Real-time detection method for fly ash particulate matter content

Abstract

The application discloses a fly ash particulate matter content real-time detection method and relates to the technical field of fly ash particulate matter detection, and comprises the following steps: in the case of flue gas backflow, based on the backflow precursor sensing sequence, combining the airflow direction change vector, the sampling channel pressure inversion rate and the particulate matter concentration mutation value, determining the sampling air path reversal behavior under the flue gas backflow condition; according to the sampling air path reversal behavior under the flue gas backflow condition, combining the particulate matter particle size change characteristic, the deviation degree between the concentration change first derivative slope and the preset concentration stability model, judging whether the current particulate matter concentration is disturbed by non-target fly ash. The application solves the problem that non-target fly ash is mixed due to flue gas backflow in the fly ash particulate matter content real-time detection process and is difficult to identify, realizes accurate determination of the sampling air path reversal, and effectively identifies, classifies and processes and dynamically avoids the interference detection data.

Description

Technical Field

[0001] This invention relates to the field of fly ash particulate matter detection technology, specifically to a method for real-time detection of fly ash particulate matter content. Background Technology

[0002] Real-time fly ash particulate matter content monitoring refers to the continuous online monitoring of the concentration of fly ash particulate matter carried in the flue gas during waste incineration. The aim is to dynamically acquire information on fly ash generation, distribution, and emissions, providing real-time data support for the operation control and environmental management of the incineration system. This monitoring is typically achieved by deploying fly ash monitoring devices at key nodes in the incineration flue gas treatment chain, such as horizontal flues, economizers, semi-dry reaction towers, bag filters, and chimney outlets. Fly ash monitoring equipment generally employs sensors based on laser scattering, beta-ray, or photoelectric sensing technologies. Combined with auxiliary parameters such as flue gas temperature, flow rate, and pressure, a series of processes including data acquisition, signal processing, concentration calculation, and data uploading are used to achieve real-time calculation and output of fly ash particulate matter mass concentration. In actual operation, it can also be integrated with the control system to monitor fly ash concentration trends and perform comprehensive analysis with other control parameters of the incineration process. This analysis guides the operation scheduling of soot blowers and the adjustment of combustion air volume ratios, thereby achieving automated system control and energy efficiency management. Meanwhile, the detection data of fly ash particulate matter can also provide front-end data support for subsequent fly ash treatment processes (such as low-temperature thermal desorption systems), helping to achieve the goals of fly ash resource utilization, volume reduction and intelligent treatment.

[0003] The existing technology has the following shortcomings:

[0004] In real-time fly ash particulate matter detection, when downstream fans stop operating or the dust removal system experiences pressure imbalance, the flue gas flow direction at the sampling point may temporarily reverse, causing flue gas backflow. At this time, particulate matter from the downstream area may mix into the flue gas sample that should have been collected from the upstream incineration process, resulting in the presence of fly ash particles from non-target sources in the sampled air. Because existing real-time fly ash particulate matter detection technologies are typically based on unidirectional isokinetic sampling, and the sampling path is assumed to be stable in the system design without a real-time monitoring or identification mechanism for changes in airflow direction, it is impossible to determine whether the current particulate matter concentration is being interfered with by non-target fly ash based on the reversed sampling air path behavior under flue gas backflow conditions. In the above situation, the system misclassifies non-target particulate matter as being emitted from the incineration process, not only causing the detection data to deviate from reality but also potentially triggering erroneous control responses, affecting the accuracy of emission determination, and interfering with subsequent control strategy optimization and intelligent learning processes based on the detection data.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a method for real-time detection of fly ash particulate matter content, so as to solve the problems in the background art mentioned above.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for real-time detection of fly ash particulate matter content, specifically comprising the following steps:

[0008] S1. By collecting fan operation signals, inlet and outlet pressures of dust removal devices, pressure changes at sampling points, and flow velocity direction disturbance values, a backflow precursor sensing sequence is constructed to determine whether flue gas backflow has occurred.

[0009] S2. In the event of flue gas backflow, based on the backflow precursor sensing sequence and combined with the airflow direction change vector, sampling channel pressure inversion rate and particulate matter concentration abrupt change value, determine the sampling air path reversal behavior under flue gas backflow conditions;

[0010] S3. Based on the reverse behavior of the sampling air path under flue gas backflow, combined with the characteristics of particulate matter size change, the slope of the first derivative of concentration change and the degree of deviation between the preset concentration stability model, determine whether the current particulate matter concentration is disturbed by non-target fly ash.

[0011] S4. Based on the judgment results, generate a particulate matter concentration confidence label for each detection period. The label types include credible, pending verification, and unreliable, and correspond to the direct access processing, correction processing, and rejection processing of concentration data, respectively.

[0012] S5. Based on the particulate matter concentration confidence label, dynamically regulate the fly ash particulate matter content detection behavior, including adjusting the sampling cycle, enabling cross-data judgment process, and recording abnormal concentration characteristic values, to achieve real-time avoidance of interfering detection data.

[0013] Preferably, S1 specifically includes the following steps:

[0014] S101. Collect fan operation signals by continuously reading fan control command change information, collect inlet and outlet pressures of dust removal device by synchronously acquiring corresponding pressure data at the inlet and outlet of dust removal device, collect pressure changes at sampling points by continuously recording micro-pressure change data at sampling points, collect flow velocity direction disturbance values ​​by real-time capture of gas flow direction change data, and align the collected data in time order.

[0015] S102. After completing the time sequence alignment, the trend of the status change of the fan operation signal, the relationship between the difference in the inlet and outlet pressures of the dust removal device, the consistency of the direction of pressure change at the sampling points, and the offset characteristics of the flow velocity direction disturbance value are jointly analyzed, and a backflow precursor sensing sequence is constructed according to the relationship of change.

[0016] S103. Perform continuous interval interpretation on the backflow precursor sensing sequence. When the backflow precursor sensing sequence simultaneously meets the following conditions: the fan operation signal deviates from the preset operating state interval, the pressure difference between the inlet and outlet of the dust removal device changes from a positive distribution to a negative distribution, the direction of pressure change at the sampling point is inconsistent with the normal smoke exhaust direction, and the flow velocity disturbance value is opposite to the preset mainstream direction, it is determined that flue gas backflow has occurred.

[0017] Preferably, S102 specifically is:

[0018] After completing the time sequence alignment, based on the data changes within the continuous sampling interval, the state change trend of the fan operation signal is extracted, the difference between the inlet pressure and outlet pressure of the dust removal device is calculated to obtain the difference change relationship, the directional consistency of the pressure change at the sampling point is determined, the directional offset feature of the flow velocity direction disturbance value is extracted, and the corresponding feature parameter set is formed.

[0019] The feature parameter set is matched according to the time index, and the relationship between the state change trend of the fan operation signal and the difference change of the inlet and outlet pressure of the dust removal device is synchronously analyzed. At the same time, the consistency of the direction of pressure change at the sampling point and the offset characteristics of the flow velocity disturbance value are analyzed collaboratively to form a change correlation relationship that reflects the state of multi-parameter collaborative change.

[0020] Based on the continuous correspondence of each characteristic parameter in the time dimension in the change correlation, the joint change state of the fan operation signal, the inlet and outlet pressure of the dust removal device, the pressure change at the sampling point, and the flow velocity direction disturbance value are sequentially encoded to construct a backflow precursor sensing sequence to describe the change trend of gas flow state.

[0021] Preferably, S2 is as follows:

[0022] In the event of flue gas backflow, the time interval corresponding to the flue gas backflow is determined based on the backflow precursor sensing sequence, and gas flow direction data is extracted within this time interval. The airflow direction change vector is constructed according to the directional change relationship between adjacent sampling times. At the same time, time series analysis is performed on the pressure change data in the sampling channel to calculate the pressure inversion rate of the sampling channel. The particulate matter concentration data is continuously sampled and compared to extract the abrupt change value of particulate matter concentration.

[0023] The airflow direction change vector, the sampling channel pressure inversion rate, and the particulate matter concentration abrupt value are matched according to the time index to analyze the synchronous relationship between the airflow direction change trend and the sampling channel pressure inversion process. At the same time, the occurrence position and persistence characteristics of the particulate matter concentration abrupt value in the backflow time interval are correlated and analyzed to form a joint feature set for characterizing the change state of the sampled airflow path.

[0024] Based on the co-occurrence relationships of the reverse distribution of airflow direction change vectors in the joint feature set, the reverse evolution of sampling channel pressure inversion rate, and the continuous occurrence of particulate matter concentration abrupt changes within the time interval, the sequence of flow direction changes of sampled air within the backflow time interval is determined to identify the path reversal behavior of sampled air under flue gas backflow conditions.

[0025] Preferably, S3 specifically includes the following steps:

[0026] S301. Within the time interval corresponding to the path reversal behavior of the sampled air under the condition of flue gas backflow, obtain particulate matter size distribution data, compare and analyze the changes in particle size distribution at continuous sampling times, extract the particle size change characteristics that reflect the particle size distribution shift, and establish the correspondence between the particle size change characteristics and the path reversal time interval.

[0027] S302. Within the time interval corresponding to the particle size change characteristics, continuously sample the particle concentration data, calculate the slope of the first derivative of the concentration change between adjacent sampling times, and perform matching analysis between the slope of the first derivative of the concentration change and the preset concentration stability model to obtain the degree of deviation of the slope of the first derivative of the concentration change relative to the preset concentration stability model.

[0028] S303. Based on the synchronous correspondence of the sampling air path reversal behavior, particulate matter size change characteristics, and the deviation between the slope of the first derivative of concentration change and the preset concentration stability model in the time dimension, the current particulate matter concentration data is comprehensively judged. When the particle size change characteristics and the deviation of concentration change occur simultaneously in the path reversal time interval, it is determined that the current particulate matter concentration is interfered with by non-target fly ash.

[0029] Preferably, S302 is as follows:

[0030] Within the time interval corresponding to the particle size variation characteristics, the particle concentration data is continuously sampled at a fixed sampling interval to form a particle concentration data sequence arranged in chronological order. The particle concentration data sequence is then denoised and smoothed to ensure the comparability of changes between adjacent sampling times.

[0031] Based on particulate matter concentration data sequences, the concentration change between adjacent sampling times is differentially calculated to obtain the slope of the first derivative of the concentration change corresponding to consecutive sampling times, and a slope sequence reflecting the evolution of the concentration change rate over time is constructed.

[0032] The slope sequence corresponding to the first derivative slope of concentration change is matched with the preset concentration stability model time by time. The deviation magnitude and duration interval between the slope sequence and the preset concentration stability model are compared and calculated to obtain the degree of deviation of the first derivative slope of concentration change relative to the preset concentration stability model.

[0033] Preferably, S4 is as follows:

[0034] Based on the judgment results formed for the current detection period, the state of particulate matter concentration data within the detection period is judged, and the judgment results are mapped as a basis for confidence judgment to characterize the reliability of particulate matter concentration data within the detection period.

[0035] Based on the credibility judgment criteria, a particulate matter concentration credibility label is generated for each detection period. The particulate matter concentration credibility label includes three types: credible, pending verification, and unreliable. Among them, credible is used to indicate that the judgment result does not show interference characteristics, pending verification is used to indicate that the judgment result has uncertain characteristics, and unreliable is used to indicate that the judgment result shows interference characteristics.

[0036] Based on the type of particulate matter concentration confidence label, processing operations are performed on the particulate matter concentration data within the corresponding detection period. Specifically, reliable corresponding concentration data is directly accessed, data awaiting verification is corrected, and unreliable corresponding concentration data is removed.

[0037] Preferably, S5 is as follows:

[0038] The confidence status of the current detection period is determined based on the confidence label of particulate matter concentration. When the confidence label of particulate matter concentration is reliable, the sampling cycle of the current fly ash particulate matter content detection behavior remains unchanged. When the confidence label of particulate matter concentration is pending verification, the sampling cycle is shortened to increase the sampling frequency. When the confidence label of particulate matter concentration is unreliable, the sampling cycle is extended to avoid interference and diffusion effects.

[0039] When the particulate matter concentration confidence label is unreliable, the cross-data judgment process is activated. Data on fly ash particulate matter content is collected in multiple detection channels or multiple detection devices. Multi-source data within the same detection period are compared synchronously, and the consistency or degree of deviation between data characteristics is identified.

[0040] During the cross-data judgment process, abnormal changes in particulate matter concentration during the detection period are recorded. The abnormal concentration feature values ​​are then matched with the corresponding particulate matter concentration confidence labels and sampling cycle adjustment records using time indexing to construct a dataset for feedback on the control process of fly ash particulate matter content detection behavior.

[0041] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0042] 1. This invention solves the problem of data interference caused by reversed sampling air paths in traditional detection systems by constructing a real-time fly ash particulate matter content detection process centered on flue gas backflow identification. Through multi-dimensional data joint sensing and analysis, including fan operation signals, dust collector pressure differences, sampling point pressure changes, and flow velocity disturbance characteristics, a backflow precursor sensing sequence can be constructed before backflow occurs, and sampling path reversal behavior can be identified during backflow, achieving accurate perception of abnormal airflow states. Based on this, by introducing synchronous matching of airflow direction change vectors, pressure inversion rates, and particulate matter concentration abrupt changes, a joint feature set for characterizing abnormal airflow behavior is further constructed, providing solid data support for the identification of interfering particulate matter. This technical solution distinguishes between target fly ash and non-target interference through a sequential judgment method, achieving accurate identification of data reliability under flue gas backflow.

[0043] 2. This invention introduces a particulate matter concentration credibility labeling mechanism after interference identification, and performs graded processing and dynamic response on the detection data based on the label status, constructing a complete data quality control and adaptive regulation system. By setting three categories of labels—credible, pending verification, and untrustworthy—the system can automatically implement data access, correction, or rejection operations, avoiding abnormal data from misleading emission judgments or control logic. Simultaneously, combined with sampling cycle adjustment, cross-channel judgment, and abnormal feature recording, the system's flexible response capability under abnormal conditions is improved, enhancing its robustness and intelligence. The overall technical path is clear, and the logical closed loop is tight, improving the accuracy of fly ash particulate matter detection and providing a stable and reliable data foundation for subsequent control strategy optimization and model training. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0045] Figure 1 This is a schematic flowchart of the real-time detection method for fly ash particulate matter content of the present invention. Detailed Implementation

[0046] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0047] This invention provides, for example Figure 1 The real-time detection method for fly ash particulate matter content shown includes the following steps:

[0048] S1. By collecting fan operation signals, inlet and outlet pressures of dust removal devices, pressure changes at sampling points, and flow velocity direction disturbance values, a backflow precursor sensing sequence is constructed to determine whether flue gas backflow has occurred.

[0049] In this embodiment, S1 specifically includes the following steps:

[0050] S101. Collect fan operation signals by continuously reading fan control command change information, collect inlet and outlet pressures of dust removal device by synchronously acquiring corresponding pressure data at the inlet and outlet of dust removal device, collect pressure changes at sampling points by continuously recording micro-pressure change data at sampling points, collect flow velocity direction disturbance values ​​by real-time capture of gas flow direction change data, and align the collected data in time order.

[0051] Collecting fan operation signals can be achieved by continuously reading the output commands of the fan control logic. Changes in fan control commands typically include start / stop status, motor frequency, and current values. Digital input / output interfaces or PLC system logs are used to capture control layer output changes, which are then timestamped to form a continuous operational status data stream. Collecting pressure data at the inlet and outlet of the dust collector typically involves using differential pressure sensors deployed at both ends of the bag filter or electrostatic precipitator. The raw pressure values ​​are obtained using pressure transmitters and then aligned with timestamps to form an inlet / outlet pressure sequence. Micro-pressure change data at sampling points is acquired by installing micro-differential pressure sensors. Sampling points are typically located in the horizontal flue or the connecting pipe section between the reaction tower and the dust collector. Millisecond-level sampling of minute pressure fluctuations is performed, and the change trend is recorded as a continuous time series. Gas flow direction change data can be acquired using bidirectional thermal gas velocity sensors. These sensors identify the main airflow direction and output directional velocity values. By comparing these values ​​with a preset mainstream direction benchmark, the velocity direction disturbance value is obtained. To achieve data correlation analysis, all collected data needs to be interpolated at a uniform time resolution and aligned sequentially based on a unified system time benchmark. This constructs a complete and consistent data base, providing a reliable foundation for subsequent trend extraction, change judgment, and multi-parameter fusion analysis. For example, before a typical backflow event, the fan control commands exhibit high-frequency start-stop switching, the inlet and outlet pressure difference continuously narrows to near reversal, the micro-pressure at the sampling point shifts from negative to near zero, and the flow velocity disturbance changes from positive to negative deviation. After time alignment, the precursory links of backflow can be clearly presented.

[0052] Fan control command change information refers to the logical signals issued by the control system to the fan for start-up, shutdown, speed adjustment, or maintenance operation. These signals are typically generated by a DCS or PLC control system and are crucial for reflecting the changing trends of the fan's operating status. Fan operation signals are data sequences representing the actual operating trend of the fan, formed by time-series encoding the state evolution of these control commands. Dust collector inlet and outlet refer to the inlet and outlet sections of the dust collector in the airflow path, corresponding to two physical interfaces connecting the reaction tower or flue, used to deploy pressure sampling points. The inlet and outlet pressures of the dust collector represent the static or dynamic pressure values ​​obtained at the two measurement points; the difference is used to characterize the equipment's load, blockage, or flow resistance status. Sampling point micro-pressure change data refers to the pressure fluctuation signals collected by micro-pressure sensors in the sampling system, reflecting whether there are abnormal airflow disturbances or reverse pressure transmission at the sampling port. Sampling point pressure changes refer to the trend of these micro-pressure data over time and can be used to identify short-term pressure reversal behavior. Gas flow direction change data is acquired through flow-direction identifiable sensors and is a key indicator for capturing deviations in the main airflow direction. The velocity direction disturbance value refers to the deviation between the gas velocity direction at a specific sampling point and the main exhaust direction of the system. It serves as a quantitative basis for determining whether there is airflow reversal or disturbance. Time sequence alignment refers to aligning the sampling nodes of raw data collected from different devices and at different frequencies using processing methods such as interpolation, delay compensation, and timestamp standardization. This ensures the temporal consistency among multiple parameters and the reliability of causal linkage analysis.

[0053] S102. After completing the time sequence alignment, the trend of the status change of the fan operation signal, the relationship between the difference in the inlet and outlet pressures of the dust removal device, the consistency of the direction of pressure change at the sampling points, and the offset characteristics of the flow velocity direction disturbance value are jointly analyzed, and a backflow precursor sensing sequence is constructed according to the relationship of change.

[0054] S103. Perform continuous interval interpretation on the backflow precursor sensing sequence. When the backflow precursor sensing sequence simultaneously meets the following conditions: the fan operation signal deviates from the preset operating state interval, the pressure difference between the inlet and outlet of the dust removal device changes from a positive distribution to a negative distribution, the direction of pressure change at the sampling point is inconsistent with the normal smoke exhaust direction, and the flow velocity disturbance value is opposite to the preset mainstream direction, it is determined that flue gas backflow has occurred.

[0055] In actual implementation, continuous interval analysis of the backflow precursor sensing sequence can effectively identify whether the system has entered an abnormal flue gas flow state. Continuous interval analysis refers to selecting multiple sets of sequentially coded data within a continuous time window and analyzing the feature combination at each moment to see if it conforms to the backflow characteristic pattern. When the fan operation signal deviates from the preset operating state range within a certain time period, i.e., the fan suddenly enters a frequent start-stop state from continuous operation, or the current load drops abnormally; at the same time, the pressure difference between the inlet and outlet of the dust removal device gradually changes from a stable positive gradient to a negative pressure difference, indicating that the downstream pressure is higher than the upstream, forming a reverse drive; combined with the trend that the pressure change direction at the sampling point is inconsistent with the normal flue gas exhaust direction, for example, the micro-pressure should be negative during normal flue gas exhaust, but if it is detected as increasing positively, it indicates reverse pressure interference; and the flow velocity direction disturbance value is opposite to the preset mainstream direction, i.e., the gas flow direction detection value shows a 180-degree offset or vector reversal superposition, then it can be comprehensively determined that there is a stable backflow trend in this continuous interval, thus confirming the occurrence of flue gas backflow. By simultaneously satisfying multiple conditions to construct a composite judgment logic, the risk of misjudgment caused by single parameter fluctuations can be significantly reduced, and the recognition accuracy can be improved.

[0056] A deviation of the fan operation signal from the preset operating range indicates a sudden change in the fan's control signal, current value, or speed data compared to its long-term stable operating state. The preset range is set based on historical stable data; for example, the fan should maintain a specific load range during normal flue gas exhaust. A change from a positive to a negative distribution in the inlet and outlet pressure difference of the dust collector indicates that the inlet pressure is usually higher than the outlet pressure, forming a positive differential pressure under positive flue gas exhaust conditions. However, when the airflow flows backward, this difference becomes negative, creating a reverse-drive phenomenon. An inconsistency between the sampling point pressure change direction and the normal flue gas exhaust direction indicates that the pressure change trend model established based on the flue gas exhaust direction deviates from the current pressure change trend. For example, the normal trend from the furnace to the chimney is decreasing; if the current trend is increasing, it indicates an abnormal direction. A flow velocity direction disturbance value opposite to the preset mainstream direction indicates that the gas flow direction collected by the flow velocity sensor is reversed from the flue gas exhaust mainstream direction set in the system. This manifests as an increased reverse velocity amplitude or a reversal of the positive and reverse velocity ratios. By constructing a composite interpretation logic based on these core features, the accurate capture of the backflow trend can be achieved, which is a prerequisite for ensuring the reliability of real-time detection of fly ash particulate matter content.

[0057] In this embodiment, S102 specifically refers to:

[0058] After completing the time sequence alignment, based on the data changes within the continuous sampling interval, the state change trend of the fan operation signal is extracted, the difference between the inlet pressure and outlet pressure of the dust removal device is calculated to obtain the difference change relationship, the directional consistency of the pressure change at the sampling point is determined, the directional offset feature of the flow velocity direction disturbance value is extracted, and the corresponding feature parameter set is formed.

[0059] After aligning the time sequence, a continuous sampling interval of a certain length is needed as the analysis window to obtain precursor signals of reverse airflow changes. A sampling interval refers to a fixed-length data segment selected at preset intervals on a unified time axis, used to represent the changes in various parameters over a period of time. Within each sampling interval, the state change trend of the fan operation signal is first extracted. This involves smoothing the continuously read fan control command values ​​and extracting their state change trend characteristics by identifying the switching frequency, frequency fluctuation rate, or number of control command jumps in the start / stop state. For the pressure data at the inlet and outlet of the dust collector, a difference sequence of inlet pressure minus outlet pressure is formed by subtracting the values ​​at each time point. This difference sequence is then subjected to a moving average or slope calculation to determine its direction of change within the current sampling interval, thus obtaining the difference change relationship. The consistency determination of the direction of pressure change at the sampling point involves extracting the trend of the first derivative of the continuous pressure data at that point and comparing it with the fan operation trend direction to determine if there is any deviation from the consistent flow direction. For example, the pressure should decrease during exhaust; if it is detected as increasing, the direction is inconsistent. The extraction of directional offset features for velocity disturbance values ​​refers to statistically analyzing the forward and reverse velocity amplitudes within a certain range of identifiable velocity signals and examining the degree of deviation of their average offset from the main direction. Four results—fan status change trend, pressure difference change relationship, consistency of pressure change direction, and velocity direction offset—are uniformly encapsulated into a set of feature parameters, serving as a precursor indicator for identifying flue gas backflow trends. This process, through local window analysis of multi-source signal characteristics, not only enhances the signal's sensitivity to backflow events but also provides a foundational data structure for subsequent time-series coding. For example, if, within a certain sampling interval, the fan command signal shows frequent start-stop cycles, the pressure difference changes from positive to negative, the micro-pressure trend shows a reverse climb, and the flow direction disturbance shifts significantly, then the set of feature parameters will uniformly exhibit precursors to backflow trends.

[0060] The feature parameter set is matched according to the time index, and the relationship between the state change trend of the fan operation signal and the difference change of the inlet and outlet pressure of the dust removal device is synchronously analyzed. At the same time, the consistency of the direction of pressure change at the sampling point and the offset characteristics of the flow velocity disturbance value are analyzed collaboratively to form a change correlation relationship that reflects the state of multi-parameter collaborative change.

[0061] Matching feature parameter sets according to time indices involves, after constructing the feature parameter set, using a unified sampling time axis as a benchmark, mapping each feature value to its generation time to ensure comparability of different types of parameters at the same point in time. Time indices are typically identified by sampling frame numbers or system clock timestamps, and the matching process is achieved through hash tables, data frame line numbers, or index alignment. After matching, a synchronous correlation analysis is needed between the trend of the fan operation signal and the difference in pressure between the inlet and outlet of the dust removal device. This involves analyzing the linkage between the two within the same time period; for example, whether the pressure difference rapidly decreases or even reverses as the fan changes from running to stopped. This analysis can be achieved by calculating the correlation coefficient between the trends and whether the trend slopes are consistent. Simultaneously, a collaborative correlation analysis is needed between the consistency of the direction of pressure changes at sampling points and the offset characteristics of flow velocity disturbance values. This involves determining whether micro-pressure fluctuations and flow direction offsets are directionally similar; for example, a rise in micro-pressure accompanied by a reversal in flow direction. Linkage features can be extracted through multivariate synchronous curve fitting and point-to-point trend comparison. When these multidimensional features establish a synchronous response relationship over time, a correlation is formed that reflects the coordinated change of multiple parameters. This relationship is the core basis for identifying abnormal trends in flue gas flow. For example, if within a certain time period there is a decrease in the fan control signal, a change in pressure difference from positive to negative, a reversal of micro-pressure changes, or a strengthening of flow direction shift, such synchronous evolution will appear as a high-confidence precursor to backflow in correlation analysis. This process, through multivariate collaborative modeling, achieves sensitive identification of complex airflow disturbances, effectively improving the robustness and accuracy of backflow trend judgment.

[0062] Based on the continuous correspondence of each characteristic parameter in the time dimension in the change correlation, the joint change state of the fan operation signal, the inlet and outlet pressure of the dust removal device, the pressure change at the sampling point, and the flow velocity direction disturbance value are sequentially encoded to construct a backflow precursor sensing sequence to describe the change trend of gas flow state.

[0063] Based on the continuous correspondence of various characteristic parameters in the time dimension of the change correlation, it is necessary to sequentially encode the joint change states of the fan operation signal, the inlet and outlet pressures of the dust removal device, the pressure changes at the sampling points, and the velocity direction disturbance values. Sequential encoding refers to discretizing the state values ​​of multiple characteristic parameters at each sampling moment with time sequence as the main axis, and using a unique identifier to represent its state type. For example, a fan in a high-frequency start-stop state can be marked as A, a pressure difference showing a reverse gradient change can be marked as B, a sampling point pressure fluctuation showing an abnormal increase can be marked as C, and a velocity direction disturbance greater than a set deviation can be marked as D. Combining these state codes in time sequence to form an encoding string can represent the evolution trajectory of airflow behavior within a specific time period. By continuously encoding the joint state changes within multiple sampling periods, a backflow precursor sensing sequence can be constructed to describe the trend of gas flow state changes. This sequence not only retains the synchronous evolution information of multi-source parameters, but can also be used as input for pattern recognition for subsequent backflow trend recognition model training or early warning rule matching. For example, when the encoded sequence is ABCDBCDDA within five consecutive sampling periods, it can be determined as a characteristic path of a high-risk backflow trend. The anti-backflow precursor sensing sequence can clearly reconstruct the entire process of the transition from normal flow to abnormal backflow within the system, serving as a key logical foundation for early warning and intervention. This method, through temporal feature compression and traceable encoding, not only improves data processing efficiency but also provides a structured representation for intelligent identification.

[0064] S2. In the event of flue gas backflow, based on the backflow precursor sensing sequence and combined with the airflow direction change vector, sampling channel pressure inversion rate and particulate matter concentration abrupt change value, determine the sampling air path reversal behavior under flue gas backflow conditions;

[0065] In this embodiment, S2 specifically refers to:

[0066] In the event of flue gas backflow, the time interval corresponding to the flue gas backflow is determined based on the backflow precursor sensing sequence, and gas flow direction data is extracted within this time interval. The airflow direction change vector is constructed according to the directional change relationship between adjacent sampling times. At the same time, time series analysis is performed on the pressure change data in the sampling channel to calculate the pressure inversion rate of the sampling channel. The particulate matter concentration data is continuously sampled and compared to extract the abrupt change value of particulate matter concentration.

[0067] In the event of flue gas backflow, the first step is to determine the corresponding time interval based on the state abrupt change points identified by the backflow precursor sensing sequence. Within this time interval, sensor data related to the gas flow direction are collected, such as ultrasonic multi-point velocity measurements or vortex flowmeter output values. By comparing the change in the flow direction sign at adjacent sampling times, a vector of airflow direction change is constructed, reflecting the process of gas evolving from the normal flow direction to the reverse direction. For pressure change data within the sampling channel, time series analysis is performed based on a continuous time-point pressure value sequence. By calculating the first derivative of the continuous pressure data and identifying the location and velocity of the sign change, the pressure inversion rate of the sampling channel can be obtained, which is used to characterize the dynamic transition characteristics of airflow from negative pressure intake to positive pressure outward push. Simultaneously, the concentration data acquired by the particulate matter concentration sensor are continuously compared at equal time intervals. The abrupt change rate of the difference is used to identify abrupt changes in particulate matter concentration, reflecting the abnormal concentration fluctuations caused by particles from non-target areas entering the sampling airflow. By extracting the change characteristics of the above three types of signals within the same time interval, a highly reliable quantitative basis can be provided for subsequent path reversal judgment.

[0068] Gas flow direction data refers to directional data about the direction of gas movement acquired through flow direction sensors, used to determine whether the airflow has reversed in a short period of time. The airflow direction change vector is a vector column constructed based on the continuous changes of this flow direction data over time, used to quantify the trend and intensity of gas direction disturbances. Time series analysis is the process of calculating and modeling parameters such as pressure, velocity, or concentration in chronological order, facilitating the capture of abrupt changes or trend shifts. The sampling channel pressure inversion rate measures the speed at which the pressure within a sampling point transitions from an inhalation state to an outward expansion state; its calculation is based on the change in the sign and absolute value of pressure per unit time, serving as a dynamic indicator for assessing gas path reversal behavior. Particulate matter concentration abrupt changes refer to sharp increases in concentration within a concentration change sequence; their formation is often related to the instantaneous intrusion of fly ash from non-target areas, and is a key trigger signal for determining data contamination. These technical features work together to form a complete data link, supporting subsequent accurate identification of sampling path reversal behavior.

[0069] The airflow direction change vector, the sampling channel pressure inversion rate, and the particulate matter concentration abrupt value are matched according to the time index to analyze the synchronous relationship between the airflow direction change trend and the sampling channel pressure inversion process. At the same time, the occurrence position and persistence characteristics of the particulate matter concentration abrupt value in the backflow time interval are correlated and analyzed to form a joint feature set for characterizing the change state of the sampled airflow path.

[0070] To identify the path reversal behavior of sampled air under flue gas backflow conditions, the airflow direction change vector, the sampling channel pressure inversion rate, and the particulate matter concentration abrupt change value need to be uniformly aligned to the same time index to ensure their direct comparability in the time dimension. This is achieved by constructing an ordered data table for each data point according to a unified timestamp, assembling the three characteristic quantities corresponding to each time node into a joint observation unit. Subsequently, a synchronization relationship analysis is performed on the rate of change, direction sign inflection point, and delayed response characteristics between the airflow direction change trend and the sampling channel pressure inversion process to determine whether they exhibit a consistent perturbation pattern within the backflow time interval. Next, information such as the first occurrence time, duration, and peak amplitude of the particulate matter concentration abrupt change value in this time interval is extracted and matched with the degree of overlap of the first two features on the time axis, forming a data structure containing the co-evolutionary relationship of the three types of characteristic variables. This joint feature set can not only characterize whether path reversal has occurred but also reflect the actual impact of path reversal on the sampled concentration, thus providing a logical basis for the final concentration assessment reliability classification.

[0071] Matching based on time indexing involves mapping different data types from the same source to a one-to-one correspondence on the same timeline, ensuring synchronization between multi-source signals. This approach avoids misjudgments caused by different sampling frequencies or time drift. The synchronization between the airflow direction change trend and the pressure inversion process in the sampling channel is crucial for determining whether a true airflow path reversal exists. This requires analyzing the temporal consistency of the two types of data in terms of change direction, change rate, and turning point. The location and duration of abrupt changes in particulate matter concentration within the reversal time interval refer to indicators such as the first time these abrupt changes occurred, their duration, and whether they were single pulses or continuous interference. These parameters directly reflect whether non-target particles were collected due to path reversal. The joint feature set used to characterize the changing state of the sampled airflow path is a dataset formed by unifying the above three types of features through time indexing and semantic analysis. It describes the entire-link physical change behavior when reversal occurs and is the core data foundation for determining path reversal.

[0072] Based on the co-occurrence relationships of the reverse distribution of airflow direction change vectors in the joint feature set, the reverse evolution of sampling channel pressure inversion rate, and the continuous occurrence of particulate matter concentration abrupt changes within the time interval, the sequence of flow direction changes of sampled air within the backflow time interval is determined to identify the path reversal behavior of sampled air under flue gas backflow conditions.

[0073] To definitively determine the reverse behavior of the sampled air path under flue gas backflow conditions, it is necessary to systematically integrate and logically determine multiple features in the joint feature set. First, the distribution characteristics of the airflow direction change vector within the backflow time interval are statistically analyzed. If this vector shows a direction opposite to the normal exhaust direction at multiple consecutive sampling points, it is considered a reverse distribution state. Second, if the pressure inversion rate of the sampling channel maintains a monotonically increasing or decreasing trend within the time interval, and the pressure direction continuously changes from positive to negative pressure or reverses, it indicates that the pressure response process is in a reverse evolution state. Simultaneously, the persistence of abrupt changes in particulate matter concentration within this time interval is determined. If multiple consecutive sampling points detect sudden increases in concentration or multiple short-period pulse-like changes, and the magnitude of the change exceeds a set threshold, it is considered that the concentration change persists. After independently confirming the three types of features, they are logically integrated point by point in chronological order to determine whether the triple conditions of airflow direction, pressure evolution, and concentration fluctuation are met at each sampling time point. If the co-occurrence relationship is met at multiple consecutive time points, it can be determined that the flow direction of the sampled air has reversed, thus confirming that the reversed path behavior of the sampled air did indeed occur within the flue gas backflow time interval.

[0074] The reverse distribution of airflow direction change vectors indicates that the gas flow direction at multiple consecutive sampling points points points in the opposite direction to the conventional emission direction, representing a shift from normal flue gas emission to reverse flow. The reverse evolution of the sampling channel pressure inversion rate indicates that the pressure value exhibits a continuous reverse change process, such as gradually changing from positive pressure to negative pressure, or from pressure decay to rise, and maintaining this trend uninterruptedly. The co-occurrence relationship of abrupt changes in particulate matter concentration values ​​within a time interval indicates that abnormal concentration values ​​not only occur frequently during the backflow period but also exhibit periodicity or superposition between multiple adjacent sampling points, forming a response signal to path anomalies. Sequential determination refers to logically comparing sampling times along the time dimension, judging point by point whether all abnormal characteristics are simultaneously satisfied, and determining the overall state accordingly. The reverse behavior of the sampling air path under flue gas backflow conditions refers to the overall reversal of the gas flow direction in the sampling path caused by downstream disturbances, resulting in interference from particulate matter from non-target areas in the sampling results; this is the core identification object of this technical problem.

[0075] S3. Based on the reverse behavior of the sampling air path under flue gas backflow, combined with the characteristics of particulate matter size change, the slope of the first derivative of concentration change and the degree of deviation between the preset concentration stability model, determine whether the current particulate matter concentration is disturbed by non-target fly ash.

[0076] In this embodiment, S3 specifically includes the following steps:

[0077] S301. Within the time interval corresponding to the path reversal behavior of the sampled air under the condition of flue gas backflow, obtain particulate matter size distribution data, compare and analyze the changes in particle size distribution at continuous sampling times, extract the particle size change characteristics that reflect the particle size distribution shift, and establish the correspondence between the particle size change characteristics and the path reversal time interval.

[0078] Within the time interval identified as flue gas backflow, particulate matter size distribution data was continuously collected using an online fly ash particle size analyzer, forming a time series sample covering the entire path reversal process. From the sampled data, consecutive time points before, during, and after the backflow were selected, and corresponding particle size distribution histograms were extracted and smoothed using probability distribution curves. Maximum frequency shift, distribution centroid drift, and fluctuations in the proportion of specific particle size segments were used as calculation indicators to quantitatively compare particle size distribution changes between different time points. If there is a concentrated distribution shift trend or a significant change in the multi-peak structure, it can be identified as a disturbance in the particle size distribution. Based on this, feature variables indicating the degree and direction of particle size distribution shift were constructed as particulate matter size change characteristics, and these were mapped one-to-one with the time interval of the sampled air path reversal for subsequent identification of non-target source particulate matter interference. The core purpose of this processing is to use abnormal changes in particle size distribution as a signal to help corroborate the phenomenon of external particle intrusion under reversed sampling path behavior, thereby enhancing the reliability of concentration anomaly judgment.

[0079] Particulate matter size distribution data refers to the distribution information of the number of particles corresponding to each size range recorded by particle size detection equipment during continuous sampling, usually presented in the form of distribution histograms or frequency functions. Particle size distribution variation refers to the structural differences in particle size data between different time points, which may manifest as a shift in the median particle size, changes in concentration, or significant fluctuations in the proportion of a certain size range. Comparative analysis involves horizontally comparing the particle size distribution at multiple consecutive sampling times to identify significant differences or trend-based changes. Particle size variation characteristics reflecting particle size distribution shifts typically include changes in the distribution centroid, the displacement of the maximum probability particle size, and changes in the 90th percentile particle size range, used to characterize the directionality and intensity of particle size disturbances. The correspondence between particle size variation characteristics and path reversal time intervals refers to binding each particle size variation characteristic value to the time of occurrence of the corresponding sampled air path reversal behavior, so that each segment of abnormal particle size change can be mapped to a specific airflow reversal scenario for subsequent identification and judgment of interfering particles.

[0080] S302. Within the time interval corresponding to the particle size change characteristics, continuously sample the particle concentration data, calculate the slope of the first derivative of the concentration change between adjacent sampling times, and perform matching analysis between the slope of the first derivative of the concentration change and the preset concentration stability model to obtain the degree of deviation of the slope of the first derivative of the concentration change relative to the preset concentration stability model.

[0081] S303. Based on the synchronous correspondence of the sampling air path reversal behavior, particulate matter size change characteristics, and the deviation between the slope of the first derivative of concentration change and the preset concentration stability model in the time dimension, the current particulate matter concentration data is comprehensively judged. When the particle size change characteristics and the deviation of concentration change occur simultaneously in the path reversal time interval, it is determined that the current particulate matter concentration is interfered with by non-target fly ash.

[0082] To determine whether the current particulate matter concentration is being interfered with by non-target fly ash, three key elements need to be synchronously matched along the time dimension: the sampled air path reversal behavior, particulate matter size change characteristics, and the degree of deviation between the slope of the first derivative of the concentration change and the preset concentration stability model. First, the time interval corresponding to the backflow is determined based on the identified sampled air path reversal behavior. Then, within this interval, the occurrence time of particle size change characteristics and the abnormal fluctuation range of the concentration change slope are retrieved. Using timestamp alignment technology, particle size offset data and slope deviation data are mapped to the same time axis to determine whether they overlap within the path reversal time interval. If there is a significant temporal overlap, meaning both occur simultaneously during the path reversal, then the particulate matter concentration can be considered to be interfered with by non-target fly ash. This multi-parameter collaborative verification mechanism can effectively avoid misjudgments caused by a single abnormal indicator, improving the accuracy and robustness of interference identification.

[0083] Particle size variation characteristics refer to the trend of particulate matter size distribution patterns deviating from historical averages within a unit of time, typically manifested as abrupt changes in particle size towards finer or coarser particles. Concentration variation deviation refers to the magnitude and duration of the numerical shift between the slope of the first derivative of particulate matter concentration change and the stable model, characterizing the degree of anomaly in current concentration fluctuations. Path reversal time intervals represent the specific time range during abnormal reversals of gas flow direction, determined by analyzing the combined characteristics of airflow direction, pressure inversion, and concentration abrupt changes in the sampling channel. When both particle size variation characteristics and concentration variation deviations are detected within the path reversal time interval, and there is significant temporal overlap, it means that the current sampling data is simultaneously affected by changes in particulate matter source and abnormal concentration fluctuations. This co-occurrence relationship becomes a crucial basis for judging concentration interference, ensuring that the analysis results have a sufficient physical basis and judgment logic.

[0084] In this embodiment, S302 specifically refers to:

[0085] Within the time interval corresponding to the particle size variation characteristics, the particle concentration data is continuously sampled at a fixed sampling interval to form a particle concentration data sequence arranged in chronological order. The particle concentration data sequence is then denoised and smoothed to ensure the comparability of changes between adjacent sampling times.

[0086] Within the time interval corresponding to the characteristics of particulate matter size changes, particulate matter concentration data can be continuously extracted from the online particulate matter concentration detection device at a preset time sampling interval, such as once per second. These data are recorded one by one according to time labels and arranged sequentially to form a continuous particulate matter concentration data sequence, reflecting the dynamic evolution of particulate matter concentration over a short period. Since the raw sampling data may exhibit spikes or discrete anomalies due to equipment response delays, airflow disturbances, and environmental fluctuations, the concentration data sequence needs to be processed. First, algorithms such as sliding window filtering or median filtering are applied to perform noise reduction, eliminating short-term non-trend disturbances. Then, methods such as local weighted regression or exponentially weighted moving average are used for smoothing to preserve the overall trend and reduce the interference of local abrupt changes on trend judgment. The purpose of this processing step is to enhance the readability of the concentration change trend, ensuring good numerical continuity and trend consistency in concentration changes between adjacent sampling times, thereby providing a stable basis for subsequent derivative slope calculations, avoiding misjudgments of the first derivative due to data noise, and ensuring that the particulate matter concentration change pattern can be accurately characterized and analyzed. Particulate matter concentration data refers to the number or mass of particles collected per unit time or unit volume; a particulate matter concentration data sequence arranged in chronological order is an ordered set of these values ​​based on the sampling time axis; denoising and smoothing are important means of optimizing and adjusting the original concentration data, aiming to improve the usability and trend expression of the data, and ensure that the subtle features of concentration changes can be accurately identified by the algorithm.

[0087] Based on particulate matter concentration data sequences, the concentration change between adjacent sampling times is differentially calculated to obtain the slope of the first derivative of the concentration change corresponding to consecutive sampling times, and a slope sequence reflecting the evolution of the concentration change rate over time is constructed.

[0088] After processing the particulate matter concentration data sequence, the concentration values ​​between any two adjacent sampling times can be differentially calculated based on the sequence. This involves subtracting the concentration value from the previous time point from the current one, thus obtaining the concentration change per unit time. Dividing this difference by the time interval between the two sampling times allows us to represent the rate of concentration change during that period, i.e., the slope of the first derivative of the concentration change. By performing differential operations on continuous time intervals and calculating the derivative slope, a set of ordered slope values ​​can be obtained, forming a slope sequence reflecting the evolution of the concentration change rate over time. This slope sequence can accurately depict the trend of particulate matter concentration change under the influence of backflow interference. For example, under the influence of reversed sampling air paths, the concentration may suddenly rise or fall, resulting in a significant positive or negative jump in the derivative slope. By analyzing the peak positions, rate of change magnitudes, and durations in the slope sequence, we can accurately determine whether there are abnormal concentration changes, thereby assisting in the identification of potential non-target fly ash interference signals. Differential computation is a fundamental algorithm for extracting local variations between data points; the slope of the first derivative of concentration change is a mathematical indicator that measures the degree of concentration change per unit time; the slope sequence is a vector composed of the first derivatives at multiple consecutive time points, used to comprehensively characterize the dynamic pattern of concentration change on the time axis. This processing method helps to transform the instantaneous characteristics of concentration change into quantifiable mathematical indicators, improving the accuracy and reliability of concentration anomaly identification.

[0089] The slope sequence corresponding to the first derivative slope of concentration change is matched with the preset concentration stability model time by time. The deviation magnitude and duration interval between the slope sequence and the preset concentration stability model are compared and calculated to obtain the degree of deviation of the first derivative slope of concentration change relative to the preset concentration stability model.

[0090] The preset concentration stability model is a baseline behavior pattern constructed from long-term particulate matter concentration data collected under undisturbed conditions. It reflects the typical range and trend of the slope of the first derivative of concentration change under normal emission conditions. This model can be established using methods such as sliding window mean smoothing, multi-segment linear fitting, or dynamic median regression to define the derivative slope characteristics when concentration fluctuations are in a stable state. The slope sequence to be detected is matched against the preset concentration stability model hourly, i.e., the slope values ​​at each time point of the two sequences are compared sequentially over time to identify numerical differences. This hourly comparison helps to discover the continuity of abnormal slopes in time distribution, thus eliminating the influence of isolated noise on the judgment results. During the matching process, the difference between the two slope sequences at each time point is calculated to obtain the offset amplitude. Simultaneously, the time period where the deviation value continuously exceeds the model's allowable threshold is recorded, forming the corresponding persistence interval. The offset amplitude reflects the abnormal intensity of the current concentration change rate, and the persistence interval reflects the time scale of the abnormal state. By jointly analyzing these two indicators, it is possible to determine whether the slope of the first derivative of the concentration change structurally deviates from the concentration stability model, thereby identifying whether the concentration data is affected by fly ash backflow interference. This method, through constructing a control model and quantifying the degree of deviation, provides a rigorous and calculable basis for determining the interference of concentration fluctuations.

[0091] S4. Based on the judgment results, generate a particulate matter concentration confidence label for each detection period. The label types include credible, pending verification, and unreliable, and correspond to the direct access processing, correction processing, and rejection processing of concentration data, respectively.

[0092] In this embodiment, S4 specifically refers to:

[0093] Based on the judgment results formed for the current detection period, the state of particulate matter concentration data within the detection period is judged, and the judgment results are mapped as a basis for confidence judgment to characterize the reliability of particulate matter concentration data within the detection period.

[0094] The judgment results for the current detection period can be determined by comprehensively considering key discrimination parameters such as the sampling air path reversal behavior, particulate matter size change characteristics, and the deviation of the slope of the first derivative of concentration change from the preset concentration stability model. After the detection system completes dynamic data acquisition and analysis for each time period, the judgment results can be divided into three categories based on preset interference identification rules: no interference, uncertain interference, and significant interference. In specific implementation, multi-dimensional index thresholds can be set. For example, if the duration of path reversal exceeds a certain critical value and both particle size change and concentration slope deviate from the model, it is considered significant interference; if only some indicators deviate, it is classified as uncertain interference. The system can convert these multi-dimensional discrimination results into qualitative label mappings, i.e., forming "credibility judgment criteria". For example, if the particle size shift is significant but path reversal does not occur in a certain detection period, the judgment result for that period will be mapped to the "pending verification" type. This process can be implemented through threshold judgment logic, fuzzy rule base, or lightweight rule engine to ensure the accuracy and real-time performance of the judgment.

[0095] The state of particulate matter concentration data within the detection period refers to the comprehensive performance of the time series trend, abrupt changes, and relationship with airflow path changes of particulate matter concentration during that period. The judgment result is a classification conclusion formed based on the linkage behavior between feature data, used to indicate whether the data is affected by non-target fly ash interference. The credibility judgment criterion is a mapping mechanism used to standardize the judgment result, aiming to transform complex and multidimensional interference judgment features into structured data credibility identifiers. Its essence is an attribution label for the interference state, usually existing in the form of rules, models, or labels output by learning engines. The detection period divides the continuous sampling process into multiple independent data processing cycle units according to time granularity, ensuring clear boundaries and timely response of data analysis. The reliability of particulate matter concentration data reflects whether the data can accurately represent the real particulate matter emission level under incineration conditions, and is an important foundation for subsequent intelligent decision-making or control feedback. Through the organic integration of these technical features, a systematic analysis and hierarchical management of the reliability of fly ash detection data sources is achieved.

[0096] Based on the credibility judgment criteria, a particulate matter concentration credibility label is generated for each detection period. The particulate matter concentration credibility label includes three types: credible, pending verification, and unreliable. Among them, credible is used to indicate that the judgment result does not show interference characteristics, pending verification is used to indicate that the judgment result has uncertain characteristics, and unreliable is used to indicate that the judgment result shows interference characteristics.

[0097] To generate a particulate matter concentration confidence label for each detection period, specific rules can be set to map the combined results of multiple key judgment parameters, such as airflow path, particle size change, and concentration slope, into three distinct label types. In practice, the system first collects feature discrimination results for each detection period and then comprehensively analyzes the performance of various parameters according to a pre-defined set of label judgment rules. For example, if no airflow path reversal is detected during the detection period, the particulate matter size change is stable, and the concentration slope is within the model's allowable range, it is judged as interference-free and labeled "confidential." If a single abnormality occurs during the detection period, or multiple feature parameters slightly deviate from the pre-defined model threshold but do not possess common interference characteristics, it is judged as pending and labeled "awaiting verification." If typical interference behaviors such as path reversal, particle size anomalies, and drastic fluctuations in concentration slope exist simultaneously during the detection period, an "unreliable" label is generated. This multi-rule joint mapping method ensures the accuracy of the judgment and the rationality of the hierarchical management of particulate matter concentration data, thus providing a standardized basis for subsequent correction processing and data screening.

[0098] The particulate matter concentration confidence label is a judgment marker used to identify the confidence level of particulate matter concentration data for each detection period, forming a core means of data quality control. "Confidential" means that within the corresponding detection period, all discrimination parameters fluctuate within a reasonable range, do not show interference characteristics related to flue gas backflow, and are data that can be directly used for control logic and model training. "Pending verification" means that the data shows a certain degree of anomaly, but lacks synchronous correlation with typical interference parameters or the interference characteristics are unclear; the data can be temporarily stored for manual review or later algorithm correction. "Unreliable" means that the detection data highly matches interference characteristics across multiple parameter dimensions, posing a clear risk of non-target source influence, and cannot be used as valid emission data. "Confidential" indicates that the judgment result does not show interference characteristics, meaning the confidence label is generated by a judgment set consisting of multiple anomaly-free parameters. "Pending verification" indicates that the judgment result has uncertain characteristics, meaning the system has identified a potential offset but cannot yet confirm whether it is a real interference. "Unreliable" indicates that the judgment result shows interference characteristics, meaning the system has identified typical interference behavior and formed a data removal suggestion. This labeling mechanism can serve as an important component of the quality assessment closed loop in fly ash detection applications.

[0099] Based on the type of particulate matter concentration confidence label, processing operations are performed on the particulate matter concentration data within the corresponding detection period. Specifically, reliable corresponding concentration data is directly accessed, data awaiting verification is corrected, and unreliable corresponding concentration data is removed.

[0100] Classifying particulate matter concentration data based on the type of confidence label can be achieved by introducing a label-based hierarchical control mechanism in the data stream processing module. Specifically, in the detection process, after the confidence level of particulate matter concentration data for each time period is determined, it is assigned a corresponding label type and then enters the corresponding processing path. If the label type is "confidential," it indicates that the data quality meets the usage standards, and the system directly includes the detection data in the pollutant emission monitoring values ​​and transmits it to the control system and model learning module. If the label type is "pending verification," the system will use interpolation, exponential smoothing, or model regression to correct the data based on the confidence concentration data and trend changes in nearby time periods, ensuring the consistency of the data in terms of structural rationality. If the label type is "unreliable," the particulate matter concentration data in that detection period is directly excluded and does not participate in statistical analysis and control decisions. For example, during continuous detection, if the concentration value suddenly increases in a 5-minute interval and is determined to be "unreliable," the system will remove the data in that interval from the valid dataset to avoid introducing non-target fly ash interference into subsequent judgments. This hierarchical processing method improves the accuracy of fly ash monitoring results and the robustness of control response.

[0101] "Direct access processing of reliable corresponding concentration data" refers to the system directly incorporating the current detection data into concentration trend calculation, compliance judgment, or feedback control logic as a valid result without additional correction when it identifies that the current detection data has no interference characteristics and all parameters are normal. "Correction processing of data to be verified" refers to the system initiating algorithms such as interpolation compensation, outlier correction, or trend consistency analysis based on historical reliable data to make limited adjustments to the values ​​of data, ensuring the data maintains continuity and accuracy in the overall trend; this processing path avoids the overall failure of data due to minor anomalies. "Removal processing of unreliable corresponding concentration data" means that the system completely excludes data with obvious interference behavior, excluding it from the concentration statistical sequence, transmission to external interfaces, and entry into the intelligent control module, fundamentally isolating the impact of abnormal data. These three processing paths, combined with the automatic generation of reliability labels, constitute a closed-loop data screening system, ensuring the stability and reliability of fly ash particulate matter detection data under complex operating conditions.

[0102] S5. Based on the particulate matter concentration confidence label, dynamically regulate the fly ash particulate matter content detection behavior, including adjusting the sampling cycle, enabling cross-data judgment process, and recording abnormal concentration characteristic values, to achieve real-time avoidance of interfering detection data.

[0103] In this embodiment, S5 specifically refers to:

[0104] The confidence status of the current detection period is determined based on the confidence label of particulate matter concentration. When the confidence label of particulate matter concentration is reliable, the sampling cycle of the current fly ash particulate matter content detection behavior remains unchanged. When the confidence label of particulate matter concentration is pending verification, the sampling cycle is shortened to increase the sampling frequency. When the confidence label of particulate matter concentration is unreliable, the sampling cycle is extended to avoid interference and diffusion effects.

[0105] After the particulate matter concentration confidence labels are determined, each label type can be bound to the corresponding sampling cycle strategy by constructing a detection behavior control mapping table. In the specific implementation, the system determines the current confidence status label based on the particulate matter concentration data and interference discrimination results within the detection period. If the label is "confident," it indicates that the detection data is not affected by interference, and the current sampling cycle can remain unchanged to ensure a balance between acquisition efficiency and system resource utilization. If the label is "pending verification," it indicates that the current concentration data fluctuates but does not constitute clear interference; the system should shorten the sampling cycle, for example, from 5 minutes to 1 minute, to improve the response capability to short-term anomalies and enhance the monitoring accuracy of concentration change trends. If the label is "unreliable," it indicates that the detection data has significantly deviated from normal characteristics, and the sampling cycle needs to be extended, for example, from 5 minutes to 15 minutes, to wait for the airflow in the sampling channel to stabilize and avoid high-risk data segments. This dynamic adjustment process can be completed through the linkage between the sampling scheduler and the label processing module, and the cycle adjustment behavior and its corresponding confidence label changes are continuously tracked through a log recording mechanism.

[0106] A particulate matter concentration confidence label is an identifier generated based on data quality assessment results, used to reflect the reliability of particulate matter data within a detection period. The confidence status is the binding result between the label and the specific detection time period, used to drive subsequent sampling logic changes. Fly ash particulate matter content detection is a continuous collection of data on changes in fly ash particulate matter concentration in the flue or sampling channel, with the sampling period as its core parameter. The sampling period adjustment strategy dynamically changes the detection rhythm based on the label value, including three modes: maintain, shorten, and extend, corresponding to different data confidence statuses, thereby achieving reasonable allocation of sampling resources and efficient avoidance of abnormal detection periods. This strategy is particularly important in flue gas fly ash detection scenarios, significantly improving the practicality and security of monitoring data.

[0107] When the particulate matter concentration confidence label is unreliable, the cross-data judgment process is activated. Data on fly ash particulate matter content is collected in multiple detection channels or multiple detection devices. Multi-source data within the same detection period are compared synchronously, and the consistency or degree of deviation between data characteristics is identified.

[0108] When a particulate matter concentration confidence label is deemed unreliable, a cross-data verification process can be implemented to further confirm whether the current detection data is distorted due to local interference. In practice, the system activates multiple independent detection channels or calls upon multiple detection devices distributed in different locations to simultaneously collect fly ash particulate matter content data during the same detection period, constructing a multi-source data comparison matrix. This matrix uses time as an index to horizontally align the real-time concentration values ​​of each channel or device, and employs a consistency comparison algorithm to identify the degree of difference between data. For example, a sliding window method is used to calculate the similarity of concentration trends, the magnitude of offsets, and the overlap rate of abrupt change points between any two channels. If most channel data are highly consistent in both temporal trends and absolute value ranges, it can be preliminarily determined that the unreliable label for a certain channel may be a misjudgment; conversely, if multiple channels show unstable characteristics, it can be further confirmed as an overall environmental interference event. This process effectively filters out misjudgments caused by abnormal local sampling paths, single-point equipment failures, or instantaneous airflow turbulence, enhancing the system's robustness and data fault tolerance.

[0109] A particulate matter concentration confidence label of "unreliable" indicates a significant anomaly in the quality of the current detection data, necessitating the implementation of a higher-dimensional data comparison mechanism. The cross-data judgment process is a verification logic based on spatial redundancy, achieving data cross-verification through the horizontal scheduling of multiple sensor resources. Multiple detection channels and multiple detection devices represent redundant acquisition deployments at the hardware level, ensuring that valid comparison data can still be obtained even if problems occur in the sampling path or individual devices. Fly ash particulate matter content data is the judgment object, and its comparison is based on whether the data possess characteristic consistency, including consistent trends, synchronized peak responses, or small overall deviations. The deviation analysis process is implemented through methods such as mean square error, sliding window volatility, or dynamic threshold difference judgment, ensuring a quantitative identification capability for data deviations. The entire process helps the system obtain supporting evidence in a timely manner when data reliability is challenged, forming an intelligent redundancy judgment system.

[0110] During the cross-data judgment process, abnormal changes in particulate matter concentration during the detection period are recorded. The abnormal concentration feature values ​​are then matched with the corresponding particulate matter concentration confidence labels and sampling cycle adjustment records using time indexing to construct a dataset for feedback on the control process of fly ash particulate matter content detection behavior.

[0111] During the cross-data judgment process, the concentration data of fly ash particles in each channel or device can be monitored and analyzed in real time by setting a concentration change rate threshold. When the concentration change rate of a certain data point or continuous data segment significantly exceeds the normal fluctuation range, it can be marked as an abnormal change value of particulate matter concentration. The system automatically records these abnormal values ​​and archives them with the corresponding particulate matter concentration confidence label and sampling cycle adjustment behavior in the current detection period. To ensure the accuracy of subsequent analysis and traceability, the system matches and associates abnormal concentration feature values, confidence labels, and sampling cycle records using a time index, constructing a data structure containing five dimensions: time, space, concentration, label, and control behavior. This dataset can be used to train subsequent control strategy models, evaluate the response effectiveness of existing control mechanisms, or assist in judging the occurrence pattern of fly ash interference events, and is one of the basic supporting means for realizing system self-learning and self-correction.

[0112] Anomaly changes in particulate matter concentration refer to sudden increases or decreases in particulate matter levels within a short period, typically identified by setting static or dynamic thresholds. The particulate matter concentration confidence label characterizes data quality and serves as the label field for anomaly identification in this dataset. Sampling period adjustment records reflect the system's real-time response to detection frequency and are a key parameter for assessing the effectiveness of control strategies. Time index matching is a mechanism for aligning and associating multiple types of time-series data along a unified timeline, ensuring clear logical relationships between different data fields. Anomaly concentration feature values, as the core content of the dataset, include dimensions such as the magnitude of change, duration of occurrence, and direction of change. The final dataset not only contains information on concentration fluctuations but also carries the context of labels and regulatory actions, serving various analysis and optimization scenarios.

[0113] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means (e.g., infrared, wireless, microwave, etc.). A computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0114] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0115] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0116] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0117] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0118] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0119] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for real-time detection of fly ash particulate matter content, characterized in that, Specifically, the following steps are included: S1. By collecting fan operation signals, inlet and outlet pressures of dust removal devices, pressure changes at sampling points, and flow velocity direction disturbance values, a backflow precursor sensing sequence is constructed to determine whether flue gas backflow has occurred. S2. In the event of flue gas backflow, based on the backflow precursor sensing sequence and combined with the airflow direction change vector, the sampling channel pressure inversion rate and the particulate matter concentration abrupt change value, the sampling air path reversal behavior under flue gas backflow is determined. S2 specifically refers to: In the event of flue gas backflow, the time interval corresponding to flue gas backflow is determined based on the backflow precursor sensing sequence, and gas flow direction data is extracted within this time interval. The airflow direction change vector is constructed according to the directional change relationship between adjacent sampling times. At the same time, time series analysis is performed on the pressure change data in the sampling channel to calculate the pressure inversion rate of the sampling channel. The particulate matter concentration data is continuously sampled and compared to extract the abrupt change value of particulate matter concentration. The airflow direction change vector, the sampling channel pressure inversion rate, and the particulate matter concentration abrupt value are matched according to the time index to analyze the synchronous relationship between the airflow direction change trend and the sampling channel pressure inversion process. At the same time, the occurrence position and persistence characteristics of the particulate matter concentration abrupt value in the backflow time interval are correlated and analyzed to form a joint feature set for characterizing the change state of the sampled airflow path. Based on the co-occurrence relationships of the reverse distribution of airflow direction change vectors in the joint feature set, the reverse evolution of sampling channel pressure inversion rate, and the continuous occurrence of particulate matter concentration abrupt values ​​within the time interval, the order of flow direction changes of sampled air within the backflow time interval is determined to identify the path reversal behavior of sampled air under flue gas backflow conditions. S3. Based on the reverse behavior of the sampling air path under flue gas backflow, combined with the characteristics of particulate matter size change, the slope of the first derivative of concentration change and the degree of deviation between the preset concentration stability model, determine whether the current particulate matter concentration is disturbed by non-target fly ash. S4. Based on the judgment results, generate a particulate matter concentration confidence label for each detection period. The label types include credible, pending verification, and unreliable, and correspond to the direct access processing, correction processing, and rejection processing of concentration data, respectively. S5. Based on the particulate matter concentration confidence label, dynamically regulate the fly ash particulate matter content detection behavior, including adjusting the sampling cycle, enabling cross-data judgment process and recording abnormal concentration characteristic values, to achieve real-time avoidance of interfering detection data. S5 specifically refers to: The confidence status of the current detection period is determined based on the confidence label of particulate matter concentration. When the confidence label of particulate matter concentration is reliable, the sampling cycle of the current fly ash particulate matter content detection behavior remains unchanged. When the confidence label of particulate matter concentration is pending verification, the sampling cycle is shortened to increase the sampling frequency. When the confidence label of particulate matter concentration is unreliable, the sampling cycle is extended to avoid interference and diffusion effects. When the particulate matter concentration confidence label is unreliable, the cross-data judgment process is activated. Data on fly ash particulate matter content is collected in multiple detection channels or multiple detection devices. Multi-source data within the same detection period are compared synchronously, and the consistency or degree of deviation between data characteristics is identified. During the cross-data judgment process, abnormal changes in particulate matter concentration during the detection period are recorded. The abnormal concentration feature values ​​are then matched with the corresponding particulate matter concentration confidence labels and sampling cycle adjustment records using time indexing to construct a dataset for feedback on the control process of fly ash particulate matter content detection behavior.

2. The method for real-time detection of fly ash particulate matter content according to claim 1, characterized in that, S1 specifically includes the following steps: S101. Collect fan operation signals by continuously reading fan control command change information, collect inlet and outlet pressures of dust removal device by synchronously acquiring corresponding pressure data at the inlet and outlet of dust removal device, collect pressure changes at sampling points by continuously recording micro-pressure change data at sampling points, collect flow velocity direction disturbance values ​​by real-time capture of gas flow direction change data, and align the collected data in time order. S102. After completing the time sequence alignment, the trend of the status change of the fan operation signal, the relationship between the difference in the inlet and outlet pressures of the dust removal device, the consistency of the direction of pressure change at the sampling points, and the offset characteristics of the flow velocity direction disturbance value are jointly analyzed, and a backflow precursor sensing sequence is constructed according to the relationship of change. S103. Perform continuous interval interpretation on the backflow precursor sensing sequence. When the backflow precursor sensing sequence simultaneously meets the following conditions: the fan operation signal deviates from the preset operating state interval, the pressure difference between the inlet and outlet of the dust removal device changes from a positive distribution to a negative distribution, the direction of pressure change at the sampling point is inconsistent with the normal smoke exhaust direction, and the flow velocity disturbance value is opposite to the preset mainstream direction, it is determined that flue gas backflow has occurred.

3. The method for real-time detection of fly ash particulate matter content according to claim 2, characterized in that, S102 specifically refers to: After completing the time sequence alignment, based on the data changes within the continuous sampling interval, the state change trend of the fan operation signal is extracted, the difference between the inlet pressure and outlet pressure of the dust removal device is calculated to obtain the difference change relationship, the directional consistency of the pressure change at the sampling point is determined, the directional offset feature of the flow velocity direction disturbance value is extracted, and the corresponding feature parameter set is formed. The feature parameter set is matched according to the time index, and the relationship between the state change trend of the fan operation signal and the difference change of the inlet and outlet pressure of the dust removal device is synchronously analyzed. At the same time, the consistency of the direction of pressure change at the sampling point and the offset characteristics of the flow velocity disturbance value are analyzed collaboratively to form a change correlation relationship that reflects the state of multi-parameter collaborative change. Based on the continuous correspondence of each characteristic parameter in the time dimension in the change correlation, the joint change state of the fan operation signal, the inlet and outlet pressure of the dust removal device, the pressure change at the sampling point, and the flow velocity direction disturbance value are sequentially encoded to construct a backflow precursor sensing sequence to describe the change trend of gas flow state.

4. The method for real-time detection of fly ash particulate matter content according to claim 1, characterized in that, S3 specifically includes the following steps: S301. Within the time interval corresponding to the path reversal behavior of the sampled air under the condition of flue gas backflow, obtain particulate matter size distribution data, compare and analyze the changes in particle size distribution at continuous sampling times, extract the particle size change characteristics that reflect the particle size distribution shift, and establish the correspondence between the particle size change characteristics and the path reversal time interval. S302. Within the time interval corresponding to the particle size change characteristics, continuously sample the particle concentration data, calculate the slope of the first derivative of the concentration change between adjacent sampling times, and perform matching analysis between the slope of the first derivative of the concentration change and the preset concentration stability model to obtain the degree of deviation of the slope of the first derivative of the concentration change relative to the preset concentration stability model. S303. Based on the synchronous correspondence of the sampling air path reversal behavior, particulate matter size change characteristics, and the deviation between the slope of the first derivative of concentration change and the preset concentration stability model in the time dimension, the current particulate matter concentration data is comprehensively judged. When the particle size change characteristics and the deviation of concentration change occur simultaneously in the path reversal time interval, it is determined that the current particulate matter concentration is interfered with by non-target fly ash.

5. The method for real-time detection of fly ash particulate matter content according to claim 4, characterized in that, S302 specifically refers to: Within the time interval corresponding to the particle size variation characteristics, the particle concentration data is continuously sampled at a fixed sampling interval to form a particle concentration data sequence arranged in chronological order. The particle concentration data sequence is then denoised and smoothed to ensure the comparability of changes between adjacent sampling times. Based on particulate matter concentration data sequences, the concentration change between adjacent sampling times is differentially calculated to obtain the slope of the first derivative of the concentration change corresponding to consecutive sampling times, and a slope sequence reflecting the evolution of the concentration change rate over time is constructed. The slope sequence corresponding to the first derivative slope of concentration change is matched with the preset concentration stability model time by time. The deviation magnitude and duration interval between the slope sequence and the preset concentration stability model are compared and calculated to obtain the degree of deviation of the first derivative slope of concentration change relative to the preset concentration stability model.

6. The method for real-time detection of fly ash particulate matter content according to claim 1, characterized in that, S4 specifically refers to: Based on the judgment results formed for the current detection period, the state of particulate matter concentration data within the detection period is judged, and the judgment results are mapped as a basis for confidence judgment to characterize the reliability of particulate matter concentration data within the detection period. Based on the credibility judgment criteria, a particulate matter concentration credibility label is generated for each detection period. The particulate matter concentration credibility label includes three types: credible, pending verification, and unreliable. Among them, credible is used to indicate that the judgment result does not show interference characteristics, pending verification is used to indicate that the judgment result has uncertain characteristics, and unreliable is used to indicate that the judgment result shows interference characteristics. Based on the type of particulate matter concentration confidence label, processing operations are performed on the particulate matter concentration data within the corresponding detection period. Specifically, reliable corresponding concentration data is directly accessed, data awaiting verification is corrected, and unreliable corresponding concentration data is removed.

Images