Crude oil isothermal oxidation reaction mechanism model construction method and system based on neural network, electronic equipment and storage medium

Abstract

The invention provides a neural network-based crude oil isothermal oxidation reaction mechanism model construction method, which comprises the following steps: carrying out a crude oil isothermal oxidation experiment by adopting a thermogravimetric analyzer to obtain isothermal oxidation weight loss curves at different temperatures, and drawing a conversion rate change curve according to the isothermal oxidation weight loss curves; constructing different reaction kinetic models, and fitting the conversion rate change curve through the reaction kinetic models to obtain a reaction rate constant; constructing a multi-layer neural network model; the neural network model comprises an input layer, a hidden layer and an output layer which are connected in sequence; determining the weight of the neural network model according to the reaction rate constant to obtain a normalized model; and screening the reaction kinetic model through the normalized model to obtain the crude oil isothermal oxidation reaction mechanism model. According to the method, reaction kinetic parameters are converted into neural network weights and the contribution rate is quantified, so that the objectivity and precision of a crude oil isothermal oxidation reaction mechanism model are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of crude oil oxidation reaction kinetics, in particular to a crude oil isothermal oxidation reaction mechanism model construction method and system based on a neural network, an electronic device and a storage medium. BACKGROUND

[0002] In recent years, scholars have conducted a large number of studies on the oxidation kinetic parameters (activation energy, frequency factor and reaction kinetic model) of different oxidation stages (low-temperature oxidation stage, fuel deposition stage and high-temperature oxidation stage) in the process of crude oil temperature oxidation. However, under actual reservoir conditions, the heterogeneity of the reservoir leads to a slow advance of the oxidation thermal front, which results in most of the crude oil being in the low-temperature oxidation stage for a long time.

[0003] At present, the existing technology calculates the kinetic parameters of the crude oil after low-temperature oxidation through an n-order reaction model, but there is a large error in solving the kinetic parameters by artificially assuming the reaction mechanism function, and the control mechanism in the isothermal oxidation process of the crude oil cannot be reflected. Therefore, it is very necessary to design a crude oil isothermal oxidation reaction mechanism model construction method and system based on a neural network, an electronic device and a storage medium. SUMMARY

[0004] The purpose of the present application is to provide a crude oil isothermal oxidation reaction mechanism model construction method and system based on a neural network, an electronic device and a storage medium, which converts reaction kinetic parameters into neural network weights and quantifies the contribution rate, so as to improve the objectivity and precision of the crude oil isothermal oxidation reaction mechanism model.

[0005] To achieve the above-mentioned purpose, the present application provides the following solutions.

[0006] A crude oil isothermal oxidation reaction mechanism model construction method based on a neural network comprises the following steps:

[0007] A thermogravimetric analyzer is used to perform a crude oil isothermal oxidation experiment, and isothermal oxidation weight loss curves at different temperatures are obtained, and a conversion rate change curve is drawn according to the isothermal oxidation weight loss curves;

[0008] Different reaction kinetic models are constructed, and the conversion rate change curve is fitted through the reaction kinetic models to obtain a reaction rate constant;

[0009] A multi-layer neural network model is constructed; the neural network model comprises an input layer, a hidden layer and an output layer connected in sequence;

[0010] The weights of the neural network model are determined according to the reaction rate constant to obtain a normalized model;

[0011] The reaction kinetic models are screened through the normalized model to obtain a crude oil isothermal oxidation reaction mechanism model.

[0012] Optionally, the conversion rate change curve is plotted according to a first conversion rate, and the first conversion rate is calculated according to the following formula: Wherein, α is the conversion rate, m is the initial mass, m t is the sample mass at t time

[0013] is the sample mass at t time f is the sample mass at t time

[0014] Optionally, the reaction kinetics model comprises a geometric shrinkage submodel, a diffusion submodel and a nucleation submodel.

[0015] Optionally, different reaction kinetics models are constructed, and the reaction rate constant of the reaction kinetics model is solved by the conversion rate change curve, specifically: the second conversion rate of the model equation in the reaction kinetics model is replaced by the first conversion rate, the model equation after replacement is solved, the slope and intercept are obtained, and the slope and intercept are integrated into the reaction rate constant.

[0016] Optionally, the input layer is built-in with one neuron, and the input of the input layer is the reaction time; the hidden layer is built-in with 10 different neurons, and the neuron is built-in with any one of the model equations; the output layer is built-in with one neuron.

[0017] Optionally, the weight of the neural network model is determined according to the reaction rate constant, and a normalized model is obtained, comprising:

[0018] The first weight between the input layer and the hidden layer is determined according to the reaction rate constant;

[0019] The weight between the hidden layer and the output is optimized by the least square method to obtain the second weight;

[0020] The second weight is normalized to obtain the normalized model.

[0021] Optionally, the first weight between the input layer and the hidden layer is determined according to the reaction rate constant, comprising:

[0022] The model equation is taken as an activation function;

[0023] The reaction rate constant is taken as the first weight;

[0024] Based on the activation function, a vector matrix is constructed according to the first weight.

[0025] A neural network-based crude oil isothermal oxidation reaction mechanism model construction system, comprising:

[0026] The data acquisition module is configured to perform an isothermal oxidation experiment on crude oil by using a thermal gravimetric analyzer to obtain isothermal oxidation weight loss curves at different temperatures, and draw a conversion rate change curve according to the isothermal oxidation weight loss curves.

[0027] The data processing module is configured to construct different reaction kinetic models, and fit the conversion rate change curve by using the reaction kinetic models to obtain a reaction rate constant.

[0028] The network construction module is configured to construct a multi-layer neural network model, and the neural network model comprises an input layer, a hidden layer and an output layer connected in sequence.

[0029] The network optimization module is configured to determine the weight of the neural network model according to the reaction rate constant to obtain a normalized model.

[0030] The model generation module is configured to screen the reaction kinetic models by using the normalized model to obtain an isothermal oxidation reaction mechanism model of crude oil.

[0031] An electronic device comprises:

[0032] At least one processor;

[0033] A memory in communication connection with the at least one processor;

[0034] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the model construction method described above.

[0035] A non-transient computer readable storage medium storing computer instructions, the computer instructions being used to enable a computer to perform the model construction method described above.

[0036] According to the specific embodiments of the present application, the present application discloses the following technical effects: the isothermal oxidation reaction mechanism model construction method of crude oil based on a neural network provided by the present application, the method comprises the following steps: performing an isothermal oxidation experiment on crude oil by using a thermal gravimetric analyzer to obtain isothermal oxidation weight loss curves at different temperatures, and drawing a conversion rate change curve according to the isothermal oxidation weight loss curves; constructing different reaction kinetic models, and fitting the conversion rate change curve by using the reaction kinetic models to obtain a reaction rate constant; constructing a multi-layer neural network model, and the neural network model comprises an input layer, a hidden layer and an output layer connected in sequence; determining the weight of the neural network model according to the reaction rate constant to obtain a normalized model; and screening the reaction kinetic models by using the normalized model to obtain an isothermal oxidation reaction mechanism model of crude oil. The method converts the reaction kinetic parameters into neural network weights and quantifies the contribution rate, thereby improving the objectivity and precision of the isothermal oxidation reaction mechanism model of crude oil. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings needed in the embodiments will be briefly introduced as follows. Obviously, the accompanying drawings in the following description only represent some embodiments of the present application, and can help those skilled in the art to obtain other drawings based on these accompanying drawings without any creative effort.

[0038] Figure 1 A flow chart of the method for constructing the isothermal oxidation reaction mechanism model of crude oil according to the present application is shown in FIG. 1.

[0039] Figure 2 An isothermal oxidation weight loss curve of the isothermal oxidation experiment of crude oil according to the embodiment of the present application is shown in FIG. 2.

[0040] Figure 3 A conversion rate change curve of the isothermal oxidation experiment of crude oil according to the embodiment of the present application is shown in FIG. 3.

[0041] Figure 4 An MLP neural network output result graph according to the embodiment of the present application is shown in FIG. 4. DETAILED DESCRIPTION

[0042] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments only represent some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without any creative effort fall within the scope of the present application.

[0043] The above-mentioned purposes, features and advantages of the present application can be more obvious and easy to understand. The present application will be described in further detail below with reference to the accompanying drawings and specific embodiments.

[0044] As shown in FIG. 1, the present application provides a method for constructing an isothermal oxidation reaction mechanism model of crude oil based on a neural network, which comprises the following steps: Figure 1 Step 100: An isothermal oxidation experiment of crude oil is performed by using a thermogravimetric analyzer to obtain isothermal oxidation weight loss curves at different temperatures, and a conversion rate change curve is drawn according to the isothermal oxidation weight loss curves.

[0045] Specifically, the crude oil in the present embodiment is thick oil, and the oxidation experiment is performed at 200℃. The obtained isothermal oxidation weight loss curve is shown in FIG. 2, in which the horizontal axis of the coordinate represents the reaction time, and the vertical axis of the coordinate represents the mass loss percentage. The conversion rate change curve drawn according to the curve is shown in FIG. 3.

[0046] Figure 2 The conversion rate change curve is a curve of the conversion rate changing with time, and the calculation formula of the conversion rate is as follows: Figure 3 The conversion rate change curve is a curve of the conversion rate changing with time, and the calculation formula of the conversion rate is as follows:​

[0047] a = 1 - (1 - a) m-m t ;

[0048] m - m f

[0049] wherein a is the conversion rate, m is the initial mass, m t is the sample mass at time t, m f is the sample mass at the end of the reaction.

[0050] Step 200: Construct different reaction kinetic models, and fit the conversion rate change curve through the reaction kinetic models to obtain the reaction rate constant;

[0051] Specifically, the reaction kinetic models include: a geometric contraction submodel, a diffusion submodel, and a nucleation submodel. The geometric contraction submodel includes three reaction models R1, R2, and R3, the diffusion submodel includes four reaction models D1, D2, D3, and D4, and the nucleation submodel includes two reaction models A m and A u , and the expressions are as follows:

[0052] R1: 1 - (1 - a) = kt + k0;

[0053] R2: 1 - (1 - a) 1 / 2 = kt + k0;

[0054] R3: 1 - (1 - a) 1 / 3 = kt + k0;

[0055] D1: a 2 = kt + k0;

[0056] D2: (1 - a) In (1 - a) + a = kt + k0;

[0057] D3: [1 - (1 - a) 1 / 3 ] 2 = kt + k0;

[0058] D4: 1 - 2 / 3 a - (1 - a) 2 / 3 = kt + k0;

[0059] A m : [-In (1 - a) 1 / m ] = kt + k0 (m = 2, 4);

[0060] A u : In a / (1 - a) = kt + k0;

[0061] wherein k and k0 are both reaction rate constants, and t is the reaction time. For each model equation, the conversion rate obtained through the experiment is substituted into the left side of its equation, and it becomes y = kt + k0, linear fitting is performed on the equation, and the obtained slope and intercept correspond to the reaction rate constant k and k0, respectively.

[0062] Step 300: constructing a multi-layer neural network model; the neural network model comprises an input layer, a hidden layer and an output layer connected in sequence;

[0063] Specifically, the neural network model of the embodiment adopts a multi-layer perception (MLP) neural network model, the input layer of which is provided with one neuron for inputting the reaction time; the hidden layer is provided with 10 different neurons, each of which corresponds to one kinetic model; and the output layer is provided with one neuron, and the output of the output layer is the conversion rate change a(t). The neural network a k The expression is:

[0064]

[0065] Wherein, f(a) is a kinetic model function, x k is a neuron of a previous layer, z kj is a weight between neurons of layers k and j, z k0 is a bias.

[0066] Step 400: determining the weight of the neural network model according to the reaction rate constant, and obtaining a normalized model;

[0067] Specifically, the input of the input layer is set as time, the reaction rate constant k and k0 are set as the weight z k1 and the bias z k0 from the input layer to the hidden layer, and the equation of the kinetic reaction model is set as an activation function, thereby constructing a vector matrix Z1i, and the expression is:

[0068] Wherein, z 21 and z 20 are the reaction rate constants k and k0 obtained by fitting the second kinetic model, z 31 and z 30 are the reaction rate constants k and k0 obtained by fitting the third kinetic model, z 41 , z 40 , z 50 and z 50 and so on. The weight z2 from the hidden layer to the output layer is optimized by the least square method, so that the error between the output result of the neural network model, i.e. the conversion rate change curve, and the conversion rate obtained according to the experiment is minimized. The optimization formula of the weight z2 is:

[0069] z2 = (X T X) -1 X T aexp ;

[0070] wherein, X = f(zx), a exp is the experimental value of the conversion rate. Finally, the weight z2 is normalized to obtain the percentage form of each kinetic contribution rate, and the calculation formula is:

[0071]

[0072] wherein, C is the contribution rate, z i is the weight z2 corresponding to different neural networks.

[0073] Step 500: screening the reaction kinetics model by the normalization model to obtain the crude oil isothermal oxidation reaction mechanism model.

[0074] Specifically, the kinetic model with a higher contribution rate in each kinetic model is selected to be combined as the crude oil isothermal oxidation reaction mechanism model at the temperature. The output result of the MLP neural network of the embodiment is as shown in Figure 4 It can be known that the isothermal oxidation process of the crude oil at 200 DEG C is mainly controlled by two reaction mechanisms, which are R3 with a contribution rate of 97.9% and D2 with a contribution rate of 2.1%, wherein R3 is the main reaction control model.

[0075] It should be noted that because the output values of all models are close to the range after the normalization processing, the scale of the output is consistent, so the weight is used to unify k and k0, thereby realizing linear fitting, and k and k0 are calculated, and the value is not directly modified, but the contribution is adjusted through the weight z2, and z2 is adaptively adjusted in the forward transmission process, thereby offsetting the difference in the order of magnitude of different k.

[0076] The application also provides a crude oil isothermal oxidation reaction mechanism model construction system based on a neural network, comprising:

[0077] A data acquisition module is used to perform a crude oil isothermal oxidation experiment by using a thermogravimetric analyzer, to obtain isothermal oxidation weight loss curves at different temperatures, and to draw conversion rate change curves according to the isothermal oxidation weight loss curves;

[0078] A data processing module is used to construct different reaction kinetics models, and to fit the conversion rate change curves through the reaction kinetics models to obtain reaction rate constants;

[0079] A network construction module is used to construct a multi-layer neural network model; the neural network model comprises an input layer, a hidden layer and an output layer connected in sequence;

[0080] A network optimization module is used to determine the weight of the neural network model according to the reaction rate constant to obtain a normalization model.

[0081] A model generation module is configured to screen the reaction kinetics model by a normalized model to obtain an isothermal oxidation reaction mechanism model of crude oil.

[0082] The application further provides an electronic device, comprising:

[0083] at least one processor;

[0084] a memory in communication connection with the at least one processor;

[0085] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the model construction method.

[0086] The application further provides a non-transient computer readable storage medium storing computer instructions, and the computer instructions are used to enable a computer to perform the model construction method.

[0087] The application has the following beneficial effects:

[0088] 1) The contribution rate of different kinetics models is automatically quantified by a neural network, subjective errors caused by artificially assuming a reaction mechanism function in a traditional method are avoided, and mechanism objectivity is improved;

[0089] 2) The reaction rate constant is used as a neural network weight, and global optimization is performed in combination with a least square method, and the solving precision of kinetics parameters is improved;

[0090] 3) The multi-layer perception (MLP) structure used is compatible with various kinetics models, and model generalization and adaptability are enhanced.

[0091] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same or similar parts of each embodiment can be mutually referred to.

[0092] The principle and implementation mode of the application are described by using specific examples in the application, and the above embodiment description is only used to help understand the method and core idea of the application;Meanwhile, for those skilled in the art, according to the idea of the application, the specific implementation mode and application range will be changed.The above description of the specification should not be understood as a limitation of the application.

Claims

1. A method for constructing a crude oil isothermal oxidation reaction mechanism model based on a neural network, characterized in that: The steps include: A crude oil isothermal oxidation experiment is conducted using a thermogravimetric analyzer to obtain isothermal oxidation weight loss curves at different temperatures, and a conversion rate change curve is drawn based on the isothermal oxidation weight loss curves; Constructing different reaction kinetic models, and fitting the conversion rate change curve using the reaction kinetic models to obtain a reaction rate constant; Constructing a multi-layer neural network model; the neural network model includes an input layer, a hidden layer, and an output layer connected in sequence; Determining the weight of the neural network model according to the reaction rate constant to obtain a normalized model; The reaction kinetics model is screened by the normalized model to obtain a crude oil isothermal oxidation reaction mechanism model.

2. The method for constructing a crude oil isothermal oxidation reaction mechanism model based on a neural network according to claim 1, characterized in that: The conversion rate change curve is drawn according to the first conversion rate, and the calculation formula of the first conversion rate is: α= m-m t ; Wherein, α is the first conversion rate, m is the starting mass mm f Quantity, m t is the sample mass at time t, m f is the sample mass at the end of the reaction.

3. The method for constructing a crude oil isothermal oxidation reaction mechanism model based on a neural network according to claim 2, characterized in that: The reaction kinetics model includes: a geometric contraction sub-model, a diffusion sub-model and a nucleation sub-model.

4. The method for constructing a crude oil isothermal oxidation reaction mechanism model based on a neural network according to claim 3, characterized in that: Constructing different reaction kinetic models, and fitting the conversion rate change curve through the reaction kinetic model to obtain the reaction rate constant, specifically: replacing the second conversion rate of the model equation in the reaction kinetic model with the first conversion rate, solving the replaced model equation to obtain the slope and intercept, and integrating the slope and intercept into the reaction rate constant.

5. The method for constructing a crude oil isothermal oxidation reaction mechanism model based on a neural network according to claim 4, characterized in that: The input layer has one built-in neuron, and the input of the input layer is the reaction time; the hidden layer has 10 different built-in neurons, and any one of the model equations is built into the neurons; and the output layer has one built-in neuron.

6. The method for constructing a crude oil isothermal oxidation reaction mechanism model based on a neural network according to claim 5, characterized in that: Determining the weight of the neural network model according to the reaction rate constant to obtain a normalized model includes: determining a first weight between the input layer and the hidden layer according to the reaction rate constant; Optimizing the weight between the hidden layer and the output by a least squares method to obtain a second weight; The second weight is normalized to obtain the normalized model.

7. The method for constructing a crude oil isothermal oxidation reaction mechanism model based on a neural network according to claim 6, characterized in that: Determining a first weight between the input layer and the hidden layer according to the reaction rate constant includes: Using the model equation as an activation function; Using the reaction rate constant as the first weight; Based on the activation function, a vector matrix is ​​constructed according to the first weights.

8. A system for constructing a crude oil isothermal oxidation reaction mechanism model based on a neural network, characterized in that: include: A data acquisition module is used to perform an isothermal oxidation experiment of crude oil using a thermogravimetric analyzer, obtain isothermal oxidation weight loss curves at different temperatures, and draw a conversion rate change curve based on the isothermal oxidation weight loss curves; A data processing module is used to construct different reaction kinetic models and fit the conversion rate change curve through the reaction kinetic models to obtain a reaction rate constant; A network construction module is used to construct a multi-layer neural network model; the neural network model includes an input layer, a hidden layer and an output layer connected in sequence; A network optimization module, configured to determine the weight of the neural network model according to the reaction rate constant to obtain a normalized model; The model generation module is used to screen the reaction kinetics model through the normalized model to obtain a crude oil isothermal oxidation reaction mechanism model.

9. An electronic device, characterized in that: include: at least one processor; a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the model building method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable a computer to execute the model building method according to any one of claims 1 to 7.

Images