An ethylene caustic wash analysis method, system, and device based on a data model

Abstract

The present application relates to the technical field of ethylene caustic washing, and discloses an ethylene caustic washing analysis method, system and device based on a data model, the method comprising: configuring a data collection task, and obtaining process data and assay analysis data of ethylene caustic washing according to the data collection task; preprocessing the process data and assay analysis data to obtain a sample data set; constructing an ethylene caustic washing analysis model, and training the ethylene caustic washing analysis model using the sample data set to obtain a trained ethylene caustic washing prediction model; and using the ethylene caustic washing prediction model to perform real-time prediction analysis or simulation prediction analysis on the ethylene caustic washing process. The present application trains an ethylene caustic washing prediction model using a large amount of data by configuring a data collection task, thereby achieving the purpose of predicting key indicators of ethylene caustic washing through a data model, avoiding taking measures after problems occur, facilitating timely and accurate analysis of the ethylene caustic washing process, and making real-time optimization according to the analysis results.

Description

Technical Field

[0001] This invention relates to the field of ethylene alkali washing technology, and specifically to an ethylene alkali washing analysis method, system, and equipment based on a data model. Background Technology

[0002] Currently, the main technology for ethylene production is the tubular furnace cracking method. The cracked gas contains a certain amount of acidic impurities such as CO2 and H2S, as well as a small amount of organic sulfides. These acidic impurities can harm cryogenic separation and distillation. To remove acidic impurities from ethylene cracked gas, alkaline washing and refining processes are commonly used to remove sulfides, cycloalkanes, etc., to eliminate their pungent odor and reduce corrosiveness. To improve alkali utilization, most ethylene plants currently use multi-stage alkaline washing, with the Lummus three-stage alkaline washing process being a common one. To reduce alkali consumption, some companies have gradually upgraded the control process of the alkaline washing tower to the long-tail soda process. The content of substances produced by alkaline washing is related to the properties of the feedstock of the ethylene plant and the operating conditions of the alkaline washing process. Due to the differences in petroleum feedstocks and process routes among different plants, the composition and content of acidic gases in ethylene cracked gas vary greatly, and the composition of waste alkaline solution also varies significantly. Improper control of the alkaline washing process can also lead to excessive consumption of NaOH or excessive production of NaHCO3. Therefore, production enterprises need to improve the optimization and control level of the alkaline washing process, reduce the amount of waste alkali discharged, and lower the treatment cost, while ensuring that the pyrolysis gas meets the standards.

[0003] With the continuous development of machine learning technology, algorithms such as neural networks, vector machines, decision trees, and regression analysis have been widely applied in the field of nonlinear system modeling in the chemical industry. This makes it possible to analyze and predict the components of each section of the alkali washing tower using data models. By employing appropriate algorithm models and learning and analyzing production process data, machines can extract features and patterns, and make predictions and provide decision suggestions based on these features and patterns. Existing ethylene alkali washing control technologies are generally based on manual testing of component concentrations in the weakly alkaline section and waste alkali solution, changes in inlet acid gas content (if an online analysis table is available), changes in unit load, and CO2 at the top outlet of the alkali washing tower. 2 Online analysis tables or observation of the color of waste alkali solution are used to determine the extent of exceedance of major components. A comprehensive assessment is then made, and historical experience is used to control the amount of new alkali solution and makeup water, aiming to reduce waste of new alkali and excessive discharge of waste alkali. Current technical solutions rely on the analysis of component concentrations in each section of the alkali washing tower, which is a delayed control method. Furthermore, there are insufficient analytical personnel to continuously analyze the ionic composition of the alkali solution in each section. Moreover, it does not consider the real-time changes in temperature, pressure, and flow rate during ethylene cracking and alkali washing reactions, making it difficult to achieve optimal timeliness and accuracy in control. Summary of the Invention

[0004] In view of this, the present invention provides a data model-based method, system and equipment for analyzing the ethylene alkaline washing process, in order to solve the problem of the inability to analyze the ethylene alkaline washing process in a timely and accurate manner.

[0005] In a first aspect, the present invention provides an ethylene alkaline washing analysis method based on a data model, the method comprising:

[0006] Configure the data acquisition task and acquire the process data and laboratory analysis data of ethylene alkali washing according to the data acquisition task;

[0007] Preprocessing of process data and laboratory analysis data yields a sample dataset.

[0008] An ethylene alkali washing analysis model was constructed, and the model was trained using a sample dataset to obtain a trained ethylene alkali washing prediction model.

[0009] The ethylene alkaline washing prediction model is used to perform real-time prediction analysis or simulation prediction analysis of the ethylene alkaline washing process.

[0010] The ethylene alkaline washing analysis method provided by this invention, by configuring data acquisition tasks and using a large amount of data to train an ethylene alkaline washing prediction model, achieves the purpose of predicting key indicators of ethylene alkaline washing through data models. This avoids taking measures after the problem of excessive acid gas composition at the top outlet of the alkaline washing tower occurs, and facilitates timely and accurate analysis of the ethylene alkaline washing process, and makes optimizations in real time based on the analysis results.

[0011] In one alternative implementation, configuring the data acquisition task includes:

[0012] The index data is configured using multiple tag numbers in the ethylene alkali washing tower as identifiers. The format of the index data includes: tag number identifier, tag number group, tag number name, tag number unit, sampling point, analysis item, and data acquisition interface type.

[0013] The format of the configuration task parameters includes: collection interface type, collection indicator data, data collection start time, data collection end time, maximum data collection time offset, collection segment duration, data collection interval, and frequency of data collection task execution.

[0014] In one optional implementation, process data and laboratory analysis data for ethylene alkaline washing are acquired according to the data acquisition task, including:

[0015] The data acquisition task is executed cyclically in each bit according to the task parameters to obtain the raw data of each bit.

[0016] The raw data was formatted according to the format of the indicator data to obtain the process data and test analysis data for each digit.

[0017] The ethylene alkali washing analysis method based on a data model provided by this invention facilitates data acquisition from databases of different systems by configuring the format of index data and task parameters. Different acquisition cycles can be set for different reference numbers, and data acquisition can be performed in parallel between reference numbers, improving the efficiency and flexibility of data acquisition. The method also unifies the format of data from different data sources, making it easier to use and manage.

[0018] In one optional implementation, the process data and laboratory analysis data are preprocessed to obtain a sample dataset, including:

[0019] Perform anomaly detection on process data and laboratory analysis data to identify normal and abnormal data;

[0020] Remove abnormal data and use a preset data supplementation method to fill in the empty spaces after the abnormal data removal to obtain supplementary data;

[0021] The normal data and supplementary data are combined to form a sample dataset.

[0022] The ethylene alkaline washing analysis method based on a data model provided by this invention improves the data quality of the sample set by performing anomaly detection, removing abnormal data, and supplementing reasonable data, thus ensuring the accuracy of the sample data. The model is trained using accurate data, resulting in a model with better performance.

[0023] In one optional implementation, the sample dataset includes at least one process data and at least one laboratory analysis data. The process data includes: alkali replenishment volume. The ethylene alkali washing analysis model is trained using the sample dataset to obtain a trained ethylene alkali washing prediction model, including:

[0024] The sample dataset is converted into a two-dimensional data structure with multiple time-aligned metrics.

[0025] Select at least one laboratory analysis data or alkali replenishment amount as output, and the remaining laboratory analysis data and process data as input. Iterate and train the ethylene alkali washing analysis model until the number of iterations reaches the iteration threshold or the model accuracy reaches the preset accuracy. Then, end the training and use the ethylene alkali washing analysis model obtained from the last iteration as the ethylene alkali washing prediction model.

[0026] The ethylene alkali washing analysis method based on a data model provided by this invention converts a columnar sample dataset into a two-dimensional data structure with multiple indicators and time alignment, which facilitates inputting data according to the parameter structure determined by the model, and uses the sample dataset to iteratively train the model, thereby improving the prediction accuracy of the model.

[0027] In one optional implementation, the ethylene alkaline washing process is predicted and analyzed in real time using an ethylene alkaline washing prediction model, including:

[0028] Acquire at least one target production data and input the target production data into the ethylene alkaline washing prediction model to obtain at least one target prediction data. The target prediction data includes at least one real-time laboratory analysis data or real-time alkali replenishment volume. The target production data includes multiple real-time laboratory analysis data and multiple real-time process data in addition to the target prediction data.

[0029] The target prediction data is compared with the corresponding preset range. If the target prediction data is not within the corresponding preset range, a corresponding warning is issued.

[0030] The ethylene alkali washing analysis method based on a data model provided by this invention uses an ethylene alkali washing prediction model to obtain predicted real-time laboratory analysis data or real-time alkali replenishment volume, thereby determining the status of non-monitored output indicators under the current operating conditions and providing early warnings accordingly. This allows production personnel to make adjustments in advance based on the warning prompts, ensuring the stability of the ethylene alkali washing process.

[0031] In one optional implementation, the ethylene alkaline washing process is simulated and predicted using an ethylene alkaline washing prediction model, including:

[0032] At least one predictive optimization data is determined based on preset optimization conditions. Multiple sets of production data are randomly generated based on the predictive optimization data, and the production data are modified to obtain multiple sets of simulated production data. The predictive optimization data includes at least one laboratory analysis data or alkali replenishment amount. The production data includes multiple laboratory analysis data and multiple process data other than the predictive optimization data.

[0033] Input each simulated production data into the ethylene alkaline washing prediction model to obtain at least one simulated prediction optimization data.

[0034] Based on preset optimization conditions and at least one simulated prediction optimization data, determine the optimal production data that satisfies the preset optimization conditions.

[0035] The ethylene alkali washing analysis method based on data models provided by this invention determines the simulation process data based on production data, then uses the ethylene alkali washing prediction model to determine the simulation prediction optimization data, optimizes the ethylene alkali washing process data based on the simulation results, adjusts the alkali replenishment amount, reduces the amount of waste alkali discharged, and lowers the ethylene alkali washing treatment cost.

[0036] Secondly, this invention provides an ethylene alkaline washing analysis system based on a data model, the system comprising:

[0037] The data acquisition task configuration module is used to configure data acquisition tasks and acquire process data and laboratory analysis data of ethylene alkali washing according to the data acquisition tasks.

[0038] The sample data generation module is used to preprocess process data and laboratory analysis data to obtain a sample dataset;

[0039] The model training module is used to construct an ethylene alkali washing analysis model and train the ethylene alkali washing analysis model using a sample dataset to obtain a trained ethylene alkali washing prediction model.

[0040] The predictive analysis module is used to perform real-time predictive analysis or simulation predictive analysis of the ethylene alkaline washing process using an ethylene alkaline washing prediction model.

[0041] Thirdly, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the method described in the first aspect or any corresponding embodiment thereof.

[0042] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to perform the method described in the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0043] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0044] Figure 1 This is a schematic flowchart of an ethylene alkaline washing analysis method based on a data model according to an embodiment of the present invention;

[0045] Figure 2 This is a schematic diagram of the process of alkaline washing in the alkaline washing tower in the ethylene alkaline washing analysis method based on the data model according to an embodiment of the present invention.

[0046] Figure 3 This is a schematic flowchart of another data model-based ethylene alkaline washing analysis method according to an embodiment of the present invention;

[0047] Figure 4 This is a schematic diagram of the data display interface after unifying the format of the collected raw data in the ethylene alkaline washing analysis method based on the data model according to an embodiment of the present invention.

[0048] Figure 5 This is a structural block diagram of an ethylene alkaline washing analysis system based on a data model according to an embodiment of the present invention;

[0049] Figure 6 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] This invention provides a data model-based method for analyzing ethylene alkali washing. By using historical data to train a prediction model and then using the prediction model to predict the index data of ethylene alkali washing, the method can achieve timely and accurate analysis of the ethylene alkali washing process.

[0052] According to an embodiment of the present invention, an embodiment of an ethylene alkaline washing analysis method based on a data model is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0053] This embodiment provides a data model-based method for ethylene alkaline washing analysis, which can be used in the aforementioned computer system. Figure 1 This is a flowchart of an ethylene alkaline washing analysis method based on a data model according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps:

[0054] Step S101: Configure the data acquisition task and acquire the process data and laboratory analysis data of ethylene alkali washing according to the data acquisition task.

[0055] Specifically, ethylene production enterprises generally establish systems such as Distributed Control System (DCS), Laboratory Information Management System (LIMS), and real-time databases to collect and store process data and laboratory analysis data from production units (such as alkali washing towers). In this embodiment, the required data is collected by communicating with the above-mentioned systems.

[0056] Since the data formats and parameters of different systems are not exactly the same, before collecting data, it is necessary to perform unified configuration management of indicator data and collection scheduling tasks from different data sources, and unify the data structure of each data source to facilitate unified processing and use.

[0057] like Figure 2 The diagram shown is a typical schematic of the Lummus three-stage alkaline washing process. Relevant process data during ethylene alkaline washing include: feed flow rate, raw material sulfur content, sulfur injection amount, furnace tube temperature, pressure, ethylene cracking gas flow rate, cracking gas CO2 and H2S concentrations, quench tower bottom liquid pH, process water tower bottom liquid pH, cracking gas compressor discharge tank condensate pH, alkaline washing tower inlet gas temperature, quench water temperature, alkaline washing tower pressure, alkaline washing tower middle section temperature, alkaline washing tower outlet gas temperature, circulating water flow rate, circulating water temperature, alkaline replenishment amount, and new alkaline concentration and flow rate. Laboratory analysis data includes: product gas CO2 and H2S concentrations, and the NaOH, Na2CO3, Na2S, NaHCO3, and NaHS contents in the strong, medium, and weak alkaline solutions of the alkaline washing tower. This is only an example and is not a limitation.

[0058] Step S102: Preprocess the process data and laboratory analysis data to obtain the sample dataset.

[0059] Specifically, due to abnormal production conditions such as short-term failures or shutdowns during the production process, some abnormal data may appear in the process data or test analysis data. Data cleaning removes the abnormal data and retains only the normal data. However, after removing the abnormal data, the corresponding data bits become empty. It is necessary to fill in the empty data bits using linear interpolation or least squares method based on the normal data to obtain the sample dataset.

[0060] When the sample data is determined, the task format is as follows: unique identifier of tag number, maximum and minimum values ​​of normal data range, data verification method, and data supplementation method. This is only an example and is not limited to this.

[0061] Step S103: Construct an ethylene alkali washing analysis model and train the ethylene alkali washing analysis model using the sample dataset to obtain a trained ethylene alkali washing prediction model.

[0062] Specifically, an ethylene alkali washing analysis model that matches the actual production situation can be built through custom configuration. The model can select algorithms including regression analysis and neural networks, and time offsets can be set between input and output indicators to reflect the process time of the production process.

[0063] The regression analysis model configuration includes: model name, input indicators, output indicators, output indicator alarm values, analysis method, and input / output time interval. The analysis method can be selected as univariate linear regression, univariate nonlinear regression, multiple linear regression, or multiple nonlinear regression.

[0064] The neural network analysis model employs a multilayer perceptron (MLP) neural network utilizing the back propagation (BP) algorithm. As shown in Table 1, the configuration of the neural network analysis model includes: model name, input metric, output metric, output metric alarm value, output-input time interval, number of neurons in each hidden layer, activation function, and whether normalization is used. The activation function can be either SIGMOID or TANH, which are only examples and are not limited to these options.

[0065] Table 1

[0066]

[0067] After training the ethylene alkali washing prediction model, the initial calibration of the model is complete. In actual production, the trained ethylene alkali washing prediction model can be periodically calibrated using actual production data to ensure its good performance. The calibration process can be achieved by executing a model calibration task. The parameters of the model calibration task are shown in Table 2, which is only an example and not a limitation.

[0068] Table 2

[0069] Calibration Model Automatic / Manual Mode Number of iterations Target accuracy Initial learning rate Data range Neural network model Manual calibration 100000 0.0001 0.1 2022.1.1-2024.6.15 Neural network model Automatic calibration (30-day cycle) 1000 0.0001 0.0001 The last 30 days

[0070] Initial calibration is generally performed manually to train the ethylene alkaline washing prediction model until its error is within a preset range. The model is then used to predict new process data, and the prediction error must also meet the prediction requirements. Subsequent calibration can be performed automatically, with the model calibration task executed every thirty days. The model is updated using the data from the most recent thirty days to achieve model calibration.

[0071] Step S104: Use the ethylene alkaline washing prediction model to perform real-time prediction analysis or simulation prediction analysis on the ethylene alkaline washing process.

[0072] Specifically, in actual production, real-time laboratory analysis data and process data of ethylene alkali washing can be acquired, and target prediction data can be obtained using the ethylene alkali washing prediction model to achieve real-time predictive analysis. Alternatively, based on the actual process data collected during production, data can be modified, such as appropriately reducing the alkali replenishment amount. The ethylene alkali washing prediction model can then be used to predict the laboratory analysis data. If the prediction results meet the requirements, it indicates that the alkali replenishment amount can be appropriately reduced, thus guiding the optimization of ethylene alkali washing. This is just an example and is not a limitation.

[0073] The data model-based ethylene alkali washing analysis method provided in this embodiment configures data acquisition tasks and uses a large amount of data to train an ethylene alkali washing prediction model. This achieves the goal of predicting key indicators of ethylene alkali washing through the data model, avoiding the need to take measures after the problem of excessive acid gas composition at the top outlet of the alkali washing tower occurs. It facilitates timely and accurate analysis of the ethylene alkali washing process and allows for real-time optimization based on the analysis results.

[0074] This embodiment provides a data model-based method for ethylene alkaline washing analysis, which can be used in the aforementioned computer system. Figure 3 This is a flowchart of an ethylene alkaline washing analysis method based on a data model according to an embodiment of the present invention, such as... Figure 3 As shown, the process includes the following steps:

[0075] Step S201: Configure the data acquisition task and acquire the process data and laboratory analysis data of ethylene alkali washing according to the data acquisition task.

[0076] Specifically, step S201 includes:

[0077] Step S2011: Using multiple tag numbers in the ethylene alkali washing tower as identifiers for indicator data, configure the format of the indicator data as follows: tag number identifier, tag number group, tag number name, tag number unit, sampling point, analysis item, and data acquisition interface type.

[0078] Specifically, both process data and laboratory analysis data include multiple data sets. Each data set is configured with the same data format, and the tag identifier is unique, arranged sequentially according to the sampling order. The tag group indicates the data source; the tag name is the specific content of the data, such as the temperature of the middle section of the alkali washing tower in the process data (this is just an example and not a limitation); the tag unit is the physical or chemical unit corresponding to the tag name; the sampling point is the specific location corresponding to the tag; the analysis item is the indicator corresponding to the tag name; and the acquisition interface type is the interface type for connecting to different data source systems. Table 3 shows an example of indicator data acquired according to a unified format.

[0079] Table 3

[0080]

[0081] Step S2012, the format of the configured task parameters includes: acquisition interface type, acquisition indicator data, data acquisition start time, data acquisition end time, data acquisition maximum time offset, acquisition segment duration, data acquisition interval time, and data acquisition task execution frequency.

[0082] Specifically, since the data acquisition task is executed automatically, task parameters need to be configured. The acquired indicator data refers to the specific content of the data, such as the NaHS content in laboratory analysis data; this is just an example and not a limitation. For data acquisition from different data sources, the same or different acquisition cycles can be set, and data acquisition from different data sources can be performed in parallel, improving data acquisition efficiency and flexibility.

[0083] Step S2013: Perform the data acquisition task in a loop according to the task parameters to obtain the raw data of each bit.

[0084] Specifically, an automated data acquisition scheduling strategy is employed to execute production data acquisition tasks. These tasks can be scheduled periodically or performed cyclically; no restriction is imposed here. Once a data acquisition task is triggered, its configuration parameters are read, and the corresponding interface method is called based on the acquisition interface type. The tag identifier list is retrieved from the indicator data configuration table, and data is collected sequentially for each tag. Based on the start time, end time, maximum data time offset (milliseconds), shard duration (days), and data interval (seconds) in the configuration task parameters, the system requests the interface multiple times according to shards to obtain device monitoring data stored in the real-time database and laboratory analysis data stored in LIMS. After each acquisition, the last acquisition time for each indicator data is recorded. If the acquisition process is interrupted, the system can continue acquiring subsequent data in the next cycle. If data acquisition fails or is interrupted, the current round of tasks will be skipped, waiting for the next cycle's data acquisition task to execute before requesting interface data again. For large-scale data, multiple acquisition tasks can be set up according to tags, enabling concurrent multi-threaded acquisition.

[0085] Step S2014: Standardize the format of the raw data to obtain the process data and laboratory analysis data for each digit.

[0086] Specifically, the raw data is formatted uniformly. This format can be the same as the indicator data, or a separate uniform data format can be set according to actual needs. For example, the format might include: tag number, analysis item, data value (monitoring process data or laboratory analysis data corresponding to the tag number), collection time, and the normalized collection time in whole minutes. The normalization method is to format the collection time into whole minutes based on the data interval period. For example, 2024-03-01 04:29:38 is rounded down to 2024-03-01 04:30:00. Figure 4The image shown is a schematic diagram of the data display interface after the format is standardized.

[0087] The data, after being formatted uniformly, is saved to the original data table in the database through batch insertion. It is stored according to different partitions of the data source and a composite index of "unique identifier + collection time" is established, which is conducive to the rapid retrieval and query of massive amounts of data.

[0088] The ethylene alkali washing analysis method based on a data model provided in this embodiment facilitates data acquisition from databases of different systems by configuring the format of index data and task parameters. Different acquisition cycles can be set for different reference numbers, and data acquisition can be performed in parallel between reference numbers, improving the efficiency and flexibility of data acquisition. The data from different data sources is formatted uniformly, making it easy to use and manage.

[0089] Step S202: Preprocess the process data and laboratory analysis data to obtain the sample dataset.

[0090] Specifically, step S202 includes:

[0091] Step S2021: Perform anomaly checks on the process data and laboratory analysis data to identify normal and abnormal data.

[0092] Specifically, outlier data can be identified using methods such as data range tests, T-tests, or Grubbs tests.

[0093] (1) Data range test: The data test is performed by verifying whether all data values ​​of the tag number fall between the corresponding minimum and maximum values. Data values ​​that are less than the corresponding minimum or greater than the corresponding maximum are considered abnormal data. Table 4 shows an example of the tag number alarm range for the data range test. HH high-high alarm, PH high alarm, PL low alarm, and LL low-low alarm are only examples and are not limited to this.

[0094] Table 4

[0095] HH PH PL LL TIC2207 47 45 38 36 AI2203 6 2 0 0 PDI2205 42 35 18 0 TI213 46 44 37 0 TI212 47 45 0 0 FI2206 1725 1575 0 0 FR206 700 640 0 0 TI212 100 46 35 33

[0096] (2) The T test is used to compare whether there is a significant difference between the mean values ​​of two sets of data. Assuming that the process data and the test analysis data are sampled according to the preset sampling period, the mean values ​​of the two sets of data X and Y at different sampling times are μX and μY, the standard deviations are sX and sY, and the sample sizes are nX and nY, respectively. Then the formula for calculating the T value is: T=(μX-μY) / sqrt(sX^2 / nX+sY^2 / nT). A preset T threshold is set in the T test. If the calculated T value is greater than the preset T threshold, it is abnormal data.

[0097] (3) The Grubbs test is used to identify outliers in a set of data. Its calculation formula is: G = max|Xi - μ| / s, where Xi is a single data point, μ is the mean, and s is the standard deviation. The specific test method is a mature existing technology and will not be elaborated here.

[0098] Step S2022: Remove abnormal data and use a preset data supplementation method to fill in the empty spaces after removing abnormal data to obtain supplementary data.

[0099] Specifically, after removing abnormal data, the empty data bits need to be filled in, which can be done using linear interpolation or least squares method.

[0100] (1) Linear interpolation is used to supplement data by drawing a straight line between two points and selecting points on it.

[0101] The calculation formula is as follows:

[0102] Y=Y1+((X-X1)*(Y2-Y1)) / (X2-X1)

[0103] Where (X1,Y1) and (X2,Y2) are two known points, X is the x-value of the new point, and Y is the y-value of the new point.

[0104] (2) The least squares method finds the best function match for the data by minimizing the sum of squared errors. For linear regression, the formula is: y=ax+b, where a and b are the parameters to be found, which are solved by minimizing the following sum of squared errors: Σ[yi-(axi+b)]^2. The specific process is a mature existing technology and will not be elaborated here.

[0105] Step S2023: Combine the normal data and the supplementary data to form a sample dataset.

[0106] Specifically, a standardized data sampling process can ensure the quality of sample data, improve processing efficiency, supplement data to ensure data integrity, and provide more reliable data support for model training.

[0107] The ethylene alkaline washing analysis method based on a data model provided in this embodiment improves the data quality of the sample set by performing anomaly detection, removing abnormal data, and supplementing reasonable data, thus ensuring the accuracy of the sample data. Using accurate data for model training results in a model with better performance.

[0108] Step S203: Construct an ethylene alkali washing analysis model and train the ethylene alkali washing analysis model using the sample dataset to obtain a trained ethylene alkali washing prediction model.

[0109] Specifically, the sample dataset includes at least one process data and at least one laboratory analysis data. The process data includes: the amount of alkali replenishment. Step S203 above includes:

[0110] Step S2031: Convert the sample dataset into a two-dimensional data structure with multiple indicators aligned to time.

[0111] Specifically, for the sample data in the sample dataset, it is necessary to extract the set of sample data required for this training based on the time range of the model configuration and the input-output time deviation, and to normalize the indicators that need to be normalized. The data is compressed to the [-1,1] interval using the following formula to accelerate the convergence speed of model training:

[0112]

[0113] Where, x norm This represents the normalized indicator data, where x represents the indicator data before normalization. min x represents the minimum value within the range corresponding to the indicator. max This indicates the maximum value within the range corresponding to the indicator.

[0114] The columnar data stored according to the "indicator + time" dimension is converted into a two-dimensional data structure with multiple indicators and time alignment. This means that the time of all process data and laboratory analysis data is aligned, ensuring that the time of each group of sample data in the sample dataset is consistent, improving the accuracy and effectiveness of the sample data, and facilitating the input of data according to the parameter structure determined by the model.

[0115] Step S2032: Select at least one laboratory analysis data or alkali replenishment amount as output, and the remaining laboratory analysis data and process data as input. Iterate and train the ethylene alkali washing analysis model until the number of iterations reaches the iteration threshold or the model accuracy reaches the preset accuracy. Then, end the training and use the ethylene alkali washing analysis model obtained from the last iteration as the ethylene alkali washing prediction model.

[0116] Specifically, iterative training is performed using a pre-selected algorithm (regression algorithm, neural network algorithm). Once the required accuracy is achieved, the iteration stops immediately, and the trained model parameters are saved, resulting in a trained ethylene-alkali washing prediction model. If the required number of iterations is reached but the training accuracy is not achieved, the iteration also stops, and the ethylene-alkali washing analysis model obtained from the last iteration is used as the ethylene-alkali washing prediction model, and this is noted in the task log.

[0117] One or more (maximum of five: the concentrations of NaOH, Na2CO3, Na2S, NaHCO3, and NaHS in each section) of laboratory analysis data or alkali replenishment volume are used as dependent variables, and the remaining laboratory analysis data and process data are used as independent variables. The values ​​of the dependent variable with different values ​​of the independent variable are found. Based on experience or the actual ethylene alkali washing process, the optimal scheme of the dependent variable is selected, and the values ​​of the ethylene alkali washing process data and independent variables are determined according to the requirements.

[0118] The ethylene alkali washing analysis method based on a data model provided in this embodiment converts the columnar sample dataset into a two-dimensional data structure with multiple indicators aligned to time. This facilitates inputting data according to the parameter structure determined by the model, and uses the sample dataset to iteratively train the model, thereby improving the model's prediction accuracy.

[0119] Step S204: Use the ethylene alkaline washing prediction model to perform real-time prediction analysis or simulation prediction analysis on the ethylene alkaline washing process.

[0120] Specifically, step S204 above utilizes an ethylene alkaline washing prediction model to perform real-time prediction and analysis of the ethylene alkaline washing process, including:

[0121] Step S2041: Obtain at least one target production data and input the target production data into the ethylene alkali washing prediction model to obtain at least one target prediction data. The target prediction data includes at least one real-time laboratory analysis data or real-time alkali replenishment amount. The target production data includes multiple real-time laboratory analysis data and multiple real-time process data in addition to the target prediction data.

[0122] Specifically, in the actual ethylene alkaline washing process, it may be necessary to monitor the alkali replenishment volume or certain laboratory analysis data in real time. Therefore, at least one target prediction data can be determined based on production monitoring needs, and the real-time process data and real-time laboratory analysis data other than the target prediction data can be used as target production data. The target production data is input into the ethylene alkaline washing prediction model to obtain the real-time predicted value of the target prediction data, thereby realizing real-time monitoring of the target prediction data.

[0123] Step S2042: Compare the target prediction data with the corresponding preset range. If the target prediction data is not within its corresponding preset range, issue a corresponding warning.

[0124] Specifically, if the predicted test analysis data or process data exceeds the corresponding preset range, an alarm will be triggered. For example, the preset range for the NaOH content of the weak alkali solution is [1%, 3%]. When the NaOH content of the weak alkali solution is not within the range of [1%, 3%], an alarm will be triggered on the monitoring page. This is just an example, but it is not a limitation.

[0125] The ethylene alkali washing analysis method based on a data model provided in this embodiment uses an ethylene alkali washing prediction model to obtain predicted real-time laboratory analysis data or real-time alkali replenishment volume, thereby determining the status of non-monitored output indicators under the current operating conditions and issuing early warnings accordingly. This allows production personnel to make adjustments in advance based on the warning prompts, ensuring the stability of the ethylene alkali washing process.

[0126] In some optional embodiments, step S204 above utilizes an ethylene alkaline washing prediction model to simulate and predict the ethylene alkaline washing process, including:

[0127] Step S2043: Determine at least one predicted optimization data according to preset optimization conditions, randomly generate multiple sets of production data based on the predicted optimization data, and modify the production data to obtain multiple sets of simulated production data. The predicted optimization data includes at least one laboratory analysis data or alkali replenishment amount, and the production data includes multiple laboratory analysis data and multiple process data other than the predicted optimization data.

[0128] Specifically, it may be necessary to find the solution with the minimum alkali replenishment or the solution with the optimal laboratory analysis data. At least one predictive optimization data point needs to be determined based on the preset optimization conditions in the production objectives. Production data includes actual production process data and laboratory data, which can be historical process data and historical laboratory data. Multiple different values ​​for each production data point are randomly generated as multiple sets of production data. Based on the experience of technical personnel, one or more data points in the process data can be appropriately modified to obtain multiple sets of simulated production data.

[0129] Step S2044: Input each simulated production data into the ethylene alkali washing prediction model to obtain at least one simulated prediction optimization data.

[0130] Specifically, each simulated production data is input into the ethylene alkaline washing prediction model to obtain multiple sets of simulated prediction optimization data corresponding to the simulated production data.

[0131] Step S2045: Determine the optimal production data that satisfies the preset optimization conditions based on the preset optimization conditions and at least one simulated prediction optimization data.

[0132] Specifically, the advantages and disadvantages of the ethylene alkaline washing process are analyzed based on simulated laboratory analysis data, and the optimal production data that meets the preset optimization conditions is determined. For example, if the NaOH content of the alkaline solution in the three stages of a certain set of simulated laboratory analysis data is 8% lower than the data collected in actual production, it indicates that the simulated process data corresponding to this simulated laboratory analysis data is better than the current process data. The data can be updated based on the simulated production data. This is just an example, but not a limitation.

[0133] The ethylene alkali washing analysis method based on data models provided in this embodiment determines simulated production data based on laboratory analysis data and process data, and then uses an ethylene alkali washing prediction model to determine simulated prediction optimization data. Based on the simulation results, the production data of the ethylene alkali washing process is optimized, the amount of alkali replenishment is adjusted, the amount of waste alkali discharged is reduced, and the cost of ethylene alkali washing is lowered.

[0134] This embodiment also provides an ethylene alkali washing analysis system based on a data model, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the systems described in the following embodiments are preferably implemented in software, hardware implementations, or a combination of software and hardware, are also possible and contemplated.

[0135] This embodiment provides an ethylene alkaline washing analysis system based on a data model, such as... Figure 5 As shown, it includes:

[0136] The data acquisition task configuration module 501 is used to configure data acquisition tasks and acquire process data and laboratory analysis data of ethylene alkali washing according to the data acquisition tasks.

[0137] The sample data generation module 502 is used to preprocess the process data and laboratory analysis data to obtain a sample dataset.

[0138] The model training module 503 is used to construct an ethylene alkali washing analysis model and train the ethylene alkali washing analysis model using the sample dataset to obtain a trained ethylene alkali washing prediction model.

[0139] The predictive analysis module 504 is used to perform real-time predictive analysis or simulation predictive analysis of the ethylene alkaline washing process using the ethylene alkaline washing prediction model.

[0140] In some optional implementations, the data acquisition task configuration module 501 includes:

[0141] The indicator data format configuration unit is used to identify indicator data according to multiple tag numbers in the ethylene alkali washing tower. The format of the indicator data includes: tag number identifier, tag number group, tag number name, tag number unit, sampling point, analysis item, and acquisition interface type.

[0142] The task parameter format configuration unit is used to configure the format of task parameters, including: acquisition interface type, acquisition indicator data, data acquisition start time, data acquisition end time, maximum data acquisition time offset, acquisition segment duration, data acquisition interval time, and frequency of data acquisition task execution.

[0143] The data cyclic acquisition unit is used to cyclically execute the data acquisition task in each bit according to the task parameters to obtain the raw data of each bit.

[0144] The data format unification unit is used to unify the format of the raw data to obtain process data and laboratory analysis data for each digit.

[0145] In some alternative implementations, the sample data generation module 502 includes:

[0146] The data verification unit is used to perform anomaly checks on process data and laboratory analysis data to identify normal and abnormal data.

[0147] The data supplementation unit is used to remove abnormal data and fill the empty spaces after the abnormal data is removed using a preset data supplementation method to obtain supplementary data.

[0148] The data combination unit is used to combine normal data and supplementary data to form a sample dataset.

[0149] In some alternative implementations, the model training module 503 includes:

[0150] Time-aligned units are used to convert sample datasets into a two-dimensional data structure with multiple time-aligned metrics.

[0151] The model training unit is used to select at least one laboratory analysis data or alkali replenishment amount as output, and the remaining laboratory analysis data and process data as input to iteratively train the ethylene alkali washing analysis model until the number of iterations reaches the iteration threshold or the model accuracy reaches the preset accuracy, then the training ends, and the ethylene alkali washing analysis model obtained from the last iteration training is used as the ethylene alkali washing prediction model.

[0152] In some alternative implementations, the predictive analytics module 504 includes:

[0153] The real-time prediction unit is used to acquire at least one target production data and input the target production data into the ethylene alkali washing prediction model to obtain at least one target prediction data. The target prediction data includes at least one real-time laboratory analysis data or real-time alkali replenishment amount. The target production data includes multiple real-time laboratory analysis data and multiple real-time process data in addition to the target prediction data.

[0154] The real-time analysis unit compares the target prediction data with the corresponding preset range. If the target prediction data is not within its corresponding preset range, a corresponding warning is issued.

[0155] The data simulation unit is used to determine at least one predictive optimization data according to preset optimization conditions, randomly generate multiple sets of production data based on the predictive optimization data, and modify the production data to obtain multiple sets of simulated production data. The predictive optimization data includes at least one laboratory analysis data or alkali replenishment amount, and the production data includes multiple laboratory analysis data and multiple process data other than the predictive optimization data.

[0156] The simulation prediction unit is used to input various simulated production data into the ethylene alkaline washing prediction model to obtain at least one simulated prediction optimization data.

[0157] The optimization analysis unit is used to determine the optimal production data that meets the preset optimization conditions based on preset optimization conditions and at least one simulated prediction optimization data.

[0158] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0159] In this embodiment, the ethylene alkali washing analysis system based on the data model is presented in the form of functional units. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0160] This invention also provides a computer device having the above-described features. Figure 5 The ethylene alkaline washing analysis system shown is based on a data model.

[0161] Please see Figure 6 , Figure 6 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 6 As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 6 Take a processor 10 as an example.

[0162] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0163] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0164] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0165] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0166] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0167] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0168] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A data model-based method for ethylene alkaline washing analysis, characterized in that, The method includes: Configure a data acquisition task and acquire process data and analytical data for ethylene alkali washing according to the task. The process data includes: alkali replenishment amount, feed flow rate of each ethylene cracking furnace, raw material sulfur content, sulfur injection amount, furnace tube temperature, pressure, ethylene cracking gas flow rate, cracking gas CO2 and H2S concentration, pH value of quench water tower bottom liquid, pH value of process water tower bottom liquid, pH value of condensate discharged from cracking gas compressor tank, alkali washing tower inlet gas temperature, quench water temperature, alkali washing tower pressure, alkali washing tower middle section temperature, alkali washing tower outlet gas temperature, circulating water flow rate, circulating water temperature, and fresh alkali concentration and flow rate. The analytical data includes: product gas CO2 and H2S concentration, NaOH content, Na2CO3 content, Na2S content, NaHCO3 content, and NaHS content in the strong, medium, and weak alkali solutions of the alkali washing tower. The process data and laboratory analysis data are preprocessed to obtain a sample dataset. The product gas CO2 concentration is selected as the output, and the remaining laboratory analysis data, alkali replenishment amount, and key process data are used as inputs. Alternatively, the NaOH content, Na2CO3 content, Na2S content, NaHCO3 content, and NaHS content in the weak alkali section are selected as the output, and the remaining laboratory analysis data, alkali replenishment amount, and key process data are used as inputs. Or, the alkali replenishment amount is selected as the output, and the remaining laboratory analysis data and key process data are used as inputs. The key process data includes at least: the feed flow rate of each ethylene cracking furnace, the sulfur content of the raw material, the sulfur injection amount, the ethylene cracking gas flow rate, the CO2 and H2S concentrations of the cracking gas, the pH value of the bottom liquid of the quench water tower, the pH value of the bottom liquid of the process water tower, and the pH value of the condensate discharged from the cracking gas compressor tank. An ethylene alkali washing analysis model is constructed, and the ethylene alkali washing analysis model is trained using the sample dataset to obtain a trained ethylene alkali washing prediction model. The system acquires real-time laboratory analysis data and real-time process data of ethylene alkali washing, and uses the ethylene alkali washing prediction model to perform real-time prediction analysis of the ethylene alkali washing process; or, it modifies the actual process data and uses the ethylene alkali washing prediction model to perform simulation prediction analysis of the ethylene alkali washing process.

2. The method according to claim 1, characterized in that, The configuration data collection task includes: The index data is configured using multiple tag numbers in the ethylene alkali washing tower as identifiers. The format of the index data includes: tag number identifier, tag number group, tag number name, tag number unit, sampling point, analysis item, and data acquisition interface type. The format of the configuration task parameters includes: collection interface type, collection indicator data, data collection start time, data collection end time, maximum data collection time offset, collection segment duration, data collection interval, and frequency of data collection task execution.

3. The method according to claim 1, characterized in that, The process data and laboratory analysis data for ethylene alkaline washing obtained according to the data acquisition task include: The data acquisition task is executed cyclically in each bit according to the task parameters to obtain the raw data of each bit. The original data was formatted according to the format of the indicator data to obtain the process data and test analysis data for each digit.

4. The method according to claim 1 or 3, characterized in that, The process data and laboratory analysis data are preprocessed to obtain a sample dataset, including: Anomaly detection was performed on the process data and laboratory analysis data to identify normal and abnormal data. The abnormal data is removed, and the empty spaces after the abnormal data removal are filled with data using a preset data filling method to obtain supplementary data; The normal data and the supplementary data are combined to form a sample dataset.

5. The method according to claim 1, characterized in that, The step of training the ethylene alkali washing analysis model using the sample dataset to obtain a trained ethylene alkali washing prediction model includes: The sample dataset is converted into a two-dimensional data structure with multiple time-aligned metrics. The product gas CO2 concentration and / or NaOH content or alkali replenishment amount are selected as outputs, and the remaining test analysis data and process data are selected as inputs. The ethylene alkali washing analysis model is iteratively trained until the number of iterations reaches the iteration threshold or the model accuracy reaches the preset accuracy. The training ends, and the ethylene alkali washing analysis model obtained from the last iteration is used as the ethylene alkali washing prediction model.

6. The method according to claim 1 or 5, characterized in that, The real-time prediction and analysis of the ethylene alkaline washing process using the ethylene alkaline washing prediction model includes: At least one target production data is obtained and the target production data is input into the ethylene alkaline washing prediction model to obtain at least one target prediction data. The target prediction data includes at least one real-time laboratory analysis data or real-time alkali replenishment amount. The target production data includes multiple real-time laboratory analysis data and multiple real-time process data other than the target prediction data. The target prediction data is compared with the corresponding preset range. If the target prediction data is not within the corresponding preset range, a corresponding warning is issued.

7. The method according to claim 1 or 5, characterized in that, The ethylene alkaline washing prediction model is used to simulate and predict the ethylene alkaline washing process, including: At least one predictive optimization data is determined based on preset optimization conditions. Multiple sets of production data are randomly generated based on the predictive optimization data, and the production data are modified to obtain multiple sets of simulated production data. The predictive optimization data includes at least one laboratory analysis data or alkali replenishment amount. The production data includes multiple laboratory analysis data and multiple process data other than the predictive optimization data. Input each simulated production data into the ethylene alkaline washing prediction model to obtain at least one simulated prediction optimization data. Based on preset optimization conditions and at least one simulated prediction optimization data, the optimal production data that satisfies the preset optimization conditions is determined.

8. A data model-based ethylene alkaline washing analysis system, characterized in that, The system includes: The data acquisition task configuration module is used to configure data acquisition tasks and acquire process data and analytical data of ethylene alkali washing according to the data acquisition tasks. The process data includes: alkali replenishment amount, feed flow rate of each ethylene cracking furnace, raw material sulfur content, sulfur injection amount, furnace tube temperature, pressure, ethylene cracking gas flow rate, cracking gas CO2 and H2S concentration, pH value of quench water tower bottom liquid, pH value of process water tower bottom liquid, pH value of condensate discharged from cracking gas compressor tank, alkali washing tower inlet gas temperature, quench water temperature, alkali washing tower pressure, alkali washing tower middle section temperature, alkali washing tower outlet gas temperature, circulating water flow rate, circulating water temperature, and fresh alkali concentration and flow rate. The analytical data includes: product gas CO2 and H2S concentration, NaOH content, Na2CO3 content, Na2S content, NaHCO3 content, and NaHS content in the strong, medium, and weak alkali solutions of the alkali washing tower. The sample data generation module is used to preprocess the process data and laboratory analysis data to obtain a sample dataset. It selects the product gas CO2 concentration as the output and the remaining laboratory analysis data, alkali replenishment amount, and key process data as inputs. Alternatively, it selects the NaOH content, Na2CO3 content, Na2S content, NaHCO3 content, and NaHS content in the weak alkali section as the output and the remaining laboratory analysis data, alkali replenishment amount, and key process data as inputs. Or, it selects the alkali replenishment amount as the output and the remaining laboratory analysis data and key process data as inputs. The key process data includes at least: the feed flow rate of each ethylene cracking furnace, the raw material sulfur content, the sulfur injection amount, the ethylene cracking gas flow rate, the CO2 and H2S concentrations of the cracking gas, the pH value of the quench tower bottom liquid, the pH value of the process water tower bottom liquid, and the pH value of the condensate discharged from the cracking gas compressor tank. The model training module is used to construct an ethylene alkali washing analysis model and train the ethylene alkali washing analysis model using the sample dataset to obtain a trained ethylene alkali washing prediction model. The predictive analysis module is used to acquire real-time laboratory analysis data and real-time process data of ethylene alkali washing, and to perform real-time predictive analysis of the ethylene alkali washing process using the ethylene alkali washing prediction model, or to modify the actual process data and perform simulation predictive analysis of the ethylene alkali washing process using the ethylene alkali washing prediction model.

9. A computer device, characterized in that, include: A memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, the processor executing the computer instructions to perform the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the method of any one of claims 1 to 7.

Images