An enterprise pollution discharge risk identification method based on multi-source data fusion

Abstract

The application discloses an enterprise pollution risk identification method based on multi-source data fusion, comprising the following steps: constructing an enterprise pollution risk correlation data set and performing time sequence correlation normalization to obtain an enterprise pollution time window record set; performing multi-source credible benchmark construction to obtain a pollution risk calculation benchmark; performing pollution risk evidence construction to obtain a pollution risk evidence set; performing risk attribution determination to obtain a risk attribution result; and performing fusion grading calculation to obtain an enterprise pollution risk grading result. The application can identify enterprises that do not exceed the standard but have pollution risks, reduce misjudgment and missed judgment caused by data anomalies, and improve the accuracy and interpretability of enterprise pollution risk grading.

Description

Technical Field

[0001] This invention belongs to the field of enterprise pollution discharge risk classification calculation technology, specifically involving an enterprise pollution discharge risk identification method, electronic device and storage medium based on multi-source data fusion. Background Technology

[0002] In the regulatory scenarios for key polluting enterprises, the pollution risks of companies in industries such as chemical, printing and dyeing, electroplating, papermaking, pharmaceuticals, building materials, and metal surface treatment are typically related to pollution discharge permit constraints, online monitoring data at discharge outlets, production load data, operational data of pollution control facilities, energy and material consumption data, historical regulatory records, regional environmental quality data, and meteorological and hydrological data. These data respectively reflect the enterprise's permitted emission boundaries, actual emission status, changes in production conditions, response status of treatment facilities, historical risk behaviors, and external environmental carrying capacity.

[0003] Currently, enterprise pollution discharge risk classification is mostly completed using methods such as judging whether the discharge outlet exceeds the standard through online monitoring, comparing the discharge permit limit, statistics of historical violations, statistics of complaints and reports, or fixed-weight comprehensive scoring. While these methods can form a risk level to some extent, they usually treat various data as independent scoring items, focusing on whether a single indicator is abnormal or the superposition of scores from multiple indicators, lacking verification of the continuous relationship between production conditions, the response of pollution control facilities, and changes in discharge outlet emissions.

[0004] However, corporate pollution risks are not solely manifested in exceeding emission standards in monitoring data. For example, situations such as increased production load without a corresponding change in the operational intensity of pollution control facilities, normal operational signals from pollution control facilities without a corresponding change in equipment power consumption or reagent consumption, consistently stable emission data despite continuous fluctuations in production conditions, and emission concentrations not exceeding standards but approaching permissible limits can all affect the assessment of pollution risk. Existing methods struggle to distinguish between genuine pollution risk, data anomaly risk, and insufficient treatment response risk, and also fail to simultaneously output risk levels, risk sources, credibility markers, and reasons for graded adjustments. Summary of the Invention

[0005] In view of the aforementioned existing problems, the present invention is proposed.

[0006] This invention provides a method, electronic device, and storage medium for identifying enterprise pollution discharge risks based on multi-source data fusion. The purpose is to at least partially solve the problems of existing enterprise pollution discharge risk classification methods that treat various types of data as independent scoring items, lack verification of the continuous relationship between production conditions, pollution control facility response, and changes in discharge outlet emissions, make it difficult to identify hidden risk states where discharge outlet concentrations do not exceed standards but treatment response is consistently insufficient, and also make it difficult to distinguish between real pollution discharge risks, data anomaly risks, and risks of insufficient treatment response.

[0007] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for identifying enterprise pollution discharge risks based on multi-source data fusion, including: Construct a dataset of enterprise pollution discharge risk association for the target enterprise and perform time-series association normalization to obtain a set of enterprise pollution discharge time window records; A multi-source reliable benchmark is constructed from the enterprise's wastewater discharge time window record set to obtain a wastewater discharge risk calculation benchmark. Evidence for pollution discharge risk is constructed based on the pollution discharge risk calculation benchmark to obtain a pollution discharge risk evidence set; The pollution discharge risk evidence set is subjected to risk attribution determination to obtain risk attribution results; The risk attribution results are integrated and graded to obtain the enterprise's pollution discharge risk classification results.

[0008] As a preferred implementation, the construction of a multi-source reliable benchmark for the enterprise's wastewater discharge time window record set to obtain a wastewater discharge risk calculation benchmark includes: Online monitoring records, production load records, pollution control facility operation records, energy and material records, and external environmental records are extracted from the enterprise's pollution discharge time window records to obtain multi-source pollution discharge records to be verified; Completeness verification, continuity verification, and fluctuation status verification are performed on the multi-source discharge records to be verified to obtain the basic quality verification results; The multi-source coupling deviation credibility algorithm is used to perform window consistency verification on the basic quality verification results to obtain multi-source credibility evaluation results; The credibility of the multi-source credibility evaluation results is written to obtain a time window record with credibility annotation; By overlaying the time window records with reliable annotations, a baseline for calculating pollution discharge risk is obtained.

[0009] As a preferred embodiment, the step of overlaying trusted time window records to obtain a pollution discharge risk calculation benchmark includes: Historical regulatory records and environmental complaint records are extracted from enterprise wastewater discharge time windows to obtain baseline supplementary records; The historical regulatory records in the baseline supplementary records are merged by combining event timeliness and rectification review status to obtain historical regulatory correction records; The environmental complaint records in the baseline supplementary records are merged by grouping similar complaints and merging complaints by time statute of limitations to obtain the environmental complaint correction records; By overlaying historical regulatory correction records, environmental complaint correction records, and credibly labeled time window records, a baseline for calculating pollution discharge risk is obtained.

[0010] As a preferred implementation, the construction of pollution risk evidence based on the pollution risk calculation benchmark to obtain a pollution risk evidence set includes: Discharge permit constraints and reliable-corrected discharge outlet monitoring records are extracted from the discharge risk calculation benchmark to obtain permit boundary verification data. Perform permit boundary verification on the permit boundary verification data to obtain permit proximity evidence records; By expanding the evidence of pollution discharge risk from the records of evidence of proximity to permits, a set of evidence of pollution discharge risk is obtained.

[0011] As a preferred implementation, the process of expanding and constructing pollution risk evidence from the permission proximity evidence records to obtain a pollution risk evidence set includes: Production load records, pollution control facility operation records, and energy and material records are extracted from the pollution discharge risk calculation benchmark to obtain governance response verification data; The production load records in the governance response verification data are quantified to obtain the operating condition change records; The pollution control facility operation records and energy and material records in the pollution control response verification data are quantified to obtain the pollution control response records; The working condition-governance response coupling deviation index algorithm is used to perform window response verification on the working condition change records and governance response records to obtain governance response deviation evidence records. Evidence records of permit proximity and governance response deviation are merged to obtain a pollution risk evidence set.

[0012] As a preferred implementation, the step of determining the risk attribution of the pollution discharge risk evidence set to obtain the risk attribution result includes: Extract records of deviation from treatment response, discharge outlet response, and permit proximity from the pollution risk evidence set to obtain input data for attribution determination; Risk attribution determination is performed on the input data for attribution determination to obtain the risk attribution results.

[0013] As a preferred embodiment, the process of performing risk attribution determination on the attribution determination input data to obtain risk attribution results includes: The consistency of operating conditions, treatment, and emissions is verified on the input data for attribution determination to obtain the consistency determination results; The attribution determination input data is combined and merged to obtain the permission governance combined state; Risk sources are stratified based on the status of the licensing governance portfolio and the consistency determination results to obtain risk attribution results.

[0014] As a preferred implementation, the risk source stratification of the license governance portfolio status and consistency determination results to obtain risk attribution results includes: The credibility status of the permitted proximity evidence records in the attribution determination input data is merged to obtain the credibility status of the permitted evidence. The credibility status of governance response deviation evidence records in the attribution determination input data is merged to obtain the credibility status of governance evidence. The credibility attribution correction is performed on the credibility status of licensing evidence, the credibility status of governance evidence, and the combined status of licensing and governance to obtain the credibility attribution results. Risk sources are stratified based on credible attribution results and consistency determination results to obtain risk attribution results.

[0015] As a preferred implementation, the process of fusing and classifying the risk attribution results to obtain the enterprise's pollution discharge risk classification results includes: Extract the risk source status and credible attribution status from the risk attribution results to obtain the hierarchical calculation input data; Constraint corrections are applied to the risk source status and regional environmental carrying capacity status in the hierarchical calculation input data to obtain environmental constraint attribution results; The attribution path gating and hierarchical algorithm is used to determine the hierarchical path of the environmental constraint attribution results, and the attribution hierarchical data determined by the path is obtained. Risk level mapping is performed on the attribution classification data determined by the path to obtain the risk level mapping results; The risk level mapping results are merged to obtain the enterprise pollution discharge risk classification results.

[0016] As a preferred embodiment, after obtaining the enterprise's pollution discharge risk classification results, the method further includes: Obtain the manual review results, enterprise rectification feedback, and actual exceedance events corresponding to the enterprise's pollution discharge risk classification results, and obtain rolling correction records; The rolling correction record is used to identify the correction object to obtain the parameter object to be updated. The parameters involved in the parameter update object are updated as follows: coupling deviation attenuation coefficient in the multi-source coupling deviation credibility algorithm, cumulative attenuation coefficient in the working condition governance response coupling deviation index algorithm, and fusion weight in the attribution path gating classification algorithm.

[0017] In a second aspect, the present invention provides an electronic device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method in the first aspect.

[0018] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method in the first aspect.

[0019] The enterprise pollution discharge risk identification method, electronic device, and storage medium provided by this invention, based on multi-source data fusion, have the following advantages compared with existing technologies: 1. Before incorporating multi-source data into risk calculation, this invention performs a window-based consistency verification of each data source based on the coupling deviation relationship between production load, pollution control facility operation, and discharge outlet emissions. The coupling credibility coefficient is then written into the corresponding time window record for subsequent evidence construction and hierarchical calculation. Therefore, even if a data source is complete and without missing data, if its direction of change contradicts upstream or downstream related data sources, it will still be marked as low credibility and its weight in risk calculation will be reduced. This helps reduce the problems of sensor malfunctions being misjudged as high-risk, stable online monitoring data being misjudged as low-risk, and data loss during equipment downtime leading to distorted risk levels.

[0020] 2. This invention does not treat treatment response deviation as a binary label. Instead, it distinguishes between occasional treatment fluctuations and persistent insufficient treatment response through a single-window treatment response deviation index and a cumulative treatment response deviation index, and uses a confidence coefficient to suppress false deviations caused by low-confidence data. Therefore, enterprises whose discharge concentrations have not yet exceeded standards but whose pollution control facility operation intensity consistently lags behind changes in production conditions can be identified to a certain extent in advance. Furthermore, short-term production fluctuations or data anomalies within a single time window will not be excessively amplified as a risk of insufficient treatment, which helps reduce the coexistence of missed risks before exceeding standards and false alarms due to occasional deviations.

[0021] 3. This invention determines the risk attribution type based on the combination of evidence of proximity to permits and evidence of deviation from remediation responses. The attribution type then selects the corresponding threshold value, fusion weight, and environmental carrying capacity correction method for tiered calculation. The tiered result simultaneously carries the risk level, risk source, credibility marker, and reason for tier adjustment. Therefore, enterprises with the same degree of proximity to permits but different causes enter different tiered calculation paths. Emission risks due to insufficient remediation and risks from process backgrounds approaching permit boundaries are no longer grouped into the same level. This helps reduce the problem of different risk types being masked by a uniform score and regulators being unable to directly locate the risk causes from the tiered results. Attached Figure Description

[0022] Figure 1 An exemplary flowchart of the enterprise pollution discharge risk identification method provided in the embodiments of the present invention.

[0023] Figure 2 A comparison diagram showing the effects of the enterprise pollution discharge risk identification method provided in this embodiment of the invention with existing technologies. Detailed Implementation

[0024] To make the technical means, creative features, and achieved objectives and effects of this invention easier to understand, the invention is further described below with reference to specific embodiments. However, the following embodiments are merely preferred embodiments of this invention and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments described herein without creative effort are all within the protection scope of this invention. Unless otherwise specified, the experimental methods in the following embodiments are conventional methods, and the materials and reagents used in the following embodiments are commercially available unless otherwise specified.

[0025] Example 1 combined Figure 1 The following is an exemplary flowchart of the enterprise pollution discharge risk identification method, and the specific implementation steps are as follows: A dataset of enterprise pollution discharge risk associations for the target enterprise is constructed and subjected to time-series association normalization to obtain a set of enterprise pollution discharge time window records.

[0026] The enterprise pollution discharge risk association dataset refers to a multi-source data set formed by associating the target enterprise's pollution discharge permit records, discharge outlet online monitoring records, production load records, pollution control facility operation records, energy and material records, historical regulatory records, environmental complaint records, regional environmental quality records, and meteorological and hydrological records one by one according to the enterprise identifier and discharge outlet number. It is used to provide a unified data source for subsequent risk calculation.

[0027] The aforementioned time-series correlation normalization refers to determining the corresponding time window length according to the data source category, mapping the original sampling timestamps of each record to a unified time window start and end interval, and aggregating and aligning multiple records within the same time window to solve the problem of inconsistent sampling frequencies from different data sources.

[0028] The enterprise discharge time window record set refers to the ordered set of complete time window records formed by various types of data from the same discharge outlet within the same time window after time-series correlation and normalization, which is used as input for the construction of multi-source reliable benchmarks.

[0029] The Unified Social Credit Code and Discharge Permit Number of the target enterprise are read, and used as the data association primary key. Based on the data association primary key, the following records are extracted from the Discharge Permit Management System, Online Monitoring System, Production Load Records, Discharge Control Facility Operation System, Energy Metering System, Regulatory Enforcement System, Complaint and Reporting System, Regional Environmental Monitoring System, and Meteorological and Hydrological System: Discharge Permit Records, Discharge Outlet Online Monitoring Records, Production Load Records, Discharge Control Facility Operation Records, Energy and Material Records, Historical Regulatory Records, Environmental Complaint Records, Regional Environmental Quality Records, and Meteorological and Hydrological Records. These records are then linked one by one according to the data association primary key and the discharge outlet number carried by each record to obtain the enterprise discharge risk association dataset.

[0030] After obtaining the enterprise pollution discharge risk association dataset, time-series association normalization is performed on the dataset. The original sampling timestamp and data source category of each record are read, and the corresponding time window length is determined according to the data source category. The original sampling timestamp of each record is mapped to the start and end interval of its respective time window. Multiple records with the same discharge outlet number and the same data source category within the same time window are aggregated to generate aggregated records based on time windows. The aggregated records of each data source category are aligned and concatenated according to the discharge outlet number and the start and end interval of the time window, so that various types of data from the same discharge outlet within the same time window form a complete time window record. All time window records are arranged in ascending order according to the start and end interval of the time window to obtain the enterprise pollution discharge time window record set.

[0031] A multi-source reliable benchmark is constructed on the enterprise's sewage discharge time window record set to obtain the sewage discharge risk calculation benchmark.

[0032] The pollution discharge risk calculation benchmark refers to the data set in each time window record that carries a coupling credibility coefficient and whose historical regulatory and environmental complaint information has been merged and written after the construction of a multi-source credible benchmark. This data set is used as input for constructing pollution discharge risk evidence.

[0033] From the enterprise's wastewater discharge time window record set, online monitoring records, production load records, pollution control facility operation records, energy and material records, and external environment records are extracted according to data source categories to obtain multi-source wastewater discharge records to be verified; the external environment records include regional environmental quality records and meteorological and hydrological records from the enterprise's wastewater discharge time window record set.

[0034] The integrity of the online monitoring records, production load records, pollution control facility operation records, energy and material records, and external environment records is verified separately. For each data source category, the number of indicators that should be collected and the number of indicators that actually arrive are read in each time window. The ratio of the number of indicators that actually arrive to the number of indicators that should be collected is taken as the integrity rate of that data source category within that time window. Subtracting the integrity rate from 1 yields the missing proportion. .

[0035] Based on the integrity verification, the continuity verification is performed on the online monitoring records, production load records, pollution control facility operation records, energy and material records, and external environment records. The observation sequence of each data source is read window by window, and the number of consecutive windows where the change in observation value within consecutive adjacent time windows is lower than a preset constant value threshold is counted. The ratio of the number of consecutive windows to the total number of verification windows is used as the continuous constant value ratio. As an implementation method, the preset constant value judgment threshold is taken as 5% of the standard deviation of the data source within the recent benchmark window, that is, when the change in the observed value between adjacent time windows is less than 0.05... It is determined to be a constant value.

[0036] Based on continuous verification, fluctuation status verification is performed on the online monitoring records, production load records, pollution control facility operation records, energy and material records, and external environment records. The variation amplitude between the observed values ​​of each data source and the observed values ​​of adjacent time windows is calculated for each time window. When the variation amplitude exceeds a preset jump judgment threshold, a fluctuation anomaly marker is generated for that time window. As an implementation method, the preset jump judgment threshold is taken as three times the standard deviation of the data source within the recent baseline window; that is, when the variation amplitude between adjacent time windows exceeds 3... If a fluctuation is identified as abnormal, the missing percentage, continuous constant value percentage, and fluctuation anomaly markers of each data source within each time window are summarized to obtain the basic quality verification results.

[0037] After the integrity verification, continuity verification, and fluctuation state verification are completed, the multi-source coupling deviation credibility algorithm is used to perform a window consistency verification on the basic quality verification results.

[0038] The multi-source coupling deviation credibility algorithm refers to the algorithm that utilizes the process coupling relationship between production load, pollution control facility operation and online monitoring of discharge outlets. It calculates the coupling deviation between the change rate of upstream data source and the change rate of downstream data source, and combines the missing proportion and the continuous constant value proportion to generate the coupling credibility coefficient of each data source. This algorithm is used to identify low-credibility data that is complete but contradicts the associated data source in the window consistency verification.

[0039] The consistency verification method utilizes the process coupling relationship between production load, pollution control facility operation, and online monitoring of discharge outlets to determine whether the change responses of each data source are consistent within the same time window.

[0040] First, read the observation sequences of each data source within the recent baseline window, and calculate the mean of each data source within the recent baseline window. and standard deviation The recent baseline window is taken as the effective time window of the same row within 14 consecutive days prior to the current time window; when the number of effective time windows is less than 20, it is extended forward to no less than 20 effective time windows or no more than 30 days, whichever condition is met first; wherein, the normalized rate of change r of the data source i within time window t. i The formula for calculating (t) is: , in, This represents the observations of data source i within time window t. This represents the mean of data source i within the recent baseline window. This represents the standard deviation of data source i within the recent baseline window. represents the zero-prevention constant; i represents the upstream data source participating in the coupling verification; t represents the time window number.

[0041] Since different coupling relationships have different focuses on the direction of change, the normalized rate of change needs to be separated according to the risk direction. Specifically, the formulas for calculating the positive rate of change of data source i are as follows: , in, This indicates the magnitude of change in data source i in the direction of increasing risk. This function represents the maximum value and is used to select the largest value from multiple values ​​or expressions given in parentheses as the output result. The formulas for calculating the reverse rate of change of the data source i are as follows: , in, This indicates the magnitude of change in data source i in the direction of risk decrease; After obtaining the positive and negative rates of change for each data source, the coupling deviation is calculated according to the coupling relationship type between the data sources. The coupling relationship between production load records and pollution control facility operation records is marked as a production-treatment coupling pair, and the coupling relationship between production load records and discharge outlet online monitoring records is marked as a production-emission coupling pair.

[0042] For the aforementioned production-treatment coupling pair, since this coupling relationship only poses a pollution discharge risk when production load increases but pollution control facilities fail to respond synchronously, the risk-increasing direction is adopted. The formula for calculating the coupling deviation of the production-treatment coupling pair is: , Where j represents a data source related to the operation of pollution control facilities. This represents the coupling deviation that occurs when production load increases while pollution control facilities fail to respond adequately. This indicates the magnitude of change in data source j in the direction of increasing risk.

[0043] For the aforementioned production-emission coupling pair, since this coupling relationship requires identifying both situations where production load increases but emissions at the outlet do not change accordingly, and situations where production load decreases but emissions at the outlet increase in the opposite direction, a bidirectional abnormality caliber is adopted. The formula for calculating the coupling deviation of the production-emission coupling pair is as follows: , in, This indicates the degree of coupling deviation between production load and discharge outlet emissions.

[0044] After obtaining the coupling deviation of each coupling pair, the missing proportion obtained from the integrity check, the continuous constant value proportion obtained from the continuity check, and the coupling deviation are combined to calculate the coupling reliability coefficient of each data source. The formula for calculating the coupling reliability coefficient of data source j within the time window t is as follows: , in, This represents the coupling reliability coefficient of data source j within the time window t. This represents the set of upstream data sources corresponding to data source j. Select according to the coupling type of the data source pair or , This indicates that the coupling deviation attenuation coefficient is greater than 0. This indicates the percentage of missing values ​​in data source j within the time window t. This represents the proportion of continuous constant values ​​of data source j within the time window t.

[0045] The coupling deviation attenuation coefficient The following method is used to determine the coupling deviation: Read the period within the recent baseline window during which the target company was in normal production and had no records of exceeding standards. Calculate the coupling deviation distribution of each coupling pair during this period, and take the 95th quantile of the coupling deviation distribution as _____. ,make This causes the coupling deviation to reach The reliability coefficient decays to 0.5 over time. For newly registered companies or companies with insufficient historical data, Take the statistical average of companies of similar size and industry as the initial value, and update it in the above manner after accumulating sufficient operating data.

[0046] The credibility of the multi-source credibility evaluation results is written, and the coupling credibility coefficient of each data source is written into the credibility field of the corresponding time window record to obtain the time window record with credibility annotation.

[0047] Benchmark overlay is performed on the time window records that have been reliably labeled. Historical regulatory records and environmental complaint records are extracted from the enterprise's wastewater discharge time window records to obtain benchmark supplementary records.

[0048] The historical regulatory records in the baseline supplementary records are merged based on event timeliness. The event occurrence time of each historical regulatory record is read, and the time interval between the event occurrence time and the current calculation time is calculated. Records with a time interval within 12 months are marked as recent records, and records with a time interval exceeding 12 months are marked as distant records. Recent records are assigned a high timeliness weight, and distant records are assigned a low timeliness weight. Next, the rectification and review status of the historical regulatory records is merged. The rectification completion status and review approval status of each record are read. The risk contribution weight of records that have completed rectification and passed review is reduced, the original risk contribution weight of records that have not completed rectification or failed review is maintained, and the risk contribution weight of records where similar problems recur is increased. The historical regulatory records after event timeliness merging and rectification and review status merging are used as historical regulatory correction records.

[0049] The environmental complaint records in the baseline supplementary records are grouped into similar complaints. The complaint category and content of each environmental complaint record are read, and multiple records with the same complaint category and a complaint time interval within a preset merging window are merged into one complaint event record. The original number of complaints before merging is recorded. Next, the environmental complaint records are merged based on complaint time. The most recent complaint time of each complaint event record is read, and timeliness weights are assigned according to the time interval from the current calculation time. The processing method is consistent with the event timeliness merging of the historical regulatory records. The environmental complaint records after similar complaint merging and complaint timeliness merging are used as environmental complaint correction records.

[0050] The historical regulatory correction records, environmental complaint correction records, and time window records with credible annotations are benchmarked and superimposed. The timeliness weight and risk contribution weight of each event in the historical regulatory correction records are written into the historical regulatory benchmark field of the corresponding enterprise identifier. The timeliness weight and the number of merged complaints of each complaint event in the environmental complaint correction records are written into the environmental complaint benchmark field of the corresponding enterprise identifier. After all fields are written, the pollution discharge risk calculation benchmark is obtained.

[0051] The pollution discharge risk calculation benchmark refers to the data set formed by overlaying historical regulatory correction records, environmental complaint correction records, and time window records with credible annotations, which is used as input for constructing pollution discharge risk evidence.

[0052] In one embodiment, the multi-source coupling deviation confidence algorithm further includes: Taking a certain electroplating company as an example, within three consecutive time windows, the production load records showed that the company's electricity load and water consumption increased by approximately 30% and 25%, respectively, indicating a significant change in production conditions. During the same period, the pollution control facility operation records showed that the operating power and chemical dosage of the wastewater treatment equipment did not change accordingly, and the online monitoring records of the discharge outlet showed that the heavy metal concentration at the discharge outlet remained constant within the aforementioned three time windows.

[0053] When calculating the coupling deviation of the production-treatment coupling pair, since the positive change rate of production load is significantly greater than that of pollution control facilities, the coupling deviation of production-treatment coupling achieves high values ​​in all three consecutive time windows, thus reducing the coupling reliability coefficient of the pollution control facility operation records. Similarly, when calculating the coupling deviation of the production-emission coupling pair, since the production load increases but the discharge outlet concentration does not respond accordingly, the coupling deviation of production-emission coupling also achieves positive values, further reducing the coupling reliability coefficient of the discharge outlet online monitoring records. Simultaneously, the discharge outlet online monitoring records show a constant concentration in all three consecutive time windows, increasing the proportion of continuously constant values, further reducing the coupling reliability coefficient of this data source.

[0054] After calculation using the multi-source coupling deviation credibility algorithm, the company's pollution control facility operation records and discharge outlet online monitoring records within the aforementioned time window were all marked as low-credibility data. The corresponding coupling credibility coefficient in the multi-source credibility evaluation results was significantly lower than that during normal operation periods. This credibility coefficient will suppress false judgments caused by low-credibility data in subsequent pollution discharge risk evidence construction and fusion classification calculations, while triggering a data credibility insufficiency marker, indicating the need for manual verification of the online monitoring equipment status and the actual operation of the pollution control facilities.

[0055] Evidence for pollution risk is constructed based on the pollution risk calculation benchmark to obtain a pollution risk evidence set.

[0056] The construction of pollution discharge risk evidence refers to generating and merging permit proximity evidence records and governance response deviation evidence records based on the pollution discharge risk calculation benchmark, which are used to provide structured evidence input for risk attribution determination.

[0057] The pollution discharge risk evidence set refers to the set of evidence records formed by aligning and merging the permitted proximity evidence records, governance response deviation evidence records, and discharge outlet emission response records according to the discharge outlet number and time window number. Each record includes the permitted proximity value, proximity status marker, governance response deviation index, normalized change rate of emission concentration, and corresponding confidence coefficient.

[0058] Discharge permit constraints and credibly corrected discharge outlet monitoring records are extracted from the discharge risk calculation benchmark to obtain permit boundary verification data. The discharge permit constraints include the permitted pollutant type, permitted emission concentration limit, permitted total emission, and implementation standard corresponding to each discharge outlet number in the discharge risk calculation benchmark. The credibly corrected discharge outlet monitoring records are the pollutant concentration, emission flow rate, and cumulative emission amount corresponding to each discharge outlet number in the discharge risk calculation benchmark, and each record carries a coupling credibility coefficient written from the multi-source credibility assessment results.

[0059] The permit boundary verification data refers to the set of discharge permit constraints extracted from the discharge risk calculation benchmark and discharge outlet monitoring records after reliable correction, which are used as input for permit boundary verification.

[0060] Permit boundary verification data is performed. For each time window, the pollutant concentration corresponding to each discharge outlet number and the corresponding permitted emission concentration limit for that pollutant are read. The ratio of the pollutant concentration to the permitted emission concentration limit is used as the concentration proximity ratio. For each time window, the cumulative emission amount corresponding to each discharge outlet number and the corresponding permitted total emission amount are read. The ratio of the cumulative emission amount to the permitted total emission amount is used as the total emission proximity ratio. For each time window record, the larger of the concentration proximity ratio and the total emission proximity ratio is used as the permit proximity value for that time window.

[0061] When the permitted proximity value is lower than the first proximity threshold, it is marked as a low proximity state, indicating that the emission level is far below the permitted boundary. For example, the COD concentration at a certain enterprise's discharge outlet is 40% of the permitted limit and the cumulative emission is less than 30% of the total permitted amount.

[0062] When the permissible proximity value reaches the first proximity threshold but is lower than the second proximity threshold, it is marked as medium proximity, indicating that the emission level has entered the range that needs attention. For example, the ammonia nitrogen concentration at a certain enterprise's discharge outlet reaches 72% of the permissible limit.

[0063] When the permissible proximity value reaches the second proximity threshold but does not exceed the permissible limit, it is marked as a high proximity state, indicating that the emission level is close to the permissible boundary but has not yet exceeded the standard. For example, the heavy metal concentration at a company's discharge outlet reaches 91% of the permissible limit and is still rising.

[0064] When the permissible proximity value exceeds the permissible limit, it is marked as an over-limit state. For example, the pollutant concentration at a company's discharge outlet has exceeded the permissible emission concentration limit or the cumulative emission has exceeded the permissible total emission amount.

[0065] As an implementation approach, the first proximity threshold is set to 0.60, and the second proximity threshold is set to 0.85. That is, a permissible proximity value below 0.60 indicates a low proximity state, 0.60 and below 0.85 indicates a medium proximity state, 0.85 and below 1.0 indicates a high proximity state, and above 1.0 indicates an over-limit state. The permissible proximity values, proximity status markers, and corresponding coupling credibility coefficients for each time window are summarized to obtain the permissible proximity evidence record.

[0066] After obtaining permission to access the evidence records, production load records, pollution control facility operation records, and energy and material records are extracted from the pollution risk calculation benchmark to obtain the governance response verification data. Each time window record in the governance response verification data carries a coupled credibility coefficient written by the multi-source credibility evaluation results. The governance response verification data refers to the set of production load records, pollution control facility operation records, and energy and material records extracted from the pollution risk calculation benchmark, which is used as input for the quantification of operating condition changes and the quantification of governance response.

[0067] The production load records in the governance response verification data are quantified for changes in operating conditions. Specifically, the production output, electricity load, water consumption, and raw material consumption in the production load records are read window by window. Using the mean and standard deviation of each indicator within the recent baseline window as a benchmark, the normalized rate of change of each production operating condition indicator within the current time window is calculated according to the normalized rate of change calculation method. Where k represents the production condition index number and t represents the time window number. The normalized rate of change of each production condition index is summarized by time window to obtain the condition change record.

[0068] The pollution control facility operation records and energy and material records in the pollution control response verification data are used to quantify the pollution control response. The operating status and parameters of each pollution control facility in the operation records are read window by window, along with the corresponding equipment power consumption and reagent consumption in the energy and material records. Using the mean and standard deviation of each indicator within the recent baseline window as a benchmark, the normalized rate of change of each pollution control facility indicator within the current time window is calculated according to the normalized rate of change calculation method. ,in This indicates the indicator number of the pollution control facility. The normalized rate of change of each pollution control facility indicator is summarized by time window to obtain the treatment response record.

[0069] After obtaining the operating condition change records and the treatment response records, the response lag window number L is determined according to the industry of the enterprise and the type of pollution control facility. For wastewater treatment facilities with equalization tanks or buffer tanks, L is 1 to 6 time windows; for direct discharge facilities or continuous treatment facilities, L is 0. The normalized rate of change of each production operating condition index in each time window of the operating condition change records is shifted backward by L time windows, so that the shifted operating condition change records and the treatment response records are aligned in the time window. The operating condition treatment response coupling deviation index algorithm is used to perform window-based response verification on the shifted operating condition change records and the treatment response records.

[0070] The aforementioned working condition governance response coupling deviation index algorithm refers to calculating the single-window governance response deviation index based on the comparison between the comprehensive change intensity of working conditions and the comprehensive governance response intensity, under the constraint of the credibility coefficient, and generating the cumulative governance response deviation index through the cumulative decay mechanism, which is used to distinguish between occasional governance fluctuations and continuous governance response in the same-window response verification.

[0071] First, the overall intensity of the operating condition change within each time window is calculated based on the operating condition change records. The formula for calculating the overall intensity of the operating condition change is: , in, This indicates the intensity of the overall change in operating conditions within the time window t. This represents the normalized rate of change of the k-th production condition indicator within the time window t. This represents the weight of the k-th production condition indicator, and m represents the number of production condition indicators. The weights of the production condition indicators are determined using an equal-weighting method within the group; when there are differences in sampling coverage, they are normalized and corrected according to the effective sampling coverage within the same benchmark window.

[0072] Accordingly, the overall governance response intensity within each time window is calculated based on the governance response records. The formula for calculating the overall governance response intensity is: , in, This indicates the overall intensity of the governance response within the time window t. Indicates the first The normalized rate of change of each pollution control facility indicator within the time window t. Indicates the first The weights of each pollution control facility indicator are assigned, where n represents the number of pollution control facility indicators. The weights of the pollution control facility indicators are determined using an equal-weighting method within the group; when there are differences in sampling coverage, they are normalized and corrected according to the effective sampling coverage within the same benchmark window.

[0073] Based on the calculation of the comprehensive change intensity of operating conditions and the comprehensive response intensity of governance, a reliability coefficient needs to be introduced to suppress spurious deviations caused by low-reliability data. (Operating condition data reliability coefficient) The reliability coefficient of governance data is obtained by merging the coupling reliability coefficients of various production condition indicators according to their weights; The coupling reliability coefficients of each pollution control facility indicator are obtained by merging them according to the indicator weights. Each coupling reliability coefficient is calculated by the multi-source coupling deviation reliability algorithm.

[0074] Based on the comprehensive change intensity of the operating conditions, the comprehensive response intensity of the governance measures, and the credibility coefficient, the single-window governance response deviation index is calculated within each time window. The formula for calculating the single-window governance response deviation index is as follows: , in, This represents the deviation index of the single-window governance response within the time window t. Represents the zero constant. This represents the reliability coefficient of the operating condition data. This represents the reliability coefficient of governance data. It is used when production conditions increase while the governance response is insufficient. Take a positive value; when the reliability of the operating data or governance data decreases, the deviation index of the single-window governance response is synchronously suppressed.

[0075] To distinguish between occasional governance response deviations and continuous governance response inadequacies, a cumulative governance response deviation index is calculated based on the single-window governance response deviation index. The formula for calculating the cumulative governance response deviation index is as follows: , in, This represents the cumulative governance response deviation index within the time window t. This indicates that the pointer function takes the value 1 when the condition is true and 0 when the condition is false. This represents the cumulative decay coefficient. When governance response deviations occur in multiple consecutive time windows, the cumulative governance response deviation index increases window by window; when the deviation is eliminated, the accumulation stops.

[0076] The cumulative decay coefficient is determined based on the time window length and the deviation of the governance response from the retention time, and the calculation formula is as follows: , in, This indicates the length of the time window in the enterprise's wastewater discharge time window record set. This indicates that the governance response deviated from the retention time. This represents an exponential function with the natural constant e as its base. The deviation of the governance response from the hold-off time is determined by the response cycle of pollution control facilities in the industry to which the enterprise belongs, regulatory trial operation samples, or manual review samples. As an implementation standard, Choose a time window length of 3 to 10, so that It is in the range of 0.72 to 0.90.

[0077] By summing up the single-window governance response deviation index, the cumulative governance response deviation index, and the corresponding credibility coefficient within each time window, a record of governance response deviation evidence is obtained.

[0078] Evidence merging is performed on the permitted proximity evidence records and governance response deviation evidence records: The emission outlet monitoring records, after credibility correction, are read from the emission risk calculation benchmark. The normalized rate of change of pollutant concentration at each outlet between adjacent time windows is calculated for each time window. Emission response records are generated according to the calculation method of the normalized rate of change, where each record includes the outlet number, time window number, normalized rate of change of emission concentration, and corresponding coupling credibility coefficient. The permitted proximity evidence records, governance response deviation evidence records, and emission response records are aligned and merged according to the outlet number and time window number, so that the permitted proximity value, proximity status marker, single-window governance response deviation index, cumulative governance response deviation index, normalized rate of change of emission concentration, and corresponding credibility coefficient for the same outlet within the same time window form a complete evidence record. All evidence records are arranged in ascending order of time window number to obtain the emission risk evidence set. The pollution discharge risk evidence set refers to the set of evidence records formed by merging the evidence records of permitted proximity and treatment response deviation according to the discharge outlet number and time window number. Each record includes the permitted proximity value, proximity status mark, treatment response deviation index, normalized change rate of emission concentration and corresponding confidence coefficient.

[0079] In one embodiment, the operating condition governance response coupling deviation index algorithm further includes: Taking a dyeing and printing company as an example, the company increased its production capacity in two phases during a certain production cycle due to increased orders. The first phase was the first to second time windows, during which the electricity load and water consumption increased by about 15%. At the same time, the operating parameters of the aeration equipment and the dosing equipment were adjusted upwards. The intensity of the comprehensive change in the operating conditions and the intensity of the comprehensive treatment response were basically the same during this phase. The single-window treatment response deviation index was close to zero, and the cumulative treatment response deviation index did not increase.

[0080] The second phase, spanning time windows 3 to 7, saw a further increase of approximately 40% in production power load and raw material consumption. However, the aeration equipment maintained its operating parameters from the first phase without adjustment, and the dosing equipment failed to increase dosing rates synchronously due to insufficient reagent inventory. During this phase, the intensity of the overall change in operating conditions consistently exceeded the intensity of the overall governance response, and the single-window governance response deviation index achieved positive values ​​in each time window. Because the deviation persisted for five consecutive time windows, the cumulative governance response deviation index increased window by window, significantly exceeding the single-window deviation index by time window 7, indicating that the company's governance fluctuations were not isolated incidents but rather a persistent inadequacy in governance response.

[0081] In the second stage described above, the coupling reliability coefficients of the enterprise's production load records and pollution control facility operation records were both within the normal range. The reliability coefficients of the operating data and the pollution control data did not suppress the deviation index, indicating that the deviation was caused by insufficient actual pollution control response rather than data anomalies. If the enterprise's pollution control facility electricity consumption data is missing or shows a constant value anomaly during the same period, the reliability coefficient will decrease, and the single-window pollution control response deviation index will be synchronously suppressed, preventing misjudgment of equipment data acquisition failure as insufficient pollution control response.

[0082] Risk attribution determination is performed on the aforementioned pollution discharge risk evidence set to obtain risk attribution results.

[0083] The aforementioned risk attribution determination refers to the consistency verification, combined state merging, and credible attribution correction based on the pollution discharge risk evidence set, including governance response deviation evidence, discharge outlet emission response, and permit proximity evidence, to determine the risk attribution type and risk source for each time window, providing attribution basis for fusion and hierarchical calculation.

[0084] The risk attribution results refer to the summary record of attribution types, risk sources, and credible identifiers for each time window after risk attribution determination. The attribution types include risk of insufficient emission control, risk of emission background being close to the target, risk of potential control mismatch, and normal operating status.

[0085] Evidence records of deviations from treatment response, discharge outlet response, and permit proximity were extracted from the pollution risk evidence set to obtain input data for attribution determination.

[0086] The emission response record of the discharge outlet is obtained by verifying the data quality and correcting the credibility of the online monitoring data of the discharge outlet. The pollutant concentration change rate between adjacent time windows is calculated for each time window, and the emission concentration normalized change rate sequence is generated according to the normalization method, and the corresponding coupling credibility coefficient is associated.

[0087] The governance response deviation evidence record includes the single-window governance response deviation index, the cumulative governance response deviation index, and the corresponding credibility coefficient within each time window. The permission proximity evidence record includes the permission proximity value, proximity status marker, and the corresponding coupling credibility coefficient within each time window.

[0088] The discharge outlet emission response record is a normalized change rate sequence of emission concentrations of each discharge outlet in the pollution risk evidence set between adjacent time windows, generated from the discharge outlet monitoring data after reliable correction according to the calculation method of the normalized change rate.

[0089] The attribution determination input data refers to the set of governance response deviation evidence records, discharge outlet emission response records, and permit proximity evidence records extracted from the pollution risk evidence set, which are used as inputs for consistency verification and combined state merging.

[0090] The input data for attribution determination is subjected to consistency verification of operating conditions, treatment, and emissions. The cumulative treatment response deviation index in the treatment response deviation evidence record and the normalized rate of change of emission concentration in the emission outlet response record are read simultaneously in time windows to determine whether there is an interpretable correspondence between the treatment response deviation state and the emission change direction. When the cumulative treatment response deviation index exceeds the deviation judgment threshold and the normalized rate of change of emission concentration is simultaneously positive, it is determined that insufficient treatment and increased emissions are consistent. When the cumulative treatment response deviation index exceeds the deviation judgment threshold but the normalized rate of change of emission concentration is zero or negative, it is determined that treatment deviation and emission response are inconsistent. When the cumulative treatment response deviation index does not exceed the deviation judgment threshold but the normalized rate of change of emission concentration remains positive, it is determined that emissions are abnormally increased. When the cumulative treatment response deviation index does not exceed the deviation judgment threshold and the normalized rate of change of emission concentration is not continuously positive, it is determined that the treatment and emissions are consistent under operating conditions. As an implementation method, the deviation judgment threshold is set to 0.30, that is, when the cumulative treatment response deviation index exceeds 0.30, it is determined to be a high treatment deviation state. The consistency determination results are obtained by summarizing the determination categories of each time window.

[0091] The input data for attribution determination is combined and merged. The proximity status markers in the permitted proximity evidence record and the cumulative governance response deviation index in the governance response deviation evidence record are read simultaneously in time windows. The proximity status markers are divided into permitted high proximity and permitted low proximity states based on whether they reach a high proximity state or an over-limit state. The cumulative governance response deviation index is divided into governance deviation and governance deviation states based on whether it exceeds the deviation determination threshold. The permitted high / low proximity states and the governance deviation states are combined in pairs to obtain the permitted governance combination state. When both permitted proximity and governance deviation are high, the permitted governance combination state is marked as insufficient governance emission risk; when both are high and low, the permitted governance combination state is marked as emission background approach risk; when both are low and high, the permitted governance combination state is marked as potential governance mismatch risk; and when both are low, the permitted governance combination state is marked as normal operation state.

[0092] After obtaining the permit governance combination status, the credibility status of the permit proximity evidence records and governance response deviation evidence records in the attribution judgment input data is merged. Specifically, the coupling credibility coefficients of each time window in the permit proximity evidence records in the attribution judgment input data are averaged over recent consecutive time windows to obtain the permit evidence credibility status. The credibility coefficients of the operating condition data for each time window in the governance response deviation evidence records in the attribution judgment input data are then read. and governance data credibility coefficient The same outlet within a recent consecutive time window and The average of the products is used to obtain the credibility status of the governance evidence.

[0093] The credibility attribution correction is performed on the combined state of the permit and governance evidence based on their credibility status. When the credibility status of the permit evidence is below a credibility judgment threshold, the combined state relying on permit proximity evidence is marked with an additional low credibility identifier for permit data; when the credibility status of the governance evidence is below a credibility judgment threshold, the combined state relying on governance response deviation evidence is marked with an additional low credibility identifier for governance data. When the combined state of the permit and governance is a risk of insufficient governance emissions and the credibility status of the governance evidence is below a credibility judgment threshold, the combined state is corrected from a risk of insufficient governance emissions to a governance risk to be verified; when the combined state of the permit and governance is a risk of near-baseline emissions and the credibility status of the permit evidence is below a credibility judgment threshold, the combined state is corrected from a risk of near-baseline emissions to a risk of near-permit to be verified. As an implementation approach, the credibility judgment threshold is set to 0.50, meaning that credibility attribution correction is triggered when either the credibility status of the permit evidence or the credibility status of the governance evidence is below 0.50. The combined state after credibility attribution correction is taken as the credibility attribution result.

[0094] Risk sources are stratified based on the credible attribution results and the consistency determination results. The attribution type in the credible attribution results is matched with the determination category in the consistency determination results for each time window. When the credible attribution result indicates insufficient treatment of emissions risk and the consistency determination result indicates that insufficient treatment is consistent with increased emissions, the risk source is determined to be emissions risk caused by insufficient treatment response. For example, a chemical company's electricity load increases, but the desulfurization equipment's operating power does not increase synchronously, and the sulfur dioxide concentration at the emission outlet increases to near the permissible limit during the same period.

[0095] When the credible attribution result indicates insufficient treatment of emissions risk, but the consistency judgment result indicates that the treatment deviation is inconsistent with the emission response, the source of risk is determined to be abnormal treatment response and questionable emission data. For example, the dosage of chemicals in the pollution control facilities of a certain printing and dyeing enterprise does not increase with the increase in production load, but the COD concentration at the discharge outlet remains constant during the same period. There is no interpretable correspondence between the stability of emission data and insufficient treatment response.

[0096] When the credible attribution result is that the emission background is close to the risk and the consistency judgment result is that the emission is consistent with the treatment under the operating conditions, the source of risk is determined to be that the process background emission is close to the permissible boundary. For example, a certain electroplating enterprise's pollution control facilities are operating normally and matching the production load, but the heavy metal concentration at the discharge outlet has been above 85% of the permissible limit for a long time. The risk of approaching the limit is not caused by insufficient treatment but is determined by the process background emission level.

[0097] When the credible attribution result is a potential risk of governance mismatch, the source of the risk is determined to be a mismatch between the governance facilities and the operating conditions and the emissions are not yet close to the permissible limit. For example, a paper company increases its output but does not adjust the operating intensity of the aeration equipment accordingly. Although the discharge concentration is not close to the permissible limit, the treatment margin has been narrowed.

[0098] When the credible attribution result indicates normal operation and the consistency judgment result indicates an abnormal increase in emissions, the source of risk is determined to be abnormal emission data. For example, if the production conditions and pollution control facilities of a pharmaceutical company have not changed significantly, but the ammonia nitrogen concentration at the discharge outlet suddenly increases within a single time window, it is necessary to investigate the status of online monitoring equipment or external interference factors.

[0099] When the credible attribution result indicates normal operating status and the consistency determination result indicates consistent emissions under operating conditions, the risk source is determined to be that there is currently no significant pollution risk.

[0100] When the credible attribution result carries a verification flag, an explanation that the data credibility is insufficient and manual verification is required is added to the corresponding risk source. For example, if an enterprise's online monitoring data has been continuously constant for more than 24 hours, and the governance response deviation index has not triggered a high deviation flag due to the suppression of the credibility coefficient, the corresponding attribution type is marked as normal operation status, but the data basis is insufficient to support the judgment.

[0101] The risk attribution results are obtained by summarizing the attribution types, risk sources, and credible identifiers for each time window.

[0102] The risk attribution results are integrated and graded to obtain the enterprise's pollution discharge risk classification results.

[0103] The aforementioned integrated hierarchical calculation refers to determining the hierarchical calculation path based on the attribution type in the risk attribution results, and calculating the comprehensive risk value of enterprise pollution discharge by combining environmental carrying capacity correction and threshold bottom line constraints. The comprehensive risk value is then mapped to a risk level, and risk sources, credibility markers, and reasons for hierarchical adjustment are generated.

[0104] The enterprise pollution discharge risk classification result refers to the set of classification records formed by summarizing the risk level, risk source, credibility mark and classification adjustment reasons of each time window after fusion classification calculation, according to the discharge outlet number and time window number.

[0105] The risk source status and credible attribution status are extracted from the risk attribution results to obtain hierarchical calculation input data. The risk source status includes the attribution type and risk source within each time window. The attribution type includes risks of insufficiently regulated emissions, risks of emissions approaching baseline levels, risks of potential governance mismatch, and normal operation status. The credible attribution status includes the credible status of permit evidence and governance evidence within each time window, as well as whether it carries a verification-pending identifier.

[0106] The risk source status and credible attribution status are obtained by decomposing the risk attribution results. The risk source status is derived from the mapping result between attribution type and risk source. The credible attribution status is composed of the credible status of permission evidence, the credible status of governance evidence, and the unverified identifier. All of the above statuses are generated by the aforementioned credible attribution correction process.

[0107] The risk source status and regional environmental carrying capacity status in the tiered calculation input data are constrained and corrected. Specifically, the regional environmental quality records and meteorological and hydrological records of the target enterprise's region are read from the pollution discharge risk calculation benchmark, and the regional environmental carrying capacity conditions are judged window by window. When the regional air quality index or water environmental quality index is close to the control target boundary, the environmental carrying capacity status of the time window is marked as carrying capacity strained; when meteorological diffusion conditions are poor, the flow of receiving water bodies is lower than the preset low flow threshold, or there are environmentally sensitive targets such as drinking water sources or sensitive sections downstream, the environmental carrying capacity status of the time window is marked as carrying capacity sensitive; when the regional environmental quality index is far below the control target and diffusion conditions are normal, the environmental carrying capacity status of the time window is marked as carrying capacity sufficient. An environmental carrying capacity correction coefficient is generated based on the environmental carrying capacity status markings. The term "load tension" corresponds to a high correction value, "load sensitivity" corresponds to a medium correction value, and "load sufficiency" corresponds to a low correction value. The value ranges from 0 to 1. The environmental carrying capacity correction coefficient is correlated with the risk source status. When the attribution type is insufficient emission risk or emission background risk and the environmental carrying capacity status is under pressure, an environmental carrying capacity constraint upgrade identifier is added to the risk source. The risk source status, credible attribution status, and environmental carrying capacity correction coefficient after environmental carrying capacity constraint correction are summarized to obtain the environmental constraint attribution result.

[0108] After obtaining the environmental constraint attribution results, an attribution path gating and grading algorithm is used to determine the grading path of the environmental constraint attribution results, resulting in attribution grading data with determined paths. The attribution path gating and grading algorithm refers to selecting the corresponding threshold value and fusion weight based on the risk attribution type, calculating the comprehensive pollution discharge risk value of enterprises under environmental carrying capacity correction and credibility constraints, and using this algorithm to guide enterprises with different attribution types into differentiated grading calculation paths.

[0109] The risk evidence used in the attribution path gating classification calculation includes the strength of evidence for permitted proximity, the strength of evidence for deviation from the governance response, and the risk intensity of historical behavior. The strength of evidence for deviation from the governance response is obtained by normalizing the cumulative governance response deviation index generated by the coupled deviation index algorithm for the operating condition governance response according to a governance response deviation evidence normalization threshold. The strength of evidence for permitted proximity is obtained by normalizing the permitted proximity degree value in the permitted proximity evidence record. The risk intensity of historical behavior is obtained by normalizing the historical regulatory benchmark field and the environmental complaint benchmark field in the pollution discharge risk calculation benchmark. All of the above evidence strengths are normalized to the range of 0 to 1.

[0110] The fusion weights and threshold values ​​are determined based on the attribution type in the environmental constraint attribution results. When the attribution type is insufficient governance emission risk, the strength of evidence for governance response deviation is used as the dominant evidence, and its fusion weight is higher than that of the strength of evidence for permissible proximity and the strength of historical behavior risk. When the attribution type is emission background approach risk, the strength of evidence for permissible proximity is used as the dominant evidence, and its fusion weight is higher than that of the strength of evidence for governance response deviation. When the attribution type is potential governance mismatch risk, the strength of evidence for governance response deviation and the data credibility status jointly participate in the determination of the hierarchical path. When the attribution type is normal operation status, the strengths of each piece of evidence participate in the calculation with balanced weights. When there are no historical review samples, the fusion weights are configured according to the dominant evidence rules corresponding to the attribution type. As an implementation metric, the fusion weights for the strength of evidence of deviation from governance response, the strength of evidence of permit proximity, and the strength of historical behavior risk corresponding to the risk of insufficient governance emissions are 0.50, 0.30, and 0.20, respectively; the fusion weights for the three categories corresponding to the risk of emission background proximity are 0.20, 0.60, and 0.20, respectively; the fusion weights for the three categories corresponding to the risk of potential governance mismatch are 0.55, 0.20, and 0.25, respectively; and the fusion weights for the three categories corresponding to the normal operating status are 0.34, 0.33, and 0.33, respectively.

[0111] When historical review samples exist, the manual review conclusion is used as the label. The hit ratio of the three types of evidence strength to the review conclusion under different attribution types is calculated. The hit ratio is normalized and then replaced with the above default fusion weight.

[0112] Risk level mapping is performed on the attribution classification data determined by the path, wherein the formula for calculating the comprehensive risk value of enterprise pollution discharge is: , in, This represents the comprehensive risk value of enterprise pollution discharge within the time window t. This represents the gate baseline value corresponding to attribution type T. This represents the fusion weight of the j-th type of risk evidence under attribution type T. Indicates the strength of evidence for the j-th type of risk. This represents the credibility coefficient of evidence for the j-th type of risk. Indicates the environmental carrying capacity correction factor. This represents the environmental carrying capacity correction factor within the time window t. Represents the zero constant. Used to limit the comprehensive risk value of enterprise pollution discharge to the range of 0 to 1. The meaning is the same as described above.

[0113] The environmental carrying capacity correction coefficient The following method is used to determine the environmental carrying capacity correction factor: Read the records of recently confirmed exceedance events that occurred in the regulatory region where the target company is located, and calculate the corresponding environmental carrying capacity correction factor at the time the exceedance event occurred. The distribution of is given by , and the median of this distribution is denoted as . ,make This causes the overall risk value to increase by approximately 20% when the environmental carrying capacity is at a moderate level of stress. When there are no historical records of events exceeding the standard, Let 0.2 be the initial value.

[0114] The threshold value Determined by the risk level mapping threshold, as an implementation caliber, the threshold value for the risk of insufficiently treated emissions is 0.60, meaning that the comprehensive risk value of the enterprise's emissions under this attribution type is not lower than the lower limit of the higher risk level; the threshold value for the risk of emissions close to the baseline is 0.30, meaning that the comprehensive risk value of the enterprise's emissions under this attribution type is not lower than the lower limit of the medium risk level; the threshold value for the risk of potential governance mismatch is 0.15; and the threshold value for the normal operation status is 0.

[0115] The risk level is determined based on the range in which the comprehensive risk value of the enterprise's pollution discharge falls. A time window in which the comprehensive risk value of the enterprise's pollution discharge is below the first-level threshold is marked as low risk, indicating that the enterprise's emissions are far below the permissible limits and that its operating conditions, treatment, and emission status are consistent. For example, if the concentrations of all pollutants in an enterprise are below 50% of the permissible limits and the operation of its pollution control facilities matches its production load.

[0116] The time window in which the comprehensive risk value of the enterprise's pollution discharge reaches the first level threshold but is lower than the second level threshold is marked as medium risk. This indicates that the enterprise's emissions are still within a controllable range, but some evidence shows signs of deviation, such as the concentration at the enterprise's discharge outlet being close to 70% of the permitted limit or the deviation index of the treatment response occasionally increasing.

[0117] The time window in which the comprehensive risk value of the enterprise's pollution discharge reaches the second-level threshold but is lower than the third-level threshold is marked as a higher risk, indicating that the enterprise has one or more of the following problems: permit approaching, insufficient treatment response, or insufficient data credibility. For example, if the concentration of a certain enterprise's discharge outlet is consistently higher than 85% of the permit limit and the intensity of the pollution control facilities' operation has not been adjusted with the increase in production load.

[0118] The time window in which the comprehensive risk value of enterprise emissions reaches the third-level threshold is marked as high risk, indicating that the enterprise's emissions have exceeded the permitted limits or multiple risk evidences have been triggered simultaneously, and environmental carrying capacity conditions are strained. For example, the concentration at an enterprise's discharge outlet exceeds the permitted limit and the air quality in the area is close to the control target boundary. As an implementation method, the first-level threshold is 0.30, the second-level threshold is 0.60, and the third-level threshold is 0.85. That is, when the comprehensive risk value of enterprise emissions is below 0.30, it is considered low risk; when it reaches 0.30 and is below 0.60, it is considered medium risk; when it reaches 0.60 and is below 0.85, it is considered relatively high risk; and when it reaches 0.85, it is considered high risk. The coupling deviation attenuation coefficient α is taken from 1.5 to 3.0, so that the confidence attenuation range corresponding to a single coupling deviation of 0.5 is within the range of 47% to 78%. The environmental carrying capacity correction strength coefficient η is taken from 0.20 to 0.50, so that the maximum upward fluctuation of the comprehensive risk value under environmental carrying capacity strain is within the range of 20% to 50%.

[0119] Based on the environmental constraint attribution results, risk sources, credibility markers, and reasons for graded adjustments are generated. The risk sources are directly written from the attribution type and risk source in the risk attribution results. The credibility markers are written from the credibility status of permissive evidence and governance evidence in the credibility attribution status; when either credibility status is below the credibility judgment threshold, it is marked as insufficient data credibility. The reasons for graded adjustments record the following information: when the threshold bottom line value imposes a bottom line constraint on the risk level, the threshold bottom line is recorded as effective and the corresponding attribution type; when the environmental carrying capacity correction coefficient causes the comprehensive risk value to rise and cross the level interval boundary, the environmental carrying capacity constraint is upgraded and the corresponding environmental carrying capacity status is recorded; when the credibility attribution status carries a pending verification identifier, it is recorded that the data credibility is insufficient and manual verification is required.

[0120] The risk level mapping results are merged as follows: the risk level, risk source, credibility mark and reason for level adjustment in each time window are summarized into a complete level record according to the discharge outlet number and time window number. All level records are arranged in ascending order according to the time window number to obtain the enterprise's pollution discharge risk level result.

[0121] After obtaining the enterprise's pollution discharge risk classification results, the corresponding manual review results, enterprise rectification feedback, and actual exceedance event results are acquired to obtain a rolling correction record. The rolling correction record is then used to identify the correction objects, resulting in the parameter objects to be updated. The coupling deviation attenuation coefficient, cumulative attenuation coefficient, and fusion weight involved in the parameter objects to be updated are updated separately. The specific process is as follows: When manual review confirms that the insufficient emission risk assessment in a certain classification result is a false alarm, the coupling deviation distribution within that time window is read, the coupling deviation corresponding to the false alarm sample is included in the normal operating condition sample set, the 95th quantile of the coupling deviation distribution is recalculated, and the coupling deviation attenuation coefficient is updated. When manual review confirms that the cumulative governance response deviation index in a certain grading result is triggered too early or too late, the governance response deviation retention time will be adjusted. And recalculate the cumulative attenuation coefficient .

[0122] When actual records of exceedance events indicate a low correlation between the dominant evidence and the exceedance result under a certain attribution type, the exceedance sample is used as a calibration sample. The hit ratio of various evidence strengths to the exceedance conclusion under that attribution type is recalculated, and the fusion weight of the corresponding attribution type is updated. The rolling correction records are archived according to correction time, correction object, and correction basis to ensure that subsequent enterprise pollution discharge risk classification calculations are more consistent with the actual operating conditions of the region, industry, and enterprise.

[0123] In one embodiment, the attribution path gating hierarchy algorithm further includes: Taking companies A and B within the same regulatory region as an example, the permit proximity values ​​of the two companies within the same time window are both 0.82, meaning that the pollutant concentrations at their discharge outlets both reach 82% of the permit limits, but the attribution types of the two companies are different.

[0124] Company A is a chemical enterprise. The risk attribution results show that its attribution type is insufficient pollution control emission risk. The reason is that the enterprise's pollution control facility operation intensity was not adjusted synchronously after the production load increased, and the cumulative pollution control response deviation index continued to grow. In the attribution path gating classification algorithm, the threshold value corresponding to insufficient pollution control emission risk is the lower limit of the higher risk level, and the strength of evidence for pollution control response deviation is used as the dominant evidence with the highest fusion weight. It was calculated that even if the strength of evidence for permissible proximity and the risk intensity of historical behavior are at a medium level, the enterprise's comprehensive pollution discharge risk value is still constrained by the threshold value and is not lower than the lower limit of the higher risk level. Ultimately, the enterprise is marked as high risk, and the risk source is recorded as emission risk caused by insufficient pollution control response.

[0125] Company B is a building materials enterprise. The risk attribution results show that its attribution type is "emission background near risk," because the company's pollution control facilities are operating normally and matching the production load, and the discharge concentration is close to the permissible limit, which is determined by the process background emission level. In the attribution path gating classification algorithm, the threshold value corresponding to emission background near risk is lower than that of insufficient treatment emission risk, and the evidence strength of permissible near risk is used as the dominant evidence. Since the company's treatment response deviation evidence strength is close to zero and its historical behavior risk strength is low, the company's comprehensive pollution discharge risk value is lower than that of Company A. Ultimately, Company B is marked as medium risk, and the risk source is recorded as process background emission close to the permissible boundary.

[0126] Within the aforementioned time window, the air quality index of the regulated area was approaching the control target boundary, and the environmental carrying capacity correction coefficient was set to a higher value. After environmental carrying capacity correction, Company A's overall risk value further increased, but because it was already within the higher risk level range, it did not cross the level threshold. Company B's overall risk value, however, increased and crossed the boundary between medium and higher risk levels, and was adjusted to higher risk. The reason for the adjustment was recorded as an upgrade in environmental carrying capacity constraints. Although the two companies had similar permit levels, their different attribution types led to different calculation paths for risk classification, resulting in differences in their final risk levels, risk sources, and reasons for adjustment.

[0127] In one simulation experiment, to verify the technical effects of the multi-source coupling deviation credibility algorithm, the working condition governance response coupling deviation index algorithm, and the risk attribution judgment and attribution path gating classification algorithm, a simulation environment for enterprise pollution discharge supervision data was built. Six typical industries—chemical, printing and dyeing, electroplating, papermaking, pharmaceutical, and building materials—were set up, with four virtual enterprises in each industry, totaling 24 enterprises. Each enterprise had one main discharge outlet. Using a 15-minute time window, 30 days of pollution discharge permit records, discharge outlet online monitoring records, production load records, pollution control facility operation records, energy and material records, historical supervision records, environmental complaint records, regional environmental quality records, and meteorological and hydrological records were continuously generated, forming a total of 69,120 time window records. The first 14 days of records served as the recent baseline window, used to determine the mean, standard deviation, coupling deviation attenuation coefficient, and risk evidence normalization parameters for each data source. The last 16 days of records were used to set test events and perform risk identification.

[0128] The testing phase included six event scenarios, each comprising 120 independent event segments, each lasting 6 to 12 time windows. The first scenario involved conflicting multi-source data, causing production load to increase by 20% to 40% within a continuous time window, while controlling changes in pollution control facility operating power and reagent consumption to within 5%, and ensuring that adjacent time windows of discharge outlet monitoring values ​​did not differ by more than 1%. The second scenario involved sensor failure, with scenarios including 15% to 25% data loss, constant output for 8 to 24 consecutive time windows, and single-point jumps exceeding the recent baseline standard deviation by 3 to 5 times. The third scenario involved the risk of exceeding limits, gradually increasing the ratio of pollutant concentration to permissible limits from 0.65 to 0.75 to 0.90 to 0.98, with a lag of 2 to 5 time windows in the treatment response, followed by a lag of 2 to 4 time windows. The first group represents the actual exceedance of limits after the window period; the second group represents the occasional fluctuation scenario, where the production load increases by 20% to 30% within 1 to 2 time windows, and the pollution control facilities complete their response within the subsequent time window, without a continuous increase in the discharge concentration; the third group represents the risk cause differentiation scenario, where both types of events have a permissible proximity of 0.78 to 0.90, with one type caused by insufficient response of the treatment facilities and the other by a high background emission level from the process; the fourth group represents the graded interpretation scenario, which superimposes environmental carrying capacity constraints, low credibility of single evidence, or increased risk of historical behavior on the aforementioned events, to verify whether the grading results can simultaneously output the risk source, credibility marker, and reason for grading adjustment. The start and end times, fault types, and risk sources of each test event are pre-recorded by the event setting process as verification labels.

[0129] In this embodiment, for wastewater treatment scenarios with equalization tanks, the response lag window number L is set to 2, and the cumulative treatment response deviation retention time is set to 6 time windows. The first proximity threshold, second proximity threshold, deviation judgment threshold, and credibility judgment threshold are set to 0.60, 0.85, 0.30, and 0.50, respectively, and the risk classification thresholds are set to 0.30, 0.60, and 0.85, respectively. The coupling deviation attenuation coefficient is calibrated according to the 95th quantile of the coupling deviation degree during the normal period of the previous 14 days. As a control, a fixed-weight comprehensive scoring method is used to directly score the permit proximity degree, historical regulatory records, and environmental complaint records independently and sum them by weight. This control method does not perform window coupling verification between production load, pollution control facilities, and discharge outlets, and does not set data credibility attenuation, cumulative treatment response deviation, or risk attribution gating. Both methods use the same input records and the same risk level range.

[0130] When verifying the test results, the ratio of the number of events that correctly identified multi-source variation contradictions and generated low-confidence labels to the total number of events in the first group is used as the multi-source data contradiction identification rate; the ratio of the number of events where sensor faults were identified as insufficient data confidence and were not directly misjudged as high-risk to the total number of events in the second group is used as the sensor fault robustness rate; the ratio of the number of events that output medium-risk or higher warnings at least two time windows before the actual exceedance to the total number of events in the third group is used as the pre-exceedance risk warning rate; the ratio of the number of events in the fourth group that were not mistakenly upgraded to higher or higher risk to the total number of events in that group is used as the random fluctuation anti-interference rate; the ratio of the number of events in the fifth group whose output risk source is consistent with the preset risk source to the total number of events in that group is used as the risk cause attribution accuracy rate; and the ratio of the number of events in the sixth group whose risk level, risk source, confidence label, and grade adjustment reason are all consistent with the preset label to the total number of events in that group is used as the grade result explainable consistency rate.

[0131] Calculations showed that the method of this invention correctly handled 109, 101, 104, 97, 106, and 111 events respectively across six indicators: multi-source data contradiction identification, sensor fault robustness, pre-excess risk warning, anti-interference of occasional fluctuations, risk cause attribution differentiation, and interpretability of grading results. The corresponding indicator values ​​were 90.8%, 84.2%, 86.7%, 80.8%, 88.3%, and 92.5%. The control method correctly handled 25, 38, 29, 52, 24, and 25 events respectively, with corresponding indicator values ​​of 20.8%, 31.7%, 24.2%, 43.3%, 20.0%, and 20.8%. These results demonstrate that the method of this invention can not only identify insufficient continuous governance response before pollutants exceed limits, but also suppress false evidence generated by constant, missing, and abrupt data using confidence coefficients, and guide events with similar permissible proximity but different risk sources into different grading paths through attribution path gating.

[0132] Taking an electroplating company incident in the third group as an example, the production load increased from 62% to 84% of the rated load. Within five consecutive time windows, the operating power of the pollution control facilities only increased by 3%, and the consumption of chemicals only increased by 2%. The ratio of pollutant concentration to the permissible limit increased from 0.74 to 0.88, and the reliability coefficients of production, treatment, and discharge data were all higher than 0.80. After aligning the response lag according to two time windows, the cumulative treatment response deviation index reached 0.37 in the fourth alignment time window, exceeding the deviation judgment threshold of 0.30. Simultaneously, the permissible proximity reached a high proximity state, thus being attributed to insufficient treatment and emission risk. Based on the threshold baseline and evidence fusion calculation, the company's comprehensive pollution discharge risk value was 0.68, and it was marked as a high risk. The incident resulted in an actual exceedance of the permissible proximity of 1.03 three time windows later. Therefore, the method of this invention outputs a warning 45 minutes earlier than the actual exceedance. The control method, due to the company's low historical regulatory risk and complaint risk, calculated a fixed weight score of 0.55, only marking it as a medium risk, and did not generate an escalation warning before exceeding the limit.

[0133] In a constant-value failure event in the second group, the production load increased by approximately 30% continuously, but the discharge outlet monitoring value remained identical for 12 consecutive time windows. The multi-source coupling deviation and the proportion of continuous constant values ​​caused the coupling reliability coefficient of the discharge outlet monitoring record to decrease to 0.34, below the reliability judgment threshold of 0.50. Therefore, the method of this invention adds an explanation to the risk source stating that the emission data reliability is insufficient and requires manual verification, and reduces the contribution of the discharge outlet monitoring record to the evidence of near-permitted access, avoiding the direct use of constant-value data as a basis for low risk. In another occasional fluctuation event in the fourth group, the production load increased by 25% in a single window, and the pollution control facilities completed the corresponding adjustment in the next time window. The maximum cumulative treatment response deviation index was 0.18, which did not exceed the deviation judgment threshold of 0.30, and therefore was not misjudged as continuous inadequate treatment. These two types of results respectively verify the inhibitory effect of reliability constraints and cumulative decay mechanisms on data failures and short-term operating condition disturbances.

[0134] The above simulation results can be compared with Figure 2 The corresponding comparison charts show the results of the proposed method in six indicators: multi-source data contradiction identification, sensor fault robustness, risk warning before exceeding limits, anti-interference of occasional fluctuations, risk cause attribution differentiation, and interpretability of grading results. The gray bar charts represent the experimental results of the fixed-weight comprehensive scoring method. As can be seen from the above event verification data, the proposed method improves the ability to identify risks not exceeding limits, suppress data faults, differentiate risk causes, and interpret grading results by forming an interconnected processing chain through multi-source reliable verification, continuous deviation accumulation, risk source attribution, and differentiated gating grading.

[0135] Combination Figure 2The chart comparing the effectiveness of the enterprise pollution discharge risk identification method with existing technologies is shown below. The black bars represent the technical effectiveness of the present invention, while the gray bars represent existing technologies. Figure 2 It can be seen that the present invention is superior to the prior art to a certain extent.

[0136] Example 2, a method for identifying enterprise pollution discharge risks based on Example 1, is as follows: A dataset of enterprise pollution discharge risk associations for the target enterprise is constructed and subjected to time-series association normalization to obtain a set of enterprise pollution discharge time windows. The length of the time window is adaptively determined according to the fluctuation frequency of the enterprise's production conditions. The frequency of change of the production load record within the recent baseline window is read. When the frequency of change is higher than a preset frequency threshold, the time window length of all data source categories is shortened to the same short period. When the frequency of change is lower than the preset frequency threshold, the time window length of all data source categories is extended to the same long period, so that various types of data from the same enterprise share a unified time window length within the same calculation period.

[0137] A multi-source reliable benchmark is constructed for the enterprise's pollution discharge time window record set to obtain the pollution discharge risk calculation benchmark. In the integrity verification, continuity verification, and fluctuation status verification, the missing proportion, continuous constant value proportion, and fluctuation anomaly frequency are weighted and summed according to preset weights to obtain the basic quality score of each data source. In the consistency verification, the coupling deviation degree of each coupling pair is calculated using the multi-source coupling deviation degree reliability algorithm. The coupling deviation degree is weighted and summed with the basic quality score to obtain the comprehensive reliability score. The comprehensive reliability score is then normalized and mapped to the interval of 0 to 1 as the coupling reliability coefficient. In the generation of historical regulatory correction records and environmental complaint correction records, the traceability period for the event timeliness consolidation is adjusted from 12 months to be determined according to the regulatory inspection cycle of the enterprise's industry. A 6-month traceability period is used for high-frequency inspection industries, and an 18-month traceability period is used for low-frequency inspection industries.

[0138] Evidence for pollution risk is constructed based on the pollution risk calculation benchmark to obtain a pollution risk evidence set. In the permit boundary verification, the permit proximity value is calculated differently according to the pollutant hazard level, with a lower proximity threshold starting point for Class I pollutants and a higher proximity threshold starting point for Class II pollutants. In the same-window response verification, the operating condition governance response coupling deviation index algorithm is used to calculate the single-window governance response deviation index and the cumulative governance response deviation index. The cumulative governance response deviation index uses a sliding window method instead of a recursive accumulation method, taking the weighted average of the single-window governance response deviation indices within the current time window and its preceding preset number of windows as the cumulative deviation. The weight of each time window within the window is distributed decreasingly according to the time interval from the current time window.

[0139] Risk attribution is performed on the pollution discharge risk evidence set to obtain risk attribution results. In the combined state merging, the permit proximity state is refined from two levels to three levels: low permit proximity, medium permit proximity, and high permit proximity. These are combined with the high and low governance deviation levels to form six permit-governance combination states. The attribution type corresponding to the medium permit proximity state and the high governance deviation state is marked as early governance mismatch risk, which is different from the potential governance mismatch risk in the low permit proximity state. In the credible attribution correction, when the credible state of both permit evidence and governance evidence is below the credible judgment threshold, the entire permit-governance combination state is corrected to a data insufficiency and verification pending state, and no longer additional verification pending labels are added according to the individual credibility levels.

[0140] The risk attribution results are fused and graded to obtain the enterprise's pollution discharge risk classification results. The attribution path gating classification algorithm is used to calculate the enterprise's comprehensive pollution discharge risk value. The environmental carrying capacity correction uses a level offset method instead of a multiplicative correction method. When the environmental carrying capacity status is "carrying capacity strained," the risk level is directly increased by one level. When the environmental carrying capacity status is "carrying capacity sensitive" and the enterprise's comprehensive pollution discharge risk value is in the upper half of the level range, the risk level is increased by one level. When the environmental carrying capacity status is "carrying capacity sufficient," the risk level is not adjusted. An extremely high risk level is added to the risk level mapping. When the enterprise's comprehensive pollution discharge risk value reaches the third classification threshold, the attribution type is "inadequate treatment emission risk," and the environmental carrying capacity status is "carrying capacity strained," the risk level is marked as extremely high risk.

[0141] The embodiments of the present invention described above are subject to modification and change of method by those skilled in the art without departing from the embodiments and broader aspects of the present invention. The appended claims are intended to include all such modifications and changes of method that do not depart from the present invention.

[0142] Those skilled in the art should understand that the embodiments of the present invention can be implemented using a pure hardware architecture, a pure software architecture, or an integrated hardware and software architecture. The present invention can be prepared as a computer program product, which can be stored in various non-volatile computer-readable storage media, including but not limited to solid-state drives, flash memory chips, mobile storage devices, optical discs, cloud storage servers, and other standardized storage media, and is not limited to traditional storage media.

[0143] Based on the foregoing description in conjunction with the accompanying drawings, those skilled in the art will understand that the embodiments of this application can also be implemented by software programs. Therefore, this application also provides a computer-readable storage medium. This computer-readable storage medium stores computer-readable instructions thereon, which, when executed by one or more processors, implement the method described above in conjunction with the accompanying drawings.

[0144] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0145] It should be noted that although the operations of the method of this application are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. On the contrary, the steps depicted in the flowchart can be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0146] It should be understood that when the terms "first," "second," "third," and "fourth," etc., are used in the claims, specification, and drawings of this application, they are used only to distinguish different objects and not to describe a specific order. The terms "comprising" and "including" as used in the specification and claims of this application indicate the presence of the described features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof.

[0147] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this specification and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this specification and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0148] Although the embodiments of this application are described above, the content is merely an example adopted for the purpose of facilitating understanding of this application and is not intended to limit the scope and application scenarios of this application. Any person skilled in the art described in this application may make any modifications and changes in the form and details of the implementation without departing from the spirit and scope disclosed in this application, but the scope of patent protection of this application shall still be determined by the scope defined in the appended claims.

Claims

1. A method for identifying enterprise pollution discharge risks based on multi-source data fusion, characterized in that, include: Construct a dataset of enterprise pollution discharge risk association for the target enterprise and perform time-series association normalization to obtain a set of enterprise pollution discharge time window records; A multi-source reliable benchmark is constructed from the enterprise's pollution discharge time window record set to obtain the pollution discharge risk calculation benchmark; Evidence for pollution risk is constructed based on the pollution risk calculation benchmark to obtain a pollution risk evidence set. Risk attribution determination is performed on the evidence set of pollution discharge risks to obtain risk attribution results; By integrating and classifying the risk attribution results, the enterprise's pollution discharge risk identification results are obtained.

2. The enterprise pollution discharge risk identification method according to claim 1, characterized in that, A multi-source reliable benchmark is constructed from the enterprise's pollution discharge time window record set to obtain the pollution discharge risk calculation benchmark, including: Online monitoring records, production load records, pollution control facility operation records, energy and material records, and external environmental records are extracted from the enterprise's pollution discharge time window records to obtain multi-source pollution discharge records to be verified; Completeness verification, continuity verification, and fluctuation status verification are performed on the multi-source discharge records to be verified to obtain the basic quality verification results; The multi-source coupling deviation credibility algorithm is used to perform window consistency verification on the basic quality verification results to obtain multi-source credibility evaluation results; The credibility of the multi-source credibility evaluation results is written to obtain a time window record with credibility annotation; By overlaying the time window records with reliable annotations, a baseline for calculating pollution discharge risk is obtained.

3. The enterprise pollution discharge risk identification method according to claim 2, characterized in that, The baseline for calculating pollution risk is obtained by overlaying trusted time window records, which includes: Historical regulatory records and environmental complaint records are extracted from enterprise wastewater discharge time windows to obtain baseline supplementary records; The historical regulatory records in the baseline supplementary records are merged by combining event timeliness and rectification review status to obtain historical regulatory correction records; The environmental complaint records in the baseline supplementary records are merged by grouping similar complaints and merging complaints by time statute of limitations to obtain the environmental complaint correction records; By overlaying historical regulatory correction records, environmental complaint correction records, and credibly labeled time window records, a baseline for calculating pollution discharge risk is obtained.

4. The enterprise pollution discharge risk identification method according to claim 1, characterized in that, Evidence for pollution risk is constructed based on the pollution risk calculation benchmark, resulting in a pollution risk evidence set, including: Discharge permit constraints and reliable-corrected discharge outlet monitoring records are extracted from the discharge risk calculation benchmark to obtain permit boundary verification data. Perform permit boundary verification on the permit boundary verification data to obtain permit proximity evidence records; By expanding the evidence of pollution discharge risk from the records of evidence of proximity to permits, a set of evidence of pollution discharge risk is obtained.

5. The enterprise pollution discharge risk identification method according to claim 4, characterized in that, The pollution risk evidence set is constructed by expanding the records of permission access evidence, resulting in a pollution risk evidence set including: Production load records, pollution control facility operation records, and energy and material records are extracted from the pollution discharge risk calculation benchmark to obtain governance response verification data; The production load records in the governance response verification data are quantified to obtain the operating condition change records; The pollution control facility operation records and energy and material records in the pollution control response verification data are quantified to obtain the pollution control response records; The working condition-governance response coupling deviation index algorithm is used to perform window response verification on the working condition change records and governance response records to obtain governance response deviation evidence records. Evidence records of permit proximity and governance response deviation are merged to obtain a pollution risk evidence set.

6. The enterprise pollution discharge risk identification method according to claim 1, characterized in that, Risk attribution determination is performed on the evidence set of pollution discharge risks to obtain risk attribution results, including: Extract records of deviation from treatment response, discharge outlet response, and permit proximity from the pollution risk evidence set to obtain input data for attribution determination; Risk attribution determination is performed on the input data for attribution determination to obtain the risk attribution results.

7. The enterprise pollution discharge risk identification method according to claim 6, characterized in that, Risk attribution is performed on the input data for attribution determination to obtain risk attribution results, including: The consistency of operating conditions, treatment, and emissions is verified on the input data for attribution determination to obtain the consistency determination results; The attribution determination input data is combined and merged to obtain the permission governance combined state; Risk sources are stratified based on the status of the licensing governance portfolio and the consistency determination results to obtain risk attribution results.

8. The enterprise pollution discharge risk identification method according to claim 7, characterized in that, The risk sources are stratified based on the status of the licensing governance portfolio and the consistency determination results to obtain risk attribution results, including: The credibility status of the permitted proximity evidence records in the attribution determination input data is merged to obtain the credibility status of the permitted evidence. The credibility status of governance response deviation evidence records in the attribution determination input data is merged to obtain the credibility status of governance evidence. The credibility attribution correction is performed on the credibility status of licensing evidence, the credibility status of governance evidence, and the combined status of licensing and governance to obtain the credibility attribution results. Risk sources are stratified based on credible attribution results and consistency determination results to obtain risk attribution results.

9. The enterprise pollution discharge risk identification method according to claim 1, characterized in that, The risk attribution results are integrated and graded to obtain the enterprise's pollution discharge risk classification results, including: Extract the risk source status and credible attribution status from the risk attribution results to obtain the hierarchical calculation input data; Constraint corrections are applied to the risk source status and regional environmental carrying capacity status in the hierarchical calculation input data to obtain environmental constraint attribution results; The attribution path gating and hierarchical algorithm is used to determine the hierarchical path of the environmental constraint attribution results, and the attribution hierarchical data determined by the path is obtained. Risk level mapping is performed on the attribution classification data determined by the path to obtain the risk level mapping results; The risk level mapping results are merged to obtain the enterprise pollution discharge risk classification results.

10. The enterprise pollution discharge risk identification method according to claim 9, characterized in that, After obtaining the enterprise's pollution discharge risk classification results, the following are also included: Obtain the manual review results, enterprise rectification feedback, and actual exceedance events corresponding to the enterprise's pollution discharge risk classification results, and obtain rolling correction records; The rolling correction record is used to identify the correction object to obtain the parameter object to be updated. The parameters involved in the parameter update object are updated as follows: coupling deviation attenuation coefficient in the multi-source coupling deviation credibility algorithm, cumulative attenuation coefficient in the working condition governance response coupling deviation index algorithm, and fusion weight in the attribution path gating classification algorithm.

Images