Method and device for predicting catalytic cracking activity value
By extracting the process parameters of catalytic cracking equipment and inputting the BP neural network model, the problem of low catalyst activity prediction accuracy in the existing technology is solved, accurate prediction of catalyst activity values and real-time adjustment of production processes are achieved, and production efficiency and economy are improved benefit.
Patent Information
- Application Number
- CN202310820058.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-05
- Publication Date
- 2025-01-07
AI Technical Summary
When the prior art predicts the activity value of catalytic cracking catalysts, the data processing is insufficient and the catalyst activity change pattern cannot be effectively captured, resulting in low prediction accuracy and affecting production efficiency and economic benefits.
By obtaining the process parameters of the catalytic cracking equipment for pre-treatment, variable parameters and characteristic parameters are extracted, such as unit addition amount of catalyst, isobutene molecular ratio, and main air volume used for unit feed, and input them into the BP neural network model for prediction, and optimize the model to Converge to historical activity data to achieve accurate prediction of catalyst activity values.
It greatly shortens the acquisition cycle of catalyst activity value, improves prediction accuracy, allows real-time adjustment of catalyst investment, reduces the use of fresh catalysts, saves resources, and realizes the long-term and stable operation of the catalytic cracking device.
Smart Images
Figure CN119274680A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of petrochemical industry, and in particular to a method and device for predicting catalytic cracking activity value, a balanced catalytic cracking system and a storage medium. Background Art
[0002] The catalytic cracking unit is the main product production unit for petroleum refining companies. It is the source of raw materials for the refining unit and has a significant impact on the entire refining enterprise. Its production efficiency also has an important impact on the company's economy.
[0003] Catalytic cracking catalysts play an important role in increasing the yield of target products and improving the rate of catalytic cracking reaction process. Catalysts change the reaction process, reduce the activation energy of the reaction, and thus increase the reaction rate, thereby promoting the reaction rate of the target product. Catalysts can play an important role in product yield and product quality. Catalytic cracking catalysts can shorten the catalytic cracking reaction cycle and thus increase the catalytic cracking reaction rate.
[0004] The catalyst of the catalytic cracking unit is different from that of other oil refining units. After the catalyst participates in the reaction in the reactor, it burns off the attached carbon deposits in the regenerator to complete regeneration, and then recycles in the reactor. It is necessary to continuously add fresh catalyst or unload the equilibrium catalyst, and monitor the reaction activity through laboratory tests to ensure the performance of the catalyst. The catalyst of other oil refining units is generally loaded into the reaction unit at one time, and when it reaches the end of its service life after a certain period of operation, it is removed and replaced at one time. Therefore, the change law of the reaction activity of the catalyst of the catalytic cracking unit is quite different from that of other units, and has certain particularities.
[0005] The model data used in the existing technology for predicting catalyst activity values using neural network models are mostly front-end data, such as: char load, regeneration temperature, reaction time, catalyst replacement rate, etc. However, the amount of front-end process parameter data in this method is large, the mining of key variable parameters cannot be properly handled, and the selection of data time periods cannot fully fit the changing laws of catalyst activity. If refinery staff can grasp the changes in the activity values of catalytic cracking catalysts in real time, they can adjust and replace catalysts or control catalyst activity indicators in a timely manner according to the different production requirements and target products. This is of great significance for improving the production efficiency and economic benefits of enterprises. Summary of the invention
[0006] The purpose of the embodiments of the present invention is to provide a method and device for predicting catalytic cracking activity value, a system for balanced catalytic cracking and a storage medium, wherein the method improves the prediction accuracy of catalytic cracking agent activity.
[0007] In order to achieve the above-mentioned purpose, an embodiment of the present invention provides a method for predicting catalytic cracking activity value, which method comprises: obtaining process parameters of the catalytic cracking equipment to be tested, preprocessing the process parameters to obtain variable parameters and characteristic parameters, wherein the characteristic parameters comprise: unit addition amount of catalyst, isobutylene molecular ratio, and main air volume used for unit feed; bringing the variable parameters and characteristic parameters into a prediction model to obtain the predicted activity value of the catalytic cracking equipment to be tested.
[0008] Optionally, the method further includes: training the prediction model, including: optimizing the prediction model according to historical catalytic cracking activity data and process parameters so that the final predicted activity value converges to the historical catalytic cracking activity data.
[0009] Optionally, the preprocessing is to remove drift and outliers from the process parameters in the same time dimension.
[0010] Optionally, the variable parameters include: reaction temperature, regenerator dense phase bottom temperature, and regenerator dense phase storage capacity.
[0011] Optionally, the unit addition amount of the catalyst at least includes the charred tank storage amount, the settler storage amount, the secret storage amount and the total storage amount; the isobutylene molecular ratio is: isobutylene content / (isobutane content+n-butane content).
[0012] Optionally, the prediction model is a BP neural network model.
[0013] Optionally, the prediction model includes: calculating hidden layer values by weighted summing of input layer nodes:
[0014] Activate the hidden layer values:
[0015] Get the predicted activity value:
[0016] Among them, k is the initial weight value,
[0017] Ff,j is the hidden layer value, j is the numerical sequence number of the hidden layer,
[0018] n is the input variable under various working conditions,
[0019] F(x) is the predicted activity value, b j is the weight value of the jth layer.
[0020] Optionally, the method further comprises: calculating an error value of the predicted activity value: t = (f(x) - fb) 2, and determine whether t is less than 0.0001. When t is less than 0.0001, the predicted activity value is the optimal predicted activity value; when t is not less than 0.0001, the two-layer weight values k and b are updated by reverse calculation, and the above operation is repeated with the updated weight values until t is less than 0.0001. The predicted activity value at convergence is the optimal predicted activity value:
[0021]
[0022]
[0023] Where t is the error value,
[0024] fb is the set catalyst activity value sample value,
[0025] X is the set learning rate,
[0026] f(x) is the predicted activity value.
[0027] On the other hand, the present invention provides a device for predicting catalytic cracking activity value, which includes: an acquisition module, used to obtain process parameters of the catalytic cracking equipment to be tested, and pre-process the process parameters to obtain variable parameters and characteristic parameters, wherein the characteristic parameters include: unit addition amount of catalyst, isobutylene molecular ratio, and main air volume used for unit feed; a processing module, used to bring the variable parameters and characteristic parameters into a prediction model to obtain the predicted activity value of the catalytic cracking equipment to be tested.
[0028] Optionally, the device further includes: training the prediction model, including: optimizing the prediction model according to historical catalytic cracking activity data and process parameters so that the final predicted activity value converges to the historical catalytic cracking activity data.
[0029] Optionally, the preprocessing is to remove drift and outliers from the process parameters in the same time dimension.
[0030] Optionally, the variable parameters include: reaction temperature, regenerator dense phase bottom temperature, and regenerator dense phase storage capacity.
[0031] Optionally, the unit addition amount of the catalyst at least includes the charred tank storage amount, the settler storage amount, the secret storage amount and the total storage amount; the isobutylene molecular ratio is: isobutylene content / (isobutane content+n-butane content).
[0032] Optionally, the prediction model is a BP neural network model.
[0033] On the other hand, the present invention provides a balanced catalytic cracking system, which includes the above-mentioned device for predicting catalytic cracking activity value and catalytic cracking equipment; the predicted activity value is obtained by the device for predicting catalytic cracking activity value; and the amount of catalytic cracking agent in the catalytic cracking equipment is set according to the predicted activity value.
[0034] On the other hand, the present invention provides a machine-readable storage medium storing machine instructions, which, when executed on a machine, enable the machine to execute the above-mentioned method for predicting catalytic cracking activity value.
[0035] The method for predicting the catalytic cracking activity value of the present invention comprises: obtaining the process parameters of the catalytic cracking equipment to be tested, pre-processing the process parameters to obtain variable parameters and characteristic parameters, wherein the characteristic parameters include: unit addition amount of catalyst, isobutylene molecular ratio, and main air volume used for unit feed; bringing the variable parameters and characteristic parameters into the prediction model to obtain the predicted activity value of the catalytic cracking equipment to be tested. The method realizes accurate prediction of catalyst activity by inputting the process parameters of the catalytic cracking equipment to be tested into the prediction model.
[0036] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:
[0038] Figure 1 1 is a schematic flow chart of a method for predicting catalytic cracking activity value according to Embodiment 1 of the present invention;
[0039] Figure 2 is a schematic diagram of a method for predicting catalytic cracking activity value according to Embodiment 3 of the present invention;
[0040] Figure 3 is a schematic diagram comparing the prediction accuracy of the second embodiment and the third embodiment of the present invention;
[0041] Figures 4a-4d This is a schematic diagram of the prediction accuracy of the model output of the second embodiment of the present invention;
[0042] Figure 5a-5d This is a schematic diagram of the prediction accuracy of the model output of the third embodiment of the present invention;
[0043] Figure 6 This is a schematic diagram of a device for predicting catalytic cracking activity value according to a fourth embodiment of the present invention.
[0044] Description of Reference Numerals
[0045] 601-Get module;
[0046] 602-Processing module. DETAILED DESCRIPTION
[0047] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.
[0048] Embodiment 1
[0049] Figure 1 : is a schematic flow diagram of a method for predicting catalytic cracking activity value according to Embodiment 1 of the present invention, such as Figure 1 As shown, step S101 is to obtain the process parameters of the catalytic cracking equipment to be tested. The catalytic cracking equipment is used for catalysis and / or cracking of petroleum refining process and is the source of raw materials for the refining device.
[0050] Step S102 is to pre-process the process parameters to obtain variable parameters and characteristic parameters, and the characteristic parameters include: unit addition amount of catalyst, isobutylene molecular ratio, and main air volume used for unit feed. According to a specific implementation method, the pre-processing is to remove drift and outliers from the process parameters in the same time dimension. The catalyst activity value and the above-treated process data are processed at the same latitude. The same latitude processing specifically refers to the unification of the time dimension of the data, that is, each set of data is data at the same time.
[0051] The specific steps of the pretreatment include:
[0052] a) According to the operating conditions of the catalytic cracking unit, determine the inspection and maintenance and shutdown time of the unit, and directly delete the relevant parameter data under abnormal conditions to ensure that each parameter data represents the normal operating conditions;
[0053] b) Cluster analysis is performed on the device process data, focusing on determining the device status classification based on variables such as device processing load, and establishing models based on different device states;
[0054] c) According to the process card and control requirements, combined with the process data under the actual working conditions of the device, the threshold value of each process parameter is calculated and determined, and the process parameters are uniformly standardized according to the determined value range, and the drift data caused by the instrument is cleaned up to ensure the business rationality of each parameter data;
[0055] d) Perform abnormal data mining analysis on all data, perform data correlation analysis and calculation on maximum and minimum values and related process parameters, determine whether each maximum and minimum value is abnormal data one by one, and retain extreme value data caused by normal operating conditions.
[0056] The variable parameters include: reaction temperature, regenerator C102 dense phase bottom temperature, regenerator dense phase storage. The catalyst unit addition amount includes at least the charred tank storage, settler storage, dense storage and total storage; the isobutylene molecular ratio is: isobutylene content / (isobutane content + normal butane content). Among them, the charred tank belongs to the regenerator of the catalytic cracking unit; the settler belongs to the reactor of the catalytic cracking unit; the dense storage refers to the weight of the catalyst in the charred tank; the total storage refers to the weight of all catalysts in the regeneration system.
[0057] Step S103 is to bring the variable parameters and characteristic parameters into the prediction model to obtain the predicted activity value of the catalytic cracking device to be tested. According to a specific implementation, the prediction model is a BP neural network model.
[0058] Specifically, the prediction model includes: calculating the hidden layer value by weighted summing the input layer nodes:
[0059] Activate the hidden layer values:
[0060] Get the predicted activity value:
[0061] Among them, k is the initial weight value, Ff,j is the hidden layer value, j is the hidden layer value sequence, n is the input variable under various working conditions, F(x) is the predicted activity value, b j is the weight value of the jth layer.
[0062] The method further comprises: calculating the error value of the predicted activity value: t = (f(x) - fb) 2 , and determine whether t is less than 0.0001. When t is less than 0.0001, the predicted activity value is the optimal predicted activity value; when t is not less than 0.0001, the two-layer weight values k and b are updated by reverse calculation, and the above operation is repeated with the updated weight values until t is less than 0.0001. The predicted activity value at convergence is the optimal predicted activity value:
[0063]
[0064]
[0065] Among them, t is the error value, fb is the set catalyst activity value sample value, X is the set learning rate, and f(x) is the predicted activity value.
[0066] The method for predicting the catalytic cracking activity value of the present invention also includes: training the prediction model, including: optimizing the prediction model according to historical catalytic cracking activity data and process parameters, so that the final predicted activity value converges to the historical catalytic cracking activity data. Specifically, the training process can be analyzed and adjusted by combining the historical data of the activity value of the catalyst actually measured with the corresponding predicted activity value obtained by the catalyst activity prediction model, and has an autonomous learning function, and enhances the reliability of the predicted activity value through forward and reverse propagation within the model. The autonomous learning function of the prediction model is a continuous optimization process, which is achieved by comparing the actual measured historical data of the activity value of the catalyst with the calculated predicted activity value, and combining the continuous optimization calculation of the input parameters to make the predicted activity value converge and continuously approach the actual measured catalyst activity value, thereby realizing the prediction training of the model.
[0067] After a large amount of data analysis and business analysis, the present invention found that the catalyst reserves need to be regularly unloaded in a certain amount according to the change of activity and production demand. Each unloading cycle is 10-15 days. The change law of the activity value in each unloading cycle has a strong correlation with the change law of the catalyst reserves. However, the correlation between the catalyst activity and the catalyst reserves is not obvious when looking at multiple unloading cycles. Therefore, the input variables of a catalytic cracking unit equilibrium catalyst micro-reaction intelligent prediction method and system are a number of time periods with a unloading cycle as the input variable. It has been verified that the prediction accuracy of catalyst activity prediction can be improved by using the data of the most recent complete unloading cycle as the input variable.
[0068] The catalyst of the catalytic cracking unit is different from that of other oil refining units. After the catalyst participates in the reaction in the reactor, it burns the attached carbon deposits in the regenerator to complete regeneration, and then recycles in the reactor. It is necessary to continuously add fresh catalyst or unload the equilibrium catalyst. The reaction activity can be monitored by laboratory tests to ensure the performance of the catalyst. The catalysts of other oil refining units are generally loaded into the reaction unit at one time, and then removed and replaced at one time when the service life reaches 3-5 years. Therefore, the change law of the reaction activity of the catalyst of the catalytic cracking unit is very different from that of other units, and has its own characteristics. Due to the particularity of the use of the catalyst of the catalytic cracking unit and the complexity of the use process, the prediction difficulty is relatively high.
[0069] The processed data is used as the input variable of the catalyst activity prediction model to train the model. The model training process can be analyzed and adjusted by combining the actual measured catalyst activity value history data with the corresponding predicted activity value obtained by the catalyst activity prediction model. It has an autonomous learning function and enhances the reliability of the predicted activity value through forward and backward propagation within the model. The autonomous learning function of the prediction model is a continuous optimization process. By comparing the actual measured catalyst activity value history data with the calculated predicted activity value, and combining the input parameters with continuous optimization calculation, the predicted activity value converges and continuously approaches the actual measured catalyst activity value, thereby realizing the prediction training of the model.
[0070] In summary, the present invention achieves at least the following technical effects:
[0071] 1) After a large amount of data analysis and business analysis, the present invention discovers for the first time the characteristic parameters of catalyst minute change, isobutylene molecular ratio, and main air volume used per unit feed, and incorporates the characteristic parameters into the BP neural network model to obtain the predicted activity value of the catalyst, which not only greatly shortens the acquisition cycle of the catalyst activity value, but also effectively improves the accuracy of catalyst activity prediction, and provides accurate catalyst micro-reaction activity prediction results for actual production on-site staff.
[0072] 2) After a large amount of data analysis and business analysis, the present invention found that the catalyst reserves need to be unloaded regularly at a certain dosage according to the change of activity and production demand. Each unloading cycle is 10-15 days. The change law of the activity value in each unloading cycle is highly correlated with the change law of the catalyst reserves. However, the correlation between the catalyst activity and the catalyst reserves over multiple unloading cycles is not obvious. Therefore, a method and system for intelligently predicting the micro-reaction activity of a balanced catalyst in a catalytic cracking unit uses an unloading cycle as an input variable and a set of time periods. It has been verified that using the data of the most recent complete unloading cycle as the input variable can improve the prediction accuracy of catalyst activity prediction.
[0073] 3) The present invention realizes timely monitoring of changes in catalyst activity in the internal environment of the reaction regeneration system of the catalytic cracking equipment during the production process without adding or replacing pollution control facilities, and can provide refinery workers with a buffer time to observe changes in catalyst activity values and regeneration reactions, so that refinery workers can better monitor the equilibrium catalyst micro-reaction activity during the catalytic cracking reaction, adjust the catalyst input in real time according to production requirements and production conditions, reduce unnecessary input of fresh catalyst, save resources, and achieve long-term stable operation of the catalytic cracking unit, which has important theoretical value and practical application value.
[0074] Embodiment 2
[0075] The catalytic cracking activity prediction method of this embodiment 1 comprises the following steps:
[0076] Step 1: Collect the process parameters of the catalytic cracking unit and perform data preprocessing. Specifically, the process parameter data can be the operating parameters and pollution production and discharge conditions of the catalytic cracking unit, including but not limited to the reaction regeneration inclined tube temperature, the regenerator C102 dense phase bottom temperature, the regenerator dense phase storage, the regeneration main air volume, the total processing volume, etc. The process parameter data includes historical product output data and real-time output data for subsequent model training and testing. This embodiment uses the catalytic cracking unit process parameter data to predict the activity of the catalyst through the algorithm model. Therefore, during the data collection process, the historical catalyst activity data in the catalytic cracking unit is also collected and recorded for subsequent algorithm model training.
[0077] Step 2. After a lot of data analysis and business analysis, it is found that the catalyst reserves need to be regularly unloaded in a certain amount according to the change of activity and production demand. Each unloading cycle is 10-15 days. The change law of the activity value in each unloading cycle is highly correlated with the change law of the catalyst reserves. However, the correlation between the catalyst activity and the catalyst reserves is not obvious when looking at multiple unloading cycles. Therefore, the input variables of a catalytic cracking unit equilibrium catalyst micro-reaction intelligent prediction method and system are a number of time periods with an unloading cycle as the input variable. It has been verified that the data of the most recent complete unloading cycle is used as the input variable to improve the prediction accuracy of catalyst activity prediction. It should be noted here that the catalyst of the catalytic cracking unit is different from other oil refining units. After the catalyst participates in the reaction in the reactor, it burns the attached carbon deposits in the regenerator to complete regeneration, and then it is recycled in the reactor, and fresh catalyst needs to be continuously added or the equilibrium catalyst is unloaded. The reaction activity can be monitored by laboratory tests to ensure the performance of the catalyst. Catalysts in other oil refining units are generally loaded into the reaction unit at one time, and then removed and replaced at one time when the life span reaches 3-5 years. Therefore, the change pattern of the catalyst activity of the catalytic cracking unit is very different from that of other units, and has its own particularity. Due to the particularity of the use of the catalyst in the catalytic cracking unit and the complexity of the use process, the prediction difficulty is relatively high.
[0078] Step three, use the processed data as the input variable of the catalyst activity prediction model to train the model. The model training process can be analyzed and adjusted by combining the actual measured catalyst activity value history data with the corresponding predicted activity value obtained by the catalyst activity prediction model. It has an autonomous learning function and enhances the reliability of the predicted activity value through forward and backward propagation within the model. The autonomous learning function of the prediction model is a continuous optimization process. By comparing the actual measured catalyst activity value history data with the calculated predicted activity value, and combining the input parameters with continuous optimization calculation, the predicted activity value converges and continuously approaches the actual measured catalyst activity value, thereby realizing the prediction training of the model.
[0079] Preferably, the catalyst activity prediction can adopt a neural network algorithm, and the algorithm model can adopt a BP neural network model. The construction process of the prediction model is as follows:
[0080] The first step is to set the initial weight values k, b.
[0081] The second step is to perform weighted summation on the input layer nodes of the model to calculate the hidden layer values. Specifically, the following formula (1) is used:
[0082]
[0083] Among them, is the hidden layer value, j is the numerical sequence number of the hidden layer, n is the input variable under various working conditions, that is, the corresponding value of each layer of nodes, and knj is the weight value corresponding to the node value n.
[0084] The third step is to execute the sigmoid activation formula (2) as follows:
[0085]
[0086] The fourth step is to combine formula (1) and formula (2) to calculate the predicted value of catalyst activity using formula (3):
[0087]
[0088] Among them, F(x) is the predicted value of catalyst activity of the catalytic cracking unit, and b is the second-layer weight value.
[0089] The fifth step is to calculate the error t = (f(x) - fb)2, which converges when the error approaches zero.
[0090] Where t is the error, fb is the set catalyst activity value sample value, and it converges when t<0.0001.
[0091] Step 6: If the error value does not meet the standard, reverse calculation is performed to update the weight values k and b of the two layers. The calculation formula (4) is as follows:
[0092]
[0093]
[0094] Among them, X is the set learning rate and f(x) is the predicted activity value.
[0095] In the seventh step, the calculations in the first to sixth steps are repeated using the updated weight values until convergence, thereby obtaining the best predicted value of the catalyst activity.
[0096] Step 4: Predict the catalyst activity in the catalytic cracking unit using the trained model.
[0097] By using the method for predicting the activity of the catalytic cracking catalyst of Example 1, the addition of the catalyst in the catalytic cracking unit and the determination of the activity value can avoid the defects of long working cycle and easy hysteresis caused by relying on manual inspection, and at the same time avoid the interference caused by the use characteristics and complexity of the catalyst in the catalytic cracking unit, and the data of the front end of the catalytic cracking reaction are not used to predict the catalyst activity.
[0098] This embodiment uses the process parameters within multiple completed unloading cycles as input variables and obtains the predicted activity value of the catalyst through an algorithm model. This not only greatly shortens the catalyst activity value acquisition cycle, but also effectively improves the accuracy of catalyst activity prediction, providing accurate catalyst micro-reaction activity prediction results for actual production on-site staff.
[0099] Embodiment 3
[0100] Figure 2 This is a schematic diagram of a method for predicting catalytic cracking activity value according to the third embodiment of the present invention. This embodiment is a more optimized embodiment based on the second embodiment. The following steps are specifically adopted:
[0101] Step 1: Collect the process parameter data of the catalytic cracking unit and perform data preprocessing. The data collection process is the same as that in step S101 in Example 2, and will not be repeated here. The data preprocessing after data collection in this embodiment specifically includes data extraction during normal operating conditions and abnormal data elimination. Among them, preferably but not restrictively, the data extraction during the normal operating period can be specifically as follows: retain the data during the normal operating period and delete the data in the shutdown state; the abnormal data elimination can be specifically as follows: according to a pre-set data threshold range, retain the collected process parameter data within the threshold range, and eliminate the data outside the threshold range. In addition, when collecting process parameter data, the actual measured value of the catalyst activity of the same period of the process parameter data is further collected and the data is processed in the same dimension. The data processing in the same dimension can facilitate calculation and comparison during the model training process.
[0102] Step 2: Based on a large amount of data analysis and business analysis, three characteristic parameters are obtained. The three characteristic parameters specifically include: the first characteristic parameter is the unit addition amount of catalyst, which is mainly the ratio of the change value of the charred tank storage, settler storage, and second-dense storage within 24 hours to the total storage, which is characterized by the factor of unit storage on catalyst activity consumption; the second characteristic parameter is the isobutylene molecular ratio, which is mainly the ratio of the isobutylene content to the sum of the isobutane and normal butane contents, which is characterized by the comprehensive influence of the catalyst activity on hydrogen transfer and isomerization reactions during the catalyst regeneration process; the third characteristic parameter is the main air volume used per unit feed, which is the ratio of the regeneration main air volume to the total feed volume, which is characterized by the oxygen used per unit feed. Compared with the regeneration main air volume and the oxygen content of the flue gas outlet at the top of the regenerator, this index parameter more accurately characterizes the oxygen used in the catalytic cracking reaction process.
[0103] Step 3: Use the three characteristic parameters in step 2 as input variables of the catalyst activity prediction model to train the model. The construction process and training process of the prediction model are the same as step 3 and will not be repeated here.
[0104] Step 4: Predict the catalyst activity in the catalytic cracking unit using the trained model.
[0105] Step 5: Control the amount of catalyst added and recycled based on the predicted results of the catalyst activity in step 4.
[0106] Through in-depth data analysis and business analysis, this embodiment finds that the complete data within multiple unloading cycles is used as the basic data, and based on this data, the characteristic parameters of the catalyst minute change, isobutylene molecular ratio, and main air volume used for unit feed discovered for the first time in this patent are used. The characteristic parameters are incorporated into the BP neural network model to obtain the predicted activity value of the catalyst. This not only greatly shortens the acquisition cycle of the catalyst activity value, but also can effectively improve the accuracy of the catalyst activity prediction, and provide accurate catalyst micro-reaction activity prediction results for actual production on-site staff.
[0107] The specific implementation contents include:
[0108] 1. Data collection and processing: The communication channel collects 400,000 pieces of raw data from March 4, 2018 to April 1, 2022 for professional analysis, and retains three product output parameters, including: reaction regeneration inclined tube temperature, regenerator C102 dense phase bottom temperature, regenerator dense phase storage and three characteristic parameters: catalyst minute change, isobutylene molecular ratio, main air volume used per unit feed, and two catalyst activity LIMS data at the beginning and end of the unloading cycle. Among the above process parameters, the reaction regeneration inclined tube temperature and the regenerator C102 dense phase bottom temperature can characterize the effect of the reaction regeneration process on the catalyst activity, and the regenerator dense phase storage is characterized by the effect of the change in storage during the catalyst system unloading cycle on the catalyst activity.
[0109] Data is deleted according to the respective value ranges of the process parameters imported into the neural network model, and the values within the value range of the process parameters are retained. A total of 1,425 data containing catalyst activity values are screened out from the processed data according to the time of obtaining the catalyst activity detection values (8 o'clock on the detection day), and the average product output data six hours before and after the detection time point is calculated. The catalyst activity value and the process data are processed at the same latitude, and a total of 200 data and 13 unloading cycles from March 10, 2021 to September 21, 2021 are selected as model training data input into the model for training and debugging the model accuracy. A total of 96 data from 7 unloading cycles from December 25, 2021 to March 30, 2022 are selected as input variables and input into the model to output the predicted value of catalyst activity, and the LIMS data of catalyst activity values from December 25, 2021 to March 30, 2022 are used as the result to compare the accuracy of the prediction model.
[0110] The model is set as Comparative Example 1-14 model, and Example 2 adopts Model 1-7. The input variables used include the reaction regeneration inclined tube temperature, the bottom temperature of the dense phase of regenerator C102, the dense phase storage of the regenerator, and two catalyst activity LIMS data at the beginning and end of the unloading cycle. A total of 96 data from 7 unloading cycles from December 25, 2021 to March 30, 2022 were used as input variables to test the model accuracy, and a total of 200 data and 13 unloading cycles from March 10, 2021 to September 21, 2021 were selected as model training data input into the model for training and debugging the model accuracy. Example 3 uses the input variables used in models 8-14, including the reaction regeneration inclined tube temperature, the bottom temperature of the dense phase of regenerator C102, the dense phase storage of the regenerator, and three characteristic parameters: the minute change in catalyst, the isobutylene molecular ratio, the main air volume used per unit feed, and two catalyst activity LIMS data at the beginning and end of the unloading cycle. A total of 96 data from 7 unloading cycles from December 25, 2021 to March 30, 2022 are used as input variables, and a total of 200 data and 13 unloading cycles from March 10, 2021 to September 21, 2021 are selected as model training data input into the model for training and debugging of model accuracy.
[0111] 2. Establishment of BP neural network structure: Comparative Examples 1-7 construct a 3-1-1 structure neural network model, i.e., a 3-input 1-output model, and Comparative Examples 8-14 construct a 6-1-1 structure neural network model, i.e., a 6-input 1-output model;
[0112] 3. BP neural network training: A total of 200 data points and 13 unloading cycles from March 10, 2021 to September 21, 2021 were used as model training data input into the model to gradually adjust the initial weights of the neural network links to improve the accuracy of the neural network to the best;
[0113] 4. BP neural network prediction output: A total of 96 data from 7 unloading cycles from December 25, 2021 to March 30, 2022 are included in the trained micro-reaction activity prediction model prediction module to output the equilibrium catalyst micro-reaction activity prediction value.
[0114] 5. Conduct accuracy test on the model prediction results.
[0115] Figure 3 is a schematic diagram comparing the prediction accuracy of the second embodiment and the third embodiment of the present invention, Figures 4a-4d This is a schematic diagram of the prediction accuracy of the model output of the second embodiment of the present invention. Figure 5a-5d Schematic diagram of the prediction accuracy of the model output of the second embodiment of the present invention. Figure 3 , Figures 4a-4d and Figure 5a-5d It can be seen that the neural network model has high prediction accuracy when the characteristic parameter 1, the minute change of the catalyst, the characteristic parameter 2, the isobutylene molecules and the characteristic parameter 3, the ratio of the main air volume used for the unit feed are not used. When the 7 groups of models in Example 2 do not adopt the three characteristic parameters, the highest prediction accuracy is 86.99%, the lowest is 84.27%, and the average is 85.72%. When the 7 groups of models in Example 3 adopt 3 characteristic parameters, the highest prediction accuracy is 97.29%, the lowest prediction accuracy is 96.69%, and the average is 97.07%.
[0116] In summary, the artificial neural network model of the characteristic parameters of catalyst minute change, isobutylene molecules and main air volume used per unit feed proposed in the present invention has higher prediction accuracy when predicting the equilibrium catalyst micro-reaction activity.
[0117] Embodiment 4
[0118] Figure 6 Schematic diagram of a device for predicting catalytic cracking activity value according to a fourth embodiment of the present invention. Figure 6As shown, the device for predicting the catalytic cracking activity value includes: an acquisition module 601, which is used to obtain the process parameters of the catalytic cracking equipment to be tested, and pre-process the process parameters to obtain variable parameters and characteristic parameters, wherein the characteristic parameters include: unit addition amount of catalyst, isobutylene molecular ratio, and main air volume used for unit feed; a processing module 602, which is used to bring the variable parameters and characteristic parameters into the prediction model to obtain the predicted activity value of the catalytic cracking equipment to be tested. The device also includes: training the prediction model, including: optimizing the prediction model according to historical catalytic cracking activity data and process parameters, so that the final predicted activity value converges to the historical catalytic cracking activity data. The pre-processing is to remove drift and outlier processing of the process parameters in the same time dimension. The variable parameters include: reaction temperature, regenerator dense phase bottom temperature, regenerator dense phase storage. The unit addition amount of catalyst includes at least the char tank storage, settler storage, dense storage and total storage; the isobutylene molecular ratio is: isobutylene content / (isobutane content+n-butane content). The prediction model is a BP neural network model.
[0119] Through in-depth data analysis and business analysis, this embodiment finds that the complete data within multiple unloading cycles is used as the basic data, and based on this data, the characteristic parameters of the catalyst minute change, isobutylene molecular ratio, and main air volume used for unit feed discovered for the first time in this patent are used. The characteristic parameters are incorporated into the BP neural network model to obtain the predicted activity value of the catalyst. This not only greatly shortens the acquisition cycle of the catalyst activity value, but also can effectively improve the accuracy of the catalyst activity prediction, and provide accurate catalyst micro-reaction activity prediction results for actual production on-site staff.
[0120] Embodiment 5
[0121] This embodiment provides a balanced catalytic cracking system, which includes the above-mentioned device for predicting catalytic cracking activity value and catalytic cracking equipment; the predicted activity value is obtained by the device for predicting catalytic cracking activity value; and the amount of catalytic cracking agent in the catalytic cracking equipment is set according to the predicted activity value.
[0122] In summary, this embodiment greatly shortens the acquisition cycle of the catalyst activity value, and can also effectively improve the accuracy of catalyst activity prediction, providing accurate catalyst micro-reaction activity prediction results for on-site staff in actual production.
[0123] The method for predicting the catalytic cracking activity value of the present invention comprises: obtaining the process parameters of the catalytic cracking equipment to be tested, pre-processing the process parameters to obtain variable parameters and characteristic parameters, wherein the characteristic parameters include: unit addition amount of catalyst, isobutylene molecular ratio, and main air volume used for unit feed; bringing the variable parameters and characteristic parameters into the prediction model to obtain the predicted activity value of the catalytic cracking equipment to be tested. The method realizes accurate prediction of catalyst activity by inputting the process parameters of the catalytic cracking equipment to be tested into the prediction model.
[0124] After a large amount of data analysis and business analysis, the present invention discovers for the first time the characteristic parameters of catalyst minute change, isobutylene molecular ratio, and main air volume used per unit feed, and incorporates the characteristic parameters into a BP neural network model to obtain the predicted activity value of the catalyst, which not only greatly shortens the acquisition cycle of the catalyst activity value, but also effectively improves the accuracy of catalyst activity prediction, and provides accurate catalyst micro-reaction activity prediction results for on-site staff in actual production.
[0125] After a large amount of data analysis and business analysis, the present invention finds that the catalyst reserves need to be regularly unloaded at a certain dosage according to the change of activity and production demand, and each unloading cycle is 10-15 days. The change law of the activity value in each unloading cycle is highly correlated with the change law of the catalyst reserves. However, the correlation between the catalyst activity and the catalyst reserves over multiple unloading cycles is not obvious. Therefore, a method and system for intelligently predicting the micro-reaction activity of a balanced catalyst in a catalytic cracking unit uses a unloading cycle as a data set of time periods as input variables. It has been verified that using the data of the most recent complete unloading cycle as input variables can improve the prediction accuracy of catalyst activity prediction.
[0126] The present invention enables timely monitoring of changes in catalyst activity in the internal environment of a reaction regeneration system of a catalytic cracking device during production without adding or replacing pollution control facilities, and can provide refinery workers with a buffer time for observing changes in catalyst activity values and regeneration reactions, so that refinery workers can better monitor the equilibrium catalyst micro-reaction activity during the catalytic cracking reaction, adjust the catalyst input in real time according to production requirements and production conditions, reduce unnecessary input of fresh catalysts, save resources, and achieve long-term stable operation of the catalytic cracking device, which has important theoretical value and practical application value.
[0127] On the other hand, an embodiment of the present invention provides a storage medium having a program stored thereon, which implements a method for predicting a catalytic cracking activity value when the program is executed by a processor.
[0128] An embodiment of the present invention provides a processor, which is used to run a program, wherein the program executes a method for predicting a catalytic cracking activity value when running.
[0129] The embodiment of the present invention provides a device, the device includes a processor, a memory and a program stored in the memory and can be run on the processor, and the processor implements the following steps when executing the program: obtaining process parameters of the catalytic cracking device to be tested, pre-processing the process parameters to obtain variable parameters and characteristic parameters, the characteristic parameters include: unit addition amount of catalyst, isobutylene molecular ratio, and main air volume used for unit feed; bringing the variable parameters and characteristic parameters into the prediction model to obtain the predicted activity value of the catalytic cracking device to be tested. The method also includes: training the prediction model, including: optimizing the prediction model according to historical catalytic cracking activity data and process parameters, so that the final predicted activity value converges to the historical catalytic cracking activity data. The pre-processing is to remove drift and outlier processing of the process parameters in the same time dimension. The variable parameters include: reaction temperature, regenerator dense phase bottom temperature, regenerator dense phase storage. The catalyst unit addition includes at least the char tank storage, settler storage, dense storage and total storage; the isobutylene molecular ratio is: isobutylene content / (isobutane content+n-butane content). The prediction model is a BP neural network model. The device in this article can be a server, PC, PAD, mobile phone, etc.
[0130] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes the following method steps: obtaining process parameters of the catalytic cracking device to be tested, preprocessing the process parameters to obtain variable parameters and characteristic parameters, the characteristic parameters including: unit addition amount of catalyst, isobutylene molecular ratio, and main air volume used per unit feed; bringing the variable parameters and characteristic parameters into a prediction model to obtain a predicted activity value of the catalytic cracking device to be tested.
[0131] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.
[0132] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0133] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0134] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0135] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0136] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.
[0137] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.
[0138] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0139] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.
Claims
1. A method for predicting catalytic cracking activity value, characterized in that: The method includes: Acquire process parameters of the catalytic cracking equipment to be tested, and pre-process the process parameters to obtain variable parameters and characteristic parameters, wherein the characteristic parameters include: unit addition amount of catalyst, isobutylene molecular ratio, and main air volume used for unit feed; The variable parameters and characteristic parameters are introduced into the prediction model to obtain the predicted activity value of the catalytic cracking device to be tested.
2. The method according to claim 1, characterized in that The method further includes: training the prediction model, including: The prediction model is optimized according to historical catalytic cracking activity data and process parameters so that the final predicted activity value converges to the historical catalytic cracking activity data.
3. The method according to claim 1, characterized in that The preprocessing is to remove drift and outliers from the process parameters in the same time dimension.
4. The method according to claim 1, characterized in that: The variable parameters include: reaction temperature, regenerator dense phase bottom temperature, and regenerator dense phase storage capacity.
5. The method according to claim 1, characterized in that The catalyst unit addition amount at least includes the charred tank storage amount, the settler storage amount, the secret storage amount and the total storage amount; The isobutylene molecular ratio is: isobutylene content / (isobutane content+n-butane content).
6. The method according to claim 1, characterized in that The prediction model is a BP neural network model.
7. The method according to claim 1, characterized in that The prediction model includes: The hidden layer value is calculated by weighted summing of the input layer nodes: Activate the hidden layer values: Get the predicted activity value: Among them, k is the initial weight value, Ff,j is the hidden layer value, j is the numerical sequence number of the hidden layer, n is the input variable under various working conditions, F(x) is the predicted activity value, b j is the weight value of the jth layer.
8. The method according to claim 7, characterized in that This method also: Calculate the error value of the predicted activity value: t = (f(x) - fb) 2 , and judging whether t is less than 0.0001, when t is less than 0.0001, taking the predicted activity value as the optimal predicted activity value; When t is not less than 0.0001, the weight values k and b of the two layers are updated by reverse calculation, and the above operation is repeated with the updated weight values until t is less than 0.0001. The predicted activity value at the time of convergence is the optimal predicted activity value: Where f(x) is the predicted activity value, t is the error value, fb is the set catalyst activity value sample value, X is the set learning rate.
9. A device for predicting catalytic cracking activity value, characterized in that: The device includes: An acquisition module is used to acquire process parameters of the catalytic cracking equipment to be tested, and preprocess the process parameters to obtain variable parameters and characteristic parameters, wherein the characteristic parameters include: unit addition amount of catalyst, isobutylene molecular ratio, and main air volume used for unit feed; The processing module is used to bring the variable parameters and characteristic parameters into the prediction model to obtain the predicted activity value of the catalytic cracking device to be tested.
10. The device according to claim 9, characterized in that The device also includes: training the prediction model, including: The prediction model is optimized according to historical catalytic cracking activity data and process parameters so that the final predicted activity value converges to the historical catalytic cracking activity data.
11. The device according to claim 9, characterized in that The preprocessing is to remove drift and outliers from the process parameters in the same time dimension.
12. The device according to claim 9, characterized in that The variable parameters include: reaction temperature, regenerator dense phase bottom temperature, and regenerator dense phase storage capacity.
13. The device according to claim 9, characterized in that The catalyst unit addition amount at least includes the charred tank storage amount, the settler storage amount, the secret storage amount and the total storage amount; The isobutylene molecular ratio is: isobutylene content / (isobutane content+n-butane content).
14. The device according to claim 9, characterized in that The prediction model is a BP neural network model.
15. A system for balanced catalytic cracking, characterized in that: The system comprises the device for predicting catalytic cracking activity value and catalytic cracking equipment as claimed in any one of claims 9 to 14; Obtaining a predicted activity value by means of the device for predicting catalytic cracking activity value; The amount of catalytic cracking agent in the catalytic cracking device is set according to the predicted activity value.
16. A machine-readable storage medium, characterized in that: Machine instructions are stored, and when the machine instructions are executed on a machine, the machine executes the method for predicting the catalytic cracking activity value as claimed in any one of claims 1 to 8.