Chemical data processing and analysis method and system for chemical production

Abstract

The present invention discloses a chemical data processing and analysis method and system for chemical production, and relates to the field of data processing technology. The method comprises: real-time collection of chemical variables in chemical production; determining a first window step size at the first moment based on the overall change of the chemical variables in each data window in a first time period before the first moment; traversing the chemical variables based on the first window step size and a preset window length to obtain the next data window, and returning to loop execution based on the overall change of the chemical variables in each data window in the first time period before the first moment, determining the first window step size at the first moment until a chemical anomaly condition is met, thereby obtaining multiple data windows; and determining the consistency of each chemical variable with the abnormal change of chemical production, as well as the control degree of each chemical variable on the abnormal change of chemical production, based on the local change of each chemical variable in each data window in a second time period before the chemical anomaly moment.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a chemical data processing and analysis method and system used in chemical production. Background Art

[0002] Safety monitoring in chemical production is crucial, as it effectively prevents accidents and safeguards both personnel and the environment. Chemical production environments are complex, often characterized by high temperatures and pressures, and involve numerous flammable, explosive, toxic, and hazardous substances. Real-time monitoring of chemical variables in chemical production can promptly identify risk signals such as equipment leaks, parameter anomalies, and process deviations, thereby preventing accidents from escalating.

[0003] Currently, a sliding window is used to measure the impact of chemical variables on chemical production anomalies based on the degree of abnormal changes in each chemical variable within the sliding window. However, this method uses only a single, fixed sliding window step size, which cannot accurately capture the actual role of different chemical variables in the chemical production anomaly process, resulting in poor accuracy in assessing the impact of chemical variables on chemical production anomalies. Summary of the Invention

[0004] The embodiments of the present invention provide a chemical data processing and analysis method and system for use in chemical production, which can improve the accuracy of evaluating the impact of chemical variables on chemical production anomalies.

[0005] A first aspect of an embodiment of the present invention provides a chemical data processing and analysis method for chemical production, the method comprising:

[0006] Real-time collection of chemical variables in chemical production, including at least one of temperature, pressure, flow, liquid level and gas concentration;

[0007] Determine a first window step size at the first moment based on the overall change of chemical variables in each data window within a first time period before the first moment, where the first moment is the start moment of the most recent data window;

[0008] Based on the first window step size and the preset window length, the chemical variables are traversed to obtain the next data window, and the loop execution is returned to determine the first window step size at the first moment based on the overall change of the chemical variables in each data window in the first time period before the first moment, until the chemical abnormality condition is met, thereby obtaining multiple data windows;

[0009] Based on the local changes of each chemical variable in each data window in the second time period before the chemical anomaly moment, the consistency of each chemical variable with the abnormal changes in chemical production, as well as the control degree of each chemical variable on the abnormal changes in chemical production are determined respectively. The chemical anomaly moment is the end moment of the data window that meets the chemical anomaly conditions.

[0010] Furthermore, the present application also proposes determining the first window step size at the first moment based on the overall change of the chemical variables in each data window in the first time period before the first moment, including:

[0011] Filtering the data windows to obtain a first data window within a first time period before a first moment;

[0012] Comparing the data distribution probability of the first data window with the corresponding reference distribution probability to determine the overall relative entropy of each first data window;

[0013] A first window step size at a first moment is determined based on the overall relative entropy of each first data window.

[0014] Furthermore, the present application also proposes determining the first window step size at the first moment based on the overall relative entropy of each first data window, including:

[0015] Subtract the overall relative entropy of the previous first data window from the overall relative entropy of each first data window to obtain a plurality of overall relative entropy differences;

[0016] The severity gradient of the problem at the first moment is determined using the number of target overall relative entropy differences and the cumulative value of the target overall relative entropy differences, where the target overall relative entropy difference is an overall relative entropy difference greater than zero;

[0017] A first window step size at the first moment is determined based on the problem severity gradient at the first moment and the target data window corresponding to the first moment.

[0018] Furthermore, the present application also proposes determining the first window step size at the first moment based on the severity gradient of the problem at the first moment and the target data window corresponding to the first moment, including:

[0019] Perform maximum and minimum normalization processing on each chemical variable in the target data window corresponding to the first moment to obtain a normalized variable value;

[0020] Perform mean processing on the values ​​of each normalized variable to obtain the target correction coefficient;

[0021] The preset window step size is corrected by using the problem severity gradient at the first moment, the overall relative entropy of the target data window, and the target correction coefficient to obtain the first window step size at the first moment.

[0022] Furthermore, the present application also proposes that, based on the first window step size and the preset window length, the chemical variables are traversed to obtain the next data window, and the loop execution is returned to determine the first window step size at the first moment based on the overall change of the chemical variables in each data window in the first time period before the first moment, until the chemical abnormality condition is met, and multiple data windows are obtained, including:

[0023] Based on the first window step size and the preset window length, traverse the chemical variables to obtain the next data window;

[0024] The overall production abnormality level of the next data window is determined by using the severity gradient of the problem and the overall relative entropy corresponding to the next data window;

[0025] Determine whether chemical abnormality conditions are met based on the overall production abnormality level;

[0026] When the chemical abnormality condition is not met, the loop is returned to execute based on the overall change of the chemical variables in each data window in the first time period before the first moment, and the first window step size at the first moment is determined until the chemical abnormality condition is met, thereby obtaining multiple data windows.

[0027] Furthermore, the present application also proposes to determine the consistency between each chemical variable and the abnormal change of chemical production, as well as the controllability of each chemical variable on the abnormal change of chemical production, based on the local changes of each chemical variable in each data window in the second time period before the chemical abnormality, including:

[0028] Filtering from each data window a second data window within a second time period before the chemical anomaly occurs; and determining the anomaly manifestation time based on the overall relative entropy of each second data window;

[0029] Compare the overall anomaly curve from the moment the anomaly is revealed to the moment the chemical anomaly occurs with the anomaly curves of each variable to determine the consistency of the abnormal changes of each chemical variable and chemical production. The overall anomaly curve is used to represent the relationship between the overall production anomaly degree of chemical production and time, and the variable anomaly curve is used to represent the relationship between the local production anomaly degree of chemical variables and time.

[0030] Based on the abnormal curves of each chemical variable and the consistency between the abnormal changes of each chemical variable and chemical production, the controllability of each chemical variable on the abnormal changes of chemical production is determined.

[0031] Furthermore, the present application also proposes determining the abnormality manifestation time based on the overall relative entropy of each second data window, including:

[0032] Obtaining a target overall relative entropy corresponding to a target second data window that first includes the second moment, and a target time interval between the start moment of the target second data window and the moment of the chemical anomaly, where the second moment is any moment in the second time period before the moment of the chemical anomaly;

[0033] Using the target overall relative entropy and the target time interval, determine the relative entropy compliance at the second moment;

[0034] Determine the abnormal analysis value at the second moment using the relative entropy compliance, the target overall relative entropy of the target second data window, and the second window step size;

[0035] The second moment with the greatest value in abnormality analysis is determined as the abnormality disclosure moment.

[0036] Furthermore, the present application proposes to compare the overall abnormal curve from the moment the abnormality is revealed to the moment the chemical abnormality occurs with the abnormal curves of each variable to determine the consistency of each chemical variable with the abnormal changes in chemical production, including:

[0037] Based on the dynamic time warping algorithm, the optimal alignment path between the overall anomaly curve and the target variable anomaly curve is determined, where the target variable anomaly curve is any variable anomaly curve;

[0038] Based on the optimal alignment path, the matching point pairs between the overall anomaly curve and the target variable anomaly curve and the corresponding point distances of the matching point pairs are determined;

[0039] The consistency between the target chemical variable and the target abnormal change of chemical production is determined by using the matching point pairs and the corresponding point distances of the matching point pairs. The target chemical variable is the chemical variable corresponding to the target variable abnormal curve.

[0040] Furthermore, the present application also proposes to determine the controllability of each chemical variable on the abnormal change of chemical production based on the abnormal curve of each chemical variable and the consistency of each chemical variable with the abnormal change of chemical production, including:

[0041] Normalizing the degree of each local production anomaly in the target variable anomaly curve to obtain multiple normalized anomaly degrees, the target variable anomaly curve being any variable anomaly curve;

[0042] Compare each normalized abnormality degree with the preset abnormality threshold to determine the target abnormality degree ratio, which is used to represent the ratio of target normalized abnormality degrees greater than the preset abnormality threshold;

[0043] The normalized abnormality degree of each target, the proportion of the target abnormality degree and the consistency of the target abnormal change are used to determine the control degree of the target abnormal change of the target chemical variable on the chemical production. The target chemical variable is the chemical variable corresponding to the target variable abnormal curve, and the consistency of the target abnormal change is the consistency between the target chemical variable and the abnormal change of chemical production.

[0044] A second aspect of an embodiment of the present invention provides a chemical data processing and analysis system for chemical production, the system comprising:

[0045] A variable acquisition module is used to collect chemical variables in chemical production in real time, where the chemical variables include at least one of temperature, pressure, flow, liquid level and gas concentration;

[0046] a step length determination module, configured to determine a first window step length at a first moment based on an overall change of chemical variables in each data window within a first time period before the first moment, wherein the first moment is a start time of a most recent data window;

[0047] a window determination module, configured to traverse the chemical variables based on the first window step size and the preset window length to obtain the next data window, and return to the loop to determine the first window step size at the first moment based on the overall change of the chemical variables in each data window in the first time period before the first moment, until the chemical abnormality condition is met, thereby obtaining multiple data windows;

[0048] The data analysis module is used to determine the consistency of each chemical variable with the abnormal changes in chemical production, as well as the control degree of each chemical variable over the abnormal changes in chemical production based on the local changes of each chemical variable in each data window in the second time period before the chemical abnormality moment. The chemical abnormality moment is the end time of the data window that meets the chemical abnormality conditions.

[0049] The present invention has the following beneficial effects:

[0050] In the chemical data processing and analysis method for chemical production provided by the present invention, after real-time collection of chemical variables, based on the overall change of chemical variables in each data window in the first time period before the first moment, the first window step size at the first moment is dynamically determined, and the window step size can be flexibly adjusted according to the actual change of chemical variables, so that the data window division is more in line with the actual chemical production. Then, based on the window step size and the preset window length, the chemical variables are traversed and the window step size is determined cyclically until the chemical abnormality condition is met, a plurality of data windows are obtained, and data is fully acquired. Finally, based on the local change of each chemical variable in each data window in the second time period before the chemical abnormality moment, the consistency and controllability of each chemical variable with the chemical production abnormality change are determined respectively, and the influence of chemical variables on chemical production abnormality is carefully analyzed from different dimensions. In this way, through this dynamic window step size determination and multi-dimensional analysis method, the actual role of different chemical variables in the chemical production abnormal process can be more accurately reflected, thereby improving the accuracy of chemical variables on chemical production abnormality impact assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 A schematic flow chart of a chemical data processing and analysis method for chemical production provided by one embodiment of the present invention;

[0053] Figure 2 A schematic diagram of the process of S102 provided in one embodiment of the present invention;

[0054] Figure 3 A schematic diagram of the process of S203 provided in one embodiment of the present invention;

[0055] Figure 4 A schematic diagram of the process of S104 provided in one embodiment of the present invention;

[0056] Figure 5 A schematic diagram of an overall abnormal curve and a target variable abnormal curve provided by one embodiment of the present invention;

[0057] Figure 6 The present invention provides a structural diagram of a chemical data processing and analysis system for use in chemical production according to an embodiment of the present invention. DETAILED DESCRIPTION

[0058] To further illustrate the technical means and effectiveness of the present invention in achieving its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail a method and system for chemical data processing and analysis in chemical production, including its specific implementation, structure, features, and effectiveness. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0059] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0060] It should be noted that the acquisition, storage, use, and processing of data in the technical solution of the present invention comply with the relevant provisions of laws and regulations.

[0061] It should be noted that in the embodiments of the present invention, certain software, components, models and other existing solutions in the industry may be mentioned. They should be regarded as exemplary and their purpose is only to illustrate the feasibility of implementing the technical solution of the present invention, but it does not mean that the applicant has or will necessarily use the solution.

[0062] Traditional chemical anomaly monitoring systems use a fixed window step size to analyze chemical variable variation patterns, but this fails to accurately reflect the differentiated characteristics of different variables in dynamic production processes. Because various chemical variables have varying time sensitivities and fluctuation amplitudes during anomaly evolution, a fixed window step size struggles to balance the high-frequency characteristics of rapidly changing parameters with the low-frequency characteristics of slowly changing parameters. This results in temporal and scale mismatches when extracting data distribution features within the window. This fixed window parameter mechanism can cause phase deviations in anomaly characterization curves, reducing the accuracy of localizing key anomaly signals within time series.

[0063] When faced with the above problems, the present application first realized that the fixed window step size leads to distortion in the feature extraction of parameters with different time sensitivity, and the fundamental contradiction is that the unified time scale cannot adapt to the dynamics of differentiated variables. To this end, the present application attempts to construct a dynamic window step size mechanism, and reversely deduce the current optimal analysis step size by examining the overall change law of multiple variables in the historical data window. Among them, the initial scheme uses a single variable change rate weighted calculation step size, but it is found that there is a subjective bias in the weight distribution when multiple variables work together; the improved scheme introduces the difference in data distribution between windows as the basis for step size adjustment, but it cannot distinguish between normal fluctuations and abnormal offsets; the final scheme combines data distribution probability comparison and relative entropy analysis, and by quantifying the overall information change intensity of the window sequence, establishes a dynamic mapping relationship between the step size and the abnormal evolution stage, and realizes abnormal tracking with adaptive window parameters.

[0064] In this regard, Figure 1 As shown, the present application proposes a chemical data processing and analysis method for chemical production, which can be applied to the server, specifically including the following S101 to S104:

[0065] S101, collecting chemical variables in chemical production in real time, where the chemical variables include at least one of temperature, pressure, flow, liquid level, and gas concentration;

[0066] S102, determining a first window step size at a first moment based on overall changes in chemical variables in each data window within a first time period before a first moment, where the first moment is the start moment of a most recent data window;

[0067] S103, based on the first window step size and the preset window length, traverse the chemical variables to obtain the next data window, and return to the loop to determine the first window step size at the first moment based on the overall change of the chemical variables in each data window in the first time period before the first moment, until the chemical abnormality condition is met, thereby obtaining multiple data windows;

[0068] S104, based on the local changes of each chemical variable in each data window in the second time period before the chemical anomaly moment, determine the consistency of each chemical variable with the abnormal changes in chemical production, as well as the control degree of each chemical variable on the abnormal changes in chemical production. The chemical anomaly moment is the end moment of the data window that meets the chemical anomaly condition.

[0069] In this embodiment, the data window refers to a fixed-length time period divided after continuous time series data collection of chemical variables. Specifically, it can be implemented using sliding window technology to dynamically segment the data stream to analyze the variable change trends in different time periods.

[0070] The window step refers to the time interval between adjacent data windows. It can be implemented using a dynamic adjustment algorithm. By analyzing the overall change trend of the historical data window in real time, the starting position of the next window is adaptively determined to capture abnormal characteristics at different stages.

[0071] Chemical abnormal conditions refer to the judgment criteria that trigger the stop of data window division. Specifically, they can be implemented by using preset abnormal thresholds or machine learning model classification results to identify irreversible abnormal conditions in the production process.

[0072] The overall change refers to the comprehensive fluctuation characteristics under the joint action of multiple chemical variables in the data window. It can be achieved by multivariate time series statistical analysis or information entropy calculation to quantify the overall stability of the production system.

[0073] Local variation refers to the individual fluctuation characteristics of a single chemical variable in a continuous data window. It can be achieved by using dynamic time warping algorithm or local outlier factor detection to extract variable-level abnormal patterns.

[0074] Abnormal change consistency refers to the degree of match between a single chemical variable and the overall production abnormality trend. It can be achieved by time series similarity calculation or correlation analysis to evaluate the correlation between variables and system abnormalities.

[0075] The abnormal change controllability refers to the influence weight of a single chemical variable on the abnormal development. It can be implemented by using a causal reasoning model or a contribution decomposition algorithm to quantify the driving role of the variable in the abnormal evolution.

[0076] The core innovation of this application lies in dynamically adjusting the data window step size and combining it with a multi-level analysis of the overall state of the chemical production system and the local changes of variables to solve the problem of abnormal assessment bias caused by traditional fixed window step size, thereby improving the accuracy of identifying the degree of impact of chemical variables on production anomalies.

[0077] As an example, temperature sensors, pressure sensors, flow sensors, liquid level sensors, and gas concentration sensors are installed at key locations in chemical production equipment. For example, temperature sensors are installed on reactors to measure reaction temperature, pressure sensors and flow sensors are installed on pipelines to measure the pressure and flow of the medium, liquid level sensors are installed on storage tanks to measure the liquid level, and gas concentration sensors are installed at gas outlets or reaction areas to measure the concentration of specific gases. These sensors convert the collected analog signals into digital signals and transmit the data to the server via wired or wireless communication. The server is responsible for receiving this data and performing preliminary preprocessing, such as data filtering to remove noise and data verification to ensure data accuracy and integrity.

[0078] The server then retrieves all data windows within a first time period (e.g., the past 10 minutes) before the first moment. Each data window contains data for all chemical variables collected within that time interval. For example, as described in Formula 1, a calculation formula for a data window is provided:

[0079] Formula 1

[0080] In formula 1, It is used to represent the data window with the end time t, s is used to represent the length of the corresponding data window, and m is used to represent a total of m chemical variables. Used to represent the data of the first chemical variable at time t-s+1, that is, the first data of the first chemical variable in this data window.

[0081] Then, for each data window, the statistical characteristics of each chemical variable, such as mean, variance, maximum, and minimum, are calculated. For example, the variance of each chemical variable in the data window is calculated, and then the average of the variances of each chemical variable is calculated to reflect the overall level and fluctuation of the chemical variable within that window. The changes in these statistical characteristics between adjacent data windows are then analyzed. The overall trend of chemical variables can be measured by calculating indicators such as the difference and ratio of statistical characteristics between adjacent windows.

[0082] Then, based on the overall change, a preset rule or algorithm is used to determine the first window step size at the first moment. For example, if the overall change is drastic (e.g., the difference in statistical features between adjacent windows exceeds a certain threshold), a smaller window step size is selected to more accurately capture the change in the chemical variable; if the overall change is relatively stable, a larger window step size is selected to reduce the amount of data processing. For example, algorithms such as fuzzy logic and neural networks can be used to comprehensively judge and determine the window step size.

[0083] Then, starting from the first moment, the next data window is generated according to the determined first window step size and the preset window length. For example, if the first moment is the current time, the first window step size is 10 seconds, and the preset window length is 1 minute, the time range of the next data window will be the data from the current time, starting 10 seconds (the first window step size) and adding 1 minute (the preset window length). The starting moment of the newly generated data window is then used as the first moment, the first window step size is re-determined, and subsequent data windows are generated.

[0084] During the loop, the system continuously checks whether chemical anomaly conditions have been met. These conditions can be set based on the actual chemical production situation. For example, if the value of a chemical variable exceeds its normal operating range, or if the trends of multiple chemical variables show an abnormal correlation, the chemical anomaly condition is considered met. Once the condition is met, the loop stops, and a series of data windows collected during the chemical production process are obtained.

[0085] Next, determine a second time period before the chemical anomaly (e.g., 30 minutes before the anomaly) and extract data for each chemical variable within each data window within this time period. For each chemical variable, calculate its local variation characteristics within each data window, such as the average rate of change (the ratio of the difference between adjacent data points to the time interval) and the direction of change (increase or decrease). For example, calculate the average rate of change of the temperature variable within each data window to reflect the speed and trend of temperature changes within that window.

[0086] Then, compare and analyze the local variation characteristics of each chemical variable with the typical variation characteristics during chemical production anomalies. Methods such as correlation analysis and pattern matching can be used to measure the degree of consistency between the local variation characteristics of chemical variables and the changes during chemical production anomalies. For example, calculate the correlation coefficient between the local variation characteristics of the temperature variable and the known temperature variation characteristics during chemical production anomalies. A higher correlation coefficient indicates a higher degree of consistency.

[0087] Finally, analyze the interrelationships between chemical variables and their impact on chemical production anomalies. Causal analysis, gray correlation analysis, and other methods can be used to determine the degree to which each chemical variable controls changes in chemical production anomalies. For example, gray correlation analysis can be used to calculate the correlation between each chemical variable and anomaly indicators. The greater the correlation, the greater the degree of control the chemical variable has over changes in chemical production anomalies, indicating that the variable plays a more critical role in the process of chemical production anomalies.

[0088] Furthermore, after obtaining the degree of consistency between each chemical variable and the abnormal change in chemical production, as well as the degree of control over the abnormal change of each chemical variable in chemical production, these can be stored and transmitted to a database. When adjustments to the chemical production process are needed, a query statement can be used to search for the degree of consistency and control over the abnormal change of different chemical variables, and the results can be visualized in the form of a table.

[0089] Through this embodiment, after collecting chemical variables in real time, based on the overall changes of chemical variables in each data window in the first time period before the first moment, the first window step size at the first moment is dynamically determined. The window step size can be flexibly adjusted according to the actual changes of chemical variables, so that the data window division is more in line with the actual chemical production. Then, based on the window step size and the preset window length, the chemical variables are traversed and the window step size is determined cyclically until the chemical anomaly condition is met, multiple data windows are obtained, and data is fully acquired. Finally, based on the local changes of each chemical variable in each data window in the second time period before the chemical anomaly moment, the consistency and controllability of each chemical variable with the chemical production anomaly change are determined respectively, and the impact of chemical variables on chemical production anomaly is carefully analyzed from different dimensions. In this way, through this dynamic window step size determination and multi-dimensional analysis method, the actual role of different chemical variables in the chemical production anomaly process can be more accurately reflected, thereby improving the accuracy of the chemical variable impact assessment on chemical production anomaly.

[0090] In some of the above-mentioned schemes of the present application, in the process of determining the window step size, the screened data window may contain historical data with low correlation with the first moment, causing the comparison result of the data distribution probability and the reference distribution probability to deviate from the actual change, thereby affecting the calculation accuracy of the overall relative entropy and ultimately reducing the adjustment accuracy of the window step size.

[0091] In this regard, Figure 2 As shown, the present application further proposes that S102 may specifically include the following S201 to S203:

[0092] S201, filtering from each data window to obtain a first data window within a first time period before a first moment;

[0093] S202, comparing the data distribution probability of the first data window with the corresponding reference distribution probability to determine the overall relative entropy of each first data window;

[0094] S203: Determine a first window step size at a first moment based on the overall relative entropy of each first data window.

[0095] In this embodiment, when screening the first data window, the historical data window is intercepted based on a preset first time period length. The first time period length can be set based on the dynamic response time or process cycle of the chemical production system, for example, to three times the average cycle of the chemical reaction. The data distribution probability is obtained by statistically analyzing the numerical distribution histogram of each chemical variable in the first data window, and the reference distribution probability is generated using a resampling-based bootstrap method. The overall relative entropy is obtained by calculating the KL divergence between the first data window and the reference distribution, which is used to quantify the overall deviation of the data distribution.

[0096] Specifically, during the screening of first data windows, a set of historical data windows prior to the first moment is intercepted using a preset first time period to ensure that the screened data windows have temporal relevance with the current moment. The probability values ​​of the numerical distribution histogram of each first data window are compared with those of the reference distribution histogram on a per-interval basis. The logarithmically weighted sum of the probability ratios for each interval is calculated to obtain the overall relative entropy value for each first data window. For example, the reference distribution probability is generated using a resampling-based bootstrap method to produce a continuous probability distribution curve. During the comparison process, if the probability of a first data window's distribution occurring in the high-pressure interval is significantly higher than that of the reference distribution, its overall relative entropy value will increase, indicating an abnormal data distribution within that window. Based on the overall relative entropy value sequence of all first data windows, the window step size adjustment range is determined by statistically analyzing the entropy change trend. When the overall relative entropy shows an increasing trend, the window step size is shortened to increase data sampling density; when the overall relative entropy remains stable, the window step size is lengthened to reduce computational overhead. This process dynamically adjusts the window step size to ensure that the data window divisions more closely align with actual production status changes, thereby improving the accuracy of subsequent anomaly detection.

[0097] As an example, first, from each data window, a first data window within a first time period before a first moment is filtered out. For example, the first time period can be set to 30 minutes, and all data windows within the last 30 minutes are filtered out as the first data window.

[0098] The data distribution probability of the first data window is then compared with the corresponding reference distribution probability to determine the overall relative entropy of each first data window. Specifically, the relative entropy can be calculated using the KL divergence. First, the chemical variable data within the first data window is normalized to obtain the data distribution probability. The data distribution under normal operating conditions is then selected as the reference distribution probability. Finally, the KL divergence between the two distributions is calculated, which is the overall relative entropy.

[0099] Finally, based on the overall relative entropy of each first data window, the first window step size at the first moment is determined. Furthermore, a weighted average approach can be employed. Recent data windows are assigned greater weights, while more distant data windows are assigned smaller weights. The overall relative entropy of the weighted average is then calculated. The window step size is then dynamically adjusted based on the overall relative entropy. The greater the relative entropy, the smaller the step size, allowing for more precise capture of abnormal changes.

[0100] This embodiment dynamically adjusts the data window step size based on actual changes in chemical variables, improving the sensitivity and accuracy of anomaly detection. This allows for more timely detection of anomalies in chemical production, effectively preventing safety incidents. Furthermore, this method avoids information loss or redundancy caused by a fixed window step size, improving the efficiency of data analysis.

[0101] In some of the above-mentioned schemes in this application, when evaluating the abnormal change trend by calculating the overall relative entropy difference, the differences in the change rates in different time intervals are not taken into account, resulting in the inability to accurately reflect the dynamic characteristics of the abnormal development during the window step adjustment process, thereby affecting the rationality of the data window division.

[0102] In this regard, Figure 3 As shown, the present application further proposes that S203 may specifically include the following S301 to S303:

[0103] S301, subtracting the overall relative entropy of the previous first data window from the overall relative entropy of each first data window to obtain a plurality of overall relative entropy differences;

[0104] S302, determining the severity gradient of the problem at the first moment using the number of target overall relative entropy differences and the accumulated value of the target overall relative entropy differences, wherein the target overall relative entropy difference is an overall relative entropy difference greater than zero;

[0105] S303 : Determine a first window step size at the first moment based on the problem severity gradient at the first moment and the target data window corresponding to the first moment.

[0106] In this embodiment, the overall relative entropy difference quantifies the rate of anomaly change by comparing the overall relative entropy differences between adjacent data windows. The target overall relative entropy difference only filters out positively changing difference sequences to prevent negative fluctuations from interfering with the gradient calculation. The severity gradient is comprehensively assessed by combining the number of positive differences and their cumulative value, dynamically measuring the persistence of the anomaly's development.

[0107] Specifically, the overall relative entropy of adjacent first data windows is firstly differentially calculated. For example, when the overall relative entropy of a data window is 0.92 and the overall relative entropy of the previous data window is 0.85, the positive difference is 0.07.

[0108] Then, add up all the positive differences and calculate their proportions, and use the following formula 2 to determine the severity of the problem:

[0109] Formula 2

[0110] In formula 2, It is used to represent the severity of the problem at time t, and N is used to represent the total number of overall relative entropy differences. Used to characterize the number of target overall relative entropy differences greater than zero, Used to characterize the target overall relative entropy difference between the overall relative entropy of the z-th first data window and the overall relative entropy of the z-1-th first data window.

[0111] Among them, the greater the proportion of the target overall relative entropy difference and the greater the cumulative value of the target overall relative entropy difference, the more gradually the overall relative entropy increases, that is, the more gradually the problems in chemical production become more serious, and the greater the degree of severity of the problems.

[0112] As an example, the server subtracts the overall relative entropy of the previous first data window from the overall relative entropy of each first data window to obtain multiple overall relative entropy differences. For example, for three consecutive first data windows A, B, and C, the overall relative entropy difference between window A and window B, and the overall relative entropy difference between window B and window C are calculated.

[0113] Then, the severity gradient of the problem at the first moment is determined by the above formula 2 using the number of target overall relative entropy differences and the accumulated value of the target overall relative entropy differences.

[0114] Finally, the first window step size at the first moment is determined based on the problem severity gradient at the first moment and the target data window corresponding to the first moment. This allows the window step size to be dynamically adjusted based on the problem severity gradient, so that a smaller window step size is used when the problem severity is high, allowing for more precise capture of abnormal changes.

[0115] This embodiment dynamically adjusts the data window step size based on the overall changes in chemical variables, improving the sensitivity of detecting chemical production anomalies. Furthermore, by calculating the severity gradient, it can more accurately reflect the abnormal development trends in the chemical production process, helping to promptly identify potential risks and take preventative measures.

[0116] In some of the above-mentioned solutions of the present application, the dimensional differences of different chemical variables are not eliminated when determining the window step size, resulting in insufficient normalization processing, affecting the accuracy of the correction coefficient, and thus reducing the accuracy of the window step size adjustment.

[0117] In this regard, the present application further proposes that S303 may specifically include:

[0118] Perform maximum and minimum normalization processing on each chemical variable in the target data window corresponding to the first moment to obtain a normalized variable value;

[0119] Perform mean processing on the values ​​of each normalized variable to obtain the target correction coefficient;

[0120] The preset window step size is corrected by using the problem severity gradient at the first moment, the overall relative entropy of the target data window, and the target correction coefficient to obtain the first window step size at the first moment.

[0121] In this embodiment, maximum-minimum normalization is performed by subtracting the historical minimum from the current value of each chemical variable and dividing the result by the historical range to eliminate the impact of different dimensions on the data distribution. Mean processing uses arithmetic averaging to aggregate the normalized results of multiple variables into a single coefficient, reflecting the overall volatility of these variables. The target correction coefficient and the severity gradient of the problem form a composite correction factor. The former adjusts the baseline value of the window step size, while the latter controls the amplitude of the step size adjustment in response to abnormal trends.

[0122] As an example, the first window step size at the first moment can be determined by the following formula 3:

[0123] Formula 3

[0124] In formula 3, It is used to represent the first window step size at time t, and T is used to represent the preset window step size. Used to characterize the target correction factor. Used to characterize the severity of the problem at time t, Used to characterize the overall relative entropy of the target data window starting at time t.

[0125] This embodiment dynamically adjusts the data window step size based on the real-time status of chemical production, improving the ability to capture abnormal changes in chemical variables. This allows for a more accurate assessment of the impact of each chemical variable on production anomalies, helping to promptly identify potential safety hazards and improve chemical production safety. Furthermore, by introducing a target correction coefficient, the combined impact of multiple chemical variables is taken into account, making the adjustment of the window step size more reasonable and better adaptable to complex chemical production environments.

[0126] In some of the above-mentioned schemes in this application, the window step size is determined by the overall relative entropy and the data window is dynamically divided. However, after the window step size is dynamically adjusted, it is impossible to effectively judge whether the data window meets the abnormal condition, resulting in a contradiction between the flexibility of window division and the accuracy of anomaly detection, and insufficient accuracy in abnormal condition judgment.

[0127] In this regard, the present application further proposes that S103 may specifically include:

[0128] Based on the first window step size and the preset window length, traverse the chemical variables to obtain the next data window;

[0129] The overall production abnormality level of the next data window is determined by using the severity gradient of the problem and the overall relative entropy corresponding to the next data window;

[0130] Determine whether chemical abnormality conditions are met based on the overall production abnormality level;

[0131] When the chemical abnormality condition is not met, the loop is returned to execute based on the overall change of the chemical variables in each data window in the first time period before the first moment, and the first window step size at the first moment is determined until the chemical abnormality condition is met, thereby obtaining multiple data windows.

[0132] In this embodiment, the overall production abnormality is obtained by weighted calculation of the problem severity gradient and the overall relative entropy. For example, the problem severity gradient weight is 0.6, and the overall relative entropy weight is 0.4. When the weighted result exceeds the preset threshold, the chemical abnormality condition is triggered; the preset window length uses a fixed time span, such as 5 minutes, and is combined with the first window step to achieve window overlapping or non-overlapping partitioning. When traversing chemical variables, the first window step controls the sliding interval of the data window. For example, when the first window step is 3 minutes, the starting time of the next data window is the starting time of the previous window plus 3 minutes. If the overall production abnormality does not exceed the threshold, the first window step continues to be adjusted to generate a new data window.

[0133] Specifically, when using the problem severity gradient and overall relative entropy to calculate the overall production abnormality, for example, if the problem severity gradient is 0.8 and the overall relative entropy is 0.6, the weighted result is 0.8×0.6+0.6×0.4=0.72. If the preset threshold is 0.7, it is determined that the chemical abnormality condition is met and the window division is stopped. If the condition is not met, for example, the weighted result is 0.65, the next data window is generated based on the new first window step size. For example, the first window step size is adjusted to 2 minutes, and the window start time is the previous time plus 2 minutes. The overall relative entropy and problem severity gradient are recalculated. By dynamically adjusting the first window step size and the iterative process of abnormality judgment, the accuracy of abnormal condition judgment is ensured, the window division deviation caused by fixed step size is avoided, and the precise control of abnormal triggering is achieved through the quantitative index of abnormality degree.

[0134] As an example, the server first traverses the chemical variables based on the first window step and the preset window length to obtain the next data window. For example, if the preset window length is 60 minutes and the first window step is 10 minutes, then starting from the start time of the current data window, the server moves back 10 minutes as the start time of the next data window, and then moves back 60 minutes as the end time to obtain the next data window.

[0135] Then, the severity gradient and overall relative entropy of the problem corresponding to the next data window are used to determine the overall production anomaly level for the next data window. Specifically, the severity gradient and overall relative entropy can be assigned weight coefficients, and the overall production anomaly level can be obtained by weighted multiplication.

[0136] Then, based on the overall production abnormality level, it is determined whether the chemical abnormality condition is met. For example, a threshold value can be set, and when the overall production abnormality level exceeds the threshold value, it is determined that the chemical abnormality condition is met.

[0137] Finally, if the chemical anomaly condition is not met, the loop returns to execute, determining the first window step size at the first moment based on the overall changes in the chemical variables in each data window during the first time period before the first moment, until the chemical anomaly condition is met, resulting in multiple data windows. This allows the window step size to be dynamically adjusted until a chemical anomaly is detected.

[0138] This embodiment dynamically adjusts the step size of the data window based on the actual conditions of chemical production, improving the sensitivity of chemical production anomalies. Furthermore, by looping until the chemical anomaly conditions are met, potential chemical production anomalies can be promptly detected, helping to prevent chemical accidents. Furthermore, this method can obtain multiple data windows, providing more comprehensive data support for subsequent chemical variable analysis, thereby improving the accuracy and reliability of chemical production anomaly analysis.

[0139] In some of the above-mentioned schemes of the present application, a method of dynamically adjusting the window step size based on the overall change situation to detect abnormal chemical conditions is proposed. However, when determining the correlation between specific chemical variables and abnormal changes in chemical production, there is a problem of being unable to accurately quantify the extent of the role of different variables in the abnormal development process.

[0140] In this regard, Figure 4 As shown, the present application further proposes that S104 may specifically include the following S401 to S404:

[0141] S401, filtering from each data window to obtain a second data window within a second time period before the chemical abnormality occurs;

[0142] S402, determining the abnormality manifestation time based on the overall relative entropy of each second data window;

[0143] S403: Compare the overall anomaly curve from the time the anomaly is revealed to the time the chemical anomaly occurs with the anomaly curves of each variable to determine the consistency of the abnormal changes of each chemical variable with the chemical production. The overall anomaly curve is used to represent the change relationship between the overall production anomaly degree of chemical production and time, and the variable anomaly curve is used to represent the change relationship between the local production anomaly degree of chemical variables and time.

[0144] S404: Determine the controllability of each chemical variable on the abnormal change of chemical production based on the abnormal curve of each chemical variable and the consistency between each chemical variable and the abnormal change of chemical production.

[0145] In this embodiment, when screening the second data window, it is necessary to trace back to the second time period based on the time of the chemical anomaly and extract all data windows within this interval as the basis for analysis. The determination of the time when the anomaly is revealed depends on the overall relative entropy of each second data window. The key time point in the initial stage of the anomaly is located by calculating the changing trend of the entropy value. The dynamic time warping algorithm is used to compare the overall anomaly curve and the variable anomaly curve, and the curve morphological differences are matched through the optimal alignment path. The controllability is determined by combining the normalization processing results and consistency indicators of the variable anomaly curve to quantify the contribution of the variable to the development of the anomaly.

[0146] Specifically, during the second time period, the overall relative entropy of each second data window is calculated to identify the moment when an anomaly surfaced. This step determines the specific time when the anomaly first appeared by analyzing the mutation point in the relative entropy. Subsequently, from the moment the anomaly surfaced to the moment the chemical anomaly occurred, the overall anomaly curve and the local anomaly curve for each variable are extracted. The overall anomaly curve and the variable anomaly curve are aligned using a dynamic time warping algorithm to generate an optimal matching path. The corresponding distances between matching point pairs are calculated to determine the degree of consistency. For example, when the morphology of the variable anomaly curve is highly similar to the overall anomaly curve, the degree of consistency is high. Furthermore, the local anomaly levels within the variable anomaly curves are normalized, and the percentage of anomaly levels exceeding a preset threshold is calculated. This is then combined with the degree of consistency to calculate the degree of control. The normalization process uses the maximum-minimum method, with the preset threshold set to 0.8. A target anomaly level exceeding 50% indicates that the variable has a significant controllable effect on the development of the anomaly. Thus, by quantifying the degree of consistency and controllability, the actual impact of each variable on the anomaly change is accurately assessed.

[0147] As an example, the server selects data windows from all collected data windows within a second time period before the chemical production anomaly. These data windows are referred to as second data windows. For each second data window, its overall relative entropy is calculated. The overall relative entropy of each second data window is analyzed over time. When the relative entropy begins to increase significantly and exceeds a preset threshold, it is considered that chemical production anomalies have begun to emerge, and this moment is the moment of anomaly emergence.

[0148] Then, using the method for calculating the overall production anomaly level described in the above embodiments of the present invention, the overall production anomaly level is calculated at each time point between the moment the anomaly is revealed and the moment the chemical anomaly occurs. An overall anomaly curve is plotted, which is used to characterize how the overall production anomaly level of chemical production changes over time. Simultaneously, for each chemical variable, using the same method for calculating the overall production anomaly level as described in the above embodiments of the present invention, the local production anomaly level is calculated at each time point between the moment the anomaly is revealed and the moment the chemical anomaly occurs. A variable anomaly curve is plotted, which is used to characterize how the local production anomaly level of that chemical variable changes over time.

[0149] Then, using methods such as correlation coefficient analysis and pattern matching, we calculate the degree of consistency between the overall abnormal curve and the abnormal curves of each variable. For example, we calculate the Pearson correlation coefficient between the two curves. The closer the correlation coefficient is to 1, the higher the degree of consistency, that is, the more consistent the chemical variable is with the abnormal changes in chemical production.

[0150] Next, the variable anomaly curve for each chemical variable is analyzed in depth to extract its characteristics, such as the amplitude, frequency, and duration of the abnormal changes. For example, the maximum deviation value of the variable anomaly curve, the number of anomalies, and the duration of each anomaly are calculated. A controllability calculation model is established by comprehensively considering the consistency between the chemical variable and the abnormal changes in chemical production, as well as the characteristics of the variable anomaly curve. For example, a weighted summation method can be used to weight the consistency and the various eigenvalues ​​of the variable anomaly curve to determine the controllability of each chemical variable over the abnormal changes in chemical production. The weights can be set based on the importance of each factor to the abnormal changes in chemical production. Based on the determined controllability calculation model, the consistency of each chemical variable and the eigenvalues ​​of the variable anomaly curve are substituted into the model to calculate the controllability of each chemical variable over the abnormal changes in chemical production. The higher the controllability, the greater the role played by the chemical variable in the abnormal changes in chemical production.

[0151] Through this embodiment, the impact of each chemical variable on chemical production anomalies can be more accurately evaluated. As a result, key variables can be effectively identified, providing an important basis for the cause analysis and prevention of chemical production anomalies. Specifically, by analyzing the data window in the second time period, early signs of anomalies can be captured. Furthermore, by comparing the overall anomaly curve and the variable anomaly curve, the consistency of each variable with the overall anomaly can be quantified. Finally, combining the consistency and the degree of abnormality of the variable itself, the controllability of each variable can be comprehensively evaluated. This multi-dimensional analysis method improves the accuracy of the assessment of the impact of chemical variables and helps to promptly discover and deal with potential risks in chemical production.

[0152] In some of the aforementioned solutions of this application, when determining the moment of anomaly manifestation based on the overall relative entropy of each second data window, relying solely on the numerical value or changing trend of the overall relative entropy may not accurately identify the initial trigger point of the anomaly. The overall relative entropy at different time points may be affected by noise interference or transient fluctuations in non-critical variables, leading to misjudgment of the moment of anomaly manifestation, which in turn affects the accuracy of subsequent consistency and controllability analyses.

[0153] In this regard, the present application further proposes that S402 may specifically include:

[0154] Obtaining a target overall relative entropy corresponding to a target second data window that first includes the second moment, and a target time interval between the start moment of the target second data window and the moment of the chemical anomaly, where the second moment is any moment in the second time period before the moment of the chemical anomaly;

[0155] Using the target overall relative entropy and the target time interval, determine the relative entropy compliance at the second moment;

[0156] Determine the abnormal analysis value at the second moment using the relative entropy compliance, the target overall relative entropy of the target second data window, and the second window step size;

[0157] The second moment with the greatest value in abnormality analysis is determined as the abnormality disclosure moment.

[0158] In this embodiment, the calculation of relative entropy compliance is combined with the target overall relative entropy and the target time interval, so that the time dimension and the entropy value change form a dual constraint condition, screening out candidate moments with both significant entropy value offset and reasonable time span. The determination of the value of anomaly analysis further integrates the relative entropy compliance, the target overall relative entropy and the second window step to form a multi-dimensional evaluation index to avoid misjudgment caused by a single parameter. The determination of the moment of anomaly exposure adopts the strategy of maximizing the value of anomaly analysis, and achieves accurate positioning of key time points by quantitatively comparing the comprehensive scores of different candidate moments.

[0159] As an example, the server obtains the target overall relative entropy corresponding to the target second data window that first contains the second moment, as well as the target time interval between the start moment of the target second data window and the moment of the chemical anomaly. The second moment is any moment within the second time period before the moment of the chemical anomaly. For example, if the moment of the chemical anomaly is 10:00:00 on May 1, 2023, and the second time period is 2 hours, then the second moment can be any moment between 08:00:00 and 10:00:00 on May 1, 2023.

[0160] Then, the relative entropy compliance at the second moment is determined using the target overall relative entropy and the target time interval. Specifically, the relative entropy compliance can be determined using the following formula 4:

[0161] Formula 4

[0162] In formula 4, Used to characterize the relative entropy compliance at time t, It is used to characterize the overall relative entropy of the target corresponding to the second data window of the target that first contains time t, This is used to characterize the target time interval between the start of the target second data window and the moment of the chemical anomaly. It should be noted that to ensure meaningful calculation results, when performing fractional operations, when the denominator is zero, a parameter adjustment factor greater than zero must be added to the denominator to prevent the denominator from being zero. The value of the parameter adjustment factor is set by the implementer based on actual conditions and is not specifically limited by this application.

[0163] Among them, the smaller the target overall relative entropy and the smaller the target time interval between the start time of the target second data window and the chemical anomaly time, the greater the relative entropy compliance.

[0164] Then, the anomaly analysis value at the second moment is determined using the relative entropy compliance, the target overall relative entropy of the target second data window, and the second window step size. Specifically, the anomaly analysis value can be determined using the following formula 5:

[0165] Formula 5

[0166] In formula 5, Used to characterize the abnormal analysis value at time t, Used to characterize the relative entropy compliance at time t, The difference between the target overall relative entropy of the target second data window and the overall relative entropy of the previous second data window is used to characterize the target second data window. The difference between the second window step size of the target second data window and the window step size of the previous second data window is used to represent the target second data window.

[0167] Among them, when the overall relative entropy of the target corresponding to the second data window that first contains the target at the second moment is greater, the overall relative entropy between adjacent data windows increases, and the step size between adjacent data windows increases smaller, the corresponding abnormality analysis value is greater.

[0168] Finally, the second moment with the greatest value for anomaly analysis is determined as the moment of anomaly manifestation. This allows us to find the moment that best reflects the onset of chemical production anomalies.

[0169] Through the above technical solution, the present application can accurately locate the moment of exposure of chemical production anomalies, and provide a reliable time reference point for the subsequent analysis of the relationship between chemical variables and chemical production anomalies. This method comprehensively considers multiple factors such as relative entropy, time interval and window step size, avoids the deviation that may be caused by a single indicator, and improves the accuracy and reliability of the judgment of the moment of abnormal exposure. At the same time, by introducing the comprehensive indicator of abnormal analysis value, it is possible to more comprehensively evaluate the suitability of each moment as the moment of abnormal exposure, thereby selecting the best moment of abnormal exposure. This provides an important time reference for subsequent chemical production abnormality analysis and early warning, which helps to improve the safety and stability of chemical production.

[0170] In some of the above-mentioned solutions of this application, the direct time point matching method based on the overall anomaly curve and the variable anomaly curve cannot effectively handle the time axis scaling problem of different time series, resulting in increased alignment error and affecting the accuracy of consistency calculation.

[0171] In this regard, the present application further proposes that S403 may specifically include:

[0172] Based on the dynamic time warping algorithm, the optimal alignment path between the overall anomaly curve and the target variable anomaly curve is determined, where the target variable anomaly curve is any variable anomaly curve;

[0173] Based on the optimal alignment path, the matching point pairs between the overall anomaly curve and the target variable anomaly curve and the corresponding point distances of the matching point pairs are determined;

[0174] The consistency between the target chemical variable and the target abnormal change of chemical production is determined by using the matching point pairs and the corresponding point distances of the matching point pairs. The target chemical variable is the chemical variable corresponding to the target variable abnormal curve.

[0175] In this embodiment, the dynamic time warping algorithm searches for the optimal alignment path by constructing a cumulative distance matrix, allowing non-equal length mapping of different time series; the optimal alignment path is composed of coordinate points corresponding to the minimum cumulative distance, and each matching point pair contains the corresponding time points of the overall anomaly curve and the target variable anomaly curve; the corresponding point distance is obtained by calculating the difference in the degree of anomaly of the two curves at the matching point, and normalization is used to eliminate dimensional differences.

[0176] Specifically, when there is a time axis offset between the overall anomaly curve and the target variable anomaly curve, the dynamic time warping algorithm generates a curved path to achieve flexible alignment of the time series. First, all possible paths in the cumulative distance matrix are traversed, and the path with the smallest total distance is selected as the optimal alignment path. Each matching point pair records the coordinates of the corresponding time points of the two curves. For example, the mth point of the overall anomaly curve and the nth point of the target variable anomaly curve constitute a matching point pair. The distance between the corresponding points is obtained by calculating the Euclidean distance of the anomaly degree of the two points. All distance values ​​are normalized to the range of 0-1 after the maximum and minimum values. Finally, the consistency between the target chemical variable and the abnormal changes in chemical production is calculated by counting the number of matching point pairs and the average corresponding point distance. For example, when the number of matching point pairs accounts for more than 80% and the average corresponding point distance is less than 0.3, it is judged to be highly consistent. This method effectively reduces the time axis scaling error and improves the accuracy of consistency assessment through the elastic time alignment mechanism.

[0177] For example, Figure 5 , a schematic diagram of an overall anomaly curve and a target variable anomaly curve is provided. Point 511 on the overall anomaly curve 510 and point 521 on the target variable anomaly curve 520 form a matching point pair; point 512 on the overall anomaly curve 510 and point 521 on the target variable anomaly curve 520 form a matching point pair; and point 513 on the overall anomaly curve 510 and point 521 on the target variable anomaly curve 520 form a matching point pair.

[0178] As an example, dynamic time warping is performed on the temperature anomaly curve and the overall anomaly curve during the time interval from the moment the anomaly is revealed to the moment the chemical anomaly occurs. First, the time axes of the two curves are nonlinearly mapped, and the optimal alignment path is searched by constructing a cumulative distance matrix. Matching point pairs are determined by the coordinates of the points along the path. For example, a matching point pair is formed when the 15th minute on the overall anomaly curve corresponds to the 13th minute on the temperature anomaly curve.

[0179] Then, for all matching point pairs, the distance difference between the target variable anomaly curve and the overall anomaly curve is determined by the following formula 6:

[0180] Formula 6

[0181] In formula 6, The distance difference between the abnormal curve of the i-th chemical variable and the overall abnormal curve is used to characterize the distance between the abnormal curve and the overall abnormal curve. The average value of the corresponding point distances used to characterize all matching point pairs, Used to characterize the p point and The corresponding point distance of the matching point pair composed of points, Used to represent the total number of matching point pairs.

[0182] Finally, the consistency between the target chemical variables and the target abnormal changes in chemical production is determined by the following formula 7:

[0183] Formula 7

[0184] In formula 7, The distance difference between the abnormal curve of the i-th chemical variable and the overall abnormal curve is used to characterize the distance between the abnormal curve and the overall abnormal curve. Used to represent the total number of matching point pairs, Used to characterize the consistency between the i-th chemical variable and the target abnormal change of chemical production.

[0185] Through the above technical solution, the present application effectively solves the problem of inaccurate matching of abnormal change patterns caused by fixed window step size, eliminates the phase offset and rate difference between different variables and overall anomalies on the time axis through the dynamic time warping algorithm, and accurately quantifies the time correlation between variable anomalies and system anomalies, thereby avoiding misjudgment or missed detection of abnormal variables caused by deviations in window division methods.

[0186] In some of the above-mentioned solutions of the present application, when determining the abnormal change controllability of chemical variables on chemical production based on a direct comparison between the overall abnormal curve and the variable abnormal curve, the accuracy of the abnormal change controllability is low.

[0187] In this regard, the present application further proposes that S404 may specifically include:

[0188] Normalizing the degree of each local production anomaly in the target variable anomaly curve to obtain multiple normalized anomaly degrees, the target variable anomaly curve being any variable anomaly curve;

[0189] Compare each normalized abnormality degree with the preset abnormality threshold to determine the target abnormality degree ratio, which is used to represent the ratio of target normalized abnormality degrees greater than the preset abnormality threshold;

[0190] The normalized abnormality degree of each target, the proportion of the target abnormality degree and the consistency of the target abnormal change are used to determine the control degree of the target abnormal change of the target chemical variable on the chemical production. The target chemical variable is the chemical variable corresponding to the target variable abnormal curve, and the consistency of the target abnormal change is the consistency between the target chemical variable and the abnormal change of chemical production.

[0191] In this embodiment, assuming that there are n time points in the target variable anomaly curve, the maximum and minimum values ​​of the local production anomaly degrees corresponding to these n time points are found, and then the local production anomaly degrees are normalized according to the maximum and minimum values ​​to obtain multiple normalized anomaly degrees.

[0192] Based on the actual needs and experience of chemical production, an appropriate preset anomaly threshold is set. This threshold is used to determine whether the degree of local production anomaly meets the abnormality standard. For example, the preset anomaly threshold can be set to 0.6. When the normalized anomaly degree is greater than 0.6, the local production anomaly degree at that point in time is considered high. Each normalized anomaly degree is compared with the preset anomaly threshold, and the number of normalized anomaly degrees greater than the preset anomaly threshold is counted to determine the target anomaly degree ratio.

[0193] Finally, the control degree of target abnormal changes of target chemical variables on chemical production is determined by the following formula 8:

[0194] Formula 8

[0195] In formula 8, It is used to characterize the control degree of abnormal changes of chemical production by the i-th chemical variable. Used to represent the target abnormality ratio. Used to characterize the target normalized abnormality degree of the i-th chemical variable at time t, It is used to characterize the number of target normalized abnormalities greater than the preset abnormality threshold. Used to characterize the consistency between the i-th chemical variable and the target abnormal change of chemical production.

[0196] Among them, when the proportion of target abnormality degrees greater than the preset abnormality threshold is greater, and the corresponding target normalized abnormality degree cumulative value is greater, the chemical variable has a greater degree of control over the abnormal changes in chemical production; the greater the consistency between the chemical variable and the target abnormal changes in chemical production, the corresponding chemical variable has a greater degree of control over the abnormal changes in chemical production.

[0197] Through this embodiment, when determining the controllability of target abnormal changes, multiple factors such as the target normalized abnormality degree, the target abnormality degree ratio, and the target abnormal change consistency are comprehensively considered, which can more comprehensively and accurately reflect the impact of the target chemical variable on the abnormal changes in chemical production.

[0198] Based on the chemical data processing and analysis method for chemical production provided by the present invention, the present invention further provides a specific embodiment of a chemical data processing and analysis system for chemical production.

[0199] like Figure 6 As shown, the chemical data processing and analysis system 600 provided by the present invention for use in chemical production includes a variable acquisition module 610 , a step size determination module 620 , a window determination module 630 and a data analysis module 640 .

[0200] Variable acquisition module 610, for real-time acquisition of chemical variables in chemical production, where the chemical variables include at least one of temperature, pressure, flow, liquid level, and gas concentration;

[0201] a step length determination module 620 for determining a first window step length at a first moment based on overall changes in chemical variables in each data window within a first time period before the first moment, where the first moment is the start time of the most recent data window;

[0202] A window determination module 630 is configured to traverse the chemical variables based on the first window step size and a preset window length to obtain a next data window, and then return to loop execution to determine a first window step size at the first moment based on the overall change of the chemical variables in each data window within a first time period before the first moment, until the chemical anomaly condition is met, thereby obtaining multiple data windows;

[0203] The data analysis module 640 is used to determine the consistency of each chemical variable with the abnormal changes in chemical production, as well as the control degree of each chemical variable over the abnormal changes in chemical production based on the local changes of each chemical variable in each data window in the second time period before the chemical abnormality moment. The chemical abnormality moment is the end time of the data window that meets the chemical abnormality conditions.

[0204] In the chemical data processing and analysis system for chemical production provided by the present invention, after real-time collection of chemical variables, based on the overall change of chemical variables in each data window in the first time period before the first moment, the first window step size at the first moment is dynamically determined, and the window step size can be flexibly adjusted according to the actual change of chemical variables, so that the data window division is more in line with the actual chemical production. Then, based on the window step size and the preset window length, the chemical variables are traversed and the window step size is determined cyclically until the chemical abnormality condition is met, a plurality of data windows are obtained, and data is fully acquired. Finally, based on the local change of each chemical variable in each data window in the second time period before the chemical abnormality moment, the consistency and controllability of each chemical variable with the chemical production abnormality change are determined respectively, and the influence of chemical variables on chemical production abnormality is carefully analyzed from different dimensions. In this way, through this dynamic window step size determination and multi-dimensional analysis method, the actual role of different chemical variables in the chemical production abnormality process can be more accurately reflected, thereby improving the accuracy of chemical variables on chemical production abnormality impact assessment.

[0205] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0206] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A chemical data processing and analysis method for chemical production, characterized in that: The method comprises: Real-time collection of chemical variables in chemical production, wherein the chemical variables include at least one of temperature, pressure, flow, liquid level and gas concentration; Determining a first window step size at a first moment based on an overall change of the chemical variables in each data window within a first time period before a first moment, where the first moment is a start moment of a most recent data window; Based on the first window step size and the preset window length, traverse the chemical variables to obtain the next data window, and return to the loop to execute the overall change of the chemical variables in each data window in the first time period before the first moment, determine the first window step size at the first moment, until the chemical abnormality condition is met, and obtain multiple data windows; Based on the local changes of each chemical variable in each data window in the second time period before the chemical anomaly moment, the consistency of each chemical variable with the abnormal change of the chemical production, as well as the control degree of each chemical variable on the abnormal change of the chemical production are determined respectively. The chemical anomaly moment is the end moment of the data window that meets the chemical anomaly condition.

2. The chemical data processing and analysis method for chemical production according to claim 1, characterized in that: Determining the first window step size at the first moment based on the overall change of the chemical variables in each data window within a first time period before the first moment includes: Filtering the data windows to obtain a first data window within a first time period before the first moment; Comparing the data distribution probability of the first data window with the corresponding reference distribution probability to determine the overall relative entropy of each first data window; Based on the overall relative entropy of each of the first data windows, a first window step size at the first moment is determined.

3. The chemical data processing and analysis method for chemical production according to claim 2, characterized in that: The determining, based on the overall relative entropy of each of the first data windows, the first window step size at the first moment includes: subtracting the overall relative entropy of the previous first data window from the overall relative entropy of each first data window to obtain a plurality of overall relative entropy differences; Determining the severity gradient of the problem at the first moment using the number of target overall relative entropy differences and the accumulated value of the target overall relative entropy differences, wherein the target overall relative entropy difference is the overall relative entropy difference greater than zero; A first window step size at the first moment is determined based on the problem severity gradient at the first moment and a target data window corresponding to the first moment.

4. The chemical data processing and analysis method for chemical production according to claim 3, characterized in that: The determining of the first window step size at the first moment based on the problem severity gradient at the first moment and the target data window corresponding to the first moment includes: Performing maximum and minimum value normalization processing on each of the chemical variables in the target data window corresponding to the first moment to obtain a normalized variable value; Performing mean processing on each of the normalized variable values ​​to obtain a target correction coefficient; The preset window step size is corrected by using the problem severity gradient at the first moment, the overall relative entropy of the target data window, and the target correction coefficient to obtain a first window step size at the first moment.

5. The chemical data processing and analysis method for chemical production according to claim 3, characterized in that: Based on the first window step size and the preset window length, the chemical variables are traversed to obtain the next data window, and the loop is returned to execute the overall change of the chemical variables in each data window in the first time period before the first moment to determine the first window step size at the first moment until the chemical abnormality condition is met, thereby obtaining multiple data windows, including: Based on the first window step size and the preset window length, traverse the chemical variables to obtain the next data window; Determine the overall production abnormality level of the next data window by using the problem severity gradient corresponding to the next data window and the overall relative entropy; Determining whether chemical abnormality conditions are met based on the overall production abnormality level; If the chemical abnormality condition is not met, the loop is returned to execute the overall change of the chemical variables in each data window in the first time period before the first moment, and the first window step size at the first moment is determined until the chemical abnormality condition is met, thereby obtaining multiple data windows.

6. The chemical data processing and analysis method for chemical production according to claim 1, characterized in that: The determination of the consistency between each chemical variable and the abnormal change of the chemical production, and the controllability of each chemical variable on the abnormal change of the chemical production based on the local change of each chemical variable in each data window within the second time period before the chemical abnormality occurs, includes: Filtering from each of the data windows to obtain a second data window within a second time period before the chemical abnormality occurs; determining a time when an anomaly is revealed based on the overall relative entropy of each of the second data windows; Compare the overall anomaly curve from the moment the anomaly is revealed to the moment the chemical anomaly occurs with the anomaly curves of each variable to determine the consistency between the abnormal changes of each chemical variable and the chemical production. The overall anomaly curve is used to characterize the relationship between the overall production anomaly degree of the chemical production and time, and the variable anomaly curve is used to characterize the relationship between the local production anomaly degree of the chemical variable and time. Based on the abnormal variable curves of each chemical variable and the consistency between each chemical variable and the abnormal changes of the chemical production, the controllability of each chemical variable on the abnormal changes of the chemical production is determined.

7. The chemical data processing and analysis method for chemical production according to claim 6, characterized in that: The determining of the abnormality manifestation time based on the overall relative entropy of each second data window includes: Obtaining a target overall relative entropy corresponding to a target second data window that first includes a second moment, and a target time interval between the start moment of the target second data window and the moment of the chemical anomaly, where the second moment is any moment in a second time period before the moment of the chemical anomaly; Determining the relative entropy compliance at the second moment using the target overall relative entropy and the target time interval; Determining the anomaly analysis value at the second moment by using the relative entropy compliance, the target overall relative entropy of the target second data window, and the second window step size; The second moment at which the abnormality analysis value is the greatest is determined as the abnormality disclosure moment.

8. The chemical data processing and analysis method for chemical production according to claim 6, characterized in that: The step of comparing the overall abnormality curve from the time when the abnormality is revealed to the time when the chemical abnormality occurs with the abnormality curves of each variable to determine the consistency between each chemical variable and the abnormal change of the chemical production includes: Determining an optimal alignment path between the overall anomaly curve and a target variable anomaly curve based on a dynamic time warping algorithm, where the target variable anomaly curve is any one of the variable anomaly curves; Based on the optimal alignment path, determining matching point pairs between the overall anomaly curve and the target variable anomaly curve and corresponding point distances between the matching point pairs; The matching point pairs and the distances between corresponding points of the matching point pairs are used to determine the consistency between the target chemical variable and the target abnormal change of the chemical production, wherein the target chemical variable is the chemical variable corresponding to the target variable abnormal curve.

9. The chemical data processing and analysis method for chemical production according to claim 6, characterized in that: Determining the controllability of each chemical variable on the abnormal change of the chemical production based on the abnormal curve of each chemical variable and the consistency between each chemical variable and the abnormal change of the chemical production includes: Normalizing each of the local production abnormality degrees in the target variable abnormality curve to obtain a plurality of normalized abnormality degrees, wherein the target variable abnormality curve is any one of the variable abnormality curves; Comparing each of the normalized abnormality levels with a preset abnormality threshold value to determine a target abnormality level ratio, where the target abnormality level ratio is used to represent the ratio of target normalized abnormality levels greater than the preset abnormality threshold value; The target abnormality change control degree of the target chemical variable on the target abnormality change of the chemical production is determined by using the normalized abnormality degree of each target, the proportion of the target abnormality degree and the consistency of the target abnormality change. The target chemical variable is the chemical variable corresponding to the target variable abnormality curve, and the target abnormality change consistency is the consistency between the target chemical variable and the abnormal change of the chemical production.

10. A chemical data processing and analysis system for chemical production, characterized in that: The system comprises: A variable acquisition module is used to collect chemical variables in chemical production in real time, wherein the chemical variables include at least one of temperature, pressure, flow, liquid level and gas concentration; a step length determination module, configured to determine a first window step length at a first moment based on an overall change of the chemical variables in each data window within a first time period before the first moment, where the first moment is the start time of a most recent data window; a window determination module, configured to traverse the chemical variables based on the first window step size and a preset window length to obtain the next data window, and return to loop execution based on the overall change of the chemical variables in each data window within a first time period before the first moment to determine the first window step size at the first moment until a chemical anomaly condition is met, thereby obtaining a plurality of data windows; The data analysis module is used to determine the consistency between each chemical variable and the abnormal change of the chemical production, as well as the control degree of each chemical variable over the abnormal change of the chemical production based on the local change of each chemical variable in each data window within a second time period before the chemical abnormality moment. The chemical abnormality moment is the end moment of the data window that meets the chemical abnormality condition.

Images