Multi-model prediction method and apparatus for thermal efficiency of ethylene cracking furnace

Abstract

This invention provides a multi-model prediction method, apparatus, and storage medium for the thermal efficiency of an ethylene cracking furnace. The multi-model prediction method includes the following steps: acquiring production process data of the ethylene cracking furnace, wherein the production process data includes feed rate data and thermal correlation data; matching the feed rate data with the operating condition ranges adapted to multiple prediction models in a multi-model database, wherein the multiple prediction models and their adapted operating condition ranges are determined through training based on historical production process data of the ethylene cracking furnace; and responding to a successful match between the feed rate data and the operating condition range adapted to any one or more of the prediction models, invoking one or more corresponding prediction models to predict the thermal efficiency of the ethylene cracking furnace within the corresponding operating condition range based on the thermal correlation data.

Description

Technical Field

[0001] This invention belongs to the field of combustion optimization technology, specifically relating to a multi-model prediction method for the thermal efficiency of an ethylene cracking furnace, a multi-model prediction device for the thermal efficiency of an ethylene cracking furnace, and a computer-readable storage medium. Background Technology

[0002] With continuous economic development, my country's GDP has been steadily increasing. However, while industry has grown rapidly, energy issues have become increasingly pressing. How to effectively utilize energy and improve thermal efficiency has become a top concern. Furthermore, environmental protection is also a significant concern. The burning of fossil fuels leads to an increase in pollutants such as nitrogen oxides (NOx), which severely pollute the atmosphere and are a major contributor to harmful pollutants like ozone. In addition, NOx reacts with water vapor in the air to form acid rain, causing serious harm to the natural ecosystem and human health.

[0003] To improve economic efficiency and reduce environmental pollution during production, it is necessary to optimize various industrial combustion processes. Ethylene plants, as leading units in the petrochemical industry and major energy consumers, require optimization of their combustion processes for significant economic benefits and environmental protection. Furthermore, in the entire ethylene cracking process, the cracking furnace accounts for approximately 60% of the total energy consumption of the ethylene plant, while fuel gas consumption accounts for approximately 70% of the total energy consumption of the cracking furnace. Therefore, combustion optimization of the ethylene cracking process can be achieved by predicting the thermal efficiency of the ethylene cracking furnace. However, in the ethylene cracking process, liquefied petroleum gas, hydrotreated tail oil, and naphtha, among other raw materials, are first cracked at high temperatures in the ethylene cracking furnace into cracked gas composed of small molecules such as ethylene, methane, propylene, and hydrogen. This cracked gas is then processed into ethylene, propylene, and various byproducts through compression, separation, and other process units. The cracking process is a complex flow, heat transfer, and mass transfer process, which is generally difficult to model mechanistically.

[0004] In order to overcome the above-mentioned defects in the existing technology, there is an urgent need in the field for a thermal efficiency prediction technology to quickly and accurately predict the thermal efficiency of ethylene cracking furnaces and lay the foundation for subsequent combustion optimization. Summary of the Invention

[0005] The following provides a brief overview of one or more aspects to offer a basic understanding of them. This overview is not an exhaustive summary of all conceived aspects, nor is it intended to identify key or decisive elements of all aspects, nor to define the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to the more detailed descriptions that follow.

[0006] To overcome the aforementioned deficiencies in the existing technology, this invention provides a multi-model prediction method for the thermal efficiency of an ethylene cracking furnace, a multi-model prediction device for the thermal efficiency of an ethylene cracking furnace, and a computer-readable storage medium, which can quickly and accurately predict the thermal efficiency of an ethylene cracking furnace and lay the foundation for subsequent combustion optimization.

[0007] Specifically, the multi-model prediction method for the thermal efficiency of an ethylene cracking furnace according to the first aspect of the present invention includes the following steps: acquiring production process data of the ethylene cracking furnace, wherein the production process data includes feed rate data and thermal correlation data; matching the feed rate data with the operating condition ranges adapted to multiple prediction models in a multi-model database, wherein the multiple prediction models and their adapted operating condition ranges are determined by training based on historical production process data of the ethylene cracking furnace; and in response to the successful matching of the feed rate data with the operating condition ranges adapted to any one or more of the prediction models, calling one or more corresponding prediction models to predict the thermal efficiency of the ethylene cracking furnace in the corresponding operating condition range based on the thermal correlation data.

[0008] Further, in some embodiments of the present invention, the step of determining the multiple prediction models in the multi-model database and their corresponding operating condition ranges includes: acquiring multiple sets of historical production process data of the ethylene cracking furnace, wherein the historical production process data includes feed rate data and thermal correlation data; designing multiple classification schemes for classifying the operating condition ranges of the multiple sets of historical production process data based on the feed rate as the standard; calculating the predicted thermal efficiency values ​​of each of the historical thermal correlation data with respect to each corresponding operating condition range in each classification scheme using multiple predetermined prediction models based on multiple thermal efficiency levels determined by pre-cluster analysis; calculating the total mean square error of each classification scheme according to the actual thermal efficiency value and the predicted thermal efficiency value corresponding to each of the historical thermal correlation data, and determining the optimal classification scheme with the smallest total mean square error; and storing the multiple operating condition ranges divided by the optimal classification scheme and their corresponding prediction models into the multi-model database.

[0009] Furthermore, in some embodiments of the present invention, before designing multiple classification schemes for classifying the operating condition ranges of the multiple sets of historical production process data based on the historical feed data, the multi-model prediction method further includes the following steps: preprocessing the acquired multiple sets of historical production process data to filter out abnormal operating conditions and / or outliers with non-steady-state distributions.

[0010] Furthermore, in some embodiments of the present invention, the step of designing multiple classification schemes for classifying the working condition ranges of the multiple sets of historical production process data based on the feed rate includes: extracting the feed rate data from the multiple sets of historical production process data respectively to determine the total range [a, b] involved in the multiple sets of historical production process data; determining the lower limit of the range based on the minimum feed rate a in the multiple sets of historical production process data, determining the upper limit of the range based on the maximum feed rate b, and constructing multiple working condition ranges of a classification scheme with a positive integer multiple i of a preset step amount Δ as the span; and traversing all positive integer multiples i of the number of working condition ranges greater than or equal to 3 to obtain i classification schemes for the total range [a, b].

[0011] Furthermore, in some embodiments of the present invention, the step of determining the lower limit of the range based on the minimum feed amount a in the multiple sets of historical production process data, determining the upper limit of the range based on the maximum feed amount b, and constructing a classification scheme with a positive integer multiple i of a preset step amount Δ as the span includes: rounding down the minimum feed amount a tons to determine the lower limit a0 tons; rounding up the maximum feed amount b tons to determine the upper limit b0 tons; and constructing n types of working condition ranges [a0, a0+Δ), [a0+Δ, a0+2Δ), ..., [a0+nΔ, b0)] of a classification scheme with a positive integer multiple i of an integer Δ tons as the span.

[0012] Furthermore, in some embodiments of the present invention, the step of designing multiple classification schemes for classifying the multiple sets of historical production process data into working condition ranges based on the feed rate as the standard further includes: determining the amount of data contained in each working condition range; and in response to any working condition range containing less than a preset quantity threshold, merging the working condition range into an adjacent working condition range with a smaller amount of data.

[0013] Further, in some embodiments of the present invention, the thermally relevant data includes the exhaust gas temperature and / or the ratio of fuel gas flow rate to feed rate. The step of determining the multiple thermal efficiency levels by cluster analysis includes: extracting the exhaust gas temperature and / or the ratio of fuel gas flow rate to feed rate from the multiple sets of historical production process data respectively; constructing a cluster analysis model and randomly generating multiple cluster centers; inputting the exhaust gas temperature and / or the ratio from each set of historical production process data into the cluster analysis model, and dividing the multiple sets of historical production process data into multiple categories according to the distance from each set of exhaust gas temperature and / or the ratio to each cluster center, wherein each category corresponds to a thermal efficiency level; repeating the operation of generating cluster centers and classifying data according to the multiple sets of historical production process data after classification, until the cluster centers no longer change; and determining the multiple thermal efficiency levels according to the finally determined multiple cluster centers.

[0014] Furthermore, in some embodiments of the present invention, the step of determining the plurality of prediction models includes: establishing an initial thermal efficiency prediction model using an inverse equilibrium method based on the operating characteristics of the ethylene cracking furnace.

[0015] η=5-α1*O2-α2*tg+α3*t a -k

[0016] Where O2 represents the oxygen content, t g The exhaust gas temperature is represented by t. a The model coefficients α1, α2, and α3 represent the preheated air temperature, and k represents the model correction coefficient, reflecting the heat loss due to heat dissipation from the surface of the ethylene cracking furnace. Based on the thermal efficiency level η corresponding to the historical production process data of each of the aforementioned operating conditions, the values ​​of the model coefficients α1, α2, and α3 and the model correction coefficient k are solved respectively to determine multiple prediction models corresponding to each of the aforementioned operating conditions.

[0017] Furthermore, in some embodiments of the present invention, the step of solving for the values ​​of the model coefficients α1, α2, α3 and the model correction coefficient k based on the thermal efficiency level η corresponding to the historical production process data of each of the said operating conditions includes: determining the overdetermined linear equation system regarding the model coefficients α1, α2, α3 and the model correction coefficient k based on the thermal efficiency prediction model.

[0018]

[0019] The least squares method is then used to solve the overdetermined linear equation system to obtain the values ​​of the model coefficients α1, α2, α3 and the model correction coefficient k.

[0020] Furthermore, in some embodiments of the present invention, the multi-model prediction method further includes the following steps: in response to the failure to match the feed amount data with the operating condition range adapted to each of the prediction models, the production process data is stored in a temporary database, and it is determined whether the amount of data in the temporary database reaches a preset data amount threshold; and in response to the amount of data in the temporary database reaching the data amount threshold, the data stored in the temporary database is used as the historical production process data, the multiple prediction models and their adapted operating condition ranges are redefined, and they are updated to the multi-model database.

[0021] Furthermore, the multi-model prediction apparatus for the thermal efficiency of an ethylene cracking furnace provided according to a second aspect of the present invention includes a memory and a processor. The processor is connected to the memory and is configured to implement the multi-model prediction method for the thermal efficiency of an ethylene cracking furnace provided according to a first aspect of the present invention.

[0022] Furthermore, the computer-readable storage medium provided according to the third aspect of the present invention stores computer instructions thereon. When the computer instructions are executed by a processor, the multi-model prediction method for the thermal efficiency of the ethylene cracking furnace provided in the first aspect of the present invention is implemented. Attached Figure Description

[0023] The above-described features and advantages of the present invention will be better understood after reading the following detailed description of embodiments of the present disclosure in conjunction with the accompanying drawings. In the drawings, components are not necessarily drawn to scale, and components having similar related characteristics or features may have the same or similar reference numerals.

[0024] Figure 1 A flowchart illustrating a multi-model prediction method for the thermal efficiency of an ethylene cracking furnace according to some embodiments of the present invention is shown.

[0025] Figure 2 A schematic diagram showing the predicted thermal efficiency of an ethylene cracking furnace according to some embodiments of the present invention is illustrated.

[0026] Figure 3 A schematic diagram showing the predicted thermal efficiency of an ethylene cracking furnace according to some embodiments of the present invention is illustrated.

[0027] Figure 4 A schematic diagram showing the predicted thermal efficiency of an ethylene cracking furnace according to some embodiments of the present invention is illustrated. Detailed Implementation

[0028] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Although the description of the present invention is presented in conjunction with preferred embodiments, this does not mean that the features of the invention are limited to these embodiments. On the contrary, the purpose of describing the invention in conjunction with embodiments is to cover other options or modifications that may be derived based on the claims of the present invention. To provide a thorough understanding of the invention, many specific details will be included in the following description. The invention may also be implemented without using these details. Furthermore, to avoid confusion or obscuring the focus of the invention, some specific details will be omitted in the description.

[0029] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0030] It is understood that although terms such as "first," "second," and "third" may be used herein to describe various components, regions, layers, and / or parts, these components, regions, layers, and / or parts should not be limited by these terms, and these terms are only used to distinguish different components, regions, layers, and / or parts. Therefore, the first components, regions, layers, and / or parts discussed below may be referred to as second components, regions, layers, and / or parts without departing from some embodiments of the present invention.

[0031] As mentioned above, in the process of ethylene cracking, liquefied petroleum gas, hydrotreated tail oil and naphtha are first cracked into cracked gas composed of small molecules such as ethylene, methane, propylene and hydrogen by an ethylene cracking furnace at high temperature. Then, it is processed into ethylene, propylene and various by-products through compression, separation and other process units. The cracking process is a complex flow, heat transfer and mass transfer process, which is generally difficult to model by mechanism.

[0032] To overcome the aforementioned deficiencies in the existing technology, this invention provides a multi-model prediction method for the thermal efficiency of an ethylene cracking furnace, a multi-model prediction device for the thermal efficiency of an ethylene cracking furnace, and a computer-readable storage medium, which can quickly and accurately predict the thermal efficiency of an ethylene cracking furnace and lay the foundation for subsequent combustion optimization.

[0033] In some non-limiting embodiments, the multi-model prediction method for the thermal efficiency of an ethylene cracking furnace provided in the first aspect of the present invention can be implemented by the multi-model prediction device for the thermal efficiency of an ethylene cracking furnace provided in the second aspect of the present invention. Specifically, the multi-model prediction device may be configured with a memory and a processor. The memory includes, but is not limited to, the computer-readable storage medium provided in the third aspect of the present invention, on which computer instructions are stored. The processor is connected to the memory and configured to execute the computer instructions stored in the memory to implement the multi-model prediction method for the thermal efficiency of an ethylene cracking furnace provided in the first aspect of the present invention.

[0034] The working principle of the above-described multi-model prediction device will be described below with reference to some embodiments of multi-model prediction methods. Those skilled in the art will understand that these embodiments of multi-model prediction methods are merely non-limiting implementations provided by the present invention, intended to clearly demonstrate the main concepts of the invention and provide specific solutions convenient for public implementation, rather than limiting all functions or all operating methods of the multi-model prediction device. Similarly, the multi-model prediction device is also only a non-limiting implementation provided by the present invention and does not limit the entities implementing the steps in these multi-model prediction methods.

[0035] Please refer to the following first. Figure 1 , Figure 1 A flowchart illustrating a multi-model prediction method for the thermal efficiency of an ethylene cracking furnace according to some embodiments of the present invention is shown.

[0036] like Figure 1 As shown, in some embodiments of the present invention, the multi-model prediction method for the thermal efficiency of an ethylene cracking furnace can be implemented in two stages: offline modeling and online prediction. Correspondingly, the multi-model prediction device provided by the present invention can also be divided into two modules: offline modeling and online prediction. In some embodiments, the offline modeling module and the online prediction module can be integrated into the same device. Optionally, in other embodiments, the offline modeling module and the online prediction module can be distributed across two or more devices, and these devices cooperate with each other to implement the multi-model prediction method for the thermal efficiency of an ethylene cracking furnace provided in the first aspect of the present invention.

[0037] Specifically, in the offline modeling stage, the multi-model prediction device provided by this invention can first acquire multiple sets of historical production process data from the ethylene cracking furnace to construct a data sample set for the ethylene cracking furnace. Here, a set of historical production process data may include feed volume data such as total feed rate, bottom fuel gas flow rate, and sidewall fuel gas flow rate, as well as heat-related data such as flue gas temperature, oxygen content, preheating air temperature, and actual thermal efficiency. A set of historical production process data typically refers to data collected at a certain point in time for each selected monitoring variable. When the dataset is presented in matrix form, one row of data (row vector) in the matrix typically represents a set of ethylene cracking furnace production process data, and each column of data (column vector) in the matrix typically corresponds to different variables. The number of variables included in the production process data (number of columns) is the dimension of the data. Furthermore, to ensure the completeness and effectiveness of the samples, the number of samples should not be too small. Historical production process data before operating condition classification is typically no less than 10,000 sets, while historical production process data for each operating condition category after classification is typically no less than 200 sets. These sample data can be randomly divided in a 7:3 ratio, with 70% used for training the thermal efficiency model and 30% used for prediction and testing of the thermal efficiency model.

[0038] Furthermore, in some embodiments, the multi-model prediction device can preferably perform preprocessing operations such as operating condition monitoring, steady-state monitoring, and data cleaning on the acquired historical production process data to improve the training efficiency and prediction accuracy of the thermal efficiency prediction model.

[0039] For example, during the preprocessing operation, the multi-model prediction device can monitor the operating conditions of the collected raw data, exclude sample data of faulty or abnormal operating conditions, and retain only historical production process data under normal operating conditions without faults.

[0040] For example, the multi-model prediction device can also use the following formula (1) to perform steady-state monitoring on the acquired sample data:

[0041]

[0042] Where t represents the start time, M represents the width of the sliding window, and v τ This represents the value of the variable at time τ. σ represents the average value of the variable from time t to t+M-1, and σ is the result value after steady-state monitoring of the variable. i This represents the critical threshold. Here, the critical threshold σ... i It can be determined as 2% of the operating range of the i-th variable.

[0043] If the data in the window satisfies the equation, the multi-model prediction device will record the corresponding time span. Otherwise, the window will move forward one data point and continue checking the data in the window until all data have been steadily monitored.

[0044] Subsequently, the multi-model prediction device can calculate the average value of the data obtained through steady-state monitoring. If the difference d between the variable value at a certain point within the steady-state range and its average value... e If the difference is greater than 20%, the variable value will be considered an outlier due to a non-steady-state distribution, and data cleaning is required. Here, the difference d... e It can be calculated using the following equation (2):

[0045]

[0046] Among them, dy vt Let the variable value be a point within the steady-state range. d is the average value over the steady-state range. e This is a calculated value.

[0047] A set of historical production process data after data preprocessing may include six variables: total feed rate, bottom fuel gas flow rate, sidewall fuel gas flow rate, exhaust gas temperature, oxygen content, and preheated air temperature, denoted as:

[0048] X = {x1, x2, ..., x} n}, x i ∈R D

[0049] Where n represents the number of samples (n = 11367 in this embodiment), and D represents the input sample dimension (D = 6), namely the total feed rate, bottom fuel gas flow rate, side wall fuel gas flow rate, exhaust gas temperature, oxygen content, and preheated air temperature.

[0050] like Figure 1As shown, after completing the preprocessing of historical production process data, the multi-model prediction device can combine clustering analysis models such as k-means and the process characteristics of the ethylene cracking furnace to classify the thermal efficiency of each group of historical production process data into five levels: "high", "relatively high", "medium", "relatively low" and "low" (corresponding to a rating of 1 to 5). The data with calculated thermal efficiency level η are then classified according to the feed rate to design multiple classification schemes.

[0051] Specifically, the process of evaluating the thermal efficiency level of historical production process data mainly includes the following three steps:

[0052] The first step involves extracting thermal correlation data, such as "exhaust gas temperature" and "fuel gas to feed rate ratio," which are highly correlated with thermal efficiency and exhibit negative correlations, from multiple sets of historical production process data. Here, the fuel gas to feed rate ratio n can be determined based on the bottom fuel gas flow rate Q. b Sidewall fuel gas flow rate Q s And the total feed amount is Q m To calculate:

[0053]

[0054] The second step is to perform k-means clustering analysis on the data extracted in the first step. Here, the input values ​​for the k-means clustering analysis model are the exhaust gas temperature and the ratio of fuel gas to feed rate. The steps and related definitions of the clustering analysis are as follows:

[0055] A. First, randomly generate five cluster centers;

[0056] B. Based on the cluster centroids, the data can be divided into five categories. The classification principle is that data is assigned to the category closest to the centroid, and the Manhattan distance is used as the definition of distance:

[0057]

[0058] C. Recalculate the cluster centers based on the categorized data;

[0059] D. Repeat steps B and C until the cluster centers no longer change.

[0060] The third step is to rate the thermal efficiency based on the five cluster centers obtained in the second step, and divide all historical production process data into five data clusters, which correspond to five thermal efficiency levels of "low", "relatively low", "medium", "relatively high" and "high", and are represented by the label η = 1 to 5.

[0061] Specifically, since the ratio of fuel gas to feed rate (n) and flue gas temperature are both highly correlated with thermal efficiency, this invention can use the k-means clustering algorithm to determine the center position of each thermal efficiency level. The algorithm input is:

[0062]

[0063] Where num represents the number of samples in the clustering algorithm (num = 11367 in this embodiment), D k This indicates the input sample dimension (D=2), which consists of two quantities: the ratio of fuel gas to feed rate n and the flue gas temperature.

[0064] Subsequently, this invention can set the number of cluster centers to 5, and the algorithm outputs cluster centers for five categories, arranged in ascending order according to the ratio n of fuel gas to feed amount as follows: (n1, t2, t3, t4, t5, t6, t7, t8, t9 ... g1 (n2, t) g2 (n3, t) g3 (n4, t) g4 (n5, t) g5 ).

[0065] Since the ratio of fuel gas to feed rate and the flue gas temperature are inversely proportional to thermal efficiency, and the data trends of the ratio of fuel gas to feed rate and the flue gas temperature are basically consistent, when n1 < n2 < n3 < n4 < n5, exactly t g1 <t g2 <t g3 <t g4 <t g5 .

[0066] This inference not only conforms to real-world logic, but can also be proven through multiple simulations.

[0067] Based on the above theory, it can be concluded that (n1, t) g1 (n2, t) g2 (n3, t) g3 (n4, t) g4 (n5, t) g5 The five data clusters with cluster centers correspond to five levels of thermal efficiency: "high", "relatively high", "medium", "relatively low" and "low", and are labeled with five values: 5, 4, 3, 2 and 1 to represent the five thermal efficiency levels.

[0068] The data sample after completing the thermal efficiency rating can be updated as follows:

[0069] X′=(x′1,x′2,...,x′ num},x′ i ∈R D′

[0070] Where num represents the number of samples (num = 11367 in this embodiment), and D represents the input sample dimension (D = 6 in this embodiment), namely, feed rate, ratio of fuel gas to total feed rate, exhaust gas temperature, oxygen content, preheated air temperature, and thermal efficiency.

[0071] In addition, the process of designing multiple classification schemes for classifying multiple sets of historical production process data based on operating condition ranges mainly includes the following three steps:

[0072] The first step involves extracting the total feed amount (in tons) and other feed quantity data from multiple sets of historical production process data to determine the total range [a, b] covered by these sets of historical production process data. Further, to simplify calculations, the multi-model prediction device can round down the minimum feed amount 'a' from the multiple sets of historical production process data to determine the lower limit of the range a0 tons, and round up the maximum feed amount 'b' to determine the upper limit of the range b0 tons, thus obtaining the integer range of feed amount [a0, b0].

[0073] The second step is to construct multiple working condition ranges for a classification scheme based on the feed rate and the positive integer multiples i of the preset step amount Δ as the span. Then, iterate through all positive integer multiples i of the number of working condition ranges greater than or equal to 3 to obtain i classification schemes for the total range [a, b].

[0074] Specifically, in some embodiments, the multi-model prediction device can preset a step size Δ = 1 ton. Thus, in the first classification scheme, each type of feed quantity ranges 1 ton, which can be divided into n1 working conditions: [a0, a0+1), [a0+1, a0+2), ..., [a0+n1, b0); in the second classification scheme, each type of feed quantity ranges 2 tons, which can be divided into n2 working conditions: [a0, a0+2), [a0+2, a0+4), ..., [a0+n2, b0); and so on, until the number of working conditions generated by the classification scheme is less than three.

[0075] The third step is to check if there are any operating conditions that need to be merged among all classification schemes. Assuming a total of S sets of historical data are obtained, if the number of data in a certain operating condition is less than 0.01*S sets, the multi-model prediction device can merge it into the adjacent operating condition with the smaller amount of data to reduce the complexity of the prediction model and the data processing load.

[0076] Specifically, taking the naphtha feed rate in the integer range of [36, 45] as an example, in the first classification scheme, the feed rate range for each working condition is 1 ton, which are 9 working conditions: [36, 37), [37, 38), [38, 39), [39, 40), [40, 41), [41, 42), [42, 43), [43, 44), and [44, 45].

[0077] In the second classification scheme, the feed range for each working condition is 2 tons, namely [36, 38), [38, 40), [40, 42), [42, 44), [44, 45], which are 5 working conditions. The last working condition is set to 1 ton because the span is less than 2 tons. Similar situations will not be described again.

[0078] In the third classification scheme, the feed rate range for each working condition is 3 tons, which can be divided into three working conditions: [36, 39), [39, 42), and [42, 45].

[0079] In the fourth classification scheme, the feed rate range for each working condition is 4 tons, which can be divided into three working conditions: [36, 40), [40, 44), and [44, 45].

[0080] When the feed rate for each operating condition is 4 tons, it is impossible to generate at least 3 operating conditions, therefore there is no fifth classification scheme. Thus, this invention can use the feed rate as the standard to design four classification schemes for classifying operating condition ranges based on multiple sets of historical production process data.

[0081] After that, as Figure 1 As shown, the multi-model prediction device can use multiple predetermined prediction models to calculate the predicted thermal efficiency values ​​of each historical thermal correlation data for each corresponding operating condition range in each classification scheme. Then, based on the actual thermal efficiency values ​​and predicted thermal efficiency values ​​corresponding to each historical thermal correlation data, the total mean square error of each classification scheme is calculated to determine the optimal classification scheme with the smallest total mean square error.

[0082] Specifically, the process of determining the optimal classification scheme mainly includes the following three steps:

[0083] The first step is to establish a thermal efficiency prediction model. The inverse equilibrium method is often used to calculate the thermal efficiency of a tubular pyrolysis furnace.

[0084] η = 1 - q1 - q2 - q3 (5)

[0085] Where q1 represents heat loss from flue gas, q2 represents heat loss from incomplete combustion, and q3 represents heat loss from surface heat dissipation. Here, the heat loss from flue gas q1 and the heat loss from incomplete combustion q2 are related to the oxygen content O2 and the flue gas temperature t. g Positively correlated with preheated air temperature t a Negative correlation. The surface heat loss q3 can be determined by selecting an appropriate value k based on the daily accumulated test data and operating load of the furnace under test.

[0086] Subsequently, this invention can combine the aforementioned thermal efficiency evaluation system with relevant knowledge of ethylene cracking furnace production processes to establish an ethylene cracking furnace thermal efficiency prediction model:

[0087] η=5-α1*O2-α2*tg+α3*t a -k (6)

[0088] Where O2 represents the oxygen content, t g The exhaust gas temperature is represented by t. a The temperature of the preheated air is represented by α1, which is the first model coefficient, α2, which is the second model coefficient, α3, which is the third model coefficient, and k is the model correction coefficient, which reflects the heat loss q3 from the surface of the ethylene cracking furnace.

[0089] The second step is to rearrange the terms in formula (6) based on the thermal efficiency prediction model obtained in the first step to calculate the model coefficients:

[0090] -α1*O2-α2*tg+α3*t a -k=η-5 (7)

[0091] Among them, O2, t g t α All of these can be obtained through processing the production process data of the ethylene cracking furnace. η can be calculated and determined through the steps of the above cluster analysis, and α1, α2, α3, and k are model coefficients. Therefore, formula (7) can be written in matrix form:

[0092]

[0093] Subsequently, the multi-model prediction device can combine the thermal efficiency level η corresponding to the historical production process data of each operating condition range, and solve the overdetermined linear equation system by the least squares method, and solve the values ​​of the model coefficients α1, α2, α3, and k of the prediction model respectively.

[0094] The third step is to calculate the total mean square error (RMSE) for each classification scheme. To evaluate the merits of different classification schemes, the thermal efficiency model established in the second step is used for prediction, and the RMSE of each prediction model for each classification scheme is calculated. The classification scheme with the smallest RMSE is then selected as the optimal classification scheme. The formula for calculating the RMSE is as follows:

[0095]

[0096] Where N is the total number of samples predicted, y ob For the true value of thermal efficiency, y pre This is the predicted value for thermal efficiency.

[0097] Please refer to Table 1, which shows the classification scheme and corresponding thermal efficiency prediction model provided according to some embodiments of the present invention.

[0098] Table 1 Classification Schemes and Models

[0099]

[0100]

[0101] The multi-model prediction device can establish thermal efficiency prediction models for a total of 20 operating conditions in the four classification schemes shown in Table 1 according to the above formula (6), observe the prediction results, and use formula (9) to calculate the total mean square error (RMSE) of each classification scheme.

[0102] The total mean square error (RMSE) of the four classification schemes calculated in this embodiment is shown in Table 2:

[0103] Table 2 Total Mean Square Error of Classification Schemes

[0104] Classification scheme 1 2 3 4 Total Mean Square Error 0.0032 0.0030 0.0029 0.0031

[0105] As shown in Table 2, classification scheme 3 has the smallest total mean square error and is therefore the optimal classification scheme. The classification prediction results of the optimal scheme are as follows: Figure 2 , Figure 3 , Figure 4 As shown. Figure 2 , Figure 3 , Figure 4 These correspond to the thermal efficiency prediction results for Models 3.1, 3.2, and 3.3, respectively. The horizontal axis of the graph represents the data sample, and the vertical axis represents the thermal efficiency. The solid line represents the predicted value, while the dashed line represents the actual value. Figure 2 , Figure 3 , Figure 4 It can be seen that the red and blue lines highly overlap, proving the high accuracy of the established model. Therefore, in this embodiment, the multi-model prediction device can adopt the [36, 39), [39, 42), and [42, 45] operating condition classification schemes, and the corresponding thermal efficiency prediction models are models 3.1, 3.2, and 3.3.

[0106] Subsequently, the multi-model prediction device can store the various working condition ranges divided by the optimal classification scheme of [36, 39), [39, 42), and [42, 45], as well as their corresponding prediction models 3.1, 3.2, and 3.3, into the multi-model database for use in the online prediction stage.

[0107] Please continue to refer to this. Figure 1In the online prediction stage of the multi-model prediction method provided by this invention, the multi-model prediction device can acquire production process data of the ethylene cracking furnace online. This production process data may include feed volume data such as total feed rate, bottom fuel gas flow rate, and sidewall fuel gas flow rate, as well as heat-related data such as flue gas temperature, oxygen content, and preheating air temperature. A set of production process data is typically collected at a certain point in time from selected monitoring variables. When the dataset is presented in matrix form, one row of data (row vector) in the matrix usually represents a set of ethylene cracking furnace production process data, and each column of data (column vector) in the matrix usually corresponds to a different variable. The number of variables included in the production process data (number of columns) is the dimension of the data.

[0108] Furthermore, in some embodiments, the multi-model prediction device can preferably perform preprocessing operations such as operating condition monitoring, steady-state monitoring, and data cleaning on the acquired production process data to improve the training efficiency and prediction accuracy of the thermal efficiency prediction model. The principle of this preprocessing operation is similar to that of the offline modeling stage described above, and will not be elaborated here.

[0109] Subsequently, for the aforementioned operating condition range classification scheme designed based on feed rate, the multi-model prediction device can match the total feed rate in the production process data with the operating condition ranges adapted to various prediction models in the multi-model database. If the total feed rate successfully matches the operating condition range adapted to any one or more of the prediction models in the multi-model database, the multi-model prediction device can call one or more corresponding prediction models to predict the thermal efficiency of the ethylene cracking furnace in the corresponding operating condition range based on thermally relevant data such as flue gas temperature, oxygen content, and preheated air temperature in the production process data.

[0110] Conversely, if the feed rate data fails to match the operating condition ranges of all prediction models in the multi-model database, the multi-model prediction device can report an error and store the production process data in a temporary database. Subsequently, the multi-model prediction device can monitor in real time whether the data volume in the temporary database reaches a preset data volume threshold 'a'. In response to the monitoring result that the data volume in the temporary database has reached the data volume threshold 'a', the multi-model prediction device can repeat the steps of the offline modeling stage described above, using the data stored in the temporary database as historical production process data to redetermine the various prediction models and their applicable operating condition ranges, and update them in the multi-model database for online prediction.

[0111] Specifically, for the embodiment with a total range of [36, 45], if the real-time collected feed rate is 38 tons (within the range of [36, 45]), the multi-model prediction device can match the corresponding operating condition and call the corresponding thermal efficiency prediction model 3.1 to predict the thermal efficiency of the ethylene cracking furnace. Conversely, if the real-time collected feed rate is 20 tons (not within the range of [36, 45]), the multi-model prediction device can determine that the feed rate exceeds the total range of the multi-model database and store all of this data in a temporary database.

[0112] Subsequently, the multi-model prediction device can also monitor the amount of data in the temporary database in real time. When the amount of data in the temporary database reaches the preset data volume threshold (0.1*S), the multi-model prediction device can determine that the current thermal efficiency prediction model can no longer meet the actual prediction requirements, and then execute the steps of the above offline modeling stage again to perform offline modeling on the 0.1*S set of data to obtain the thermal efficiency prediction model corresponding to the new operating condition, and update it to the multi-model database.

[0113] Thus, by using cluster analysis to adaptively determine multiple thermal efficiency levels, and combining the total mean square error (RMSE) of each classification scheme to adaptively determine the classification scheme for each operating condition, and then combining the relevant knowledge of the thermal efficiency evaluation system and the ethylene cracking furnace production process, multiple thermal efficiency prediction models corresponding to each operating condition range are established. This invention can quickly and accurately predict the thermal efficiency of the ethylene cracking furnace and lay the foundation for subsequent combustion optimization.

[0114] Although the methods described above are illustrated and depicted as a series of actions for the sake of simplicity, it should be understood and appreciated that these methods are not limited by the order of the actions, as some actions may occur in a different order and / or concurrently with other actions from the illustrations and descriptions herein or not illustrated and described herein but which may be understood by those skilled in the art, according to one or more embodiments.

[0115] Although the multi-model prediction device for the thermal efficiency of the ethylene cracking furnace described in the above embodiments can be implemented through a combination of software and hardware, it is understood that the multi-model prediction device can also be implemented independently in software or hardware. For hardware implementation, the multi-model prediction device can be implemented using one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, other electronic devices for performing the above functions, or a selection of combinations of the above devices. For software implementation, the multi-model prediction device can be implemented using independent software modules such as procedures and functions running on a general-purpose chip, each module performing one or more functions and operations described herein.

[0116] The various illustrative logic modules and circuits described in conjunction with the embodiments disclosed herein may be implemented or performed using a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. The general-purpose processor may be a microprocessor, but in alternatives, it may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors cooperating with a DSP core, or any other such configuration.

[0117] The prior description of this disclosure is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to this disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not intended to be limited to the examples and designs described herein, but should be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A multi-model prediction method for the thermal efficiency of an ethylene cracking furnace, characterized in that, Includes the following steps, Acquire production process data from an ethylene cracking furnace, wherein the production process data includes feed rate data and thermal correlation data; The process involves matching the feed rate data with the operating condition ranges adapted to various prediction models in a multi-model database. The step of determining the various prediction models in the multi-model database and their adapted operating condition ranges includes: acquiring multiple sets of historical production process data from the ethylene cracking furnace, including feed rate data and thermal correlation data; designing multiple classification schemes for classifying the operating condition ranges of the multiple sets of historical production process data using the feed rate as a standard; calculating the predicted thermal efficiency values ​​of each historical thermal correlation data for each corresponding operating condition range in each classification scheme using multiple predetermined prediction models based on multiple thermal efficiency levels determined by pre-cluster analysis; calculating the total mean square error of each classification scheme based on the actual thermal efficiency values ​​and the predicted thermal efficiency values ​​corresponding to each historical thermal correlation data; determining the optimal classification scheme with the smallest total mean square error; and storing the multiple operating condition ranges divided by the optimal classification scheme and their corresponding prediction models in the multi-model database. In response to a successful match between the feed rate data and the operating condition range of any one or more of the prediction models, the corresponding one or more prediction models are invoked to predict the thermal efficiency of the ethylene cracking furnace in the corresponding operating condition range based on the thermal correlation data.

2. The multi-model prediction method as described in claim 1, characterized in that, Before designing multiple classification schemes for classifying the operating condition ranges of the multiple sets of historical production process data based on the historical feed data, the multi-model prediction method further includes the following steps: The acquired sets of historical production process data are preprocessed to filter out abnormal operating conditions and / or outliers with non-steady-state distributions.

3. The multi-model prediction method as described in claim 1, characterized in that, The steps of designing multiple classification schemes for classifying the operating condition ranges of the multiple sets of historical production process data based on the feed rate include: The feed rate data is extracted from the multiple sets of historical production process data to determine the total scope covered by the multiple sets of historical production process data. ]; Based on the minimum feed rate from the multiple sets of historical production process data Determine the lower limit of the range based on the maximum feed rate. Determine the upper limit of the range, and construct a classification scheme for multiple working condition ranges using a span of i, which is a positive integer multiple of a preset step size Δ; and Iterate through all positive integer multiples i of the number of working condition ranges greater than or equal to 3 to obtain the total range. [i classification schemes].

4. The multi-model prediction method as described in claim 3, characterized in that, The minimum feed rate is based on the multiple sets of historical production process data. Determine the lower limit of the range based on the maximum feed rate. The steps for constructing multiple operating condition ranges for a classification scheme, based on determining the upper limit of the range and using a positive integer multiple of the preset step size Δ as the span, include: For the minimum feed rate Round down to the nearest ton to determine the lower limit of the range. ton; For the maximum feed rate Round up to the nearest ton to determine the upper limit of the range. tons; and Using positive integer multiples of integer Δ tons as the span, construct a classification scheme with n working condition ranges. , ), [ , ), ..., [ , ).

5. The multi-model prediction method as described in claim 3, characterized in that, The step of designing multiple classification schemes for classifying the operating condition ranges of the multiple sets of historical production process data based on the feed rate also includes: Determine the amount of data contained in each of the aforementioned operating condition ranges; and In response to any operating condition range containing less than a preset quantity threshold, the operating condition range is merged into an adjacent operating condition range with a smaller data volume.

6. The multi-model prediction method as described in claim 1, characterized in that, The thermal correlation data includes exhaust gas temperature and / or the ratio of fuel gas flow rate to feed rate. The step of determining the multiple thermal efficiency levels through cluster analysis includes: Extract the exhaust gas temperature and / or the ratio of fuel gas flow rate to feed rate from the multiple sets of historical production process data; Construct a cluster analysis model and randomly generate multiple cluster centers; The exhaust gas temperature and / or the ratio in each group of historical production process data are input into the cluster analysis model. Based on the distance from each group of exhaust gas temperature and / or the ratio to each cluster center point, the multiple groups of historical production process data are divided into multiple categories, wherein each category corresponds to a thermal efficiency level. Based on the categorized historical production process data, the process of generating cluster centers and classifying data is repeated until the cluster centers no longer change; and The multiple thermal efficiency levels are determined based on the finally determined multiple cluster centers.

7. The multi-model prediction method as described in claim 6, characterized in that, The steps for determining the plurality of prediction models include: Based on the operating characteristics of the ethylene cracking furnace, an initial thermal efficiency prediction model is established using the inverse equilibrium method. in, Indicates oxygen content. Indicates the exhaust gas temperature. Indicates the preheated air temperature. , , The model coefficients are represented by , k represents the model correction coefficients, reflecting the heat loss due to surface heat dissipation in the ethylene cracking furnace; and Based on the thermal efficiency rating corresponding to the historical production process data for each of the aforementioned operating conditions. Solve the model coefficients respectively , , The model correction coefficient k is used to determine multiple prediction models corresponding to each of the said operating conditions.

8. The multi-model prediction method as described in claim 7, characterized in that, The thermal efficiency level corresponding to the historical production process data of each of the aforementioned operating conditions. Solve the model coefficients respectively , , The steps for determining the value of the model correction coefficient k include: Based on the thermal efficiency prediction model, determine the coefficients of the model. , , and the overdetermined linear equations of the model correction coefficient k ;as well as The overdetermined linear equations are solved using the least squares method to obtain the model coefficients. , , And the value of the model correction coefficient k.

9. The multi-model prediction method as described in claim 1, characterized in that, It also includes the following steps: In response to the failure of the feed rate data to match the working condition range of each of the prediction models, the production process data is stored in a temporary database, and it is determined whether the amount of data in the temporary database has reached a preset data amount threshold. as well as In response to the data volume in the temporary database reaching the data volume threshold, the data stored in the temporary database is used as the historical production process data, and multiple prediction models and their applicable operating condition ranges are redefined and updated to the multi-model database.

10. A multi-model prediction device for the thermal efficiency of an ethylene cracking furnace, characterized in that, include: Memory; as well as A processor, connected to the memory, and configured to implement the multi-model prediction method for the thermal efficiency of an ethylene cracking furnace as described in any one of claims 1 to 9.

11. A computer-readable storage medium storing computer instructions thereon, characterized in that, When the computer instructions are executed by the processor, the multi-model prediction method for the thermal efficiency of the ethylene cracking furnace as described in any one of claims 1 to 9 is implemented.

Images