Method for predicting sulfur content in petroleum coke, storage medium, electronic device, and apparatus

Abstract

The application relates to a method for predicting the sulfur content in petroleum coke, a storage medium, an electronic device and an apparatus, and belongs to the field of delayed coking for producing petroleum coke. The method comprises the following steps: firstly, detecting the sulfur content in petroleum coke products and indexes of various influencing factors in raw materials, thereby establishing an original database of the sulfur content in petroleum coke and the influencing factors in raw materials; secondly, adopting an improved neural network, taking the original database as the basis, and establishing a prediction model of the sulfur content in petroleum coke and the influencing factors in raw materials; and finally, inputting the values of the influencing factors in raw materials into the prediction model, thereby obtaining the prediction value of the sulfur content in petroleum coke generated by the batch of raw materials. The application significantly improves the prediction accuracy, and makes the model capable of accurately predicting the sulfur content in petroleum coke by using only two variables of the sulfur content and the carbon residue content in raw materials. In addition, the application uses a particle swarm optimization algorithm into the neural network model, reversely trains unknown parameters in the model to evaluate the unknown parameters, and improves the prediction accuracy of the model.

Description

Technical Field

[0001] This invention pertains to the delayed coking production of petroleum coke in the petroleum refining field, specifically a method, storage medium, electronic equipment, and apparatus for predicting the sulfur content in petroleum coke. Background Technology

[0002] Delayed coking is characterized by mature technology, lower investment, lower operating costs, and strong adaptability to raw materials, making it a primary processing method for low-quality heavy oil. Petroleum coke is the main product of delayed coking units, with a yield typically between 25% and 30%. Petroleum coke has a wide range of applications, including as fuel in the glass, cement, and CFB boilers of thermal power plants; as an electrode material for electrolytic aluminum and steel production; as a carbonaceous reducing agent for silicon carbide and industrial silicon; and as a feedstock for coal gasification in the chemical industry. Different industries have different requirements for the quality grade of petroleum coke. The main indicator for classifying petroleum coke quality is its sulfur content; the lower the sulfur content, the higher the quality grade. The price difference between different quality grades of petroleum coke is significant, meaning that sulfur content is the primary basis for evaluating petroleum coke prices.

[0003] Producing low-sulfur petroleum coke can significantly improve the economic efficiency of delayed coking units. Establishing a model to predict the sulfur content of petroleum coke is fundamental to guiding refineries in optimizing their processes to produce low-sulfur petroleum coke. However, there is currently no mature technology for accurately predicting the sulfur content in petroleum coke. Therefore, developing a method for predicting the sulfur content of petroleum coke is of great significance. Summary of the Invention

[0004] To address the lack of existing technologies for predicting sulfur content in petroleum coke, this invention provides a method, storage medium, electronic device, and apparatus for predicting sulfur content in petroleum coke, offering guidance for oil refining enterprises to optimize process conditions and produce low-sulfur petroleum coke.

[0005] The technical solution adopted by this invention to solve the above-mentioned technical problems is as follows: a method for predicting the sulfur content in petroleum coke, comprising the following steps:

[0006] 1) Detect the sulfur content of petroleum coke products and the indicators of various influencing factors in its raw materials, thereby establishing a raw database of petroleum coke sulfur content and influencing factors in raw materials;

[0007] 2) An improved neural network was used to establish a predictive model for the sulfur content of petroleum coke and the influencing factors in the raw materials, based on the original database.

[0008] 3) Detect the values ​​of influencing factors in the raw materials and input the values ​​into the prediction model in step 2) to obtain the predicted value of sulfur content in the petroleum coke produced from this batch of raw materials.

[0009] As an optimized scheme of the above-mentioned method for predicting the sulfur content in petroleum coke, in step 1), the influencing factors in the production raw materials refer to the sulfur content and residual carbon content.

[0010] As another optimized scheme of the above-mentioned method for predicting the sulfur content in petroleum coke, the specific operation of step 1) is as follows: In industrial practice or experiment, first detect the content of various components in the raw material, then detect the sulfur content in the petroleum coke product produced from the raw material, organize the component content data in the raw material and the sulfur content data in the product, determine the components that affect the sulfur content, label them as influencing factors, and then establish an original database of petroleum coke sulfur content and influencing factors in the raw material.

[0011] As another optimized scheme of the above-mentioned method for predicting sulfur content in petroleum coke, the specific operation of establishing a predictive model for the sulfur content of petroleum coke and the influencing factors in the raw materials is as follows:

[0012] ① Establish a typical neural network model

[0013] This neural network model consists of an input layer, hidden layers, and an output layer. The three-layer neural network completes the mapping from input to output.

[0014] ② Normalized data processing

[0015] The input variables are normalized according to formula (1) and mapped to the range [0, 1].

[0016]

[0017] In the formula, x max x is the maximum value among the input variables. min Let y be the minimum value among the input variables. max The maximum value after normalization is 1, y min The minimum value of the normalized data is 0, x * These are the input variables after normalization;

[0018] ③ Set the activation function of the hidden layer

[0019] The neurons in the hidden layer use the linear activation function purelin of formula (2), and the neurons in the output layer also use the linear activation function purelin of formula (2).

[0020] f(x)=x (2)

[0021] ④ Determine the number of neurons in the hidden layer

[0022] If the input layer has m neurons and the hidden layer has h neurons, then according to Kolmogorov's theorem, the number of neurons in the hidden layer is h = 2m + 1.

[0023] ⑤ Establish a neural network model

[0024] The input layer is defined as X = (x1, x2, ..., x...). m The hidden layer is represented as Net = (net1, net2, ..., net2). h The output layer is represented as Y = (y1, y2, ..., y...). n )express;

[0025] The calculation formula for the parameters in the hidden layer is shown in formula (3):

[0026]

[0027] In the formula, x i This represents the i-th parameter of the input layer, i∈(1,m), net j α represents the j-th parameter of the hidden layer, j∈(1,h); i,j Let A be a factor in the exponential matrix, where the exponential matrix A = (α... 1,1 ,...,α 1,h ;...;α m,1 ,...,α m,h ), has m×h parameters; w i,j Let W represent the hidden layer weight matrix, where W = (w 1,1 ,...,w 1,h ;...;w m,1 ,...,w m,h It has m×h parameters;

[0028] The calculation formula for the parameters in the output layer is shown in formula (4):

[0029]

[0030] In the formula, y represents the predicted sulfur content; j∈(1,h); t j Let T be a factor in the output layer weight vector, where T = (t1, ..., t2) h The system has h parameters; and W, A and T need to be estimated, so the number of parameters to be estimated is 2×m×h+h.

[0031] As another optimization scheme for the above-mentioned method for predicting the sulfur content in petroleum coke, in step ①, the input variables of the neural network model are the sulfur content and residual carbon content of the raw material, and the output variable is the sulfur content of the petroleum coke.

[0032] As another optimization scheme for the above-mentioned method for predicting sulfur content in petroleum coke, the operation for estimating the unknown parameters in step ⑤ is as follows: extract the sulfur content and residual carbon content in the raw materials and the corresponding sulfur content in the products from the original database established in step 1), and use them as the input variables of step ① and the output variables of step ⑤, respectively. Then, use the particle swarm optimization algorithm to estimate the unknown parameters in step ⑤, and the estimation needs to minimize the objective function Φ:

[0033]

[0034] In the formula, i represents the i-th raw material, and y represents the detected or experimental value of the sulfur content of coke. This represents the predicted sulfur content of coke, and nfeed represents the number of raw materials.

[0035] As another optimization scheme for the above-mentioned method for predicting sulfur content in petroleum coke, the specific operation of using the particle swarm optimization algorithm to estimate the unknown parameters in step ⑤, and minimizing the objective function Φ during the estimation, is as follows:

[0036] A) Select a set of influencing factors from the original database of petroleum coke sulfur content and influencing factors in raw materials, and substitute them into the neural network model; at the same time, use the particle swarm optimization algorithm to randomly generate an initial particle swarm, and assign its values ​​to the unknown parameters in step ⑤, and then calculate according to formula (4)

[0037] B) The result calculated in step A The actual detected sulfur content y corresponding to the group of influencing factors stored in the original database is substituted into formula (5) to calculate the objective function Φ;

[0038] C) Determine if the objective function Φ satisfies the termination condition. If yes, define the set of data as the global optimum and assign its value to the unknown parameters in the neural network model; otherwise, proceed to step D).

[0039] D) Update the particle swarm and repeat steps A-C until the global optimum is obtained and its value is assigned to the unknown parameters in the neural network model.

[0040] A machine-readable storage medium storing executable code that, when executed by a processor of an electronic device, enables the processor to implement at least one of the aforementioned prediction methods.

[0041] An electronic device for recording decision-making behavior includes a memory and a processor, wherein the memory stores an executable program, instructions, or code, which, when executed by the processor, enables the processor to perform execution according to any of the aforementioned prediction methods.

[0042] An optimized solution for the aforementioned electronic device that records decision-making behavior includes a processor, a memory, an external device, and an I / O interface. The memory stores programs, instructions, or code executed by the processor. The processor executes any of the aforementioned prediction methods based on the programs, instructions, or code stored in the memory. The I / O interface receives or sends data under the control of the processor. The external device inputs data to the processor and displays data output by the processor.

[0043] A petroleum coke sulfur content prediction device, the sulfur content prediction device having a prediction module that uses any of the above prediction methods to predict sulfur content.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] 1) This invention uses an improved neural network to establish a model for predicting the sulfur content of petroleum coke, which significantly improves the prediction accuracy. The model can accurately predict the sulfur content of petroleum coke using only two variables: the sulfur content of the raw material and the residual carbon content. In addition, this invention uses a particle swarm optimization algorithm to be substituted into the neural network model and back-trained to estimate the unknown parameters in the model, thereby improving the prediction accuracy of the model.

[0046] 2) The method for predicting the sulfur content of petroleum coke in this invention only requires data on the sulfur content and residual carbon content of the raw materials to predict the sulfur content of petroleum coke. It requires less analysis of the raw materials and has lower analysis and time costs. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the topological structure of the neural network model of the present invention;

[0048] Figure 2 A schematic diagram illustrating the process of estimating unknown parameters in a neural network model. Detailed Implementation

[0049] The technical solution of the present invention will be further described in detail below with reference to specific embodiments. The parts not described in the following embodiments of the present invention are all technologies well known to those skilled in the art, such as the topology of neural networks, particle swarm optimization algorithms, etc.

[0050] Example 1

[0051] A method for predicting the sulfur content in petroleum coke includes the following steps:

[0052] 1) Detect the sulfur content of petroleum coke products and the indicators of various influencing factors in its raw materials, thereby establishing a raw database of petroleum coke sulfur content and influencing factors in raw materials;

[0053] Research has revealed that the sulfur content and residual carbon content of raw materials are the main factors affecting the sulfur content of petroleum coke. Therefore, the influencing factors in the production raw materials mentioned in this step refer to the sulfur content and residual carbon content. This step involves organizing industrial test data to obtain a large amount of data containing the sulfur content of raw materials, residual carbon content, and petroleum coke sulfur content. Specifically, in industrial practice or experiments, the content of various components in the raw materials is first tested, and then the sulfur content in the petroleum coke product produced from the raw materials is tested. The component content data in the raw materials and the sulfur content data in the product are organized to identify the components that affect the sulfur content and label them as influencing factors. In this way, a raw database of petroleum coke sulfur content and influencing factors in the raw materials is established.

[0054] 2) An improved neural network was used to establish a predictive model for the sulfur content of petroleum coke and the influencing factors in the raw materials, based on the original database.

[0055] The specific steps are as follows:

[0056] ① Establish a typical neural network model

[0057] This neural network model consists of an input layer, hidden layers, and an output layer. The three layers of the neural network complete the mapping from input to output, such as... Figure 1 As shown;

[0058] In this step, the input variables of the neural network model are the sulfur content and residual carbon content of the raw material, and the output variable is the sulfur content of the petroleum coke.

[0059] ② Normalized data processing

[0060] The input variables are normalized according to formula (1) and mapped to the range [0, 1].

[0061]

[0062] In the formula, x max x is the maximum value among the input variables. min Let y be the minimum value among the input variables. max The maximum value after normalization is 1, y min The minimum value of the normalized data is 0, x * These are the input variables after normalization;

[0063] ③ Set the activation function of the hidden layer

[0064] The neurons in the hidden layer use the linear activation function purelin of formula (2), and the neurons in the output layer also use the linear activation function purelin of formula (2).

[0065] f(x)=x (2)

[0066] ④ Determine the number of neurons in the hidden layer

[0067] Assume the input layer has m neurons, as shown in the attached diagram. Figure 1 As shown, m takes the value 2, and the number of neurons in the hidden layer is h. Therefore, according to Kolmogorov's theorem, the number of neurons in the hidden layer is h = 2m + 1, as shown in the appendix. Figure 1 As shown, h takes the value 5;

[0068] ⑤ Establish a neural network model

[0069] The input layer is defined as X = (x1, x2, ..., x...). m The hidden layer is represented as Net = (net1, net2, ..., net2). h The output layer is represented as Y = (y1, y2, ..., y...). n )express;

[0070] The calculation formula for the parameters in the hidden layer is shown in formula (3):

[0071]

[0072] In the formula, x i This represents the i-th parameter of the input layer, i∈(1,m), net j α represents the j-th parameter of the hidden layer, j∈(1,h); i,j Let A be a factor in the exponential matrix, where the exponential matrix A = (α... 1,1 ,...,α 1,h ;...;α m,1 ,...,α m,h ), has m×h parameters; w i,j Let W represent the hidden layer weight matrix, where W = (w 1,1 ,...,w 1,h ;...;w m,1 ,...,w m,h It has m×h parameters;

[0073] The calculation formula for the parameters in the output layer is shown in formula (4):

[0074]

[0075] In the formula, y represents the predicted sulfur content value, as shown in the appendix. Figure 1 As shown; j∈(1,h); t j Let T be a factor in the output layer weight vector, where T = (t1, ..., t2) h The system has h parameters; and W, A, and T require parameter estimation calculations, so the number of parameters to be estimated is 2×m×h+h.

[0076] 3) Detect the values ​​of influencing factors in the raw materials and input the values ​​into the prediction model in step 2) to obtain the predicted value of sulfur content in the petroleum coke produced from this batch of raw materials.

[0077] Example 2

[0078] This embodiment is a further optimization based on Embodiment 1. Its main body is the same as Embodiment 1, but the optimization lies in the following: the operation for estimating the unknown parameters in step ⑤ involves extracting the sulfur content and residual carbon content of the raw materials and the corresponding sulfur content of the products from the original database established in step 1), using these as the input variables for step ① and the output variables for step ⑤, respectively. Then, the particle swarm optimization algorithm is used to estimate the unknown parameters in step ⑤, and the estimation requires minimizing the objective function.

[0079]

[0080] In the formula, i represents the i-th raw material, and y represents the detected or experimental value of the sulfur content of coke. This represents the predicted sulfur content of coke, and nfeed represents the number of raw materials.

[0081] Example 3

[0082] This embodiment is a further optimization based on Embodiment 2. Its main body is the same as Embodiment 2, but the optimization lies in the fact that the particle swarm optimization algorithm is used to estimate the unknown parameters in step ⑤, and the specific operation of minimizing the objective function Φ during the estimation is as follows: Figure 2 As shown:

[0083] A) Select a set of influencing factors from the original database of petroleum coke sulfur content and influencing factors in raw materials, and substitute them into the neural network model; at the same time, use the particle swarm optimization algorithm to randomly generate an initial particle swarm, and assign its values ​​to the unknown parameters in step ⑤, and then calculate according to formula (4)

[0084] B) The result calculated in step A The actual detected sulfur content y corresponding to the group of influencing factors stored in the original database is substituted into formula (5) to calculate the objective function Φ;

[0085] C) Determine if the objective function Φ satisfies the termination condition. If yes, define the set of data as the global optimum and assign its value to the unknown parameters in the neural network model; otherwise, proceed to step D).

[0086] D) Update the particle swarm and repeat steps A-C until the global optimum is obtained and its value is assigned to the unknown parameters in the neural network model.

[0087] Example 4

[0088] A machine-readable storage medium storing executable code, which, when executed by a processor of an electronic device, enables the processor to implement at least the prediction method described in any one of the embodiments 1-3 above.

[0089] Example 5

[0090] An electronic device for recording decision-making behavior includes a memory and a processor. The memory stores an executable program, instructions, or code. When the executable program, instructions, or code is executed by the processor, the processor is able to perform the execution according to any one of the methods in embodiments 1-3 above.

[0091] Example 6

[0092] This embodiment is an improvement on embodiment 5. Its main functions and results are the same as in embodiment 5, but the improvements are as follows: it includes a processor, memory, external devices, I / O interfaces, a bus, and necessary network devices. The memory includes random access memory, cache memory, read-only memory, and a memory cell array consisting of at least one memory cell, used to store programs or instructions executed by the processor. The processor is used to execute the methods described in any one of embodiments 1-3 according to the programs or instructions stored in the memory. The I / O interface is used to receive or send data under the control of the processor. The external devices are used to input data to the processor and display the data output by the processor, including but not limited to any electronic product that can interact with the user via a keyboard, remote control, touchpad, or voice control device, such as a computer, smartphone, ordinary mobile phone, tablet computer, etc. The network devices include but are not limited to computers, network hosts, single network servers, multiple network server sets, or a cloud consisting of multiple servers.

[0093] Example 7

[0094] A petroleum coke sulfur content prediction device, the sulfur content prediction device having a prediction module for predicting sulfur content using any one of the prediction methods in Examples 1-3 above.

Claims

1. A method for predicting the sulfur content in petroleum coke, characterized in that, Includes the following steps: 1) Detect the sulfur content of petroleum coke products and the indicators of various influencing factors in its raw materials, thereby establishing a raw database of petroleum coke sulfur content and influencing factors in raw materials. The specific operation is as follows: In industrial practice or experiments, the content of various components in the raw materials is first tested, and then the sulfur content in the petroleum coke product produced from the raw materials is tested. The data on the content of components in the raw materials and the sulfur content in the product are compiled, the components that affect the sulfur content are identified, and they are labeled as influencing factors. Then, an original database of the sulfur content of petroleum coke and the influencing factors in the raw materials is established. 2) Using an improved neural network and based on the original database, a predictive model for the sulfur content of petroleum coke and its influencing factors in raw materials is established. The specific operation is as follows: ① Establish a neural network model This neural network model consists of an input layer, hidden layers, and an output layer. The three-layer neural network completes the mapping from input to output. ② Normalized data processing The input variables are normalized according to formula (1) and mapped to the range [0, 1]. （1） In the formula, The maximum value among the input variables. The minimum value among the input variables. This represents the maximum value after normalization, which is 1 in this case. The minimum value of the normalized data is 0. These are the input variables after normalization; ③ Set the activation function of the hidden layer The neurons in the hidden layer use the linear activation function purelin of formula (2), and the neurons in the output layer also use the linear activation function purelin of formula (2). （2） ④ Determine the number of neurons in the hidden layer Assuming the input layer has m neurons and the hidden layer has h neurons, then according to Kolmogorov's theorem, the number of neurons in the hidden layer is... ; ⑤ Establish a neural network model Input layer with This indicates that the hidden layer is... This indicates that the output layer uses express; The calculation formula for the parameters in the hidden layer is shown in formula (3): （3）； In the formula, This represents the i-th parameter in the input layer. , This represents the j-th parameter of the hidden layer. ; Let be a factor in the exponent matrix, the exponent matrix ,have One parameter; Represents the hidden layer weight factors, hidden layer weight matrix ,have One parameter; The calculation formula for the parameters in the output layer is shown in formula (4): （4）； In the formula, y represents the predicted sulfur content value. ; For a factor in the output layer weight vector, the output layer weight vector... ,have One parameter; and , and Parameter estimation calculations are required; the number of parameters to be estimated is [number missing]. indivual; 3) Detect the values ​​of influencing factors in the raw materials and input the values ​​into the prediction model in step 2) to obtain the predicted value of sulfur content in the petroleum coke produced from this batch of raw materials.

2. The method for predicting sulfur content in petroleum coke according to claim 1, characterized in that: In step 1), the influencing factors in the production raw materials refer to sulfur content and residual carbon content.

3. The method for predicting the sulfur content in petroleum coke according to claim 1, characterized in that, In step ①, the input variables of the neural network model are the sulfur content and residual carbon content of the raw material, and the output variable is the sulfur content of the petroleum coke.

4. The method for predicting the sulfur content in petroleum coke according to claim 1, characterized in that, The operation of estimating the unknown parameters in step ⑤ is as follows: Extract the sulfur content and residual carbon content of the raw materials and the corresponding sulfur content of the products from the original database established in step 1), and use them as the input variables of step ① and the output variables of step ⑤, respectively. Then, use the particle swarm optimization algorithm to estimate the unknown parameters in step ⑤, minimizing the objective function during the estimation. : （5） In the formula, Indicates the first One raw material, This indicates the detected or experimental value of the sulfur content in coke. This indicates the predicted value of sulfur content in coke. Indicates the number of raw materials.

5. The method for predicting the sulfur content in petroleum coke according to claim 4, characterized in that, The particle swarm optimization algorithm is used to estimate the unknown parameters in step ⑤, and the estimation requires minimizing the objective function. The specific operation is as follows: A) Select a set of influencing factors from the original database of petroleum coke sulfur content and influencing factors in raw materials, and substitute them into the neural network model; at the same time, use the particle swarm optimization algorithm to randomly generate an initial particle swarm, and assign its values ​​to the unknown parameters in step ⑤, and then calculate according to formula (4) ; B) The result calculated in step A The actual measured sulfur content corresponding to this set of influencing factors stored in the original database. Substitute into formula (5) to calculate the objective function. ; C) Determine the objective function If the termination condition is met, define the set of data as the global optimum and assign its value to the unknown parameters in the neural network model; otherwise, proceed to step D. D) Update the particle swarm and repeat steps A-C until the global optimum is obtained and its value is assigned to the unknown parameters in the neural network model.

6. A machine-readable storage medium, characterized in that: The machine-readable storage medium stores executable code that, when executed by a processor of an electronic device, enables the processor to perform at least the method of any one of claims 1 to 5.

7. An electronic device for recording decision-making behavior, comprising a memory and a processor, wherein the memory stores executable programs, instructions, or code, characterized in that: When the executable program, instructions, or code is executed by a processor, the processor is enabled to perform the operation according to any one of claims 1 to 5.

8. An electronic device for recording decision-making behavior according to claim 7, characterized in that: The device includes a processor, a memory, external devices, and an I / O interface. The memory is used to store programs, instructions, or code executed by the processor. The processor is used to execute the method of any one of claims 1 to 5 according to the programs, instructions, or code stored in the memory. The I / O interface is used to receive or send data under the control of the processor. The external devices are used to input data to the processor and display data output by the processor.

9. A petroleum coke sulfur content prediction device, characterized in that: The sulfur content prediction device has a prediction module that uses the prediction method described in any one of claims 1-5 to predict the sulfur content.

Images