Online monitoring and alarming method and system for smoke components in carbonization process

Abstract

The invention relates to the technical field of industrial control, and discloses an online monitoring and alarming method and system for smoke components in the carbonization process, and the method comprises the steps: carrying out the multi-dimensional data cleaning of the smoke component data in the carbonization process, and obtaining the standardized component data; performing time sequence trend analysis on the key smoke component concentration of the standardized component data to obtain a component concentration trend; based on the component concentration trend, performing cooperative dynamic calibration on a historical concentration threshold value in the carbonization process to obtain a target dynamic threshold value; on the basis of the target dynamic threshold value, carrying out situation study and judgment on the key smoke component concentration to obtain a real-time state evaluation result; comprehensively evaluating the abnormal state of the carbonization process based on the real-time state evaluation result to obtain an abnormal state identifier; based on a preset alarm rule knowledge base, performing grade mapping on the abnormal state identifier to obtain a graded alarm instruction of the carbonization process; the online monitoring and alarming efficiency of the smoke components in the carbonization process can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial control, and in particular to a method and system for online monitoring and alarming of carbonization process flue gas components. BACKGROUND

[0002] In the field of online monitoring of carbonization process flue gas components, the existing technology has significant shortcomings in the preprocessing of flue gas component data. Most technologies can only perform simple outlier screening or single-level noise suppression on the data, failing to cover the multi-dimensional data purification needs, resulting in problems such as residual frequency domain interference and random fluctuations in the original data directly passing to the subsequent analysis link, seriously affecting the authenticity and stability of the concentration data. At the same time, due to the lack of a unified dimensionless normalization mechanism, the concentration data of different flue gas components cannot be effectively compared and integrated due to unit differences, further weakening the reference value of the monitoring data and failing to provide a reliable data basis for subsequent concentration trend analysis.

[0003] The adaptability and accuracy of the existing technology in threshold calibration and abnormal alarm are also insufficient. The traditional scheme generally uses a fixed historical concentration threshold as the basis for abnormal judgment, without dynamic adjustment in combination with the real-time time series change trend of the component concentration in the carbonization process, so that the threshold cannot adapt to the concentration changes caused by process fluctuations. When the process is upgraded, abnormal omission is easily caused by a too wide threshold, and when the process is stable, invalid alarms are easily caused by a too strict threshold, greatly reducing the practical value of the monitoring system. In addition, the evaluation of abnormal state is mostly focused on a single concentration deviation index, without systematic analysis of the pattern type, impact degree and duration of the abnormality, resulting in a lack of scientific basis for alarm level classification, difficulty in matching corresponding disposal strategies, and inability to provide accurate guidance for the safety control of the carbonization process, which is not conducive to the stability and safety of industrial production. SUMMARY

[0004] The present application provides a method and system for online monitoring and alarming of carbonization process flue gas components to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides a method for online monitoring and alarming of carbonization process flue gas components, comprising:

[0006] S1. Multi-dimensional data cleaning of the flue gas component data of the carbonization process to obtain standardized component data of the carbonization process;

[0007] S2. Time series trend analysis of the key flue gas component concentration of the standardized component data to obtain the component concentration trend of the carbonization process;

[0008] S3. Based on the component concentration trend, the historical concentration threshold of the carbonization process is dynamically calibrated in a cooperative manner to obtain the target dynamic threshold of the carbonization process;

[0009] S4. Based on the target dynamic threshold, the key flue gas component concentration is analyzed to obtain the real-time state evaluation result of the carbonization process;

[0010] S5. Based on the real-time state evaluation result, the abnormal state of the carbonization process is comprehensively evaluated to obtain the abnormal state identifier of the carbonization process;

[0011] S6. Based on the preset alarm rule knowledge base, the abnormal state identifier is mapped to obtain the graded alarm instruction of the carbonization process.

[0012] In a preferred embodiment, the flue gas component data of the carbonization process is subjected to multi-dimensional data cleaning to obtain the standardized component data of the carbonization process, comprising:

[0013] The flue gas component data is subjected to outlier rejection to obtain the preliminary purified data of the carbonization process;

[0014] The preliminary purified data is subjected to frequency domain noise suppression to obtain the smoothed data of the carbonization process;

[0015] The smoothed data is subjected to dimensionless normalization to obtain the standardized component data of the carbonization process.

[0016] In a preferred embodiment, the key flue gas component concentration of the standardized component data is subjected to time series trend analysis to obtain the component concentration trend of the carbonization process, comprising:

[0017] The key flue gas component concentration is subjected to sliding window segmentation to obtain the concentration subsequence of the carbonization process;

[0018] The concentration subsequence is subjected to change vector analysis to obtain the trend direction of the carbonization process;

[0019] The trend direction is subjected to trend mode identification to obtain the typical trend pattern of the carbonization process;

[0020] The typical trend pattern is subjected to time series coherence fusion to obtain the component concentration trend of the carbonization process.

[0021] In a preferred embodiment, the historical concentration threshold of the carbonization process is dynamically calibrated based on the component concentration trend to obtain the target dynamic threshold of the carbonization process, comprising:

[0022] Based on the component concentration trend, the reference threshold curve of the historical concentration threshold is subjected to trend coordination alignment to obtain the threshold calibration vector of the carbonization process;

[0023] offset the threshold calibration vector to obtain a threshold correction parameter of the carbonization process;

[0024] based on the threshold correction parameter, asymptotically converge the reference threshold curve to obtain a target dynamic threshold of the carbonization process.

[0025] In a preferred embodiment, the threshold correction parameter of the carbonization process is obtained by dynamically offsetting the threshold calibration vector, and the calculation formula of the threshold correction parameter is as follows:

[0026] ;

[0027] wherein, is the threshold correction parameter, is an amplitude adjustment coefficient of the threshold calibration vector, is a direction control factor of the threshold calibration vector, is a preset nonlinear gain coefficient, is a value of the reference threshold curve at time point is a preset historical weight decay coefficient, is a natural constant, is a continuous running time of the carbonization process, is a preset time decay constant, is a preset trend dynamic response coefficient, is a change rate of the amplitude adjustment coefficient, is a direction sign function of the direction control factor.

[0028] In a preferred embodiment, based on the target dynamic threshold, the trend of the key flue gas component concentration is analyzed to obtain a real-time state evaluation result of the carbonization process, including:

[0029] based on the target dynamic threshold, the dynamic gap of the key flue gas component concentration is analyzed to obtain a real-time deviation vector of the carbonization process;

[0030] time domain characteristics are coupled to the real-time deviation vector to obtain a continuous period characteristic of the carbonization process;

[0031] multi-parameter associated characteristics are mined from the real-time deviation vector to obtain a coupling tensor of the carbonization process;

[0032] based on the continuous period characteristic, a state trajectory of the coupling tensor is deduced to obtain a state migration path of the carbonization process;

[0033] ​Performing situation assessment on the state transition path to obtain a real-time state assessment result of the carbonization process.

[0034] In a preferred embodiment, the real-time deviation vector is deconstructed in time domain characteristics to obtain a sustained period characteristic of the carbonization process, including:

[0035] The real-time deviation vector is deconstructed in time domain characteristics to obtain a sustained period characteristic of the carbonization process, including:

[0036] The cumulative effect representation is distributed structure interpreted to obtain an extreme value distribution mode of the carbonization process.

[0037] The extreme value distribution mode is periodically structured to obtain a sustained period characteristic of the carbonization process.

[0038] In a preferred embodiment, based on the real-time state assessment result, the abnormal state of the carbonization process is comprehensively evaluated to obtain an abnormal state identification of the carbonization process, including:

[0039] Based on the real-time state assessment result, the abnormal state of the carbonization process is identified in abnormal mode to obtain an abnormal mode type of the carbonization process.

[0040] The abnormal mode type is quantified in feature parameters to obtain an abnormal degree index and an abnormal duration parameter of the carbonization process.

[0041] The abnormal mode type, the abnormal degree index and the abnormal duration parameter are structured and coded to obtain an abnormal state identification of the carbonization process.

[0042] In a preferred embodiment, based on a preset alarm rule knowledge base, the abnormal state identification is mapped in levels to obtain a hierarchical alarm instruction of the carbonization process, including:

[0043] Based on a preset alarm rule knowledge base, the abnormal state identification is matched in multiple levels to obtain an alarm level of the carbonization process.

[0044] The alarm level is mapped in response strategies to obtain an alarm content and a disposal suggestion of the carbonization process.

[0045] The alarm content, the disposal suggestion and the alarm level are packaged in instructions to obtain a hierarchical alarm instruction of the carbonization process.

[0046] In order to solve the above problems, the present application also provides a carbonization process flue gas component online monitoring and alarm system, the system comprises:

[0047] The data cleaning and standardization module is used for multi-dimensional data cleaning of flue gas component data of the carbonization process, so as to obtain standardization component data of the carbonization process.

[0048] The time sequence trend analysis module is used for time sequence trend analysis of the concentration of the key flue gas component of the standardization component data, so as to obtain the component concentration trend of the carbonization process.

[0049] The threshold value cooperative dynamic calibration module is used for cooperative dynamic calibration of the historical concentration threshold value of the carbonization process based on the component concentration trend, so as to obtain the target dynamic threshold value of the carbonization process.

[0050] The situation research and state evaluation module is used for situation research of the concentration of the key flue gas component based on the target dynamic threshold value, so as to obtain the real-time state evaluation result of the carbonization process.

[0051] The abnormal state evaluation and identification module is used for comprehensive evaluation of the abnormal state of the carbonization process based on the real-time state evaluation result, so as to obtain the abnormal state identification of the carbonization process.

[0052] The grade mapping and alarm instruction module is used for grade mapping of the abnormal state identification based on a preset alarm rule knowledge base, so as to obtain the hierarchical alarm instruction of the carbonization process.

[0053] Compared with the prior art, the present application has the following beneficial effects:

[0054] 1. The present application implements multi-dimensional data cleaning on the flue gas component data of the carbonization process, sequentially completes outlier rejection, frequency domain noise suppression and dimensionless operation, effectively improves the purity and standardization level of the flue gas component data, and provides high-quality data support for subsequent key flue gas component concentration analysis; at the same time, time sequence trend analysis is carried out by means of sliding window segmentation, change vector analysis and trend mode identification, which can accurately capture the dynamic change law of the concentration of the key flue gas component, and in combination with the trend, the historical concentration threshold value is calibrated in a cooperative manner, so that the target dynamic threshold value can closely fit the real-time working condition of the carbonization process, and the accuracy of the flue gas component online monitoring and the reliability of the state evaluation result are significantly improved.

[0055] 2.The application is based on the target dynamic threshold to judge the trend of the key flue gas component concentration, through dynamic gap analysis, time domain feature coupling and multi-parameter correlation feature mining, the real-time running state of the carbonization process can be comprehensively and deeply mastered; for abnormal state, through abnormal mode identification, abnormal degree and time quantization and structured coding to realize comprehensive evaluation, combined with the preset alarm rule knowledge base to complete the grade mapping and hierarchical alarm instruction generation of abnormal state identification, not only can accurately identify the core attributes of abnormal state, but also can output the corresponding disposal suggestion, which provides accurate and feasible guidance for the safety control of carbonization process, and effectively guarantees the stable and safe operation of carbonization process. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 A flowchart of a carbonization process flue gas component online monitoring and alarm method provided by an embodiment of the application is shown in

[0057] Figure 2 A functional module diagram of a carbonization process flue gas component online monitoring and alarm system provided by an embodiment of the application is shown in

[0058] The implementation, functional features and advantages of the application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0059] It should be understood that the specific embodiments described herein are only used to explain the application, and are not used to limit the application.

[0060] The embodiment of the application provides a carbonization process flue gas component online monitoring and alarm method. The execution subject of the carbonization process flue gas component online monitoring and alarm method includes but is not limited to at least one of the electronic devices capable of being configured to execute the method provided by the embodiment of the application, such as a server, a terminal and the like. In other words, the carbonization process flue gas component online monitoring and alarm method can be executed by software or hardware installed in a terminal device or a server device. The server includes but is not limited to a single server, a server cluster, a cloud server or a cloud server cluster and the like. The server can be a stand-alone server, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks (CDN), and big data and artificial intelligence platforms and the like basic cloud computing services.

[0061] Referring to Figure 1 A flowchart of a carbonization process flue gas component online monitoring and alarm method provided by an embodiment of the application is shown in the figure. In this embodiment, the carbonization process flue gas component online monitoring and alarm method includes:

[0062] S1, multi-dimensional data cleaning is performed on the flue gas component data of the carbonization process to obtain standardized component data of the carbonization process;

[0063] In the embodiment of the present application, the multi-dimensional data cleaning on the flue gas component data of the carbonization process to obtain the standardized component data of the carbonization process comprises:

[0064] The flue gas component data is subjected to outlier rejection to obtain preliminary purified data of the carbonization process;

[0065] The preliminary purified data is subjected to frequency domain noise suppression to obtain smoothed data of the carbonization process;

[0066] The smoothed data is subjected to dimensionless normalization to obtain the standardized component data of the carbonization process.

[0067] Specifically, when the flue gas component data of the carbonization process is subjected to outlier rejection, first, based on the process technical requirements of the carbonization process and the historical normal operation stage accumulated flue gas component concentration data, the normal fluctuation range of each flue gas component concentration is determined, which needs to cover the reasonable concentration variation interval allowed by the process, then each data point in the flue gas component data is checked one by one to determine whether the component concentration corresponding to each data point is within the determined normal fluctuation range, if the concentration of a certain data point exceeds the range, the data point is determined as an outlier, then all determined outliers are removed from the original flue gas component data, and the remaining flue gas component data set after removing the outliers is the preliminary purified data of the carbonization process.

[0068] Further, when the preliminary purified data is subjected to frequency domain noise suppression, first, time series analysis is performed on the preliminary purified data to extract the frequency characteristics of the data changing with time, through which the low-frequency signal representing the real concentration change of the flue gas component and the high-frequency noise signal generated by the measurement equipment interference, environmental fluctuation and other factors in the data are distinguished, then the distinguished high-frequency noise signal is processed, the low-frequency signal in the preliminary purified data is retained, and the strength of the high-frequency noise signal is weakened, until the influence of the high-frequency noise signal on the overall trend of the data is minimized, after the high-frequency noise signal processing is completed, the processed time series data is arranged into a new data set, which is the smoothed data of the carbonization process.

[0069] Further, when dimensionally normalizing the smoothed data, first, each key flue gas component contained in the smoothed data is sorted out, and the original dimension of each component concentration data is determined, while the maximum and minimum values of each key flue gas component concentration during normal operation of the carbonization process are collected as reference benchmarks for normalization. Then, for each key flue gas component, each concentration data point of the component is processed one by one, the concentration value of the data point is subtracted from the minimum concentration value of the corresponding component to obtain a concentration difference, and then the concentration difference is divided by the difference between the maximum and minimum concentration values of the corresponding component, so that the concentration data of different dimensions are uniformly converted to the same numerical interval. After the concentration data of all components is converted, the data set formed is the standardized component data of the carbonization process.

[0070] In summary, the core of the abnormal value elimination of the flue gas component data is to determine a reasonable concentration fluctuation range according to the process requirements and historical normal data, remove the abnormal data outside the range through point-by-point checking, and finally obtain the preliminary purified data. This process can effectively eliminate invalid data in the original data caused by abnormal conditions or measurement errors, and provide pure basic data for subsequent data processing.

[0071] In summary, the core of the frequency domain noise suppression of the preliminary purified data is to distinguish the real concentration signal from the interference noise through time series frequency domain analysis, to weaken the high-frequency noise and to retain the low-frequency real signal, and finally to obtain the smoothed data. This process can eliminate the interference of non-target factors on the data, ensure the stability and reliability of the concentration change trend reflected by the data, and provide accurate fluctuation characteristic data for subsequent time series trend analysis.

[0072] In summary, the core of the dimension normalization of the smoothed data is to determine the original dimension of each component and take the historical normal concentration extreme value as the benchmark, and to convert the data of different dimensions to the same interval through a unified calculation method, and finally to obtain the standardized component data. This process can eliminate the influence of dimension difference on the comparison and integration of multi-component data, and provide unified dimension data support for subsequent key component concentration collaborative analysis.

[0073] S2, performing time series trend analysis on the key flue gas component concentration of the standardized component data to obtain a component concentration trend of the carbonization process;

[0074] In the embodiment of the present application, the time series trend analysis on the key flue gas component concentration of the standardized component data to obtain a component concentration trend of the carbonization process comprises:

[0075] performing sliding window segmentation on the key flue gas component concentration to obtain a concentration sub-sequence of the carbonization process;

[0076] performing change vector analysis on the concentration sub-sequence to obtain a trend direction of the carbonization process.

[0077] trend mode recognition is performed on the trend directions to obtain a typical trend pattern of the carbonization process;

[0078] time sequence coherence fusion is performed on the typical trend pattern to obtain a component concentration trend of the carbonization process.

[0079] Specifically, when the key flue gas component concentration is segmented by a sliding window, first, the time length of the sliding window is determined according to the normal change period of the flue gas component concentration in the carbonization process. The time length needs to be able to completely cover a basic fluctuation process of the concentration, while avoiding loss of detailed information due to being too long or producing fragmented data due to being too short. Then, starting from the starting position of the time sequence of the key flue gas component concentration, the window is fixed at the position and all concentration data corresponding to the time points in the window are intercepted to form a first data set. Then, the window is moved at a fixed time interval, and the concentration data in the current window are intercepted after each movement to form a new data set. The above moving and intercepting operations are repeated until the window covers the entire time sequence of the key flue gas component concentration. Each data set obtained by interception is a concentration sub-sequence of the carbonization process.

[0080] Further, when the concentration sub-sequences of the carbonization process are analyzed for change vectors, for each concentration sub-sequence, first, all concentration data arranged in time sequence in the sub-sequence are extracted, and the concentration values corresponding to adjacent two time points are determined. Then, the difference between the concentration value of the latter time point and the concentration value of the former time point is calculated, and the change direction of the concentration is determined by the difference. If the difference is positive, it indicates that the concentration increases in the time period. If the difference is negative, it indicates that the concentration decreases in the time period. If the difference is zero, it indicates that the concentration remains stable in the time period. The concentration of all adjacent time points in each concentration sub-sequence is subjected to the above difference calculation and direction determination, and finally the overall change direction corresponding to each concentration sub-sequence is obtained. The set of these change directions is the trend direction of the carbonization process.

[0081] Further, when identifying the trend mode of the trend direction of the carbonization process, first, all obtained trend directions are collected and classified according to the continuous characteristics of the trend direction. Trend directions that are all rising are classified into one category, defined as an upward trend group. Trend directions that are all falling are classified into one category, defined as a downward trend group. Trend directions that are all stable are classified into one category, defined as a stable trend group. Trend directions that change from rising to falling, from falling to rising, or from stable to rising / falling are classified into one category, defined as a turning trend group. Then, for each trend group, the duration of the trend direction and the cumulative amplitude of the concentration change are extracted to determine the unified characteristic performance of each trend group. Each trend group with unified characteristic performance is defined as a standard mode, and these standard modes are the typical trend modes of the carbonization process.

[0082] Further, when fusing the typical trend modes of the carbonization process in time sequence, first, all typical trend modes are arranged in time sequence according to the time sequence of the key flue gas component concentration, ensuring that the time interval of each typical trend mode accurately corresponds to the time interval of the original concentration data. Then, the time connection between adjacent two typical trend modes is checked. If there is a gap between the time intervals of adjacent two typical trend modes, a transition mode that conforms to the natural change law of the concentration is supplemented according to the ending characteristics of the previous typical trend mode and the starting characteristics of the next typical trend mode, so that adjacent typical trend modes are continuously connected in time. Then, the logical association between all arranged typical trend modes is analyzed to determine whether the previous typical trend mode can reasonably transition to the next typical trend mode. If there is a logical gap, the characteristics of the transition mode are adjusted to ensure that the overall trend conforms to the process rules of the carbonization process. Finally, all time-continuous and logically-associated typical trend modes are integrated into a complete trend sequence, which is the component concentration trend of the carbonization process.

[0083] In summary, the key flue gas component concentration is divided by a sliding window. The core is to determine the window time length according to the normal change period of the flue gas component concentration in the carbonization process. From the starting position of the concentration time sequence, the window is moved at a fixed time interval, and the concentration data in each window is intercepted. Through the interception operation of completely covering the entire time sequence, the continuous concentration time series data is disassembled into multiple independent data sets, and finally the concentration sub-sequences of the carbonization process are obtained. This process provides basic data units for subsequent analysis of concentration change characteristics by unit.

[0084] In summary, the core of vector analysis of concentration subsequences is to extract the concentration data sorted by time for each subsequence, calculate the difference between concentration values ​​at adjacent time points, and determine the trend of concentration change within that time period based on whether the difference is positive, negative, or zero. By integrating the trend of change at all adjacent time points within each subsequence, the overall trend of change corresponding to each concentration subsequence is clarified, and the trend direction of the carbonization process is finally obtained. This process achieves accurate capture of the change characteristics of concentration subsequences.

[0085] In summary, the core of trend modality identification for trend direction is to collect all trend directions and classify them according to their continuous characteristics. Trend directions that continuously show an upward, downward, or stable trend are grouped into corresponding groups, and situations where the trend direction changes are grouped into a turning point group. Then, features such as the duration and cumulative magnitude of concentration change of each trend direction group are extracted to clarify the unified performance. Each trend group with unified characteristics is defined as a standard form, and finally, the typical trend form of the carbonization process is obtained. This process completes the structured classification of trend directions.

[0086] In summary, the core of integrating typical trend patterns in a time-series manner is to arrange all typical trend patterns on a timeline according to the chronological order of the concentrations of key flue gas components, check the temporal connection between adjacent patterns, and supplement transitional patterns that conform to the natural laws of concentration change to eliminate time intervals. At the same time, the transitional features at logical breaks are adjusted to ensure reasonable correlation between patterns. By integrating all typical trend patterns that are continuous in time and logically related, a complete trend sequence is formed, and finally the component concentration trend of the carbonization process is obtained. This process achieves the overall integration of dispersed typical patterns.

[0087] S3. Based on the component concentration trend, perform collaborative dynamic calibration on the historical concentration threshold of the carbonization process to obtain the target dynamic threshold of the carbonization process;

[0088] In this embodiment of the invention, the step of performing collaborative dynamic calibration on the historical concentration threshold of the carbonization process based on the component concentration trend to obtain the target dynamic threshold of the carbonization process includes:

[0089] Based on the component concentration trend, the baseline threshold curve of the historical concentration threshold is trend-co-aligned to obtain the threshold calibration vector of the carbonization process.

[0090] The threshold calibration vector is used to perform offset deduction to obtain the threshold correction parameters for the carbonization process;

[0091] Based on the threshold correction parameter, the baseline threshold curve is asymptotically converged to obtain the target dynamic threshold of the carbonization process.

[0092] The dynamic offset of the threshold calibration vector is derived to obtain the threshold correction parameter of the carbonization process, and the calculation formula of the threshold correction parameter is as follows:

[0093] ;

[0094] In the formula, is the threshold correction parameter, is the amplitude adjustment coefficient of the threshold calibration vector, is the direction control factor of the threshold calibration vector, is a preset nonlinear gain coefficient, is the value of the reference threshold curve at the time point , is a preset historical weight decay coefficient, is a natural constant, is the duration of the carbonization process, is a preset time decay constant, is a preset trend dynamic response coefficient, is the change rate of the amplitude adjustment coefficient, is the direction sign function of the direction control factor.

[0095] Specifically, when the reference threshold curve of the historical concentration threshold is aligned with the trend of the component concentration, the component concentration trend of the key flue gas component in the carbonization process is first obtained, which includes the change direction and amplitude of the key flue gas component concentration corresponding to different time nodes. At the same time, the reference threshold curve in the historical concentration threshold is extracted, which includes the historical concentration threshold standard value corresponding to each time node. Then, the time axis of the component concentration trend is completely corresponding to the time axis of the reference threshold curve, ensuring that the component concentration data and the threshold data of each same time node are matched one by one. Then, according to the rising, falling or stable change rule of the concentration in the component concentration trend, the threshold standard value trend of the corresponding time node of the reference threshold curve is adjusted, so that the overall change trend of the reference threshold curve is consistent with the change rule of the component concentration trend. After the adjustment process, the threshold calibration vector for subsequent threshold correction is generated, which includes the direction and amplitude information of the reference threshold curve adjustment of each time node.

[0096] Further, when the offset of the threshold calibration vector is deduced, the adjustment direction and amplitude of each time node corresponding to the threshold calibration vector are analyzed one by one to determine the offset difference between the reference threshold curve and the component concentration trend at the node. In combination with the actual change characteristics of the key flue gas component concentration in the carbonization process, such as the continuity and relevance of the concentration change, the size of the offset that needs to be further corrected at each time node is determined. For the period of continuous concentration rise, the increasing or decreasing rule of the offset is determined according to the accumulation of the offset difference in the early stage. For the period of stable concentration, the stability of the offset is maintained. The offsets of all time nodes are integrated according to the time sequence and the correlation of the concentration change, and finally the threshold correction parameter that can accurately correct the reference threshold curve is determined. The parameter directly corresponds to the specific value of the reference threshold that needs to be adjusted at each time node.

[0097] Further, when the threshold correction parameter is used to adjust the reference threshold curve, the reference threshold is adjusted for the first time according to the correction value corresponding to the starting time node of the reference threshold curve in the threshold correction parameter. The matching degree of the adjusted threshold at the node and the component concentration trend at the same period is compared. Then, the next time node is entered, and the threshold is continuously adjusted according to the corresponding correction value in the threshold correction parameter on the basis of the previous adjustment, while ensuring that the difference between the adjustment amplitude of this time and the adjustment amplitude of the last time is within a reasonable range to avoid sudden changes in the threshold. The process is repeated to adjust the threshold of each time node in turn until the overall change trend of the reference threshold curve is completely coordinated with the component concentration trend, and the adjustment amplitude of the threshold at the subsequent time node is less than the stability standard. At this time, the threshold curve that can match the change of the flue gas component in the carbonization process in real time is obtained, which is the target dynamic threshold of the carbonization process.

[0098] Specifically, the amplitude adjustment coefficient of the threshold calibration vector is derived from the threshold calibration vector, which is obtained by trend coordination alignment of the reference threshold curve of the historical concentration threshold based on the component concentration trend of the carbonization process. The amplitude adjustment coefficient is a feature parameter extracted from the threshold calibration vector and is specially used to represent the adjustment strength of the threshold calibration vector in the amplitude dimension. The value directly corresponds to the amplitude attribute of the threshold calibration vector, ensuring complete matching with the characteristics of the threshold calibration vector.

[0099] Further, the direction control factor of the threshold calibration vector is also derived from the threshold calibration vector. After the threshold calibration vector is obtained by trend coordination alignment of the reference threshold curve based on the component concentration trend, the feature parameter in the direction dimension is extracted from the threshold calibration vector, which is the direction control factor Its value is used to characterize the directional control property of the threshold calibration vector, and corresponds one-to-one with the directional characteristics of the threshold calibration vector.

[0100] Furthermore, the preset nonlinear gain coefficient The determination is based on the process technology requirements of the carbonization process, historical operating data, and the variation characteristics of flue gas component concentration. Before system deployment, the correlation between flue gas component concentration and threshold calibration under different operating conditions during the carbonization process is analyzed, and the allowable nonlinear adjustment range of the process is combined to pre-set the parameters. The fixed value remains unchanged in subsequent calculations and is used only to enhance the amplitude adjustment coefficient. The nonlinear effect on the threshold correction parameter.

[0101] Furthermore, the baseline threshold curve at time point The value of The baseline threshold curve is derived from historical concentration thresholds. This baseline threshold curve is constructed based on concentration threshold data accumulated during the historical normal operation phase of the carbonization process. When calculating the threshold correction parameters, the duration of the current carbonization process is taken into account. Find the relationship between time and the baseline threshold curve. The exact corresponding concentration threshold value is... This ensures that Ht reflects the actual threshold level of the baseline threshold curve at a specific time point.

[0102] Furthermore, the preset historical weight decay coefficient It is determined by combining the influence of historical data during the carbonization process on the current threshold calibration. By analyzing the reference value of historical concentration thresholds on real-time monitoring results at different operating stages, the weight decay law of historical data over time is determined, and then the threshold is pre-set. A fixed value is used to control the baseline threshold curve over time. The value of The influence weight of the threshold correction parameter remains unchanged during the calculation.

[0103] Furthermore, the duration of the carbonization process This is obtained by recording the start time and current time of the carbonization process in real time. Timing is started from the beginning of the carbonization process, and the runtime is continuously monitored. The time difference between the current time and the start time is the [time value]. Its value is updated in real time as the carbonization process continues to run, ensuring that it can accurately reflect the current operating stage of the carbonization process.

[0104] Furthermore, the preset time decay constant is determined based on the decay law of the concentration threshold value in the carbonization process with the running time, the decay trend of the baseline threshold curve with time is obtained by statistical analysis of historical operation data, the correlation between the decay rate and the running time is analyzed, and the fixed value of is set in advance to control the decay effect of the time factor in the exponential term on the threshold correction parameter, and the value is fixed in the calculation process.

[0105] Further, the preset trend dynamic response coefficient is determined by combining the change speed of the component concentration trend in the carbonization process and the response requirement of the threshold calibration, the coefficient value that can make the threshold calibration quickly adapt to the trend change is determined by analyzing the response effect of the threshold adjustment under different concentration trend change rates, and the fixed value of is set in advance to enhance the dynamic response effect of the change rate of the amplitude adjustment coefficient on the threshold correction parameter, and the value is unchanged in the calculation process.

[0106] Further, the change rate of the amplitude adjustment coefficient is obtained by continuously monitoring the value change of the amplitude adjustment coefficient During the operation of the carbonization process, the real-time value of the amplitude adjustment coefficient is recorded at a fixed time interval, the difference between the values of in the adjacent two time intervals is calculated, and then the difference is divided by the time length of the two time intervals to obtain the change rate of the amplitude adjustment coefficient , which is dynamically updated with the real-time change of , and reflects the change speed of with time.

[0107] Further, the parameter of the direction symbol function of the direction control factor comes from the threshold calibration vector, and the use process is as follows: first, judge the positive and negative of the value of the direction control factor , if the value of is greater than zero, the result of takes 1; if the value of is less than zero, the result of takes -1; if the value of is equal to zero, the result of takes 0, and the final value of the direction symbol function is determined through the judgment process, and then participates in the calculation of the threshold correction parameter.

[0108] Further, the natural constant It is a fixed constant in the field of mathematics. Its value is a fixed number. In the calculation process, this fixed value is directly used without additional calculation or adjustment. It is only used as the base of the exponent operation to participate in the calculation of the exponent term in the formula, ensuring the mathematical rigor of the exponent operation.

[0109] In summary, in the process of obtaining a threshold calibration vector by trend-co-aligning the component concentration trend based on the historical concentration threshold with the baseline threshold curve, the variation patterns of key flue gas components included in the component concentration trend at different stages are first clarified. These patterns need to be combined with the actual data characteristics of online monitoring of flue gas components. Then, the component concentration trend is precisely correlated with the time dimension of the baseline threshold curve to ensure that the two are completely matched at the time nodes. Subsequently, the change direction of the corresponding stage of the baseline threshold curve is adjusted according to the rising, falling, or stable trend of the component concentration trend, so that the overall trend of the baseline threshold curve is coordinated with the component concentration trend. Finally, a threshold calibration vector is generated for subsequent calibration, which can directly serve the dynamic calibration of flue gas component concentration thresholds.

[0110] In summary, in the process of extrapolating the threshold calibration vector to obtain the threshold correction parameters, the adjustment direction and magnitude of the baseline threshold curve corresponding to each stage in the threshold calibration vector are first analyzed. Combining the continuity and correlation of the actual changes in the concentration of key flue gas components during carbonization, the degree of offset that needs further correction between the baseline threshold curve and the component concentration trend at each stage is determined. Then, the offset requirements of each stage are integrated according to the logical relationship of flue gas component changes to determine the specific parameters that can accurately correct the baseline threshold curve. Finally, the threshold correction parameters are obtained, which can be directly used for the precise adjustment of the baseline threshold curve.

[0111] In summary, in the process of asymptotically converging and adjusting the baseline threshold curve based on the threshold correction parameters to obtain the target dynamic threshold, starting from the initial stage of the baseline threshold curve, the threshold is adjusted according to the correction requirements of the corresponding stage in the threshold correction parameters. Then, the baseline thresholds of each subsequent stage are adjusted sequentially to ensure that the magnitude of each adjustment is within a reasonable range to avoid threshold abrupt changes. The adjustment continues until the overall trend of the baseline threshold curve is completely synchronized with the trend of component concentration, and the subsequent adjustment magnitude reaches a stable standard. Finally, the target dynamic threshold that can adapt to the changes in flue gas composition during the carbonization process in real time is obtained. This target dynamic threshold can be directly used for online monitoring and anomaly judgment of flue gas composition during the carbonization process.

[0112] S4. Based on the target dynamic threshold, the concentration of the key flue gas components is assessed to obtain the real-time status evaluation result of the carbonization process.

[0113] In the embodiment of the present application, based on the target dynamic threshold, the trend of the key flue gas component concentration is judged to obtain the real-time state evaluation result of the carbonization process, which includes:

[0114] Based on the target dynamic threshold, the dynamic gap of the key flue gas component concentration is analyzed to obtain the real-time deviation vector of the carbonization process;

[0115] The time domain characteristics of the real-time deviation vector are coupled to obtain the sustained period characteristics of the carbonization process;

[0116] The multi-parameter correlation characteristics of the real-time deviation vector are mined to obtain the coupling tensor of the carbonization process;

[0117] Based on the sustained period characteristics, the state trajectory of the coupling tensor is deduced to obtain the state migration path of the carbonization process;

[0118] The state migration path is evaluated to obtain the real-time state evaluation result of the carbonization process.

[0119] The time domain characteristics of the real-time deviation vector are coupled to obtain the sustained period characteristics of the carbonization process, which includes:

[0120] The time domain trend of the real-time deviation vector is converged to obtain the cumulative effect characteristic quantity of the carbonization process;

[0121] The distribution structure of the cumulative effect characteristic quantity is interpreted to obtain the extreme value distribution mode of the carbonization process;

[0122] The periodic structure of the extreme value distribution mode is analyzed to obtain the sustained period characteristics of the carbonization process.

[0123] Specifically, based on the target dynamic threshold, the dynamic gap of the key flue gas component concentration is analyzed, first, the real-time value of the target dynamic threshold at the current monitoring time is obtained, which is determined by the coordinated dynamic calibration of the historical concentration threshold, which can fit the current working condition of the carbonization process; then the real-time data of the key flue gas component concentration at the same monitoring time is extracted, which comes from the standardized component data after multi-dimensional data cleaning; then the difference between the real-time data of the key flue gas component concentration and the real-time value of the target dynamic threshold is calculated, which is the dynamic gap value at a single monitoring time; according to the order of monitoring time, the dynamic gap values of all monitoring times are arranged in turn to form a vector data structure containing time dimension and gap value dimension, which is the real-time deviation vector of the carbonization process.

[0124] Further, when the real-time deviation vector is time-domain feature coupled, the real-time deviation vector is first unfolded in the order of monitoring time, and the deviation value corresponding to each time point and the change of the deviation value of adjacent time points are combed. Then, the repeatedly appearing deviation change patterns in the real-time deviation vector are identified, for example, continuous rise followed by stability, periodic fluctuation and the like. These patterns are time-domain features. The correlation between different time-domain features is analyzed to determine whether there is feature superposition or mutual influence, for example, a certain fluctuation feature and another rising feature coincide in time and jointly act. The time-domain features with correlation are integrated, and the continuous performance of these features in the time dimension is focused on. The time interval of the repeated features and the change law of the feature intensity are extracted. The collection of these laws is the sustained periodic feature of the carbonization process.

[0125] Further, when the real-time deviation vector is multi-parameter correlation feature mined, first, the deviation data corresponding to all key flue gas components involved in the real-time deviation vector is determined, and the deviation data of each key flue gas component is taken as an independent parameter. Then, the change correlation between different parameters is analyzed, for example, whether the deviation values of other components increase, decrease or remain stable when the deviation value of a certain component increases. By comparing the deviation change trends of different parameters in the same time interval, the positive correlation, negative correlation or no correlation between parameters is determined. Then, all parameter combinations with correlation are collected, and the correlation strength of each combination and the time range of the correlation occurrence are recorded. These correlation relationships are structured and integrated according to the parameter dimension and the time dimension to construct a multi-dimensional data structure containing multi-parameter correlation information. The multi-dimensional data structure is the coupling tensor of the carbonization process.

[0126] Further, based on the sustained periodic feature, when the coupling tensor is state trajectory deduced, the time period of the repeated features and the deviation change law of each parameter in each period are extracted from the sustained periodic feature, and these laws are taken as the time reference and change reference of trajectory deduction. Then, taking the coupling tensor data at the current time as the starting point, the deviation change of each parameter in each subsequent period is calculated in turn according to the time period determined by the sustained periodic feature. The correlation and change amplitude of each parameter in the sustained periodic feature are strictly followed during the calculation process. The coupling tensor data at each calculated time is connected in time sequence to form a path reflecting the change of the coupling tensor with time. At the same time, if the parameter correlation relationship in the coupling tensor is slightly adjusted during the deduction process, the adjustment direction needs to be corrected based on the overall law of the sustained periodic feature to ensure that the path conforms to the process operation logic of the carbonization process. The final path is the state migration path of the carbonization process.

[0127] Further, when performing situation assessment on the state transition path, a standard state transition path library under normal operation of the carbonization process is first constructed, and the standard paths in the path library are formed based on a large amount of historical normal operation data and cover normal state change laws under different working conditions. Then, the current obtained state transition path is compared with the paths in the standard state transition path library segment by segment, the deviation degree of the current path and the standard path at each time node, the time range of the deviation and the number of affected parameters are analyzed. Then, according to the deviation degree, the deviation range and the number of affected parameters, the deviation level of the current state transition path from the normal state is determined, for example, slight deviation, moderate deviation or serious deviation. Finally, combined with the process safety requirements of the carbonization process, the deviation level is further interpreted to judge whether the current state is in the safe operation interval and whether there is a potential abnormal risk. The set of these judgment results is the real-time state evaluation result of the carbonization process.

[0128] Specifically, when performing time domain situation convergence on the real-time deviation vector, the real-time deviation vector is first completely unfolded in the order of monitoring time to clearly show the deviation value corresponding to each time point and the change amplitude of the deviation value of adjacent time points. The real-time deviation vector is derived from the result of dynamic gap analysis of the concentration of the key flue gas component based on the target dynamic threshold, and contains the difference information of the concentration of the key flue gas component and the target dynamic threshold at each monitoring time. Then, the superposition effect of the deviation value in the continuous time interval is analyzed. If the deviation values of continuous multiple time points are all positive or all negative, the cumulative sum of these deviation values is calculated. If the deviation values alternate between positive and negative, the net cumulative result of the deviation values is calculated to reflect the cumulative influence of the deviation in the time dimension. At the same time, the overall trend of the deviation change in the cumulative process is recorded, for example, the cumulative value continuously rises, first rises and then stabilizes, or periodically fluctuates. The comprehensive data structure containing the cumulative sum and the cumulative trend is defined as the cumulative effect representation of the carbonization process.

[0129] Further, when performing distribution structure interpretation on the cumulative effect representation, all the value data in the cumulative effect representation are first extracted, which contains the size and trend information of the deviation accumulation. Then, the maximum cumulative value and the minimum cumulative value are selected from the value data, which are defined as the cumulative maximum value and the cumulative minimum value respectively. Then, the specific time nodes of the cumulative maximum value and the cumulative minimum value in the monitoring time range are counted, and the distribution characteristics of these time nodes are analyzed, for example, whether they are concentrated in a fixed time period, whether they are evenly distributed with running time or whether there is a specific interval rule. Then, the transition data between the cumulative maximum value and the cumulative minimum value is observed to judge whether the change path from the maximum value to the minimum value or vice versa has similarity. The set containing the extreme value position, distribution rule and transition path characteristics is defined as the extreme value distribution mode of the carbonization process.

[0130] Further, when the extreme value distribution pattern is periodically structured, the time intervals of the cumulative extreme maximum or the cumulative extreme minimum in the extreme value distribution pattern are first sorted out, and the time lengths between adjacent two cumulative extreme maximums or between adjacent two cumulative extreme minimums are recorded. Then, the consistency of these time lengths is compared. If most of the time lengths remain consistent or fluctuate within a small range, the time length is preliminarily determined as a potential period. Then, the stability of the potential period is verified. By observing the change characteristics of the extreme values in multiple consecutive periods, such as whether the numerical range of the maximum value, the numerical range of the minimum value and the transition path in each period are consistent, if the characteristics are highly consistent, the potential period is confirmed as a stable period. Finally, the length of the stable period, the change law of the extreme values in each period and the overall trend of the deviation accumulation in the period are integrated to form structured data reflecting the repetition law of the extreme value distribution. The structured data is the duration period characteristic of the carbonization process.

[0131] In summary, the dynamic gap analysis of the concentration of the key flue gas components based on the target dynamic threshold value is to obtain the real-time value of the target dynamic threshold value and the real-time data of the concentration of the key flue gas components at the same period, calculate the difference between the two at each monitoring time, arrange all the difference values in the order of monitoring time to form a vector data structure, and finally obtain the real-time deviation vector of the carbonization process. This process realizes the quantification and time sequencing of the difference between the concentration of the key flue gas components and the dynamic threshold value, and provides basic difference data for subsequent situation judgment.

[0132] In summary, the time domain characteristics of the real-time deviation vector are coupled. The core is to sort out the time sequence deviation values of the real-time deviation vector, identify the repeatedly appearing deviation change patterns and analyze the correlation between the patterns, integrate the patterns with correlation and extract the time interval and intensity change law of the continuous repetition, and finally obtain the duration period characteristic of the carbonization process. This process mines the time sequence law of the deviation data, and provides a time reference and a change reference for subsequent state trajectory deduction.

[0133] In summary, the multi-parameter correlation characteristics of the real-time deviation vector are mined. The core is to take the deviation data of each key flue gas component in the real-time deviation vector as an independent parameter, analyze the change correlation and correlation strength of different parameters in the same time interval, collect the parameter combinations with correlation and structure and integrate them according to the parameter dimension and time dimension, and finally obtain the coupling tensor of the carbonization process. This process integrates the correlation information of multiple parameters, and provides multi-dimensional correlation data for subsequent state transition path deduction.

[0134] In summary, the core of state trajectory extrapolation of coupled tensors based on continuous periodic characteristics is to use the time period and variation law of continuous periodic characteristics as a benchmark, take the current coupled tensor data as the starting point, calculate the deviation changes of parameters in subsequent periods, connect the coupled tensor data at each extrapolation time, and at the same time correct the minor adjustments of parameter correlation to conform to the process logic, and finally obtain the state transition path of the carbonization process. This process realizes the prediction of the trend of coupled tensor changes over time.

[0135] In summary, the core of assessing the state transition path is to compare the current state transition path segment by segment with a standard state transition path library built based on historical normal data, analyze the degree, range, and number of influencing parameters of the path deviation to determine the deviation level, and then combine the process safety requirements to determine the safe range and potential risks of the current state, ultimately obtaining the real-time state assessment results of the carbonization process. This process completes a comprehensive judgment of the operating status of the carbonization process and provides an accurate basis for subsequent anomaly assessment.

[0136] In summary, the core of time-domain situational aggregation of real-time deviation vectors is to unfold the real-time deviation vectors in the order of monitoring time, clarify the deviation values ​​at each time point and the magnitude of adjacent deviation changes, and reflect the cumulative impact of time dimension by calculating the cumulative sum or net cumulative result of deviation values ​​within a continuous time interval. At the same time, the overall trend of deviation accumulation is recorded. The comprehensive data structure containing the cumulative sum and cumulative trend is defined as the cumulative effect characterization quantity of the carbonization process. This process realizes the effect aggregation of real-time deviation vectors in the time dimension, providing basic cumulative data for subsequent interpretation of distribution structure.

[0137] In summary, the core of interpreting the distribution structure of the cumulative effect characterization quantity is to extract all the values ​​of the cumulative effect characterization quantity, screen out the cumulative maximum and minimum values, statistically analyze the specific time nodes of the occurrence of the two types of extreme values ​​and their distribution characteristics, and observe the similarity of the change path of the transition data between extreme values. The set containing the extreme value position, distribution law and transition path characteristics is defined as the extreme value distribution pattern of the carbonization process. This process explores the extreme value distribution attributes of the cumulative effect characterization quantity and provides key extreme value data for subsequent periodic structure analysis.

[0138] In summary, the core of periodic structural analysis of extreme value distribution patterns is to sort out the time intervals of similar extreme values ​​in the extreme value distribution patterns, compare the consistency of the intervals to preliminarily determine the potential cycle, confirm the stable cycle by verifying the consistency of extreme value change characteristics in multiple consecutive cycles, and then integrate the duration of the stable cycle, the extreme value change pattern within the cycle, and the cumulative trend of deviation within the cycle to form structured data reflecting the repetitive pattern of extreme value distribution and define it as the continuous periodic characteristics of the carbonization process. This process completes the extraction of the periodic pattern of extreme value distribution patterns and provides a time-series periodic basis for subsequent carbonization process state judgment.

[0139] S5、based on the real-time state evaluation result, the abnormal state of the carbonization process is comprehensively evaluated, and the abnormal state identification of the carbonization process is obtained;

[0140] In the embodiment of the application, based on the real-time state evaluation result, the abnormal state of the carbonization process is comprehensively evaluated, and the abnormal state identification of the carbonization process is obtained, which comprises:

[0141] Based on the real-time state evaluation result, the abnormal state of the carbonization process is abnormal mode recognition, and the abnormal mode type of the carbonization process is obtained;

[0142] The abnormal mode type is quantified by a feature parameter, and the abnormal degree index and the abnormal duration parameter of the carbonization process are obtained;

[0143] The abnormal mode type, the abnormal degree index and the abnormal duration parameter are structured and coded, and the abnormal state identification of the carbonization process is obtained.

[0144] Specifically, based on the real-time state evaluation result, the abnormal state of the carbonization process is abnormal mode recognition, first, the key information deviating from the normal state contained in the real-time state evaluation result is extracted, these information includes the key flue gas component category deviating from the normal range, the change trend of the component concentration, the time node of the deviation occurrence and the influence of the deviation on other associated parameters; at the same time, the abnormal mode library of the carbonization process is constructed in advance, the library is established based on all abnormal event data of the carbonization process occurred in history, each abnormal mode in the library corresponds to an explicit abnormal performance, including specific component deviation characteristics, change trend type and associated influence mode; then the deviation information in the real-time state evaluation result is compared with each abnormal mode in the abnormal mode library one by one, during the comparison process, whether the key flue gas component category is matched, whether the concentration change trend is consistent and whether the associated influence mode is similar are focused on, if the real-time deviation information and each feature of a certain abnormal mode are completely consistent, the type corresponding to the abnormal mode is determined as the type of the current carbonization process abnormal state, and the type is the abnormal mode type of the carbonization process.

[0145] Further, when the characteristic parameter of the abnormal pattern type is quantified, the core characteristic parameter dimension corresponding to the type of abnormality is determined according to the determined abnormal pattern type. For the "single component concentration sudden rise" type of abnormality, the core characteristic parameter dimension includes the amplitude of the concentration exceeding the normal range and the speed of the concentration rising. For the "multi-component collaborative deviation" type of abnormality, the core characteristic parameter dimension includes the average amplitude of the deviation of each component from the normal range and the synchronization of the deviation between components. Then, for each core characteristic parameter dimension, the specific information related to the dimension in the real-time monitoring data is collected. For example, when calculating the amplitude of the concentration exceeding the normal range, the difference between the real-time concentration value of the abnormal component and the upper limit value of the normal range is obtained, and the relative deviation amplitude is determined by combining the interval width of the normal range of the component. The relative deviation amplitude is the core component of the abnormality degree index. At the same time, the time length from the time when the abnormal state is first monitored to the time when the characteristic of the abnormal pattern type is met is recorded, and the time length is the abnormal duration parameter of the carbonization process. Through the above process, the abnormality degree index and the abnormal duration parameter of the carbonization process are determined and extracted respectively.

[0146] Further, when the abnormal pattern type, the abnormality degree index and the abnormal duration parameter are structured and coded, a unique character identifier is assigned to each abnormal pattern type. The character identifier can directly correspond to a specific abnormal pattern, such as a specific capital letter corresponding to different types of "single component concentration sudden rise" and "multi-component collaborative deviation". Then, the abnormality degree index is classified and coded. According to the value range of the abnormality degree index, different levels are set, and each level corresponds to a unique lowercase letter identifier. For example, the abnormality degree is divided into three levels of slight, moderate and severe, which correspond to different lowercase letters respectively. The abnormal duration parameter is divided into intervals and coded. According to the duration range, different intervals are set, and each interval corresponds to a unique symbol identifier. For example, the duration is divided into three intervals of short, medium and long, which correspond to different symbols respectively. Finally, according to the fixed order of "abnormal pattern type code-abnormality degree index code-abnormal duration parameter code", the three codes are combined in turn to form a continuous and fixed structure string, which is the abnormal state identifier of the carbonization process.

[0147] Overall, based on the real-time state evaluation result, the abnormal state of the carbonization process is identified, the key information deviating from the normal state in the real-time state evaluation result is extracted, including the key flue gas component type involved in the abnormality, the concentration change trend, the deviation occurrence time and the associated impact, and then compared with the abnormal mode library constructed based on the historical abnormal event data, the mode type corresponding to the current abnormality is determined by matching the key characteristics, and finally the abnormal mode type of the carbonization process is obtained, which realizes the definition of abnormal state from phenomenon to clear type and provides clear direction for subsequent quantitative analysis.

[0148] Overall, the feature parameter quantization of the abnormal mode type is performed, the core is to determine the corresponding core feature parameter dimension according to the determined abnormal mode type, such as the concentration deviation amplitude for single component abnormality and the deviation synchronization for multi-component abnormality, and then calculate the index reflecting the severity of the abnormality combined with the real-time monitoring data, and record the duration of the abnormality from the first identification to the current quantization time, respectively obtain the abnormality degree index and the abnormality duration parameter of the carbonization process, which converts the qualitative abnormal type into quantitative data and provides integrable information for subsequent structured coding.

[0149] Overall, the abnormal mode type, the abnormality degree index and the abnormality duration parameter are structured and coded, the core is to assign a unique identifier to each of them, such as a dedicated character for the abnormal mode type, a level code for the abnormality degree index, and an interval code for the abnormality duration parameter, then combined in a fixed order to form a continuous structured string, and finally obtain the abnormal state identifier of the carbonization process, which realizes the standardization and integration of abnormal multi-dimensional information, and facilitates subsequent quick matching of alarm rules.

[0150] S6, based on the preset alarm rule knowledge base, the abnormal state identifier is mapped to obtain the hierarchical alarm instruction of the carbonization process.

[0151] In the embodiment of the application, based on the preset alarm rule knowledge base, the abnormal state identifier is mapped to obtain the hierarchical alarm instruction of the carbonization process, which comprises:

[0152] Based on the preset alarm rule knowledge base, the abnormal state identifier is matched with multiple levels of rules to obtain the alarm level of the carbonization process;

[0153] The alarm level is mapped to a response strategy to obtain the alarm content and disposal suggestion of the carbonization process;

[0154] The alarm content, the disposal suggestion and the alarm level are packaged into an instruction to obtain the hierarchical alarm instruction of the carbonization process.

[0155] Specifically, based on the preset alarm rule knowledge base, when performing multi-level rule matching on the abnormal state identifier, first, the construction basis and content of the preset alarm rule knowledge base are determined. The knowledge base is established based on historical abnormal event processing data of the carbonization process, process safety standards, and flue gas composition monitoring risk level division requirements. The knowledge base is divided into a basic rule layer, an association rule layer, and a priority rule layer according to rule levels. The basic rule layer corresponds to the preliminary matching of the abnormal mode type. The association rule layer corresponds to the combined matching of the abnormal degree index and the abnormal duration parameter. The priority rule layer corresponds to the grade adjustment rule under the superposition of multiple factors. Then, the structured code of the abnormal state identifier is disassembled, and the abnormal mode type, the abnormal degree index, and the abnormal duration parameter are extracted. These information are used as matching inputs.

[0156] Further, the abnormal mode type is compared with the rule entries of the basic rule layer one by one to determine the initial grade range corresponding to the mode. Then, the abnormal degree index and the abnormal duration parameter are substituted into the association rule layer to filter out the intermediate grade that meets the degree and duration combination conditions within the initial grade range. Finally, whether there is a superposition of multiple abnormal factors is determined by combining the priority rule layer. If there is, the intermediate grade is adjusted according to the rule. If not, the intermediate grade is directly determined as the final grade. The final grade is the alarm grade of the carbonization process.

[0157] Further, when performing response strategy mapping on the alarm grade, it is determined that the preset alarm rule knowledge base stores a response strategy entry corresponding to each alarm grade. Each entry includes alarm content description and disposal suggestion scheme for the abnormality of the grade. The alarm content description clearly explains the core attributes and potential risk range of the abnormality. The disposal suggestion scheme clearly specifies the operation steps. Then, according to the obtained alarm grade of the carbonization process, the corresponding response strategy entry is located in the alarm rule knowledge base. The text description under the "alarm content" field in the entry is extracted. The description accurately reflects the key information of the current abnormality to ensure that the operator quickly knows the abnormality. At the same time, the operation guide under the "disposal suggestion" field in the entry is extracted. The guide needs to be consistent with the carbonization process operation specification, and clearly specifies the operation subject, operation object, and operation sequence. Through the above extraction process, the alarm content and disposal suggestion of the carbonization process are obtained respectively.

[0158] Further, when the alarm content, the treatment suggestion and the alarm level are packaged into an instruction, a standard format of the instruction is determined first, which needs to meet the identification requirements of the carbonization process monitoring system or the control terminal. The format structure is "alarm level identification-exception core information-treatment operation guidance" in turn. The alarm level determined is marked at the beginning of the instruction to ensure that the level information is eye-catching. Then, the core information such as the abnormal mode type, the influence range corresponding to the abnormal degree index and the continuous state corresponding to the abnormal duration parameter in the alarm content is integrated in logical order to form a coherent abnormal situation description. The operation steps in the treatment suggestion are sorted according to the execution order, and the matters needing attention in the operation process are supplemented to ensure the executability of the treatment suggestion. Finally, the marked alarm level, the integrated abnormal situation description and the sorted treatment operation guidance are combined into a complete instruction text according to the preset format. The instruction text needs to be simple in language and complete in information, and can be directly received and executed by the relevant equipment or the operator of the carbonization process. The instruction text is the hierarchical alarm instruction of the carbonization process.

[0159] In general, the multi-level rule matching of the abnormal state identification based on the preset alarm rule knowledge base is the core. The multi-level rule base is constructed by relying on the historical abnormal event processing data of the carbonization process, the process safety standards and the flue gas composition monitoring risk level division requirements. The structured coding of the abnormal state identification is first disassembled to extract the abnormal mode type, the abnormal degree index and the abnormal duration parameter. Then, the process is matched step by step according to the process of "basic rule layer matching initial level range-association rule layer screening intermediate level-priority rule layer adjusting final level". Finally, the alarm level of the carbonization process is determined. This process realizes the accurate conversion of the abnormal state from structured information to explicit level, and provides a clear basis for subsequent response strategy matching.

[0160] In general, the response strategy mapping of the alarm level is the core. The response strategy entries corresponding to each alarm level in the preset alarm rule knowledge base are used. According to the determined alarm level of the carbonization process, the corresponding entry is located in the knowledge base. The alarm content clearly describing the abnormal core attributes and the potential risk range in the entry is extracted. At the same time, the treatment suggestion conforming to the carbonization process operation specification, the clear operation subject and the steps in the entry is extracted. This process converts the abstract alarm level into specific understandable abnormal information and executable operation guidance.

[0161] Overall, the alarm content, treatment suggestion and alarm level are packaged into instructions, the core is to mark the alarm level according to the standard format recognizable by the carbonization process monitoring system or control terminal to ensure that the information is eye-catching, then integrate the abnormal key information in the alarm content to form a coherent explanation, sort out the operation steps of the treatment suggestion and supplement the execution precautions, and finally combine into instruction text with simple language, complete information and direct execution, that is, the hierarchical alarm instruction of the carbonization process, which realizes the standardized integration of abnormal response information and ensures that the instruction can effectively support the abnormal treatment of the carbonization process.

[0162] As Figure 2 shown, it is a functional module diagram of a carbonization process flue gas component online monitoring and alarm system provided by an embodiment of the application.

[0163] The carbonization process flue gas component online monitoring and alarm system 100 can be installed in an electronic device. According to the functions implemented, the carbonization process flue gas component online monitoring and alarm system 100 can include a data cleaning and standardization module 101, a time series trend analysis module 102, a threshold value cooperative dynamic calibration module 103, a situation judgment and state evaluation module 104, an abnormal state evaluation and identification module 105, and a level mapping and alarm instruction module 106. The modules of the application can also be referred to as units, which refer to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

[0164] In this embodiment, the functions of each module / unit are as follows:

[0165] The data cleaning and standardization module 101 is used for multi-dimensional data cleaning of the flue gas component data of the carbonization process to obtain standardized component data of the carbonization process.

[0166] The time series trend analysis module 102 is used for time series trend analysis of the key flue gas component concentration of the standardized component data to obtain the component concentration trend of the carbonization process.

[0167] The threshold value cooperative dynamic calibration module 103 is used for cooperative dynamic calibration of the historical concentration threshold of the carbonization process based on the component concentration trend to obtain the target dynamic threshold of the carbonization process.

[0168] The situation judgment and state evaluation module 104 is used for situation judgment of the key flue gas component concentration based on the target dynamic threshold to obtain the real-time state evaluation result of the carbonization process.

[0169] The abnormal state evaluation and identification module 105 is configured to comprehensively evaluate the abnormal state of the carbonization process based on the real-time state evaluation result, and obtain an abnormal state identification of the carbonization process.

[0170] The grade mapping and alarm instruction module 106 is configured to perform grade mapping on the abnormal state identification based on a preset alarm rule knowledge base, and obtain a graded alarm instruction of the carbonization process.

[0171] In several embodiments provided in the present application, it should be understood that the disclosed method and system can be implemented in other ways. For example, the system embodiments described above are merely illustrative, for example, the division of the modules is only a logical function division, and actual implementation can have another division manner.

[0172] The modules described as separate components can or can not be physically separated, and the components displayed as modules can or can not be physical units, that is, they can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs.

[0173] In addition, each functional module in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of hardware plus software functional modules.

[0174] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0175] The embodiments of the present application can acquire and process related data based on artificial intelligence technology. Among them, artificial intelligence is to use digital computers or digital computer controlled machines to simulate, extend and expand human intelligence, perceive environment, acquire knowledge and use knowledge to obtain the best results.

[0176] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for online monitoring and alarm of flue gas composition during carbonization process, characterized in that, The method includes: S1. Perform multi-dimensional data cleaning on the flue gas composition data of the carbonization process to obtain the standardized composition data of the carbonization process; S2. Perform time-series trend analysis on the concentrations of key flue gas components in the standardized component data to obtain the component concentration trends during the carbonization process; S3. Based on the component concentration trend, perform collaborative dynamic calibration on the historical concentration threshold of the carbonization process to obtain the target dynamic threshold of the carbonization process; S4. Based on the target dynamic threshold, the concentration of the key flue gas components is assessed to obtain the real-time status evaluation result of the carbonization process; S5. Based on the real-time status assessment results, a comprehensive assessment of the abnormal status of the carbonization process is performed to obtain the abnormal status identifier of the carbonization process; S6. Based on a preset alarm rule knowledge base, perform a level mapping on the abnormal state identifiers to obtain a graded alarm instruction for the carbonization process.

2. The method for online monitoring and alarm of flue gas composition in a carbonization process as described in claim 1, characterized in that, The process involves multi-dimensional data cleaning of the flue gas composition data from the carbonization process to obtain standardized composition data for the carbonization process, including: Outlier removal is performed on the flue gas composition data to obtain preliminary purification data for the carbonization process; Frequency domain noise suppression is performed on the preliminary purification data to obtain smooth data of the carbonization process; The smoothed data is normalized to obtain the standardized composition data of the carbonization process.

3. The method for online monitoring and alarm of flue gas composition in a carbonization process as described in claim 1, characterized in that, The time-series trend analysis of the key flue gas component concentrations in the standardized component data to obtain the component concentration trends during the carbonization process includes: The concentrations of the key flue gas components are divided into concentration subsequences by a sliding window to obtain the concentration subsequences of the carbonization process; Variation vector analysis was performed on the concentration subsequence to obtain the trend direction of the carbonization process; Trend mode identification is performed on the trend direction to obtain the typical trend pattern of the carbonization process; By performing temporal coherence fusion on the typical trend patterns, the component concentration trend of the carbonization process is obtained.

4. The method for online monitoring and alarm of flue gas composition in a carbonization process as described in claim 1, characterized in that, The step of performing collaborative dynamic calibration on the historical concentration thresholds of the carbonization process based on the component concentration trend to obtain the target dynamic threshold of the carbonization process includes: Based on the component concentration trend, the baseline threshold curve of the historical concentration threshold is trend-co-aligned to obtain the threshold calibration vector of the carbonization process. The threshold calibration vector is used to perform offset deduction to obtain the threshold correction parameters for the carbonization process; Based on the threshold correction parameter, the baseline threshold curve is asymptotically converged to obtain the target dynamic threshold of the carbonization process.

5. The method for online monitoring and alarm of flue gas composition in a carbonization process as described in claim 4, characterized in that, The threshold calibration vector is dynamically offset to obtain the threshold correction parameter for the carbonization process. The calculation formula for the threshold correction parameter is as follows: ； In the formula, The threshold correction parameter, The magnitude adjustment coefficient of the threshold calibration vector. The direction control factor of the threshold calibration vector. The preset nonlinear gain coefficient, The baseline threshold curve at time point The value of , The preset historical weight decay coefficient, For natural ripening, The continuous operating time of the carbonization process. The preset time decay constant, The preset trend dynamic response coefficient, The rate of change of the amplitude adjustment coefficient. The direction sign function is the direction control factor.

6. The method for online monitoring and alarm of flue gas composition in a carbonization process as described in claim 1, characterized in that, The step of assessing the concentration of key flue gas components based on the target dynamic threshold to obtain a real-time status evaluation result of the carbonization process includes: Based on the target dynamic threshold, the concentration of the key flue gas components is dynamically gap analyzed to obtain the real-time deviation vector of the carbonization process; By coupling the real-time deviation vector with temporal features, the continuous periodic characteristics of the carbonization process are obtained; Multi-parameter correlation feature mining is performed on the real-time deviation vector to obtain the coupling tensor of the carbonization process; Based on the aforementioned periodic characteristics, the state trajectory of the coupling tensor is deduced to obtain the state transition path of the carbonization process; A situation assessment is performed on the state transition path to obtain the real-time state assessment results of the carbonization process.

7. The method for online monitoring and alarm of flue gas composition in a carbonization process as described in claim 6, characterized in that, The step of performing time-domain feature deconstruction on the real-time deviation vector to obtain the continuous periodic features of the carbonization process includes: The real-time deviation vector is converged in the time domain to obtain the cumulative effect characterization of the carbonization process; The distribution structure of the cumulative effect characterization quantity is interpreted to obtain the extreme value distribution pattern of the carbonization process; Periodic structural analysis of the extreme value distribution pattern yields the continuous periodic characteristics of the carbonization process.

8. The method for online monitoring and alarm of flue gas composition in a carbonization process as described in claim 1, characterized in that, The step of comprehensively evaluating the abnormal states of the carbonization process based on the real-time state assessment results to obtain an abnormal state identifier for the carbonization process includes: Based on the real-time state assessment results, abnormal modes are identified for the abnormal states of the carbonization process to obtain the abnormal mode types of the carbonization process. The abnormal mode type is quantified by feature parameters to obtain the abnormality degree index and abnormal duration parameter of the carbonization process; The abnormal mode type, the abnormality degree index, and the abnormal duration parameter are structured and encoded to obtain the abnormal state identifier of the carbonization process.

9. The method for online monitoring and alarm of flue gas composition in a carbonization process as described in claim 1, characterized in that, The pre-set alarm rule knowledge base maps the abnormal state identifiers to levels to obtain graded alarm instructions for the carbonization process, including: Based on a preset alarm rule knowledge base, multi-level rule matching is performed on the abnormal state identifiers to obtain the alarm level of the carbonization process; The alarm levels are mapped to response strategies to obtain the alarm content and handling suggestions for the carbonization process; The alarm content, the handling suggestions, and the alarm level are encapsulated into instructions to obtain the graded alarm instructions for the carbonization process.

10. An online monitoring and alarm system for flue gas composition during carbonization processes, characterized in that, The system includes: The data cleaning and standardization module is used to perform multi-dimensional data cleaning on the flue gas composition data of the carbonization process to obtain the standardized composition data of the carbonization process. The time-series trend analysis module is used to perform time-series trend analysis on the concentration of key flue gas components in the standardized component data to obtain the component concentration trend of the carbonization process. The threshold collaborative dynamic calibration module is used to perform collaborative dynamic calibration on the historical concentration threshold of the carbonization process based on the component concentration trend, so as to obtain the target dynamic threshold of the carbonization process. The situation assessment and status evaluation module is used to assess the concentration of the key flue gas components based on the target dynamic threshold, and obtain the real-time status evaluation results of the carbonization process. An abnormal state assessment and identification module is used to comprehensively assess the abnormal state of the carbonization process based on the real-time state assessment results, and obtain the abnormal state identification of the carbonization process. The level mapping and alarm instruction module is used to perform level mapping on the abnormal state identifier based on a preset alarm rule knowledge base to obtain the level alarm instructions for the carbonization process.

Images