A Method and System for Failure Detection and Diagnosis of Protective Layer in Petroleum Refining and Chemical Units Based on Multi-Source Data Fusion

Abstract

This invention belongs to the field of industrial control system safety technology and discloses a method for detecting protective layer failure in petroleum refining units based on multi-source data fusion. This invention obtains a normalized process monitoring data training matrix and a statistical alarm information matrix by normalizing and adjusting alarm levels using data collected from petroleum refining unit simulation software. Following the principle of maximum correlation and minimum redundancy, the variable set for protective layer failure detection and diagnosis is selected. The invention then fuses the protective layer failure detection and diagnosis results obtained from the statistical alarm information data using a discrete Bayesian classifier and from the protective layer failure detection and diagnosis results obtained from the process monitoring data training matrix using a continuous Bayesian classifier, resulting in the final protective layer failure detection and diagnosis results. Compared with existing methods, this invention creatively integrates statistical alarm information data, eliminating the interference of noise in the process monitoring data on the detection and diagnosis of protective layer failure.

Description

Technical Field

[0001] This invention belongs to the field of industrial control system security technology, and in particular relates to a method and system for detecting the failure of the protective layer of an oil refining unit based on multi-source data fusion. Background Technology

[0002] The petrochemical industry is inherently a high-risk industry characterized by flammable and explosive substances. Its production processes involve complex technologies and chemical reactions, harboring numerous safety risks. These risks not only pose a significant threat to human life but can also impact production efficiency. Therefore, ensuring the safe operation of petroleum refining units is fundamental to improving their overall effectiveness.

[0003] Layer-of-protection analysis is a typical semi-quantitative hazard analysis method, usually comprising multiple independent layers of protection, such as the basic process control system layer, the alarm and personnel intervention layer, the safety instrumented system layer, the physical protection layer, and the emergency response layer. These layers of protection are independent, effective, and auditable, and are widely used in petroleum refining plants. In particular, the basic process control system layer, the alarm and personnel intervention layer, and the safety instrumented system layer are the three most critical independent layers in the operation of petroleum refining plants. Therefore, designing failure detection and diagnosis methods for these three layers is especially important.

[0004] Currently, methods for detecting and diagnosing protective layer failures in petroleum refining units primarily rely on process monitoring data. CN201610131396.2 utilizes process monitoring data to implement a method for detecting and diagnosing process failures in refining. While process monitoring data can reflect the operating status of the unit to some extent, noise data can interfere with the detection and diagnosis of protective layer failures during the occurrence, recovery, and switching between different protective layer failure states, leading to incorrect judgments. Statistical alarm information, due to its discrete characteristics, can effectively reduce the interference of noise data and is often used to solve this problem. CN201610133660.6 provides a dynamic alarm threshold generation method; however, relying solely on statistical alarm information often makes it difficult to accurately determine the specific type of protective layer failure. Furthermore, neither of the above two inventions analyzes petroleum refining unit failures from the perspective of the protective layer. Therefore, relying solely on statistical alarm information cannot meet the needs of protective layer failure detection and diagnosis in petroleum refining units.

[0005] Based on the above analysis, the problems and shortcomings of the existing technology are as follows:

[0006] Methods for detecting and diagnosing protective layer failures in petroleum refining units primarily rely on process monitoring data. While process monitoring data can reflect the operating status of the unit to some extent, noise data can interfere with the detection and diagnosis of protective layer failures during the occurrence, recovery, and transition between different protective layer failure states, leading to erroneous judgments. Relying solely on statistical alarm information is often insufficient to accurately determine the specific type of protective layer failure. Summary of the Invention

[0007] To address the problems existing in the prior art, this invention provides a method for detecting the failure of the protective layer of an oil refining unit based on multi-source data fusion.

[0008] This invention is implemented as follows: A method for detecting the failure of protective layers in petroleum refining and chemical plants based on multi-source data fusion includes:

[0009] Step 1: Determine the sliding window length and sampling interval;

[0010] Step 2: Collect process monitoring data and update the process monitoring data matrix based on the step size;

[0011] Step 3: Perform column vector normalization on the process monitoring data;

[0012] Step 4: Generate statistical alarm information by normalizing process monitoring data;

[0013] Step 5: Filter the variable set according to the maximum relevance-minimum redundancy index;

[0014] Step 6: Using statistical alarm information data, a discrete Bayesian classifier is used to obtain the protection layer failure detection and diagnosis results. ;

[0015] Step 7: Using process monitoring data, obtain the protection layer failure detection and diagnosis results using a continuous Bayesian classifier. ;

[0016] Step 8, Fusion and The final protective layer failure detection and diagnosis results are obtained. .

[0017] Furthermore, the process monitoring data acquired based on the sliding window length and sampling interval includes:

[0018] The process monitoring data training matrix at the initial sampling time is determined as follows: Where n is the total number of variables monitored in the process, and w is the length of the sampling window;

[0019] After k samplings, the process monitoring data training matrix is ​​updated as follows: ;in The sampling interval;

[0020] The process monitoring data training matrix is ​​centered to generate a normalized process monitoring data training matrix, including:

[0021] Through the formula: Determine the training matrix for normalized process monitoring data ;

[0022] Statistical alarm information data is normalized from process monitoring data using the formula:

[0023] Generate; where For the first The standard deviation of each variable This is the alarm level constant. The higher the alarm level, the greater the deviation of the current operating state of the oil refining unit from the normal state.

[0024] Furthermore, the statistical alarm information data training matrix generated after k sampling updates of the process monitoring data is as follows: .

[0025] Furthermore, following the principle of maximum relevance and minimum redundancy, variable selection is performed based on the normalized process monitoring data training matrix, including:

[0026] Determine the number m of features to be retained in the target set;

[0027] Based on formula Determine the set most relevant to the label ,in ;

[0028] The above formula is derived from the formula Approximate calculation;

[0029] Using formula Determine the element with the least redundancy within the set;

[0030] By combining the most relevant metrics to the tags and the metrics with the least redundancy within the set, through

[0031] Determine the feature set for training .

[0032] Furthermore, the discrete Bayesian classifier follows the formula:

[0033]

[0034] Calculate the probability of failure for each type of protective layer; therefore, based on the formula: This yields the case with the highest probability of protective layer failure. .

[0035] Furthermore, the continuous Bayesian classifier, according to the formula:

[0036] Calculate the probability of failure for each type of protective layer;

[0037] Approximation using the multivariate kernel density estimation method ,use replace ,

[0038] in ,

[0039] Where K is the kernel function, and this implementation adopts the multivariate normal kernel function, that is: ;

[0040] It is a symmetric positive definite matrix:

[0041] ;

[0042] get ;

[0043] then, Therefore, based on the formula:

[0044] The case with the highest probability of protective layer failure. ;

[0045] The results of the discrete Bayesian classifier and the continuous Bayesian classifier are combined based on the formula:

[0046] Obtain the final protective layer failure detection results .

[0047] Another objective of this invention is to provide a failure detection system for the protective layer of an oil refining unit based on multi-source data fusion, comprising:

[0048] The parameter setting module is used to determine parameters such as the sliding window length and sampling interval.

[0049] The data acquisition module is used to acquire the values ​​of control variables from each control loop in the oil refining unit simulation software;

[0050] The data update module is used to update the process monitoring data training matrix based on the sampling interval in order to achieve the dynamic effect of the system.

[0051] The normalization module is used to normalize the column vectors of process monitoring data using a formula.

[0052] The statistical alarm information calculation module is used to train a matrix using normalized process monitoring data to generate statistical alarm information data.

[0053] The variable selection module is used to select the set of variables most relevant to the detection and diagnosis of protective layer failure based on the maximum relevance-minimum redundancy index.

[0054] The Discrete Bayes classifier module is used to obtain the detection and diagnosis results of the Discrete Bayes classifier protection layer failure by utilizing statistical alarm information data;

[0055] The Continuous Bayes Classifier module is used to obtain the detection and diagnosis results of the continuous Bayes Classifier protection layer failure by utilizing process monitoring data;

[0056] The protection layer failure determination module is used to integrate modules 7 and 8 to obtain the final protection layer failure detection and diagnosis results.

[0057] Another object of the present invention is to provide a computer device including a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the steps of the method for detecting the failure of the protective layer of an oil refining unit based on multi-source data fusion.

[0058] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method for detecting the failure of the protective layer of an oil refining unit based on multi-source data fusion.

[0059] Another objective of this invention is to provide an information data processing terminal for implementing the oil refining unit protective layer failure detection system based on multi-source data fusion.

[0060] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by this invention are as follows:

[0061] First, this invention, through normalization and alarm levels, utilizes data collected from petroleum refining unit simulation software to obtain a normalized process monitoring data training matrix and a statistical alarm information matrix. Following the principle of maximum correlation and minimum redundancy, it selects the variable set for protective layer failure detection and diagnosis. It then integrates the protective layer failure detection and diagnosis results obtained using a discrete Bayesian classifier on the statistical alarm information data and the protective layer failure detection and diagnosis results obtained using a continuous Bayesian classifier on the process monitoring data training matrix, resulting in the final protective layer failure detection and diagnosis results. Compared with existing methods, this invention creatively integrates statistical alarm information data, eliminating the interference of noise in the process monitoring data on protective layer failure detection and diagnosis.

[0062] Secondly, the expected benefits and commercial value of the technical solution of this invention after its transformation are as follows: it can provide domestic oil refining plants with protective layer failure detection and diagnosis technology, reducing the significant hazards and economic losses caused by protective layer failure. The price of protective layer failure analysis varies from hundreds of thousands to millions of yuan depending on the number of on-site operation points. Currently, there are hundreds of oil refineries in my country. Considering that this invention achieves practical accuracy in protective layer failure detection and diagnosis, and is more convenient, it can serve as an alternative solution, thus possessing enormous commercial value.

[0063] The technical solution of this invention fills a technological gap in the industry both domestically and internationally: Currently, failure analysis of petroleum refining and chemical plants focuses more on the failure of specific controllers, actuators, and sensors in the basic control system layer, neglecting the failure states of other protection layers. The technical solution proposed in this invention expands the failure analysis of petroleum refining and chemical plants to the basic process control system layer, the alarm and personnel intervention layer, and the safety instrumented system layer, thus filling the technological gap.

[0064] Does the technical solution of this invention solve a technical problem that people have long desired to solve but have never been able to successfully address? (1) Existing technologies mostly focus on the failure of specific controllers, actuators, and sensors in the basic control system layer, ignoring the failure states of other protection layers. The technical solution proposed in this invention expands the failure analysis of petroleum refining units to the basic process control system layer, alarm and personnel intervention layer, and safety instrument system layer, filling the technical gap. (2) Existing failure analysis of petroleum refining units is mostly based on process monitoring data, which is difficult to cope with the impact of noise data on failure analysis during the failure state transition process. This invention uses a fusion strategy of process monitoring data and statistical alarm information to make up for this deficiency.

[0065] The technical solution of this invention overcomes technical bias: failure analysis of protective layers in petroleum refining units is mostly based on process monitoring data, because process monitoring data is the most easily obtained and explicit data reflecting the operating status of petroleum refining units. This invention introduces statistical alarm information into protective layer failure analysis, aiming to comprehensively obtain non-intuitive implicit characteristics, thereby overcoming the impact of noise data during the long-neglected failure state transition process on failure analysis. Attached Figure Description

[0066] Figure 1 This is a flowchart of a method for detecting the failure of the protective layer of an oil refining unit based on multi-source data fusion, provided in an embodiment of the present invention.

[0067] Figure 2 This is a structural block diagram of a petroleum refining unit protective layer failure detection system based on multi-source data fusion provided in an embodiment of the present invention.

[0068] Figure 3 This is an example diagram of the original data collected for process monitoring of the control variable TI1150 in the protection layer of the basic control process system of the petroleum refining unit provided in this embodiment of the invention.

[0069] Figure 4 This is an example diagram of the normalized process monitoring data of the control variable TI1150 in the protection layer of the basic control process system of the petroleum refining unit provided in the embodiment of the present invention.

[0070] Figure 5 This is an example diagram of statistical alarm information generated from the normalized process monitoring data of the control variable TI1150 in the protection layer of the basic control process system of the petroleum refining unit provided in this embodiment of the invention. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0072] like Figure 1 As shown in the figure, the method for detecting the failure of the protective layer of an oil refining unit based on multi-source data fusion provided by this embodiment of the invention includes the following steps:

[0073] S101, Determine the sliding window length and sampling interval;

[0074] S102, collect process monitoring data and update the process monitoring data matrix based on the step size;

[0075] S103, perform column vector normalization on process monitoring data;

[0076] S104, generates statistical alarm information by normalizing process monitoring data;

[0077] S105, filter the variable set according to the maximum correlation-minimum redundancy index;

[0078] S106, using statistical alarm information data, a discrete Bayesian classifier is used to obtain the protection layer failure detection and diagnosis results. ;

[0079] S107, using process monitoring data, a continuous Bayesian classifier is used to obtain the detection and diagnosis results of protective layer failure. ;

[0080] S108, Fusion and The final protective layer failure detection and diagnosis results are obtained. .

[0081] In the first step (S101) of this embodiment of the invention, by setting the length of the sliding window and the sampling interval, continuous process data can be obtained during the monitoring process. The length of the sliding window determines the time span of each data analysis, while the sampling interval determines the frequency of data acquisition. The proper selection of these two parameters is crucial for subsequent data updates, processing, and analysis, ensuring the real-time nature and continuity of the monitoring data.

[0082] Step S102 involves collecting process monitoring data and updating the process monitoring data matrix based on a set step size. This step aims to dynamically acquire and update the monitoring data matrix, enabling the system to reflect the real-time operating status of the protective layer in the oil refining unit. The updated data matrix will be used for subsequent data processing and analysis, providing a basis for protective layer failure detection.

[0083] In step S103, column vector normalization is performed on the process monitoring data. The purpose of normalization is to unify the dimensions of each variable, enabling comparisons of monitoring data from different magnitudes under the same standard. This process eliminates the influence of dimensional differences in the original data, thus providing a more reliable foundation for subsequent data analysis.

[0084] Step S104 generates statistical alarm information based on the normalized process monitoring data. By statistically analyzing the monitoring data, the system can detect abnormal fluctuations or abnormal patterns and generate alarm signals. These alarm messages can reflect the potential failure risk of the protective layer of the oil refining unit in real time and provide preliminary early warning for subsequent failure detection and diagnosis.

[0085] In step S105, the maximum correlation-minimum redundancy (mRMR) metric is used to filter the variable set, selecting the feature variables most correlated with the protection layer failure and removing redundant information to improve data processing efficiency. Next, in step S106, these feature variables and statistical alarm information are used to detect and diagnose protection layer failure using a discrete Bayesian classifier. The discrete Bayesian classifier allows for modeling the probability of failure and yielding preliminary diagnostic results.

[0086] In step S107, a continuous Bayesian classifier is used for further failure detection and diagnosis based on process monitoring data. This classifier considers the continuity of the data and can more accurately identify potential failure risks. Finally, in step S108, the detection results of the discrete Bayesian classifier and the continuous Bayesian classifier are fused to obtain the final protection layer failure detection and diagnosis results. By fusing multi-source data, the accuracy and robustness of detection can be further improved, ensuring the safety and stability of oil refining units.

[0087] The monitoring data collected based on the sliding window length and sampling interval provided in this embodiment of the invention includes:

[0088] The process monitoring data training matrix at the initial sampling time is determined as follows: Where n is the total number of variables monitored in the process, and w is the length of the sampling window;

[0089] After k samplings, the process monitoring data training matrix is ​​updated as follows: ;in The sampling interval;

[0090] The process monitoring data training matrix is ​​centered to generate a normalized process monitoring data training matrix, including:

[0091] Through the formula: Determine the training matrix for normalized process monitoring data ;

[0092] Statistical alarm information data is normalized from process monitoring data using the formula:

[0093] Generate; where For the first The standard deviation of each variable This is the alarm level constant. The higher the alarm level, the greater the deviation of the current operating state of the oil refining unit from the normal state.

[0094] The statistical alarm information data training matrix generated after k sampling updates of the process monitoring data provided in this embodiment of the invention is as follows: .

[0095] The present invention provides a method for selecting variables based on a normalized process monitoring data training matrix, following the maximum correlation-minimum redundancy guiding principle. This includes:

[0096] Determine the number m of features to be retained in the target set;

[0097] Based on formula Determine the set most relevant to the label ,in ;

[0098] The above formula is derived from the formula Approximate calculation;

[0099] Using formula Determine the element with the least redundancy within the set;

[0100] By combining the most relevant metrics to the tags and the metrics with the least redundancy within the set, through

[0101] Determine the feature set for training .

[0102] The discrete Bayesian classifier provided in this embodiment of the invention is based on the formula:

[0103]

[0104] Calculate the probability of failure for each type of protective layer; therefore, based on the formula: This yields the case with the highest probability of protective layer failure. .

[0105] The continuous Bayesian classifier provided in this embodiment of the invention is based on the formula:

[0106] Calculate the probability of failure for each type of protective layer;

[0107] Approximation using the multivariate kernel density estimation method ,use replace ,

[0108] in ,

[0109] Where K is the kernel function, and this implementation adopts the multivariate normal kernel function, that is: ;

[0110] It is a symmetric positive definite matrix:

[0111] ;

[0112] get ;

[0113] then, Therefore, based on the formula:

[0114] The case with the highest probability of protective layer failure. ;

[0115] The results of the discrete Bayesian classifier and the continuous Bayesian classifier are combined based on the formula:

[0116] Obtain the final protective layer failure detection results .

[0117] like Figure 2 As shown in the figure, an embodiment of the present invention provides a failure detection system for the protective layer of an oil refining unit based on multi-source data fusion, comprising:

[0118] The parameter setting module is used to determine parameters such as the sliding window length and sampling interval.

[0119] The data acquisition module is used to acquire the values ​​of control variables from each control loop in the oil refining unit simulation software;

[0120] The data update module is used to update the process monitoring data training matrix based on the sampling interval in order to achieve the dynamic effect of the system.

[0121] The normalization module is used to normalize the column vectors of process monitoring data using a formula.

[0122] The statistical alarm information calculation module is used to train a matrix using normalized process monitoring data to generate statistical alarm information data.

[0123] The variable selection module is used to select the set of variables most relevant to the detection and diagnosis of protective layer failure based on the maximum relevance-minimum redundancy index.

[0124] The Discrete Bayes classifier module is used to obtain the detection and diagnosis results of the Discrete Bayes classifier protection layer failure by utilizing statistical alarm information data;

[0125] The Continuous Bayes Classifier module is used to obtain the detection and diagnosis results of the continuous Bayes Classifier protection layer failure by utilizing process monitoring data;

[0126] The protection layer failure determination module is used to integrate modules 7 and 8 to obtain the final protection layer failure detection and diagnosis results.

[0127] Another object of the present invention is to provide a computer device including a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the steps of the method for detecting the failure of the protective layer of an oil refining unit based on multi-source data fusion.

[0128] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method for detecting the failure of the protective layer of an oil refining unit based on multi-source data fusion.

[0129] Another objective of this invention is to provide an information data processing terminal for implementing the oil refining unit protective layer failure detection system based on multi-source data fusion.

[0130] Specific implementation of the present invention:

[0131] This invention provides a method for detecting and diagnosing protective layer failures in petroleum refining units based on multi-source data fusion, including:

[0132] Step 1: Determine the sliding window length and sampling interval. Using the specific petroleum refining unit simulation software, determine the sliding window length and sampling interval based on the data update frequency. The window length indicates how many data samples are included in the process monitoring data matrix, and the sampling interval indicates how often the data acquisition script executes the data acquisition operation.

[0133] Step 2: Collect process monitoring data and update the process monitoring data matrix based on the sampling step size. Taking the sampling results of some control variables in the protection layer of the basic control process system of the reaction regeneration module in an oil refining unit as an example, the process monitoring data collected in this step is as follows: Each column of data represents: the main riser outlet temperature measuring point TI1149 (degrees Celsius), the main riser outlet temperature measuring point TI1150 (degrees Celsius), the main riser outlet temperature measuring point TI1151 (degrees Celsius), and the regeneration slide valve opening TV1101 (percentage). The five rows of data represent a sliding window size of 5. This is only used here to help those skilled in the art better understand the embodiments of the present invention and is not a fixed sliding window size or variable set. In practical applications, the sliding window size will be much larger than 5. During process data acquisition, all controlled variables in the control loop should be collected as much as possible for screening.

[0134] Based on the above data update method, the process monitoring data matrix after the 45th update is as follows: ,in Rows 1 to 5 contain the data obtained from the 221st to the 225th sampling.

[0135] Step 3: Perform column vector normalization on the process monitoring data. In this step, the column vector is normalized using the formula: Determine the normalized process monitoring data training matrix, in order to For example, after normalization, it becomes: .

[0136] Step 4: Generate statistical alarm information using normalized process monitoring data. In this step, the statistical alarm information data is generated using the formula: Generate; where For the first The standard deviation of each variable As an alarm level constant; Taking the generated statistical alarm data as an example, the resulting statistical alarm information matrix is ​​as follows: ,in The value is 3, here. Values ​​that are appropriate for practical applications.

[0137] Step 5: Filter the variable set according to the maximum correlation-minimum redundancy index. In this step, the maximum correlation index is calculated using the following formula: Determine the set most relevant to the label ,in Since the joint probability density of continuous variables is usually difficult to obtain, the above equation is derived from formula... Approximate calculation; the minimum redundancy index is calculated using the formula: Identify the element with the lowest redundancy within the set; combine the most relevant metric to the label and the metric with the lowest redundancy within the set, based on... Determine the feature set for training .

[0138] Step 6: Using statistical alarm information data, a discrete Bayesian classifier is used to obtain the protection layer failure detection and diagnosis results. In this step, the discrete Bayesian classifier for statistical alarm information uses the statistical alarm information to train a matrix and calculates the probability of each type of protection layer failure: Therefore, based on the formula: where This yields the case with the highest probability of protective layer failure. Here This is a type of protective layer failure condition, such as "Basic control process protective layer failure - main riser pipe regeneration slide valve fixed at a large opening" and "Basic control process protective layer failure - smoke machine fixed at a small speed".

[0139] Step 7: Using process monitoring data, a continuous Bayesian classifier is used to obtain the protection layer failure detection and diagnosis result F2. In this step, a matrix is ​​trained using process monitoring data to calculate the probability of each type of protection layer failure: Due to reality Since this information is often unavailable, this step uses a multivariate kernel density estimation method to approximate it. ,use replace ,in Where K is the kernel function, and this implementation adopts the multivariate normal kernel function, i.e.: ; It is a symmetric positive definite matrix: ;get

[0140] Finally, the formula is used: The classification results of the continuous Bayesian classifier for the process monitoring data are obtained. Therefore, based on the formula: This yields the case with the highest probability of protective layer failure. , here This is a type of protective layer failure condition, such as "Basic control process protective layer failure - main riser pipe regeneration slide valve fixed at a large opening" and "Basic control process protective layer failure - smoke machine fixed at a small speed".

[0141] Step 8: Combine the protective layer failure detection and diagnosis results obtained in Step 6. The protective layer failure detection and diagnosis results obtained in step 7 The final protective layer failure detection and diagnosis results are obtained. In this step, based on the formula:

[0142] Obtain the final protective layer failure detection results Furthermore, here For the search space of protective layer failure detection and diagnosis, for One possible value state is: "Basic control process protection layer failure - main riser regeneration slide valve fixed at a large opening", "Basic control process protection layer failure - smoke machine fixed at a small speed", etc.

[0143] like Figure 2 As shown, this embodiment of the invention provides a system for detecting and diagnosing the failure of the protective layer of an oil refining unit based on multi-source data fusion, comprising:

[0144] Module 1, Parameter Setting Module, is used to determine parameters such as sliding window length and sampling interval.

[0145] Module 2, the data acquisition module, is used to acquire the values ​​of control variables for each control loop from the petroleum refining unit simulation software.

[0146] Module 3, the data update module, is used to update the process monitoring data training matrix based on the sampling interval to achieve the dynamic effect of the system.

[0147] Module 4, the normalization module, is used to perform column vector normalization on process monitoring data through calculation.

[0148] Module 5, Statistical Alarm Information Calculation Module, is used to train a matrix using normalized process monitoring data to generate statistical alarm information data.

[0149] Module 6, Variable Selection Module, is used to select the set of variables most relevant to the detection and diagnosis of protective layer failure based on the maximum relevance-minimum redundancy index.

[0150] Module 7, Discrete Bayes Classifier Module, is used to obtain the detection and diagnosis results of the Discrete Bayes Classifier protection layer failure by utilizing statistical alarm information data.

[0151] Module 8, Continuous Bayes Classifier Module, is used to obtain the failure detection and diagnosis results of the continuous Bayes classifier protection layer using process monitoring data.

[0152] Module 9, the protection layer failure determination module, is used to integrate Module 7 and Module 8 to obtain the final protection layer failure detection and diagnosis results.

[0153] In addition, module 3, the parameter update module, is specifically used to: determine the process monitoring data for training as follows: Where n is the total number of process monitoring variables, and w is the sampling window length; at every sampling interval After updating the process monitoring data once, the process monitoring data matrix after k updates is as follows: ;

[0154] To enable those skilled in the art to better understand the present invention, an example of process monitoring data for a single variable is given below. Here, the illustrative embodiments and descriptions of the present invention are only for explaining the invention and are not intended to limit the invention. Figure 3As shown, this embodiment uses the temperature TI1150 of the main riser pipe of the protective layer in the basic control process of the reaction regeneration module of the petroleum refining unit as a sampling example for monitoring variables, with a sliding window size of 4496. At times 11, 1609, and 3006, the regeneration slide valve TV1101 controlling this variable is manually deactivated. Reflected on the process monitoring data curve, the data at these points deviate significantly from the mean. Furthermore, the process monitoring data during these failure times differs significantly from the process monitoring data during normal operation; therefore, this data can be used to detect and diagnose protective layer failure scenarios in petroleum refining units.

[0155] In addition, module 4, the normalization module, is specifically used for: using formulas Determine the training matrix for normalized process monitoring data .

[0156] To enable those skilled in the art to better understand the present invention, an example of normalizing process monitoring data for a single variable is given below. Here, the illustrative embodiments and descriptions of the present invention are only for explaining the invention and are not intended to limit the invention. Figure 4 As shown, this embodiment normalizes the process monitoring data values ​​of the TI1150 into a distribution centered at 0. (Comparison) Figure 4 and Figure 3 Normalization does not lose the abnormal performance of process monitoring data within the three protection layer failure times, but the normalized data of each process monitoring variable has smaller numerical differences, which will reduce the interference caused by the large numerical differences of different variables due to different units.

[0157] In addition, module 5, the statistical alarm information calculation module, is specifically used for: [using formulas] Generate statistical alarm information data; among which For the first The standard deviation of each variable The alarm level is a constant; the generated statistical alarm information data is as follows: .

[0158] To enable those skilled in the art to better understand the present invention, an embodiment for generating statistical alarm information based on process monitoring data is given below. Here, the illustrative embodiments and descriptions of the present invention are only used to explain the invention and are not intended to limit the invention. Figure 5 As shown, the TI1150 obtains discretized statistical alarm information through function calculation. Compared with process monitoring data, the discretized statistical alarm information does not contain noise data caused by the protection layer failure process, thus mitigating the impact of noise data during the generation and recovery stages of protection layer failure on the detection and diagnosis of protection layer failure.

[0159] In addition, module 6, the variable selection module, is specifically used for: based on formulas: Determine the set most relevant to the label ,in Since the joint probability density of continuous variables is usually difficult to obtain, the above equation is derived from formula... Approximate calculation; the minimum redundancy index is calculated using the formula: Identify the element with the lowest redundancy within the set; combine the most relevant metric to the label and the metric with the lowest redundancy within the set, and utilize... Determine the feature set for training .

[0160] In addition, module 7, the Discrete Bayes classifier module, is specifically used to calculate the probability of each type of protective layer failure. Based on the formula: The case with the highest probability of protective layer failure. , here This is a mapping of a type of protective layer failure condition, such as The protective layer failure condition is described as "Basic control process protective layer failure - main riser regeneration slide valve fixed at a large opening". The protective layer failure condition is described as "Basic control process protective layer failure - the smoke machine is fixed at a relatively low speed," etc.

[0161] In addition, module 8, the continuous Bayesian classifier module, is specifically used to calculate the probability of each type of protective layer failure. Due to reality Since this information is often unavailable, this step uses a multivariate kernel density estimation method to approximate it. ,use replace ,in Where K is the kernel function, and this implementation adopts the multivariate normal kernel function, i.e.: ; It is a symmetric positive definite matrix: ;get The formula for calculating the probability of failure for each type of protective layer is obtained as follows: Based on the formula: This yields the case with the highest probability of protective layer failure. , here This is a mapping of a type of protective layer failure condition, such as The protective layer failure condition is described as "Basic control process protective layer failure - main riser regeneration slide valve fixed at a large opening". The protective layer failure condition is described as "Basic control process protective layer failure - the smoke machine is fixed at a relatively low speed," etc.

[0162] In addition, module 9, the protection layer failure determination module, is specifically used for: based on the formula: Obtain the final protective layer failure detection results , here This is a mapping of a type of protective layer failure condition, such as For the protective layer failure condition "Basic control process protective layer failure - main riser regeneration slide valve fixed at a large opening", such as The protective layer failure condition is described as "Basic control process protective layer failure - the smoke machine is fixed at a relatively low speed," etc.

[0163] Based on the above methods, the present invention also provides a program, a computer device, and a non-transient computer-readable storage medium storing computer instructions for executing the above methods, based on a multi-source data fusion method for detecting and diagnosing the failure of the protective layer of an oil refining unit. These will not be elaborated upon here.

[0164] I. Specific application areas or related products of this invention.

[0165] Example 1: Failure Analysis of Protective Layer of Temperature Control Module in Reaction Regeneration Unit of Petroleum Refining Plant

[0166] 1. System Overview

[0167] The reaction regeneration unit is the core component of residue catalytic cracking, and its stable operation directly affects product quality and yield. Temperature is a crucial control variable in the reaction regeneration unit. The task of the temperature control module is to maintain a stable reaction temperature.

[0168] 2. Protective Layer Model Structure

[0169] The temperature-related properties of each protective layer are as follows:

[0170] The basic process control system layer includes the temperature control loop of the horizontal section of the main riser, the temperature control loop of the horizontal section of the auxiliary riser, and the control loop of the slide valve of the external heat return pipe.

[0171] The alarm and personnel intervention layer functions as over-temperature alarms and under-temperature alarms;

[0172] The safety instrumented system layer is designed to trigger interlocking conditions and perform critical variable recovery upon receiving a temperature alarm signal.

[0173] 3. Types of protective layer failure

[0174] The most widespread failure types are found at the basic process control system level, including failures of specific controllers, actuators, and sensors, which cause critical control variables to exceed safety thresholds.

[0175] There are two main types of failures at the alarm and personnel intervention level: false alarms and missed alarms. A false alarm means that the basic process control system level has not failed, but the alarm and personnel intervention level has triggered an alarm. A missed alarm means that the basic process control system level has failed, but the alarm and personnel intervention level has not triggered an alarm.

[0176] The main types of safety instrumented system (SAS) level failures include: cascading trigger failures, cascading action execution failures, and cascading reset failures. A cascading trigger failure refers to the SAS level failing to effectively activate a safety cascading mechanism after an alarm is triggered; a cascading action execution failure refers to the SAS level failing to effectively execute a safety cascading action; and a cascading reset failure refers to the SAS level failing to restore critical variables to normal even after executing a safety cascading action.

[0177] II. Evidence related to the technical effects obtained by the embodiments of the present invention.

[0178] Example 1: Failure Analysis of Protective Layer of Temperature Control Module in Reaction Regeneration Unit of Petroleum Refining Plant

[0179] Figure 3 The process monitoring value of the main riser outlet temperature TI1150 was used as the control variable. The regeneration slide valve TV1101 was manually set to fail at times 11, 1609, and 3306 using a simulation system, resulting in process monitoring data curves. The protective layer failure detection method for petroleum refining units, based on multi-source data fusion, was used in conjunction with the variable selection of the regeneration slide valve TV1101 opening and multiple measuring point temperature values ​​to detect protective layer failure. Figure 4 The results show that normalization can amplify the numerical differences between the failed state and the normal state. Figure 5 The discrete statistical alarm information displayed lacks noise data generated during failure state transitions. Finally, by fusing the results of two Bayesian classifiers, two types of failures were identified: "Basic control process protection layer failure—main riser regeneration valve fixed at a large opening" and "Basic control process protection layer failure—main riser regeneration valve fixed at a small opening." Furthermore, since the TI1150 automatically returns to its normal temperature range after each failure, the classifier can determine whether the alarm and personnel intervention layer and the safety instrumented system layer have failed.

[0180] It should be noted that embodiments of the present invention can be implemented in hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuitry such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., or by software executed by various types of processors, or by a combination of the above-described hardware circuitry and software, such as firmware.

[0181] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for detecting failure of a protective layer of a petroleum refining device based on multi-source data fusion, characterized in that, The method comprises the following steps: Step 1, determining the length of a sliding window and a sampling interval; Step 2, collecting process monitoring data and updating the process monitoring data matrix based on a step; Step 3, performing column vector normalization on the process monitoring data; Step 4, generating statistical alarm information through the normalized process monitoring data; Step 5, screening a variable set according to a maximum correlation-minimum redundancy index; Step 6, Using statistical alarm information data, using a discrete Bayesian classifier to obtain a protective layer failure detection, diagnosis result c a ; Step 7, Using process monitoring data, use a continuous Bayesian classifier to obtain protection layer failure detection, diagnostic results c x ; Step 8, fusion c a and c x , resulting in final protective layer failure detection, diagnostic results c f ; The process monitoring data is collected based on the length of the sliding window and the sampling interval, comprising: The process monitoring data training matrix for determining the initial sampling time is: X0∈R n×w ; wherein n is the total number of process monitoring variables, and w is the length of the sampling window. After k times of sampling, the process monitoring data training matrix is updated as: X kΔt ; wherein Δt is the sampling interval; The process monitoring data training matrix is subjected to centering processing to generate a normalized process monitoring data training matrix, comprising: The normalized process monitoring data training matrix is determined by the formula: ​ The statistical alarm information data is generated according to the normalized process monitoring data by a formula: generated; wherein σ j is the standard deviation of the jth variable, and p is an alert level constant, the larger the p, the higher the alert level, that is, the greater the degree of deviation of the current operating state of the petroleum refining device from the normal state. The statistical alarm information data training matrix generated after k times sampling of the process monitoring data is: A kΔt ∈R n×w ; The discrete Bayesian classifier is according to a formula: calculating the failure probability of each type of protective layer; thus, based on the formula: obtaining the type of situation c with the highest protective layer failure probability a ; The continuous Bayesian classifier is according to a formula: calculating a probability of failure of each type of protective layer; P(x(t) | c) is approximated by a multivariate kernel density estimation method, using k P(x(t) | c) is approximated by a multivariate kernel density estimation method, using P(x(t) | c) is approximated by a multivariate kernel density estimation method, using k P(x(t) | c) is approximated by a multivariate kernel density estimation method, using where K is a kernel function, and the present embodiment adopts a multivariate normal kernel function, i.e.: H is a symmetric positive definite matrix: obtained Thus, Thus, based on the formula: The case c with the highest probability of failure of the protective layer is obtained x ; The results of the discrete Bayesian classifier and the continuous Bayesian classifier are fused based on a formula: Obtaining final protective layer failure detection result c f .

2. The method according to claim 1, wherein the method further comprises: determining the failure of the protective layer of the oil refining plant based on the fusion result of the multi-source data. According to the maximum correlation-minimum redundancy guiding principle, the variable selection is performed based on the normalized process monitoring data training matrix, comprising: The number m of features to be retained in the target set is determined; Based on the formula determining the most relevant set S to the tag m wherein The above formula is derived from the formula Approximate calculation; Using the formula determining the element with the least degree of internal redundancy within the set; The index most relevant to the label and the index with the minimum internal redundancy are combined to select the variable set by determining a set of features S for training m .

3. A multi-source data fusion based petrochemical plant protective layer failure detection system for implementing the multi-source data fusion based petrochemical plant protective layer failure detection method according to any one of claims 1-2, characterized in that, The petroleum refining device protection layer failure detection system based on multi-source data fusion comprises: A parameter setting module is configured to determine parameters such as the length of a sliding window and a sampling interval; A data collection module is configured to collect values of control variables of each control loop from a petroleum refining device simulation software; A data updating module is configured to update the process monitoring data training matrix based on the sampling interval to realize the dynamic effect of the system; A normalization processing module is configured to perform column vector normalization on the process monitoring data by a formula; A statistical alarm information calculation module is configured to generate statistical alarm information data by using the normalized process monitoring data training matrix; A variable selection module is configured to select a variable set most relevant to protection layer failure detection and diagnosis according to a maximum correlation-minimum redundancy index; A discrete Bayesian classifier module is configured to obtain a discrete Bayesian classifier protection layer failure detection and diagnosis result by using the statistical alarm information data; A continuous Bayesian classifier module is configured to obtain a continuous Bayesian classifier protection layer failure detection and diagnosis result by using the process monitoring data; A protection layer failure determination module is configured to fuse the discrete Bayesian classifier module and the continuous Bayesian classifier module to obtain a final protection layer failure detection and diagnosis result.

4. A computer device, comprising: The computer device comprises a memory and a processor, and the memory stores a computer program, and the computer program is executed by the processor to enable the processor to perform the steps of the petroleum refining device protection layer failure detection method based on multi-source data fusion in any one of claims 1-2.

5. A computer readable storage medium storing a computer program, the computer program being executed by a processor to enable the processor to perform the steps of the petroleum refining device protection layer failure detection method based on multi-source data fusion in any one of claims 1-2.

6. An information data processing terminal, characterized by The information data processing terminal is used to implement the petroleum refining device protection layer failure detection system based on multi-source data fusion in claim 3.

Images