions at high temperature at different pyrolysis final temperatures can be obtained, and the mass fraction of each substance can be calculated according to formula (4).

[0193] (3);

[0194] (4);

[0195] The changes in the mass fractions of the various substances determined in the examples are shown in Table 1.

[0196] Table 1 Mass changes of various substances in high temperature pyrolysis experiment

[0197]

[0198] Step 104, combining the activation energy distribution hypothesis, using the Gaussian activation energy distribution method to separate the kerogen reaction and the heavy hydrocarbon reaction, and fitting to determine the reaction rate distributions of kerogen and heavy hydrocarbons respectively.

[0199] According to the experimental results of pyrolysis weight loss of low-middle-mature shale, the conversion rate curve was fitted using the least squares method based on the double parallel Gaussian distribution formula (5), and the fitting accuracy was calculated using formula (6). The fitting accuracy of the pyrolysis conversion rate curve of organic matter with a kerogen to heavy hydrocarbon mass ratio of 20 at a heating rate of 5K / min was 98.53%, and the fitting accuracy of the pyrolysis conversion rate curve of organic matter with a kerogen to heavy hydrocarbon mass ratio of 5 was 99.08%. Finally, the kerogen reaction and heavy hydrocarbon reaction were separated, and the various coefficients in the double parallel Gaussian distribution formula were determined, as shown in Table 2. In this way, the pyrolysis conversion rate distributions of kerogen and heavy hydrocarbons were determined. The reaction rate fitting curves of low-middle-mature shale composed of two organic matters are shown in Figure 2. Figure 4~Figure 5 shown.

[0200] (5);

[0201] (6);

[0202] In the formula, , represents the peak amplitude of the distribution, , represents the distribution peak center temperature, , Reflects the ratio of peak width to reaction temperature range. is the fitting accuracy between experimental data and calculated data; is the experimental value, is the calculated value, is the average value, as shown in Table 2.

[0203] Table 2 Distribution parameters of kerogen and heavy hydrocarbon pyrolysis reaction rates

[0204]

[0205] Step 105, based on the kerogen and heavy hydrocarbon reaction distribution data, the respective reaction mechanism functions are determined using the main graph method.

[0206] Based on the conversion rate data of kerogen and heavy hydrocarbon reactions, the ordinate value Y1 of the two reactions is calculated using the left side of the equal sign in formula (7). The ordinate value Y2 of the commonly used reaction mechanism function is calculated using the right side of the equal sign in formula (7), where the expressions of each mechanism function are summarized in Table 3. The Y(α)-α curve is drawn, and the mechanism function with the highest accuracy in fitting the experimental result Y1 is determined by combining the nonlinear least squares method, which is the mechanism function that most accurately describes the reaction of kerogen and heavy hydrocarbons. The fitting accuracy between the experimental data and the function calculation data is calculated using formula (6).

[0207] (7);

[0208] In the formula, is the conversion rate, is the conversion rate when the conversion rate is 0.5; is the differential form of the reaction mechanism function, It is the integral form of the reaction mechanism function, as shown in Table 3.

[0209] Table 3 Summary of mechanism function expressions

[0210]

[0211] The fitting curves of the pyrolysis reaction mechanism function of medium- and low-maturity shale composed of two organic matters are shown in Figure 6 and 7 As shown in the figure, the fitting results of the kerogen reaction and heavy hydrocarbon reaction mechanism functions under the two organic matter compositions are consistent. That is, for the kerogen reaction, the reaction level F2 model has the highest fitting accuracy, averaging 98.85%, while for the heavy hydrocarbon reaction, the geometric contraction R3 model and the diffusion D3 model have the highest R2, averaging 97.99%.

[0212] Step 106, summarizes the pyrolysis reaction of organic matter in the in-situ transformation process of low-middle-mature shale, based on the mass change data of various substances obtained from the high-temperature pyrolysis experiment of low-middle-mature shale, and uses data augmentation and adversarial generation network to expand the above mass change data of various substances to obtain merged data, combines the reaction mechanism function and Arrhenius formula, and uses genetic algorithm to determine the reaction kinetic equation and parameters.

[0213] In this embodiment, a genetic algorithm is used to determine the reaction kinetics equation, and its algorithm is given, which specifically includes:

[0214] (1) Based on the reactants and pyrolysis products in the high-temperature pyrolysis experiment of medium- and low-maturity shale, the main reactions in the in-situ transformation process of medium- and low-maturity shale are summarized as follows:

[0215] ①Kerogen pyrolysis reaction: (E1, A1);

[0216] ②Heavy hydrocarbon cracking reaction: (E2, A2);

[0217] ③ Pyrolysis oil cracking reaction: (E3, A3);

[0218] Wherein, Kerogen, HHC, Oil, Gas, Coke represent kerogen, heavy hydrocarbon, pyrolysis oil, pyrolysis gas, and coke, respectively. f is the stoichiometric coefficient of each substance in the reaction equation. E and A are the activation energy and frequency factor of the pyrolysis reaction, respectively.

[0219] The upper and lower limits of each stoichiometric coefficient, activation energy and frequency factor are set: f∈(0,1), E∈(100,300), A∈(1.0×1010,1.0×1016), allowing the optimal solution to be found within the parameter value range. In order to improve the calculation speed, the mass changes of kerogen, heavy hydrocarbons, pyrolysis oil and pyrolysis gas are fitted, and the sum of the stoichiometric coefficients in each reaction is equal to 1 as a constraint, so as to determine the stoichiometric coefficient of coke in the chemical equation.

[0220] (2) Set the population size in the genetic algorithm to 200, the number of iterations to 300, the crossover probability to 0.8, and the mutation probability to 0.2. Randomly generate the initial population.

[0221] (3) The kerogen reaction mechanism function determined in step 105 is a reaction second-order function, and the heavy hydrocarbon reaction is characterized by the geometric contraction R3 model and the diffusion D3 model. In one embodiment of the present invention, the geometric contraction R3 model is used for calculation. Combining the reaction rate formula and the Arrhenius formula, the mass changes of kerogen, heavy hydrocarbons, pyrolysis oil and pyrolysis gas under the current kinetic parameters are calculated by solving nonlinear differential equations. Compare the calculation results with the experimental results, and use formula (6) to obtain the fitting accuracy.

[0222] Reaction level F2 model: ;

[0223] Geometry Shrinkage R3 Model: ;

[0224] Reaction rate characterization: ;

[0225] Arrhenius formula: ;

[0226] In the formula, α is the conversion rate; t is the reaction time, h; k is the reaction rate constant, h -1 ; E is the activation energy, kJ / mol; A is the frequency factor, h-1 ; T is the reaction temperature, K; R is the universal gas constant, which is 8.314 J / (mol·K).

[0227] (4) Based on the current calculation results, the next generation population is generated through selection, crossover and mutation operations.

[0228] (5) In one embodiment of the present invention, when the fitting accuracy is higher than 90%, the operation is stopped to obtain the final reaction kinetic stoichiometric coefficients (fH, fO1, fO2, fG1, fG2, fG3, fC1, fC2, fC3), reaction activation energy (E1, E2, E3) and frequency factor (A1, A2, A3); otherwise, return to step (3).

[0229] The genetic algorithm was used to fit the mass changes of various substances in the pyrolysis reaction of medium and low-maturity shale under different organic matter compositions. The fitting results are shown in Figure 8 and Fig. 9 The obtained kinetic models of pyrolysis reaction of medium-low mature shale are shown in Tables 4 and 5. The fitting accuracy of the kinetic model of pyrolysis reaction of medium-low mature shale with kerogen to heavy hydrocarbon mass ratio of 20 and 5 is 92.6% and 91.8%, respectively.

[0230] Table 4 Summary of kinetic models for pyrolysis of medium- and low-maturity shales with a kerogen to heavy hydrocarbon mass ratio of 20

[0231]

[0232] Table 5 Summary of kinetic models for pyrolysis of medium- and low-maturity shales with a kerogen to heavy hydrocarbon mass ratio of 5

[0233]

[0234] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for constructing a pyrolysis kinetic model for in-situ conversion and exploitation of medium- and low-maturity shale oil, characterized in that: The following steps are involved: S10, preparing low-mature shale samples with different organic matter compositions; S20. Carry out pyrolysis weight loss experiments on low-mature shale at different heating rates; S30. Conduct high-temperature pyrolysis experiments on low-mature shale with different pyrolysis final temperatures, collect pyrolysis oil and gas products, and quantitatively characterize the mass changes of kerogen, heavy hydrocarbons, pyrolysis oil, pyrolysis gas and coke in the pyrolysis reaction, which are recorded as the mass change data of the experimental materials; S40, constructing a pyrolysis kinetics data augmentation model to generate an expanded data set based on the experimental substance mass change data; S50, establishing a generative adversarial network model, wherein the generator of the generative adversarial network model is responsible for generating theoretical material mass change data that conforms to physical laws, and the discriminator of the generative adversarial network model is responsible for distinguishing the difference between the generated data and the real experimental data; S60, using the expanded data set to train a generative adversarial network, and optimizing the generation effect of theoretical material mass change data by minimizing the loss function between the generator and the discriminator; S70, merging the theoretical material mass change data generated by the generative adversarial network with the experimental material mass change data to form merged data; S80, based on the combined data, combined with the Arrhenius formula and the reaction mechanism function, using a genetic algorithm to iteratively optimize the fitting accuracy of the prediction results and the experimental data to be no less than 90%, output the final reaction kinetic equation, reaction activation energy and frequency factor as the pyrolysis kinetic model for in-situ conversion and exploitation of medium- and low-maturity shale oil; The process of preparing low-mature shale samples with different organic matter compositions includes extracting low-mature shale powder samples with dichloromethane, rotary evaporating the extract to obtain heavy hydrocarbon samples, and acid washing the extracted solid powder to obtain kerogen samples; The mass calculation of the kerogen is specifically expressed as follows: ; In the formula, is the mass of remaining kerogen in the residue after the pyrolysis experiment, in g; is the mass of the residue after removing heavy hydrocarbons after pyrolysis experiments at different temperatures, in g; is the organic carbon content in shale residue, unit: % is the organic carbon content of the original kerogen, in %; The calculation of the coke quality is specifically expressed as follows: ; In the formula, is the mass of coke, in g; is the total mass of kerogen and heavy hydrocarbons in the initial sample, in g; is the remaining kerogen mass, in g; is the mass of remaining heavy hydrocarbons, in g; is the mass of pyrolysis oil, in g; is the mass of pyrolysis gas, in g; The calculation of the conversion rate is specifically expressed as follows: ; In the formula, is the conversion rate; is the initial mass, in g; is the mass at time t, in g; is the mass of the reaction endpoint, in g; The calculation of the conversion rate is specifically expressed as follows: ; In the formula, is the conversion rate, which indicates the proportion of reactants that have been converted, in units of min -1 ; is the heating rate, in K / min; is the weight loss rate, in g / K; is the initial mass, in g; The reaction activation energy is calculated by distributing a double parallel Gaussian distribution function, which is specifically expressed as follows: ; In the formula, is the distribution peak amplitude, ranging from 0 to 1; is the center temperature of the distribution peak, in K; is the standard deviation, reflecting the ratio of peak width to reaction temperature range; is the temperature variable, the unit is K; Wherein, the mechanism function is specifically expressed as follows: Reaction level F2 model: ; Geometry Shrinkage R3 Model: ; Reaction rate characterization: ; Arrhenius formula: ; In the formula, α is the conversion rate; t is the reaction time; k is the reaction rate constant; E is the activation energy; A is the frequency factor; T is the reaction temperature; R is the universal gas constant, which is 8.314 J / (mol·K); The integral form of the mechanism function is specifically expressed as follows: ; In the formula, is the integral form of the mechanism function; is the differential form of the mechanism function; is the conversion rate; The pyrolysis kinetic model for in-situ conversion and exploitation of medium- and low-maturity shale oil is specifically expressed as follows: ; In the formula, is the pre-exponential factor; is the activation energy, in kJ / mol; is the gas constant, which is taken as 8.314 J / (mol·K); is the temperature, in K; is the diffusion coefficient; is the space coordinate.

2. The method for constructing a pyrolysis kinetic model for in-situ conversion and exploitation of medium- and low-maturity shale oil according to claim 1, characterized in that: The pyrolysis kinetics data augmentation model includes a pyrolysis kinetics basic equation, a pyrolysis kinetics perturbation equation, a temperature field distribution equation, a component diffusion equation, a Gaussian noise perturbation model, a temperature gradient transformation model, and a component ratio adjustment model.

Images