A Method and System for Detecting Anomalies in Enterprise Exhaust Gas Emissions Based on Multi-Source Data

Abstract

This invention discloses a method and system for detecting abnormal enterprise exhaust emissions based on multi-source data, belonging to the field of multi-source data fusion and analysis technology. The method includes: integrating multi-source sensors of the enterprise and deploying an asynchronous iterative sampling mode that uses a combination of processing-side and emission-side sensors to determine a binary array; configuring an anomaly detector with a sampler and constraint condition library accordingly, embedding it into the industrial control system to establish a communication network; and finally triggering sampling drive, dual-channel data processing, and counterfactual derivation to determine the anomaly detection result and perform industrial control alarm management in the central control system. This invention solves the technical problems of traditional enterprise exhaust emission detection lacking multi-source data integration, processing, and correlation analysis capabilities, failing to accurately correlate processing and emission characteristics, resulting in incomplete data, delayed anomaly detection, and difficulty in meeting the needs of precise supervision and efficient management. It achieves the technical effect of accurate detection and efficient alarm of enterprise exhaust emission anomalies, meeting the needs of precise supervision and efficient management.

Description

Technical Field

[0001] This invention relates to the field of multi-source data fusion and analysis technology, and in particular to a method and system for detecting abnormal enterprise exhaust emissions based on multi-source data. Background Technology

[0002] In enterprise production processes, monitoring of exhaust emissions is crucial for environmental supervision and compliant operation, with accurate detection of abnormal emissions being a core element. Current technologies for detecting enterprise exhaust emissions largely rely on sensors with single or limited data sources, processing and monitoring data through conventional sampling and analysis. While these methods are effective under stable operating conditions, their shortcomings become apparent in complex industrial scenarios as environmental requirements increase. Because enterprise production involves multiple processes and emission points, traditional detection methods struggle to integrate multi-source data and accurately correlate processing and emission characteristics, resulting in incomplete data and delayed anomaly detection, failing to meet the demands for precise supervision and efficient management of enterprise exhaust emissions. Summary of the Invention

[0003] This application provides a method and system for detecting abnormal enterprise exhaust emissions based on multi-source data. It is used to solve the technical problems of traditional enterprise exhaust emission detection lacking the ability to integrate, process and analyze multi-source data, failing to accurately correlate processing and emission characteristics, resulting in incomplete data, delayed anomaly detection, and difficulty in meeting the needs of accurate supervision and efficient management.

[0004] The first aspect of this application provides a method for detecting abnormal enterprise exhaust emissions based on multi-source data. The method includes: integrating multi-source sensors from enterprise projects; deploying an asynchronous iterative sampling mode based on binary arrays, wherein the binary array is determined by any combination of sensors from the processing side and the emission side; configuring a sampler and an anomaly detector according to the asynchronous iterative sampling mode, embedding them in an industrial control system, and establishing a communication network between the sampler, multi-source sensors, and the anomaly detector, wherein the anomaly detector has an embedded constraint library based on energy conservation and mass conservation; triggering a continuous binary combination and sampling drive based on the sampler, determining the sampled data, and triggering anomaly detection based on the asynchronous iterative sampling mode. The system employs dual-channel data processing of a normal detector and counterfactual deduction based on abnormal emissions to determine anomaly detection results, enabling industrial control alarm management at the central control system's response end. The anomaly detector configuration includes: using time-varying causality based on process stages, where process characteristics are the cause and emission characteristics are the effect, including exhaust gas type and exhaust gas flow rate; constructing a first deduction branch based on the time-varying causality, where the first deduction branch performs deduction from process characteristics to emission characteristics; constructing a second sampling branch using real-time array integration; and parallelizing the first deduction branch and the second data branch, introducing counterfactual deduction based on lateral interaction to construct the anomaly detector.

[0005] In a possible implementation, for the multi-source sensing end, a set of combination methods is determined; the set of combination methods is traversed, and a constraint condition library based on mass conservation and energy conservation is determined with the combination sensing as the guide; the constraint condition library is built into the anomaly detector.

[0006] In possible implementations, the process chain of the enterprise project is determined, and multi-source sensing terminals are integrated, wherein the multi-source sensing terminals include distributed built-in and external sensors; a binary array is introduced, wherein the binary array contains metadata based on the processing side and binary metadata based on the emission side; sampling based on the binary array is performed for the multi-source sensing terminals to determine the sampling array.

[0007] In a possible implementation, the industrial control system receives the sampling array, imports the metadata in the sampling array into the first derivation branch, performs causal derivation under the constraints of mass conservation and energy conservation, and determines the first emission data; imports the sampling data into the second sampling branch, integrates and determines the second array; and verifies the first emission data and the second array according to the lateral interaction channel. If they are consistent, an emission standard status identifier is generated.

[0008] In possible implementations, if there is a discrepancy, a non-standard emission condition identifier is generated; static detection data is determined based on the standard emission condition identifier or the non-standard emission condition identifier; and anomaly tracing and location based on counterfactual deduction is performed based on the non-standard emission condition identifier to determine the anomaly source.

[0009] In a possible implementation, for the sampling array, a first processing and acquisition node and a second emission acquisition node are determined; using the first processing and acquisition node and the second emission acquisition node as the positioning range, a deductive decision is made to determine the source of the anomaly.

[0010] In a possible implementation, a sampling frequency is set; with the sampling frequency as a constraint, a combination of binary arrays is used for iterative acquisition and decision-making based on the anomaly detector, and an asynchronous array-based detection chain is integrated and determined; based on the detection chain, the dynamic trend of gas emissions is analyzed and determined; according to the dynamic trend, dynamic detection data is generated, wherein the dynamic detection data is a standard condition trend or a non-standard condition trend; the static detection data and the dynamic detection data are added to the anomaly detection results.

[0011] In possible implementations, an early warning mode is deployed based on the source of the anomaly. If the source of the anomaly belongs to the processing side, a first early warning mode is triggered; if the source of the anomaly belongs to the emission side, a second early warning mode is triggered. Multiple early warning levels are deployed according to the anomaly level. Based on the early warning mode and the multiple early warning levels, exhaust gas emission anomaly early warning management based on the operating condition system is performed.

[0012] The second aspect of this application provides an enterprise exhaust gas emission anomaly detection system based on multi-source data. The system includes: an asynchronous iterative sampling mode deployment module, used to integrate multi-source sensors from enterprise projects and deploy an asynchronous iterative sampling mode based on binary arrays, wherein the binary array is determined by any combination of sensors from the processing side and the emission side; a communication network construction module, used to configure samplers and anomaly detectors according to the asynchronous iterative sampling mode, embedding them in the industrial control system, and establishing a communication network of samplers, multi-source sensors, and anomaly detectors, wherein the anomaly detectors have an embedded constraint library based on energy conservation and mass conservation; and an industrial control alarm management execution module, used to trigger continuous binary combination and sampling drive based on the samplers. The process involves determining sampled data and triggering dual-channel data processing based on an anomaly detector and counterfactual deduction based on abnormal emissions to determine the anomaly detection result. Industrial control alarm management is then implemented at the response end of the central control system. Specifically, the communication network construction module configures the anomaly detector using the following steps: First, based on time-varying causality at the process stage, where process characteristics are the cause and emission characteristics are the effect, including exhaust gas type and exhaust gas flow rate; second, constructing a first deduction branch based on the time-varying causality, where the first deduction branch performs deduction from process characteristics to emission characteristics; third, constructing a second sampling branch by integrating real-time arrays; and fourth, parallelizing the first deduction branch and the second data branch, introducing counterfactual deduction based on lateral interaction to construct the anomaly detector.

[0013] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0014] This application integrates multi-source sensors from enterprise projects, deploys an asynchronous iterative sampling mode based on binary arrays, and performs dual-path data processing (real-time integration of causal derivation in the first derivation branch and the second sampling branch) and lateral interactive counterfactual derivation through an anomaly detector. It also verifies the results using a constraint library based on energy and mass conservation to determine the anomaly detection results and implement industrial control alarm management. This allows for accurate detection of abnormal enterprise exhaust emissions, making the detection results more comprehensive and reliable. The application achieves the technical effect of accurate detection and efficient alarming of abnormal enterprise exhaust emissions, meeting the needs of precise supervision and efficient control. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the enterprise exhaust gas emission anomaly detection method based on multi-source data provided in this application embodiment.

[0017] Figure 2 This is a schematic diagram of the structure of the enterprise exhaust gas emission anomaly detection system based on multi-source data provided in the embodiments of this application.

[0018] Figure labeling: Asynchronous iterative sampling mode deployment module 1, communication network construction module 2, industrial control alarm management execution module 3. Detailed Implementation

[0019] This application provides a method and system for detecting abnormal enterprise exhaust emissions based on multi-source data. It is used to solve the technical problems of traditional enterprise exhaust emission detection lacking the ability to integrate, process and analyze multi-source data, failing to accurately correlate processing and emission characteristics, resulting in incomplete data, delayed anomaly detection, and difficulty in meeting the needs of accurate supervision and efficient management.

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0021] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices.

[0022] Example 1, as Figure 1 As shown, a method for detecting abnormal enterprise exhaust emissions based on multi-source data includes:

[0023] Step A100: Integrate the multi-source sensors of the enterprise project and deploy an asynchronous iterative sampling mode based on binary arrays, wherein the binary array is determined by any combination of sensors from the processing side and the emission side.

[0024] In this embodiment, a binary array refers to a data set formed by combining primary data from the processing side and secondary data from the emission side in the enterprise's process chain. The asynchronous iterative sampling mode refers to setting a sampling frequency, using this frequency as a constraint, and iteratively collecting data from different combinations of the aforementioned binary arrays sequentially. Each time, only one set of processing-side and emission-side sensor data is sampled before switching to the next set, continuously iterating to form a continuous asynchronous array detection chain. Enterprise waste gas includes... Smoke , , wait.

[0025] Specifically, firstly, in enterprise waste gas emission monitoring, to achieve data correlation between the processing and emission stages, it is necessary to integrate multi-source sensors from the enterprise project. These sensors are not of a single type, but rather encompass two categories: one is built-in sensors, i.e., sensors mounted on the production equipment itself, such as the real-time temperature sensor built into the processing reactor, which can collect temperature data within the range of 10-300℃, and the operating current sensor built into the machine tool, which monitors the operating current of 0-50A; the other is external sensors, i.e., additional supplementary monitoring equipment, such as an external raw material feed rate metering sensor on the processing side with an accuracy of ±0.1kg / h, and an external waste gas composition spectrometer on the emission side, which can identify... Smoke , , The concentration of waste gas from enterprises, with a resolution of 0.1 ppm, and flow rate sensors within the flue, etc.

[0026] These sensors are distributed along the process chain, covering raw material pretreatment, reaction synthesis, and finished product processing on the processing side, and waste gas collection, purification, and end-of-pipe emission on the emission side, forming a complete sensor coverage network. This integration allows for the simultaneous acquisition of key process parameters (such as reaction temperature and raw material consumption) on the processing side and waste gas characteristic data (such as pollutant concentration and emission flow rate) on the emission side. This provides a comprehensive and corresponding data source for subsequent sampling based on binary arrays, enabling a direct correlation between processing and emission data.

[0027] Next, based on this sensor coverage network, when determining the composition of the binary array, it is necessary to establish a one-to-one association combination logic based on the sensor data from the processing side and the emission side. The processing side sensor data comes from multi-source sensors in each process stage, covering core parameters of equipment operation, such as real-time pressure of the reactor, raw material feed rate, and motor speed; the emission side sensor data comes from sensors in the waste gas treatment and emission stages, including enterprise waste gas concentration, particulate matter size distribution, and waste gas flow rate at the emission outlet.

[0028] The rule for constructing binary arrays is as follows: select one item from any single set of data on the processing side and pair it with any single set of data on the emission side to form a binary array. For example, the real-time pressure of the reactor and the enterprise's exhaust gas concentration can be combined into one binary array; the raw material feed rate and particulate matter size distribution can be combined into another binary array; the motor operating speed and the exhaust gas flow rate can be combined into yet another binary array, and so on. This achieves comprehensive pairing of all single sets of sensor data on the processing and emission sides, laying the foundation for subsequent data correlation analysis of the impact of the processing stage on the emission status.

[0029] Subsequently, an asynchronous iterative sampling mode is deployed, setting a reasonable sampling frequency, such as once per second, and sampling is performed sequentially according to the combination method of the binary arrays described above. Each sampling only targets the currently selected set of processing and emission side sensor data. After completion, the sampling switches to the next set of combinations and continues iterating to form a continuous binary array sampling sequence.

[0030] Through the above steps, effective integration and correlation sampling of multi-source data were achieved, providing a coherent data foundation covering the processing and emission stages for subsequent anomaly detection, while avoiding the burden of synchronous processing of massive amounts of data.

[0031] Step A200: According to the asynchronous iterative sampling mode, configure the sampler and the anomaly detector, embed them in the industrial control system, and establish a communication network of sampler-multi-source sensor-anomaly detector. The anomaly detector has an embedded constraint library based on energy conservation and mass conservation.

[0032] In this embodiment, the sampler is embedded in the industrial control system, forming a communication network with multi-source sensors and anomaly detectors to collect binary array data from the processing and emission sides. The anomaly detector is used to determine the anomaly detection results and perform anomaly tracing and location. The industrial control system integrates the sampler and anomaly detector, receives the sampled array data and processes it, realizing the communication network of the sampler, multi-source sensors, and anomaly detectors, and finally performs industrial control alarm management based on the anomaly source and anomaly level at the response end.

[0033] Optionally, the sampler configuration must strictly adapt to the deployed asynchronous iterative sampling mode. First, based on the combinational logic and iteration order of the binary array, a corresponding sampling sequence is set for the sampler. For example, sampling can be performed cyclically in the order of processing-side sensor point A → emission-side sensor point a, and processing-side sensor point B → emission-side sensor point b. The sampling frequency parameter is also configured, such as completing one set of binary array sampling per second, to ensure continuous acquisition of the associated data between the processing and emission sides. Simultaneously, a data format conversion module is integrated into the sampler to uniformly convert heterogeneous data (such as voltage signals and digital signals) output from multiple sensor sources into a preset array format for easier subsequent processing.

[0034] After configuring the sampler, embed it into the industrial control system along with the built anomaly detector. The construction of the anomaly detector is described in detail in A210-A270. This process requires establishing a physical connection between the device and the industrial control system through hardware interface adapters (such as RS485 or Ethernet interfaces), and loading the corresponding drivers so that the industrial control system can recognize and schedule the sampler and the anomaly detector. For example, the central controller of the industrial control system can send start / stop sampling commands to the sampler and receive intermediate processing results returned by the anomaly detector.

[0035] Subsequently, a communication network was established connecting the sampler, multi-source sensors, and anomaly detectors. LoRa wireless sensor network technology was used for data exchange between the sampler and the multi-source sensors, ensuring that the sampler could acquire real-time data from each sensor point with a transmission latency controlled within 50ms. A dedicated line was established between the sampler and the anomaly detector via an industrial Ethernet line, transmitting the sampled binary array data in packet form. Each packet contained key information such as timestamps, processing-side data, and emission-side data. The anomaly detector maintained communication with the industrial control system via an internal bus, providing feedback on data processing status and anomaly detection results, forming a complete data transmission link.

[0036] By configuring samplers adapted to asynchronous iterative mode, embedding deployment devices, and establishing a multi-node communication network, the orderly collection and efficient flow of multi-source data were realized, providing a stable hardware and communication foundation for collaborative processing of anomaly detection.

[0037] Step A300: Trigger the continuous binary combination and sampling drive based on the sampler, determine the sampled data, and trigger dual-channel data processing based on the anomaly detector and counterfactual deduction based on the anomaly emission to determine the anomaly detection result. Perform industrial control alarm management at the response end of the central control system.

[0038] In one embodiment of this application, firstly, a sampling frequency is set, and the combination iterative acquisition of binary arrays and anomaly detector decision are performed with the constraint of the frequency. The detection chain based on asynchronous array is integrated and determined, dynamic trends are analyzed to generate standard or non-standard dynamic detection data, and static and dynamic detection data are added to the anomaly detection results. The specific steps are described in detail in A381-A385.

[0039] Next, the process chain of the enterprise project is determined and a multi-source sensor terminal with built-in and external sensors of distributed equipment is integrated. A binary array containing data on the processing side and data on the emission side is introduced. Sampling based on the binary array is performed on the multi-source sensor terminal to determine the sampling array. The specific steps are explained in detail in A310-A330.

[0040] Then, dual-path data processing based on anomaly detectors is triggered. After the industrial control system receives the sampled array, it imports the primary data into the first derivation branch to determine the first emission data by causal derivation under the constraints of mass conservation and energy conservation. The sampled data is then imported into the second data branch to integrate and determine the second array. After verification through the lateral interaction channel, the emission standard status identifier is generated. The specific steps are explained in detail in A340-A360.

[0041] If the calibration is inconsistent, a non-standard emission condition label is generated. Static detection data is determined based on the standard or non-standard emission condition label. Anomaly tracing and location based on counterfactual inference is performed to determine the source of the anomaly. The specific steps are explained in detail in A371-A373.

[0042] Finally, industrial control alarm management is performed at the central control system response end. This involves deploying corresponding early warning modes based on the source of the anomaly (processing side or emission side), deploying multiple early warning levels based on the anomaly level, and then performing exhaust emission anomaly early warning management based on the operating condition system. The specific steps are explained in detail in A391-A393.

[0043] Furthermore, step A200 in the method provided in this application embodiment includes:

[0044] A210: Time-varying causality based on process stage, wherein process characteristics are the cause and emission characteristics are the effect, the emission characteristics including waste gas type and waste gas flow rate.

[0045] A220: Based on the time-varying causality, construct a first derivation branch, wherein the first derivation branch performs the derivation from process characteristics to emission characteristics.

[0046] A230: Construct a second data branch using real-time array integration.

[0047] A240: Parallelize the first derivation branch and the second data branch, introduce counterfactual derivation based on lateral interaction, and construct an anomaly detector.

[0048] In this embodiment, time-varying causality is a dynamic causal relationship based on changes in the process stages of enterprise production, with process characteristics as the cause and emission characteristics as the result.

[0049] Specifically, when configuring anomaly detectors, the time-varying causal relationship based on the process stage should be clarified first. That is, according to the different process stages of enterprise production, such as raw material pretreatment, high-temperature reaction, and finished product cooling, the dynamic correspondence between process characteristics and emission characteristics should be determined, as shown in Table 1. That is, process characteristics are the cause, covering reaction temperature, raw material input rate, equipment operating power, etc.; emission characteristics are the effect, including waste gas type and waste gas flow rate.

[0050] Based on the aforementioned time-varying causality, a first derivation branch is constructed. This branch takes process characteristic data as input and performs derivation through a built-in causal model: for example, inputting reaction temperature and raw material input rate, and combining energy conservation (balance between reaction energy consumption and energy carried by waste gas) and mass conservation (ratio of raw material consumption to waste gas generation) constraints, the theoretical emission characteristics are derived. Concentration and flow rate form the derivation results.

[0051] Furthermore, the causal model is constructed with the time-varying causal relationships of each process stage as its core framework. First, the correlation between process characteristics and emission characteristics in different process stages is analyzed to clarify which process characteristic parameters (such as reaction temperature and feed rate) directly affect emission characteristics (such as waste gas type and flow rate) at a specific stage. Based on this, the constraints of energy and mass conservation are transformed into quantitative logic within the model. For example, the energy consumed in the reaction must be balanced with the energy carried by the waste gas, and the amount of feed consumed must have a reasonable proportional relationship with the amount of waste gas generated. By integrating these correlation patterns and conservation constraints, a derivation path from process characteristic parameters to emission characteristic parameters is formed. This allows the model to automatically deduce the theoretically expected emission characteristics under a given process state based on the preset correlation logic and conservation rules when specific process characteristic data for a particular process stage are input, thus achieving causal derivation from process characteristics to emission characteristics.

[0052] Simultaneously, when constructing the second data branch, its core function is clearly defined as integrating real-time collected actual data. This branch establishes direct communication with the sampler, continuously receiving binary arrays collected by the sampler in an asynchronous iterative mode. These arrays consist of real-time process parameters from the processing side (such as the operating status data of a certain reaction stage) and real-time monitoring data from the emission side (such as the exhaust gas emission data of the corresponding stage), with each data set accompanied by a collection time identifier. During the integration process, the branch will correlate and integrate binary array data corresponding to the same process stage within the same time period based on the time series and the correlation of process stages. For example, it will match the reaction status data from the processing side within a certain minute with the exhaust gas data from the emission side during the same period, forming a structured real-time array containing the actual process parameters and corresponding emission data for that time period. This integration is not a simple splicing, but rather ensures the correspondence between the actual process parameters and emission data in terms of time and process logic, thereby accurately reflecting the actual process and emission correlation under the current production state, providing a reliable actual monitoring basis for subsequent comparison with the theoretical results of the first derivation branch.

[0053] Subsequently, the theoretical derivation results of the first derivation branch and the actual monitoring results of the second data branch are processed in parallel, and the data are compared through a lateral interaction channel. When there is a discrepancy between the two, counterfactual derivation is introduced, that is, assuming that the process characteristics remain unchanged, the deviation of the emission data is analyzed to see if it conforms to the laws of energy conservation and mass conservation. If it does not conform, it is judged as an anomaly. Through this dual-branch parallel and counterfactual derivation logic, an anomaly detector that can dynamically correlate process and emissions and accurately identify anomalies is finally constructed.

[0054] By clarifying time-varying causality, constructing a dual-branch system, and introducing counterfactual derivation, this anomaly detector achieves dynamic correlation analysis between process characteristics and emission characteristics, thereby improving the accuracy of anomaly detection in enterprise exhaust emissions.

[0055] Table 1: Time-varying causal relationship table of process stages

[0056]

[0057] Furthermore, step A200 in the method provided in this application embodiment includes:

[0058] A250: For the multi-source sensing terminals, determine the combination method set.

[0059] A260: Traverse the set of combination methods and determine the constraint condition library based on mass conservation and energy conservation, guided by combination sensing.

[0060] A270: Integrate the constraint library into the anomaly detector.

[0061] Optionally, firstly, a set of combination methods is determined for multi-source sensing ends. These combination methods originate from the association of sensing data between the processing side and the emission side. For example, the combination of a raw material feed rate sensor on the processing side and a sensor of a certain pollutant concentration in the exhaust gas on the emission side, or the combination of an equipment energy consumption sensor on the processing side and an exhaust gas flow and temperature sensor on the emission side, etc., covering all sensor data pairings that can reflect the association between processing and emission, forming a comprehensive set of combination methods.

[0062] Subsequently, when iterating through the set of combinations of multi-source sensors, specific constraints need to be derived for each combination of sensor data from the processing and emission sides, based on the laws of conservation of mass and energy. For the mass conservation constraint, taking the combination of raw material consumption data from the processing side and the concentration data of a certain pollutant from the emission side as an example: the total amount of the pollutant contained in the raw materials, after deducting reasonable losses during processing, should match the total amount of pollutants emitted into the exhaust gas. This can be transformed into an upper limit constraint on the concentration of the pollutant in the exhaust gas, meaning the concentration detected on the emission side should not exceed the theoretical value calculated based on the ratio of raw material consumption to losses.

[0063] For energy conservation constraints, taking the combination of equipment energy consumption data on the processing side and exhaust gas temperature and flow rate data on the emission side as an example: the portion of the energy consumed by the equipment that is converted into exhaust gas heat energy must be within a reasonable range. Combining the physical properties of the exhaust gas (such as specific heat capacity and density), the correlation constraint between exhaust gas temperature and flow rate can be derived, that is, the exhaust gas temperature at a specific flow rate should not exceed the theoretical value calculated based on the energy conversion ratio. By iterating through all sensor combinations, the derived conservation constraints are compiled and summarized to construct a constraint condition library.

[0064] Finally, the constraint library is built into the anomaly detector. This process involves integrating various constraints (such as conservation relationships corresponding to different sensor combinations) into the core processing unit of the anomaly detector in the form of algorithm modules or data models, making them the basic judgment criteria that the detector can directly call upon during runtime. After being built in, when the anomaly detector processes sampled data, it will automatically refer to these constraints to verify the rationality of the data, such as determining whether process parameters and emission data conform to the conservation laws of mass or energy, providing a unified and scientific benchmark for subsequent anomaly detection and judgment.

[0065] By defining the set of combination methods, traversing and deriving conservation constraints and embedding them, a judgment basis based on scientific laws is provided for the anomaly detector, ensuring the objectivity and accuracy of anomaly detection in exhaust gas emissions.

[0066] Furthermore, step A300 in the method provided in this application embodiment includes:

[0067] A310: Determine the process chain of the enterprise project and integrate multi-source sensing terminals, wherein the multi-source sensing terminals include distributed built-in devices and external sensors.

[0068] A320: Introduce a binary array, wherein the binary array contains metadata based on the processing side and binary metadata based on the emission side.

[0069] A330: For the multi-source sensing end, perform sampling based on a binary array to determine the sampling array.

[0070] Specifically, determining the process chain of a company's project involves first outlining the entire process from raw material input to finished product output. For example, in a chemical company, this includes raw material pretreatment, catalytic reaction, product separation, and finished product storage, clarifying the process parameters and equipment distribution for each stage. Based on this, multi-source sensors are integrated. Built-in sensors include temperature sensors on the reactor and pressure sensors on the pump, while external sensors include mass flow meters at the raw material inlet and vibration sensors on the pipeline. These sensors are distributed across various process stages, forming a sensor network covering the entire processing chain.

[0071] When introducing binary arrays, data combinations are constructed based on the parameter correlation between the processing side and the emission side: the univariate data on the processing side is a single process parameter, such as the stirring motor power in the reaction stage; the binary data on the emission side is two waste gas parameters of the corresponding stage, such as the particulate matter concentration and emission flow rate before waste gas treatment. Through this 1+2 combination, a direct data correlation between the processing state and the emission state is established.

[0072] When performing sampling based on binary arrays at multi-source sensors, data is collected once per second at a fixed frequency, following a preset binary combination order. The resulting sampling array is a structured data type, containing timestamps, processing-side metadata fields, and emission-side binary metadata fields, such as {collection time: t01, processing parameters: stirring power 25kW, emission parameters: particulate matter concentration 30mg / m³, flow rate 10m / s}. Through continuous sampling, a series of sampling arrays with a unified format are generated, serving as the basic data type for subsequent data processing.

[0073] By defining the process chain for integrated sensing, introducing binary arrays, and sampling to form a structured sampling array, the precise correlation and orderly collection of data between the processing and emission stages were achieved, providing a standardized and traceable basic data source for subsequent anomaly detection.

[0074] Furthermore, step A300 in the method provided in this application embodiment includes:

[0075] A340: The industrial control system receives the sampling array, imports the metadata in the sampling array into the first derivation branch, and performs causal derivation under the constraints of mass conservation and energy conservation to determine the first emission data.

[0076] A350: Import the sampled data into the second data branch and integrate to determine the second array.

[0077] A360: Based on the lateral interaction channel, verify the first emission data with the second array. If they match, generate an emission standard status identifier.

[0078] In this embodiment, the lateral interaction channel is an information interaction path connecting the first derivation branch and the second data branch within the anomaly detector, used to realize data transmission and verification between the two branches.

[0079] Specifically, the industrial control system and the sampler establish a data transmission link through adapted industrial communication protocols, such as Modbus and Profinet, specifically for receiving the sampled arrays collected by the sampler. These sampled arrays are arranged sequentially along a time axis, and each array contains two core pieces of information: first, a single process parameter from the processing side, i.e., primary data, such as equipment operating power and raw material reaction temperature; second, two waste gas parameters from the corresponding emission side, i.e., secondary data, such as the concentration of specific pollutants in the enterprise's waste gas and the emission velocity. During transmission, the system's built-in CRC cyclic redundancy check mechanism verifies the integrity of each transmitted array. If data loss or error is detected, retransmission is triggered, thus ensuring the accuracy of the received sampled arrays and providing a reliable raw data foundation for subsequent dual-channel data processing.

[0080] After receiving the data, the industrial control system extracts single data from the processing side of the sampling array, namely, single parameters reflecting the process status, such as raw material feed rate and equipment operating power, and imports them into the first derivation branch of the anomaly detector. This branch uses mass conservation and energy conservation as core constraints to carry out causal derivation: For raw material-related parameters on the processing side, such as the feed rate of a certain raw material, according to the principle of mass conservation, the total amount of a specific substance contained in the raw material, after deducting reasonable losses in the process, should theoretically be entirely converted into that substance in the waste gas. Combined with the preset waste gas flow range, the theoretical concentration range of that substance in the waste gas can be derived, thereby determining the first emission data; at the same time, for the equipment energy consumption parameters on the processing side, according to the principle of energy conservation, the portion of the energy consumed by the equipment that is converted into waste gas heat energy must be within a reasonable range, thereby further verifying whether the aforementioned derivation of the first emission data conforms to the process law and ensuring the rigor of the theoretical derivation.

[0081] Meanwhile, the industrial control system imports complete sampled data, including raw values ​​from both the processing and emission sides, into the second data branch. This branch first standardizes the data format, converting various signals output from different sensors, such as voltage and digital signals, into directly comparable numerical forms. Next, it performs time-series alignment, precisely associating processing-side parameters with emission-side parameters for the same time period based on the data acquisition timestamps. Finally, it integrates these into a second array containing actual processing parameters (such as equipment operating status data) and corresponding emission parameters (such as exhaust gas composition and flow rate data), i.e., the actual monitoring values.

[0082] Subsequently, the theoretically derived values ​​of the first emission data and the actual monitored values ​​of the second array are checked item by item through the lateral interaction channel, i.e., the data interaction interface inside the anomaly detector. If the deviation of the key parameters (such as pollutant concentration and flow rate) between the two is within the preset threshold, which can be +5%, the emission status is determined to be in accordance with the process law and conservation constraints, and an emission standard status label is generated, indicating that there is no anomaly in the current emission.

[0083] Through dual-channel data processing and lateral calibration using theoretical derivation and actual monitoring, the status of exhaust gas emissions was verified, providing a reliable basis for accurately determining whether emissions are abnormal.

[0084] Furthermore, step A360 in the method provided in this application embodiment includes:

[0085] A371: If inconsistent, generate a non-standard emission condition label.

[0086] A372: Determine static detection data based on the emission standard condition label or the emission non-standard condition label.

[0087] A373: Based on the aforementioned non-standard emission condition identifier, perform anomaly tracing and location based on counterfactual reasoning to determine the source of the anomaly.

[0088] In this embodiment, the non-standard condition identifier is used to mark emissions that deviate from normal process patterns during dual-channel data processing by the abnormal detector. Static detection data, determined based on the emission standard condition identifier or the emission non-standard condition identifier, reflects the emission status at a specific moment and includes information such as process parameters, emission parameters, and whether an anomaly exists at that moment.

[0089] Specifically, when the dual-channel data processing results of the anomaly detector are inconsistent, that is, when the theoretical emission data obtained from the first derivation branch deviates from the actual monitoring data integrated by the second data branch by more than a preset threshold of ±5%, the system automatically generates an emission non-standard condition identifier to mark the current emission status as deviating from the normal process pattern.

[0090] Subsequently, when determining static monitoring data based on standard or non-standard emission conditions, the static monitoring data focuses on the emission status record at a specific moment. If it is a standard condition, the matching relationship between the process parameters (such as equipment operating status data) and the corresponding emission parameters (such as exhaust gas composition and flow data) at that moment is recorded and marked as normal emission; if it is a non-standard condition, the same process parameters, actual emission parameters, and the deviation between the two are recorded and marked as abnormal emission, thus forming a static data record that can intuitively reflect whether the emission at that moment conforms to the process rules.

[0091] Finally, based on the non-standard emission condition identification, anomaly tracing and location are performed based on counterfactual deduction. This requires determining the first processing collection node and the second emission collection node for the sampling array, and using these two nodes as the location range for deduction and decision-making to determine the anomaly source. The specific steps are explained in detail in A373-1-A373-2.

[0092] By generating non-standard condition identifiers, determining static detection data, and using counterfactual reasoning to trace the source, the system enables real-time marking, status recording, and precise location of abnormal exhaust emissions, providing a clear basis for subsequent alarms and handling.

[0093] Furthermore, step A373 in the method provided in this application embodiment includes:

[0094] A373-1: For the sampling array, determine the first processing acquisition node and the second emission acquisition node.

[0095] A373-2: Using the first processing and data acquisition node and the second emission data acquisition node as the location range, perform deduction and decision-making to determine the source of the anomaly.

[0096] In one embodiment, when determining the first processing acquisition node and the second emission acquisition node based on the sampling array, it is necessary to first parse the source identifiers of the processing-side and emission-side data in the sampling array. The processing-side metadata contained in the sampling array corresponds to specific processing stage sensing devices. For example, the reactor temperature data comes from the built-in temperature sensor of reactor No. 1 in workshop, and the raw material feed rate data comes from the external flow meter of the feed pipeline. The location of these sensing devices is the first processing acquisition node. The emission-side metadata (such as the concentration in the enterprise's waste gas and the emission flow rate) corresponds to the emission stage sensing devices. For example, the enterprise's waste gas concentration data comes from the online monitoring instrument at the outlet of the waste gas treatment tower, and the flow rate data comes from the flow velocity sensor of the exhaust stack. The location of these devices constitutes the second emission acquisition node, thereby clarifying the specific physical scope of anomaly tracing.

[0097] Next, when making inference decisions based on the first processing and data acquisition node and the second emission data acquisition node determined in the above steps, the status of each node is verified step by step based on counterfactual logic. First, it is assumed that the first processing and data acquisition node is normal, meaning that the process parameters of this node (such as equipment operating status and raw material supply) meet the standards. If the emission-side parameters (such as pollutant concentration and flow rate) deviate significantly from the theoretically derived values, then the focus shifts to the second emission data acquisition node. This involves checking whether its sensing equipment has calibration deviations, data transmission failures, or whether the corresponding waste gas treatment equipment (such as purification devices and conveying pipelines) is malfunctioning. If it is assumed that the second emission data acquisition node is normal, meaning that the emission-side sensor data is accurate and the treatment equipment is functioning correctly, then the first processing and data acquisition node is traced back to check whether its process parameters have unrecorded fluctuations, whether the raw material composition is abnormal, or whether the equipment has hidden faults. Through this method of fixing one end and verifying the other, nodes with normal status are gradually eliminated, ultimately pinpointing the source of the anomaly, which may be a sensor malfunction on the emission side or an abnormal operation of the equipment on the processing side.

[0098] By identifying specific processing and emission collection nodes and making counterfactual inference decisions within their scope, the source of anomalies was traced precisely from abstract data deviations to specific equipment or sensors, providing clear direction for subsequent targeted processing.

[0099] Furthermore, step A300 in the method provided in this application embodiment includes:

[0100] A381: Set the sampling frequency.

[0101] A382: With the sampling frequency as a constraint, perform combined iterative acquisition of binary arrays and decision-making based on the anomaly detector, and integrate to determine the detection chain based on the asynchronous array.

[0102] A383: Based on the detection chain, analyze and determine the dynamic trend of gas emissions.

[0103] A384: Generate dynamic detection data based on the dynamic trend, wherein the dynamic detection data is a standard condition trend or a non-standard condition trend.

[0104] A385: Add the static detection data and dynamic detection data to the anomaly detection results.

[0105] Optionally, triggering continuous binary combination and sampling drive based on samplers requires first setting a reasonable sampling frequency according to the production rhythm and exhaust emission characteristics of the enterprise's process chain. For example, for chemical processes with fast reaction rates, the sampling frequency can be set to once per second to ensure that parameter changes in a short period of time can be captured; for links with a slow production rhythm, the sampling frequency can be set to once every 5 seconds to reduce redundant data while ensuring data validity.

[0106] Once the sampling frequency is determined, iterative sampling of binary arrays is performed with that frequency as a constraint. Specifically, the sampling is performed according to a preset combination order of processing-side and emission-side sensors, such as the reactor temperature sensor on the processing side and the emission-side sensor... The sampler sequentially collects data from various binary arrays, including concentration sensors, raw material feed rate sensors on the processing side, and exhaust gas velocity sensors on the emission side. After completing one set, it immediately switches to the next set, repeating the process. During this process, the anomaly detector simultaneously processes and decides on each collected binary array, such as determining whether the current data meets the constraints of mass and energy conservation. The continuously collected and processed arrays are integrated according to time sequence to form a detection chain based on asynchronous arrays. Due to the time difference between the collection of different combinations, the arrays in the detection chain exhibit asynchronous characteristics, but can still fully reflect the continuous correlation between the process and emissions.

[0107] Next, based on the constructed detection chain, a trend analysis algorithm is used to extract the dynamic variation patterns of gas emissions. The trend analysis algorithm first receives detection chain data based on asynchronous arrays. This data is arranged in a time series, and each group contains processing-side parameters (such as reaction temperature) and corresponding emission-side parameters (such as...). (Concentration, flow rate). The algorithm first determines the stability of the processing-side parameters by calculating the fluctuation range of a preset number of consecutive sets of data. For example, if the standard deviation is ≤5℃ when the reaction temperature is 300℃, it is considered stable, thus determining whether the process is in a stable stage. When the process is stable, the algorithm extracts the sequence features of the corresponding emission-side parameters, including the mean range and fluctuation degree. At the same time, it refers to the conservation constraint library built into the anomaly detector to verify whether the current emission parameters are within the theoretically reasonable range. If the emission parameters are continuously within the statistical feature range and meet the conservation constraints, it is determined to be a standard condition trend; if the emission parameters suddenly exceed the statistical range, the fluctuation range increases sharply, and they do not meet the conservation constraints, and this continues for multiple sets of data, it is determined to be a non-standard condition trend. For example, in 20 consecutive sets of detection chain data, if the processing-side reaction temperature is stable at 300℃, the emission-side... If the concentration remains consistently between 100-120 mg / m³ and the flow rate remains stable at 80-90 m³ / h, then the trend can be analyzed and determined to be that of standard conditions; if the reaction temperature is the same, If the concentration suddenly rises to 200 mg / m³ and continues to fluctuate, it is considered a non-standard trend.

[0108] Finally, based on the above dynamic trends, corresponding dynamic detection data is generated, namely standard condition trend or non-standard condition trend, and integrated with the static detection data obtained by a single detection in step A371 (such as the standard condition or non-standard condition identifier at a certain moment) to form a component of the anomaly detection result.

[0109] By setting the sampling frequency, executing combined iterative acquisition to form a detection chain, analyzing dynamic trends, and integrating static and dynamic data, continuous dynamic monitoring of enterprise exhaust emissions was achieved, improving the comprehensiveness and timeliness of abnormal detection results.

[0110] Furthermore, step A300 in the method provided in this application embodiment includes:

[0111] A391: Deploy an early warning mode based on the source of the anomaly. If the source of the anomaly belongs to the processing side, the first early warning mode is triggered. If the source of the anomaly belongs to the emission side, the second early warning mode is triggered.

[0112] A392: Deploy multiple early warning levels based on the anomaly level.

[0113] A393: Based on the aforementioned warning mode and multiple warning levels, implement exhaust gas emission anomaly warning management based on the operating condition system.

[0114] In one embodiment, when managing industrial control alarms at the response end of the central control system, the attribution of the anomaly source is first determined based on the anomaly tracing and location results. If the anomaly source originates from the processing side, such as reactor temperature runaway or abnormal raw material feeding, a first warning mode is triggered. If the anomaly source belongs to the emission side, such as exhaust gas treatment equipment malfunction or emission monitoring sensor anomaly, a second warning mode is activated. For example, when the anomaly source is a leak in the raw material feed valve on the processing side, the system automatically activates the first warning mode and associates it with the equipment control module of that processing stage. If the anomaly source is a stoppage in the desulfurization tower spray pump on the emission side, the second warning mode is triggered, and the emergency components of the exhaust gas treatment system are activated.

[0115] Subsequently, multiple warning levels are assigned based on the severity of the anomaly. Typically, three warning levels can be set: Level 1 (Minor) indicates emission parameters exceed standards by 5%-10%, without affecting the overall process; Level 2 (Moderate) indicates exceedances by 10%-20%, requiring timely intervention; Level 3 (Severe) indicates exceedances by more than 20%, potentially triggering compliance risks. Different levels correspond to different response intensities. For example, a Level 1 warning only displays a yellow alert box on the central control system interface and records the anomaly information; a Level 2 warning triggers an audible and visual alarm in the workshop and simultaneously sends a notification to on-duty technical personnel; a Level 3 warning, in addition to an audible and visual alarm, automatically sends an emergency notification to the company's environmental management department and key personnel, and prepares to activate the emergency emission reduction plan.

[0116] Finally, specific management of abnormal exhaust emissions is implemented by combining the early warning mode and the early warning level. If it is a Level 1 abnormality on the processing side, such as a reaction temperature fluctuation exceeding the allowable range by 5%, the first early warning mode will display the location of the abnormal equipment and real-time parameters, prompting operators to make fine adjustments. If it is a Level 3 abnormality on the emission side, such as a pollutant concentration in the exhaust gas exceeding the standard by 30%, the second early warning mode will immediately cut off the main emission channel, switch to the emergency treatment pipeline, and simultaneously trigger the highest level of audible and visual alarms and multi-level personnel notifications.

[0117] By deploying early warning modes according to the source of the anomaly, classifying early warning levels according to the severity, and combining the two to implement targeted management, accurate early warning and graded response to abnormal exhaust emissions have been achieved, improving the timeliness and effectiveness of anomaly handling.

[0118] In summary, the enterprise exhaust gas emission anomaly detection method based on multi-source data provided in this application has the following technical effects:

[0119] This application integrates multi-source sensors from enterprise projects and deploys an asynchronous iterative sampling mode based on binary arrays. Through the configuration and communication networking of samplers and anomaly detectors, continuous binary combination and sampling are triggered. Anomaly detection results are obtained through dual-channel data processing and counterfactual deduction. Combined with the industrial control alarm management of the central control system, the abnormal situation of enterprise exhaust gas emissions can be accurately detected, making the detection results of abnormal enterprise exhaust gas emissions more accurate and reliable. This achieves the technical effect of accurate detection and efficient alarm of abnormal enterprise exhaust gas emissions, meeting the needs of precise supervision and efficient management.

[0120] Example 2, as Figure 2 As shown, based on the same inventive concept as in Embodiment 1 above, this application provides an enterprise exhaust gas emission anomaly detection system based on multi-source data, the system comprising:

[0121] The asynchronous iterative sampling mode deployment module 1 is used to integrate the multi-source sensing terminals of the enterprise project and deploy an asynchronous iterative sampling mode based on a binary array, wherein the binary array is determined by any combination of sensors on the processing side and the emission side.

[0122] The communication network construction module 2 is used to configure the sampler and the anomaly detector according to the asynchronous iterative sampling mode, embed them in the industrial control system, and establish a communication network of sampler-multi-source sensor-anomaly detector. The anomaly detector has an embedded constraint condition library based on energy conservation and mass conservation.

[0123] The industrial control alarm management execution module 3 is used to trigger the continuous binary combination and sampling drive based on the sampler, determine the sampled data, and trigger the dual-channel data processing based on the anomaly detector and the counterfactual deduction based on the anomaly emission to determine the anomaly detection result, and perform industrial control alarm management at the response end of the central control system.

[0124] Furthermore, the communication network construction module 2 is used to perform the following steps:

[0125] Using time-varying causality based on process stages, where process characteristics are the cause and emission characteristics are the effect, the emission characteristics include waste gas type and waste gas flow rate; a first derivation branch is constructed based on the time-varying causality, wherein the first derivation branch performs derivation from process characteristics to emission characteristics; a second data branch is constructed by integrating real-time arrays; the first derivation branch and the second data branch are parallelized, and counterfactual derivation based on lateral interaction is introduced to construct an anomaly detector.

[0126] Furthermore, the communication network construction module 2 is used to perform the following steps:

[0127] For the multi-source sensing end, a set of combination methods is determined; the set of combination methods is traversed, and a constraint condition library based on mass conservation and energy conservation is determined with the combination sensing as the guide; the constraint condition library is built into the anomaly detector.

[0128] Furthermore, the industrial control alarm management execution module 3 is used to perform the following steps:

[0129] The process chain of the enterprise project is determined, and multi-source sensing terminals are integrated, wherein the multi-source sensing terminals include distributed built-in and external sensors; a binary array is introduced, wherein the binary array contains primary data based on the processing side and secondary data based on the emission side; sampling based on the binary array is performed for the multi-source sensing terminals to determine the sampling array.

[0130] Furthermore, the industrial control alarm management execution module 3 is used to perform the following steps:

[0131] The industrial control system receives the sampling array, imports the primitive data in the sampling array into the first derivation branch, performs causal derivation under the constraints of mass conservation and energy conservation, and determines the first emission data; imports the sampling data into the second data branch, integrates and determines the second array; and verifies the first emission data and the second array according to the lateral interaction channel. If they are consistent, an emission standard status identifier is generated.

[0132] Furthermore, the industrial control alarm management execution module 3 is used to perform the following steps:

[0133] Based on the emission standard condition identifier or the emission non-standard condition identifier, determine the static detection data; based on the emission non-standard condition identifier, perform anomaly tracing and location based on counterfactual reasoning to determine the anomaly source.

[0134] Furthermore, the industrial control alarm management execution module 3 is used to perform the following steps:

[0135] For the sampling array, a first processing and acquisition node and a second emission acquisition node are determined; using the first processing and acquisition node and the second emission acquisition node as the positioning range, inference decisions are made to determine the source of the anomaly.

[0136] Furthermore, the industrial control alarm management execution module 3 is used to perform the following steps:

[0137] Set a sampling frequency; using the sampling frequency as a constraint, perform combined iterative acquisition of binary arrays and decision-making based on the anomaly detector, and integrate to determine a detection chain based on an asynchronous array; based on the detection chain, analyze and determine the dynamic trend of gas emissions; generate dynamic detection data according to the dynamic trend, wherein the dynamic detection data is a standard condition trend or a non-standard condition trend; add the static detection data and the dynamic detection data to the anomaly detection results.

[0138] Furthermore, the industrial control alarm management execution module 3 is used to perform the following steps:

[0139] Based on the source of the anomaly, a warning mode is deployed, wherein if the source of the anomaly belongs to the processing side, a first warning mode is triggered, and if the source of the anomaly belongs to the emission side, a second warning mode is triggered; based on the anomaly level, multiple warning levels are deployed; and based on the warning mode and the multiple warning levels, exhaust gas emission anomaly warning management based on the operating condition system is implemented.

[0140] The enterprise exhaust gas emission anomaly detection system based on multi-source data provided in the embodiments of the present invention can execute the enterprise exhaust gas emission anomaly detection method based on multi-source data provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.

[0141] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0142] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for detecting abnormal enterprise exhaust emissions based on multi-source data, characterized in that, The method includes: The multi-source sensing terminals of the integrated enterprise project are deployed in an asynchronous iterative sampling mode based on binary arrays, wherein the binary array is determined by any combination of sensors from the processing side and the emission side. According to the asynchronous iterative sampling mode, a sampler and anomaly detector are configured and embedded in the industrial control system to establish a communication network of sampler-multi-source sensor-anomaly detector. The anomaly detector has an embedded constraint library based on energy conservation and mass conservation. Trigger the continuous binary combination and sampling drive based on the sampler, determine the sampled data, and trigger dual-channel data processing based on the anomaly detector and counterfactual deduction based on the anomaly emission to determine the anomaly detection result, and perform industrial control alarm management at the response end of the central control system; The configuration of the anomaly detector includes: The time-varying cause and effect is based on process stage, wherein process characteristics are the cause and emission characteristics are the effect, and the emission characteristics include waste gas type and waste gas flow rate. Based on the time-varying causality, a first derivation branch is constructed, wherein the first derivation branch performs the derivation from process characteristics to emission characteristics; By integrating real-time arrays, a second sampling branch is constructed; The first derivation branch and the second data branch are parallelized, and counterfactual derivation based on lateral interaction is introduced to construct an anomaly detector.

2. The method for detecting abnormal enterprise exhaust emissions based on multi-source data as described in claim 1, characterized in that, The anomaly detector has an embedded constraint library based on energy conservation and mass conservation, including: For the aforementioned multi-source sensing terminals, a set of combination methods is determined; Traverse the set of combination methods and determine the constraint condition library based on mass conservation and energy conservation, guided by the combination sensing. The constraint library is built into the anomaly detector.

3. The method for detecting abnormal enterprise exhaust emissions based on multi-source data as described in claim 1, characterized in that, Sampling-driven, determining the sampled data, including: Determine the process chain of the enterprise project and integrate multi-source sensing terminals, wherein the multi-source sensing terminals include distributed built-in devices and external sensors; A binary array is introduced, wherein the binary array contains metadata based on the processing side and binary metadata based on the emission side; For the multi-source sensing end, sampling based on a binary array is performed to determine the sampling array.

4. The method for detecting abnormal enterprise exhaust emissions based on multi-source data as described in claim 3, characterized in that, Triggering dual-path data processing based on anomaly detectors includes: The industrial control system receives the sampling array, imports the primitive data in the sampling array into the first derivation branch, and performs causal derivation under the constraints of mass conservation and energy conservation to determine the first emission data; The sampled data is imported into the second sampling branch and integrated to determine the second array; Based on the lateral interaction channel, the first emission data and the second array are checked. If they match, an emission standard status identifier is generated.

5. The method for detecting abnormal enterprise exhaust emissions based on multi-source data as described in claim 4, characterized in that, If there is a discrepancy, a non-standard emission condition label will be generated; Static detection data are determined based on the emission standard condition label or the emission non-standard condition label; Based on the non-standard emission condition identifier, perform anomaly tracing and location based on counterfactual reasoning to determine the source of the anomaly.

6. The method for detecting abnormal enterprise exhaust emissions based on multi-source data as described in claim 5, characterized in that, Performing counterfactual reasoning-based anomaly tracing and localization includes: For the aforementioned sampling array, determine the first processing acquisition node and the second emission acquisition node; Using the first processing and data acquisition node and the second emission data acquisition node as the location range, inference decisions are made to determine the source of the anomaly.

7. The method for detecting abnormal enterprise exhaust emissions based on multi-source data as described in claim 6, characterized in that, Continuous binary combination and sampling-driven, including: Set the sampling frequency; With the sampling frequency as a constraint, perform combined iterative acquisition of binary arrays and decision-making based on the anomaly detector, and integrate to determine the detection chain based on the asynchronous array; Based on the detection chain, the dynamic trend of gas emissions is analyzed and determined; Based on the dynamic trend, dynamic detection data is generated, wherein the dynamic detection data is a standard condition trend or a non-standard condition trend; The static and dynamic detection data are added to the anomaly detection results.

8. The method for detecting abnormal enterprise exhaust emissions based on multi-source data as described in claim 1, characterized in that, Industrial control alarm management is performed at the response end of the central control system, including: Based on the source of the anomaly, a warning mode is deployed. If the source of the anomaly belongs to the processing side, the first warning mode is triggered. If the source of the anomaly belongs to the emission side, the second warning mode is triggered. Deploy multiple early warning levels based on the anomaly level; Based on the aforementioned warning mode and multiple warning levels, implement abnormal exhaust emission warning management based on the operating condition system.

9. An enterprise exhaust gas emission anomaly detection system based on multi-source data, characterized in that, The system is used to implement the enterprise exhaust gas emission anomaly detection method based on multi-source data as described in any one of claims 1-8, the system comprising: The asynchronous iterative sampling mode deployment module is used to integrate multi-source sensors in enterprise projects and deploy an asynchronous iterative sampling mode based on binary arrays, wherein the binary array is determined by any combination of sensors from the processing side and the emission side. The communication network construction module is used to configure the sampler and the anomaly detector according to the asynchronous iterative sampling mode, embed and deploy them in the industrial control system, and establish a communication network of sampler-multi-source sensor-anomaly detector. The anomaly detector has an embedded constraint condition library based on energy conservation and mass conservation. The industrial control alarm management execution module is used to trigger the continuous binary combination and sampling drive based on the sampler, determine the sampled data, and trigger the dual-channel data processing based on the anomaly detector and the counterfactual deduction based on the anomaly emission to determine the anomaly detection result and perform industrial control alarm management at the response end of the central control system. The configuration of the anomaly detector in the communication network construction module includes the following steps: The time-varying cause and effect is based on process stage, wherein process characteristics are the cause and emission characteristics are the effect, and the emission characteristics include waste gas type and waste gas flow rate. Based on the time-varying causality, a first derivation branch is constructed, wherein the first derivation branch performs the derivation from process characteristics to emission characteristics; By integrating real-time arrays, a second sampling branch is constructed; The first derivation branch and the second data branch are parallelized, and counterfactual derivation based on lateral interaction is introduced to construct an anomaly detector.

Images