A method and system for carbon credit project data consistency review

Abstract

The present application belongs to the technical field of carbon emission reduction project data verification, in particular to a carbon credit project data consistency review method and system, which comprises: obtaining multi-source original data and preliminary verification, triggering rules to locate abnormal subjects; relying on the material balance equation to lock the associated data with physical dependence, directional review and reverse tracing to generate a trace evidence chain and mark the missing link; segmenting the index history sequence according to the recorded process change nodes, determining the timing anomaly by the ratio of the index and the associated production index change rate, and calculating the residual; combining the material balance sensitivity coefficient to feedback adjust the review threshold and close loop review; multi-dimensional aggregation of abnormal output review report. The present application relies on material balance to determine the targeted verification range, builds a closed-loop verification mechanism for timing anomaly feedback threshold reduction, solves the problems of low efficiency of traditional verification, difficulty in identifying gradual drift and other problems, and improves the accuracy and objectivity of greenhouse gas voluntary emission reduction project data verification.

Description

Technical Field

[0001] This invention belongs to the technical field of carbon emission reduction project data verification, and in particular relates to a method and system for verifying the consistency of carbon credit project data. Background Technology

[0002] Voluntary greenhouse gas emission reduction projects are an important component of China's carbon market. The authenticity and consistency of project emission reduction accounting data are the core foundation for project emission reduction registration, trading, and supervision. If there are deviations, alterations, or logical contradictions in the project's activity data such as fuel, production, and emission factors, it will directly lead to distorted emission reduction accounting and disrupt the carbon market order. Therefore, an automated and accurate data consistency review mechanism is of critical practical significance for ensuring the credibility of emission reduction data and standardizing project verification work.

[0003] The current industry standard review method mainly relies on manual verification combined with basic fixed rules: the inspectors collect multi-source data such as project ledgers, measurement, and laboratory tests, and compare each single indicator with fixed thresholds to see if it exceeds the standard; after an anomaly is found, the inspectors manually review the upstream and downstream vouchers to complete the traceability; all data adopts a unified judgment standard, and the data is judged to be compliant based on the values ​​of a single reporting period. Finally, the anomaly risk level is classified based on the inspectors' experience, and a written review report is completed.

[0004] The existing technology has the following shortcomings:

[0005] 1. Existing verification schemes mostly adopt single-point data compliance verification or full data traversal review mode. They do not build objective physical dependencies between data based on the material conservation law of the production process. The full review mode requires traversing all data items, which has high computing power consumption and low verification efficiency, and is difficult to adapt to the routine review of large-scale projects. The single-point verification mode only judges a single indicator independently and cannot identify cross-data item correlation deviations. The rate of missed detection of related data fraud or systemic deviations is high.

[0006] 2. Existing technologies typically employ fixed thresholds and fixed benchmarks for judgment, which cannot automatically update the judgment benchmarks as project process modifications, equipment upgrades, or other filing changes occur. The misjudgment rate increases significantly after changes in operating conditions. Furthermore, threshold verification is only performed on data from a single reporting period, which cannot identify hidden data anomalies such as compliance of single-period values ​​or long-term gradual drift. The detection of data deviations or human tampering is significantly delayed.

[0007] 3. Existing evidence verification simply distinguishes whether vouchers are retained, without considering the impact weight of various data in carbon accounting to quantify the risk caused by the lack of quantification. Anomaly classification relies entirely on the subjective experience of the verification personnel, and the judgment criteria are inconsistent among different personnel and different projects, making the review conclusions difficult to reproduce and lacking objectivity.

[0008] 4. In the existing solution, historical data statistical analysis and current data consistency verification are two independent steps. The trend anomalies identified at the time series level cannot be transmitted back to the current verification rules. It is impossible to optimize the current verification standards through historical risk characteristics. The verification system is in a static and passive state, making it difficult to continuously improve the review accuracy through data accumulation. It also has limited ability to identify weak signal risks such as slight correlation deviations and abnormal working conditions. Summary of the Invention

[0009] The purpose of this invention is to address the problems raised in the background section by providing a method and system for verifying the consistency of carbon credit project data.

[0010] To achieve the above objectives, the present invention adopts the following technical solution: a carbon credit project data consistency review method, applicable to voluntary greenhouse gas emission reduction projects with material input, output, or loss accounting chains, executed by computer equipment, comprising the following steps:

[0011] Acquire current multi-source raw data and perform preliminary consistency verification. When a preset verification rule is triggered, generate an exception event record and determine the raw data item or intermediate calculation result that triggered the rule as the exception subject.

[0012] Based on the material conservation equation of the project's production process, a set of associated data items that have a physical dependency relationship with the abnormal entity is determined, and a corresponding sensitivity coefficient is determined for each data item in the set of variables of the material conservation equation.

[0013] The associated data item set is subject to targeted review, and an evidence chain is generated by tracing back along the data collection record chain.

[0014] Extract the historical numerical sequence of the abnormal subject and associated data item set, segment the historical numerical sequence according to the process change time point, calculate the change rate of the abnormal subject and the change rate of the associated production indicator within the segment, and obtain the change rate ratio. When the change rate ratio exceeds the constraint interval corresponding to the data type of the abnormal subject, mark the time series abnormality and calculate the indicator residual.

[0015] Multiply the residual of the index by the sensitivity coefficient of the abnormal subject in the material conservation equation and the preset threshold reduction conversion coefficient to obtain the threshold reduction amount, reduce the review judgment threshold of the set of related data items, and the reduced judgment threshold is not lower than the preset minimum threshold lower limit. Based on the reduced threshold, the set of related data items is re-examined and reversed, and the evidence chain and abnormal event records are updated.

[0016] Summarize abnormal event records within the same review period, aggregate them according to preset dimensions, and output a consistency review report.

[0017] Furthermore, the preset verification rules are a set of automated data verification rules based on applicable project methodologies, monitoring plans, and data quality requirements. The multi-source raw data includes data output from on-site metering equipment, production control systems, and laboratory testing systems. The sensitivity coefficient represents the degree of influence of a unit change in the material conservation equation on the conservation deviation; specifically, it is the absolute value of the partial derivative of the material conservation equation with respect to the anomalous subject. The index residual is the deviation of the rate of change ratio from the nearest boundary value of the constraint interval: when the rate of change ratio is higher than the upper limit of the constraint interval, the deviation is the measured ratio minus the upper limit; when it is lower than the lower limit, the deviation is the lower limit minus the measured ratio. The preset threshold reduction conversion coefficient is a positive value, in percentage points, and is determined for each downstream verification rule based on the difference between the rule's original judgment threshold and the preset minimum threshold lower limit. When it is necessary to uniformly lower the threshold for all downstream verification rules of the associated data item set, the minimum allowable threshold reduction for each rule is taken as the uniform reduction magnitude. The preset dimensions include the anomalous subject, the associated data item set, missing links in the evidence chain, and time-series anomaly markers.

[0018] Furthermore, based on the material conservation equation, a set of associated data items with physical dependencies on the anomaly is determined, including:

[0019] Material input, output and loss data items for each production stage are extracted in advance from project design documents and original ledgers. Material conservation equations for each production stage are established and stored as an equation library.

[0020] When the abnormal subject is an intermediate calculation result, the source variable of the intermediate calculation result is parsed and mapped to the input, output or loss in the material conservation equation;

[0021] When the abnormal subject is a raw data item, the material conservation equation corresponding to its production process is directly located.

[0022] Data items belonging to the same set of variables in the material conservation equation are identified as having physical dependencies and thus constitute a set of related data items.

[0023] Furthermore, a targeted review of the associated data item set will be conducted, including:

[0024] Starting from the abnormal subject, the downstream verification rules are identified in the forward direction along the emission reduction calculation path corresponding to the set of associated data items. The abnormal intensity value of the abnormal subject is calculated. The abnormal intensity value is the absolute value of the relative deviation between the measured value and the preset standard value when the preset verification rule is triggered.

[0025] The abnormal intensity value is multiplied by the methodological measurement error attenuation coefficient of the data item corresponding to each downstream verification rule to obtain the transmission intensity value. When a downstream verification rule corresponds to multiple data items, the transmission intensity value of each data item is calculated separately, and the maximum transmission intensity value is taken as the transmission intensity value of the downstream verification rule. The methodological measurement error attenuation coefficient is determined as follows: when the project methodology specifies the maximum permissible measurement error rate, the attenuation coefficient is 1 minus the error rate; when not specified, it is determined according to the maximum permissible error rate in the calibration certificate of the metrological equipment.

[0026] In the downstream verification rules, rules whose transmission strength value is greater than or equal to the preset anomaly judgment threshold are identified as rules that actually trigger a review, and targeted reviews are only performed on rules that actually trigger a review.

[0027] Furthermore, an evidence chain is generated by tracing back along the data acquisition and recording chain. The data acquisition and recording chain is the traceability path of data from on-site collection, equipment transmission, system recording to final archiving, including:

[0028] Collect the collection time, device number, operator identification, and original voucher number for each data item along the traceability path;

[0029] Identify missing links in the data collection and recording chain within the chain of evidence;

[0030] The weight of the original suspicious point is determined based on the distance between the missing link and the abnormal subject in the data collection and recording chain. The weight of the original suspicious point is negatively correlated with the distance between the intervals.

[0031] When the missing data item belongs to the set of related data items and is used as an input, output or loss in the material conservation equation, the original suspicious point weight is multiplied by the enhancement coefficient to obtain the corrected suspicious point weight. For missing items that do not meet the aforementioned enhancement conditions, the corrected suspicious point weight is equal to the original suspicious point weight. The enhancement coefficient is equal to the ratio of the sensitivity coefficient of the data item to the preset baseline sensitivity coefficient. The preset baseline sensitivity coefficient is a baseline value that is greater than zero, and is the larger of the arithmetic mean, median or preset minimum baseline value of all non-zero sensitivity coefficients of the item.

[0032] The anomaly level of the abnormal event record is determined based on the anomaly intensity value, the conduction intensity value, and the corrected suspicious point weight.

[0033] Furthermore, the rate of change is the year-on-year change rate between adjacent complete reporting periods; the constraint interval is primarily determined based on the allowable deviation of the project methodology; when the project methodology does not specify, it is determined based on the material conservation equation and the statistical characteristics of historical anomaly-free data; when the anomaly is an intermediate calculation result, its historical numerical sequence is recalculated based on the source variables of the corresponding reporting period; the process change time point is determined based on the project design documents, monitoring plan, or confirmed process change records; the historical numerical sequence is segmented according to the process change time point, including:

[0034] The first complete reporting period after project filing without process changes and within the initial temporary constraint range is defined as the normal data window for the initial segment;

[0035] When a process change event is detected as recorded in the project design documents, monitoring plan, or a confirmed process change record, a breakpoint is set at the corresponding change time point.

[0036] The first complete reporting period after the change that does not trigger a timing anomaly will be defined as the first anomaly-free reporting period after the change.

[0037] Based on the average of the ratios of change rates during the first reporting period without abnormalities after the change, and the maximum permissible deviation specified by the project methodology, the constraint intervals for subsequent segments are redefined.

[0038] When a project is in the initial verification stage and the historical numerical series is insufficient to construct a statistical benchmark, the initial temporary constraint interval is formed by combining the benchmark parameters in the project design documents with the allowable deviation of the project methodology.

[0039] When a process change occurs, if there is no abnormality reporting period that meets the conditions after the change, the most recent effective segmented constraint interval before the change is used as a temporary reference for the transition period. If there is no segmented constraint interval before the change, the baseline parameters in the project design documents are combined with the allowable deviation of the project methodology to form a temporary constraint interval for the transition period. If the current complete reporting period after the change does not exceed the temporary constraint interval for the transition period, it is determined as the first abnormality reporting period after the change, and the adaptive segmented constraint interval based on the average of that reporting period is switched.

[0040] Furthermore, when the abnormal entity is an emission factor, the associated production indicator is the corresponding material property detection indicator. The linkage between the emission factor and the material property detection indicator is determined based on the carbon conservation law. The change rate ratio is the ratio of the emission factor change rate to the material property detection indicator change rate. The constraint interval is determined based on the historical detection ratio statistics of the same material category, the same detection method, and within the same process segment in the same project, or the allowable deviation specified by the project methodology. When the absolute value of the material property detection indicator change rate is less than the preset small change threshold, the change rate ratio is not calculated, and the absolute value of the emission factor change rate is compared with the preset stability threshold for judgment. The preset small change threshold is determined based on the accuracy level of the metering equipment or the allowable fluctuation range specified by the project methodology. The preset stability threshold is determined based on the allowable deviation of the project methodology, the accuracy level of the metering equipment, or the fluctuation range of historical data without abnormalities.

[0041] Furthermore, when the abnormal subject is active data, the associated production indicator is output, and the rate of change ratio is the ratio of the rate of change of active data to the rate of change of output. The constraint interval satisfies that the absolute deviation of the ratio from the theoretical ratio of active data and output determined by the material conservation equation does not exceed a preset coordination threshold. The preset coordination threshold is calculated based on the allowable deviation range of active data and output specified in the project methodology, combined with the uncertainty propagation relationship of the material conservation equation. If there is no explicit methodology, it can be preset based on the project methodology, monitoring plan, or historical fluctuation range of data without abnormalities. When the absolute value of the output change rate is less than the preset small change threshold, the rate of change ratio is not calculated, and instead, the absolute value of the active data change rate is compared with the preset stability threshold. The preset small change threshold is determined based on the accuracy level of the metering equipment or the allowable fluctuation range specified in the project methodology. The preset stability threshold is determined based on the allowable deviation of the project methodology, the accuracy level of the metering equipment, or the fluctuation range of historical data without abnormalities.

[0042] This invention also provides a carbon credit project data consistency review system for implementing the above method, including a data access interface, an equation library, a rule library, and a report output interface, and:

[0043] The initial review module is used to obtain the current multi-source raw data and perform preliminary consistency verification. When a preset verification rule is triggered, an abnormal event record is generated to determine the original data item or intermediate calculation result that triggered the rule as the abnormal subject.

[0044] The association and location module is used to determine the set of associated data items that have physical dependencies on the abnormal entity based on the material conservation equation of the project's production process.

[0045] The targeted review module is used to identify downstream verification rules in a forward manner starting from the abnormal subject, calculate the abnormality intensity value and the transmission intensity value, filter the rules that actually trigger the review, and execute the targeted review task.

[0046] The evidence chain tracing module is used to generate an evidence chain by tracing back along the data acquisition record chain. It determines the original suspicious point weight based on the interval distance of the missing links in the data acquisition record chain. When the missing link involves related data items and belongs to the input, output or loss of the material conservation equation, the suspicious point weight is corrected with an enhancement coefficient, and the abnormality level is determined by combining the abnormality intensity value and the transmission intensity value.

[0047] The timing feedback module is used to extract the historical numerical sequence of the abnormal subject and related data item set, segment it according to the time point of process change, mark the timing abnormality when the ratio of change rate in the segment exceeds the constraint interval, and calculate the index residual. Multiply the index residual by the sensitivity coefficient of the abnormal subject in the material conservation equation and the preset threshold reduction conversion coefficient to obtain the threshold reduction amount, reduce the review judgment threshold of the related data item set, and ensure that the reduced judgment threshold is not lower than the preset minimum threshold lower limit, and trigger the targeted review module and the evidence chain tracing module to re-perform targeted review and reverse tracing based on the reduced threshold;

[0048] The aggregation report module is used to summarize abnormal event records within the same review period, and output a consistency review report after being aggregated and associated according to preset dimensions.

[0049] The data access interface provides raw data to the initial review module, the equation library is called by the correlation positioning module and the time-series feedback module, the rule library is called by the initial review module and the targeted review module, and the report output interface connects to the external verification and supervision platform to output review documents.

[0050] Compared with existing technologies, the advantages of this invention are:

[0051] 1. This invention uses the material conservation equation to lock down all related indicators associated with anomalies, and only conducts targeted reviews on the screened target content. At the same time, it relies on measurement error to filter out meaningless verification tasks, which not only reduces the large amount of repetitive work caused by full verification, but also helps to identify deviations and systematic data fraud caused by the linkage of multiple indicators. It balances review efficiency and the ability to identify related anomalies, and reduces the amount of data processing required for full traversal.

[0052] 2. This invention breaks down historical data by filing process change nodes, dynamically updates the data fluctuation range to adapt to the current working conditions, and uses the ratio of changes in indicators over multiple periods for verification. It can adapt to production changes after equipment modification, and effectively identify hidden data anomalies such as single-period compliance and long-term slow deviations, reducing misjudgments and delayed problem detection.

[0053] 3. This invention calculates risk scores by layering the location of voucher links and the influence coefficient of material balance, and determines the level of abnormality by relying on unified quantitative rules. It no longer relies solely on the subjective experience of auditors to classify risks. The main evaluation criteria are based on the physical laws of production, which helps to improve the consistency of the verification conclusions of different personnel and different projects, and the results can be repeatedly verified.

[0054] 4. This invention automatically tightens the current period's verification threshold by calculating the deviation range through time-series anomalies, and performs a second review of related data, opening up a linkage channel between historical trend analysis and current period verification, forming a self-optimizing verification closed loop. It can identify slight deviations and operational condition adaptation anomalies that are difficult to detect with traditional fixed thresholds, which helps to improve the sensitivity of data review and reduce the number of invalid review tasks. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating the overall structure of a carbon credit project data consistency review method provided by the present invention.

[0056] Figure 2 This is a flowchart of the targeted review and evidence chain risk assessment process in a carbon credit project data consistency review method provided by the present invention. Detailed Implementation

[0057] The technical solution of the present invention will be described in detail below with reference to specific embodiments. It should be understood that the following embodiments are only some implementations of the present invention, and all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] It should be noted that the carbon credit projects referred to in this invention include voluntary greenhouse gas emission reduction projects and other data verification projects that target emission reduction credits.

[0059] Example 1

[0060] like Figure 1 As shown, a method for verifying the consistency of carbon credit project data includes the following steps:

[0061] S1. Obtain the current multi-source raw data and perform preliminary consistency verification. When the preset verification rule is triggered, generate an abnormal event record and determine the raw data item or intermediate calculation result that triggered the rule as the abnormal subject.

[0062] Specifically, the multi-source raw data consists of comprehensive basic data from various data acquisition systems, ledger management systems, and testing systems of the project. It covers the entire business chain of production, operation, and carbon accounting for voluntary greenhouse gas emission reduction projects. Specifically, it includes fuel or raw material procurement ledger data, on-site weighing and metering data, material inventory statistics, fuel or raw material testing data, unit production and operation log data, energy consumption statistics, greenhouse gas emission monitoring data, and metering equipment calibration data, thus achieving multi-dimensional coverage of accounting data sources.

[0063] The preset verification rules are a standardized and automated set of rules, which differs from the traditional manual and experience-based verification mode. They include at least five categories: compliance rules for methodological parameters, boundary rules for monitoring plan data, rules for the operating status of metering equipment, rules for the integrity of original vouchers, and rules for the standardization of unit conversion. They can cover various legal unit conversion verification scenarios and realize multi-dimensional automated verification of the compliance, integrity, and standardization of accounting data.

[0064] Abnormal event logs are standardized and traceable records automatically generated after the system triggers an anomaly determination. They possess unique identification, statistical comparability, and traceability, and at least include a unique identifier for the triggering rule, an anomaly trigger timestamp, an identifier for the data item involved, the original data item or intermediate calculation result, the anomaly type, and an anomaly description information. This provides standardized data support for subsequent correlation analysis, source tracing verification, risk rating, and review report output.

[0065] The anomaly subjects are the core analysis objects of this round of anomaly verification, and are specifically divided into two categories: one is the raw data items directly collected on site, including fuel consumption, raw material usage, equipment running time, and grid-connected electricity in a single reporting period; the other is the intermediate calculation results obtained based on the raw data, including greenhouse gas emissions in a single link, emission factor per unit energy consumption, system operating efficiency, and equivalent fuel consumption, comprehensively covering anomaly scenarios throughout the entire carbon accounting process.

[0066] After data collection is completed, the system performs a preliminary consistency check on the current multi-source raw data. It automatically compares the collected raw data with the project filing methodology parameters, fixed parameters in the project design documents, and the standard specifications of the routine monitoring plan. The verification dimensions cover the range of data parameter values, data format specifications, legal units of measurement, equipment monitoring accuracy, data collection frequency, etc., to achieve preliminary screening of the basic compliance of raw data, eliminate obviously non-compliant data, and reduce the computing power consumption and unnecessary workload of subsequent in-depth verification.

[0067] In this embodiment, the system connects to the various business data source systems of the project through a standardized data interface, automatically collects all the original data of the current period in batches according to the preset review cycle, and traverses and executes all preset verification rules. The measured value of each data item is compared and verified with the preset standard value, reasonable threshold range and specification requirements one by one. If any preset verification rule is not met and the anomaly judgment mechanism is triggered, the system automatically generates a standardized anomaly event record and accurately marks the corresponding original data item or intermediate calculation result as the anomaly subject, which serves as the sole starting point for subsequent correlation analysis and in-depth verification, ensuring the uniqueness and accuracy of anomaly tracing.

[0068] S2. Based on the material conservation equation of the project production process, determine the set of associated data items that have a physical dependency relationship with the abnormal subject, and determine the corresponding sensitivity coefficient for each data item in the variable set of the material conservation equation.

[0069] Specifically, the material conservation equation is a standardized material conservation relationship constructed based on the law of conservation of mass, used to characterize the balance between material input, output and loss in each production stage of the project. Its general form is that the input quantity equals the sum of the output quantity and the loss quantity. Physical dependency refers to multiple sets of data items belonging to the same variable system of the material conservation equation, and the data have fixed and objective physical constraint relationships. The set of associated data items is the set of all data items that have physical dependencies on the abnormal subject.

[0070] Upon completion of project registration, the system initiates a pre-modeling process, extracting material input, output, and loss data items for each production stage from project design documents and original ledgers. Standardized material conservation equations are then established for each production stage and stored in an equation library. These material conservation equations are continuously reused in subsequent review cycles, eliminating the need for re-establishment after each anomaly trigger. Once the anomaly is identified, the system directly retrieves the corresponding material conservation equation for the production stage to which the anomaly belongs from the equation library, completing variable mapping and aggregation of associated data items, and simultaneously calculating the corresponding sensitivity coefficient for each data item in the equation variable set.

[0071] The system employs differentiated correlation analysis strategies for different types of abnormal entities: if the abnormal entity is an intermediate calculation result, it reverse-analyzes all its source variables and maps the source variables to the input, output, or loss quantities in the material conservation equation; if the abnormal entity is an original data item, it directly locates the material conservation equation corresponding to its production process, thereby achieving accurate locking of abnormal correlation data.

[0072] In this embodiment, the system first determines the type of abnormal subject, matches the corresponding association parsing logic based on the type of abnormal subject, collects all variable data items under the corresponding material conservation equation, and forms a closed-loop association data item set. This eliminates the limitations of traditional single-point verification, effectively avoids the problems of missed or incorrect detection of association data, and accurately defines the boundary range of subsequent in-depth verification.

[0073] S3. Conduct targeted review of the set of related data items and trace back along the data collection record chain to generate an evidence chain.

[0074] Specifically, such as Figure 2 As shown, targeted review is a targeted review mode that only targets the set of related data items and the corresponding downstream verification rules. It is different from the repeated verification of all data. It can achieve targeted focus of the review scope by accurately identifying the abnormal transmission path. The evidence chain is a standardized traceability file generated by the traceability, which includes metadata, original vouchers, and records of missing links.

[0075] The system starts with the anomalous entity and identifies downstream verification rules forward along the emission reduction calculation path corresponding to the set of associated data items. Based on the identified downstream verification rules, it generates a targeted review task list and executes it. The emission reduction calculation path is a standardized sequence of calculation steps from raw monitoring data, through intermediate calculation results, stage emissions, baseline emissions, to the final emission reduction. Downstream verification rules are those that use data items directly or indirectly involved in the calculation by the anomalous entity as core input parameters. When a downstream verification rule corresponds to multiple data items, the transmission strength value of each data item is calculated separately, and the maximum transmission strength value is used as the transmission strength value of the downstream verification rule.

[0076] During the targeted review process, the system calculates the anomaly intensity value of the abnormal entity. The anomaly intensity value is the absolute value of the relative deviation between the measured value and the preset standard value when the preset verification rule is triggered. The anomaly intensity value is multiplied by the methodological measurement error attenuation coefficient of the data item corresponding to each downstream verification rule to obtain the transmission intensity value. The methodological measurement error attenuation coefficient is determined as follows: when the project methodology specifies a maximum permissible measurement error rate, the attenuation coefficient is 1 minus that error rate; when not specified, it is determined according to the maximum permissible error rate in the calibration certificate of the metrological equipment. The system identifies rules in the downstream verification rules whose transmission intensity value is greater than or equal to the preset anomaly judgment threshold as the rules that actually trigger the review, and only performs targeted reviews on the rules that actually trigger the review, filtering out invalid anomaly transmission caused by normal measurement errors.

[0077] Furthermore, the system traces back along the data acquisition record chain to generate an evidence chain. It collects the acquisition time, equipment number, operator identification, and original voucher number for each data item along the tracing path, marks missing links in the data acquisition record chain, and determines the original suspicious point weight based on the distance between the missing link and the abnormal entity in the data acquisition record chain. The original suspicious point weight is negatively correlated with the distance. When the data item involved in the missing link belongs to a set of related data items and is used as an input, output, or loss in the material conservation equation, the original suspicious point weight is multiplied by an enhancement coefficient to obtain the corrected suspicious point weight. For missing links that do not meet the aforementioned enhancement conditions, the corrected suspicious point weight is equal to the original suspicious point weight. The enhancement coefficient is equal to the ratio of the sensitivity coefficient of the data item to the preset baseline sensitivity coefficient. The preset baseline sensitivity coefficient is a baseline value greater than zero, taken as the larger of the arithmetic mean, median, or preset minimum baseline value of all non-zero sensitivity coefficients of the item.

[0078] The system integrates anomaly intensity value, conduction intensity value, and corrected suspicious point weight to determine the anomaly level of an abnormal event record. To ensure the composability of parameters with different dimensions, the system adopts a unified normalization rule: the anomaly intensity value and conduction intensity value are divided by their corresponding preset benchmark values, mapping them to a dimensionless risk score in the 0-1 range; the corrected suspicious point weight itself is a value in the 0-1 range and is directly used as a risk score in the calculation. The final anomaly level is determined according to a preset risk level mapping table: when the comprehensive risk score reaches the first threshold, it is judged as high risk; when it reaches the second threshold, it is judged as medium risk; otherwise, it is judged as low risk. The risk level mapping table is preset based on project methodological requirements or historical verification data statistics.

[0079] This step is performed entirely within the aforementioned set of related data items. Both forward rule identification and reverse link tracing are carried out based on the boundaries of the related set to ensure the targeting of the review and the completeness of the tracing, thereby achieving refined and low-redundancy anomaly verification and risk assessment.

[0080] S4. Extract the historical numerical sequence of the abnormal subject and associated data item set, divide the historical numerical sequence into segments according to the process change time point, calculate the change rate of the abnormal subject and the change rate of the associated production indicator within the segment, and obtain the change rate ratio. When the change rate ratio exceeds the constraint interval corresponding to the data type of the abnormal subject, mark the time series as abnormal and calculate the indicator residual.

[0081] Specifically, the historical numerical sequence consists of continuous and valid data records across multiple reporting periods; the process change time point is the official effective date of equipment upgrades, process modifications, or operating parameter adjustments as specified in the project design documents, monitoring plans, or confirmed process change records; time-series anomalies are abnormal states where the ratio of the rate of change of an indicator deviates from the constraint interval within a segmented interval; the indicator residual is the deviation of the ratio of the rate of change from the nearest boundary value of the constraint interval: when the ratio of the rate of change is higher than the upper limit of the constraint interval, the deviation is the measured ratio minus the upper limit value; when it is lower than the lower limit, the deviation is the lower limit value minus the measured ratio; related production indicators are production parameters that belong to the same physical constraint system as the indicator to be measured and have a quantitative linkage relationship. The rate of change is the year-on-year rate of change between adjacent complete reporting periods; the constraint interval is primarily determined based on the allowable deviation of the project methodology; when the project methodology does not specify, it is determined based on the material conservation equation and the statistical characteristics of historical data without anomalies; when the anomaly is an intermediate calculation result, its historical numerical sequence is recalculated based on the source variables of the corresponding reporting period.

[0082] The system designates the first complete reporting period after project registration without process changes and within the initial temporary constraint interval as the normal data window for the initial segment. When a process change event is detected as recorded in the project design documents, monitoring plan, or confirmed process change record, a segmentation breakpoint is set at the corresponding change time point. The first complete reporting period after the change that does not trigger a time-series anomaly is designated as the first anomaly-free reporting period after the change. Based on the mean of the change rate ratio within the first anomaly-free reporting period after the change, and the maximum permissible deviation specified by the project methodology, the constraint intervals for subsequent segments are redefined. When the project is in the initial verification phase and the historical numerical sequence is insufficient to construct a statistical benchmark, the initial temporary constraint interval is formed by combining the baseline parameters in the project design documents with the permissible deviation of the project methodology. When a process change occurs, if there is no abnormality reporting period that meets the conditions after the change, the most recent effective segmented constraint interval before the change is used as a temporary reference for the transition period. If there is no segmented constraint interval before the change, the baseline parameters in the project design documents are combined with the allowable deviation of the project methodology to form a temporary constraint interval for the transition period. If the current complete reporting period after the change does not exceed the temporary constraint interval for the transition period, it is determined as the first abnormality reporting period after the change, and the adaptive segmented constraint interval based on the average of that reporting period is switched.

[0083] For different types of data metrics, the constraint intervals and related production metrics are configured differently:

[0084] When the anomaly is an emission factor, the associated production indicator is the corresponding material property detection indicator. The linkage between the emission factor and the material property detection indicator is determined based on the law of carbon conservation. The change rate ratio is the ratio of the emission factor change rate to the material property detection indicator change rate. The constraint interval is determined based on the statistical value of historical detection ratios for the same material category, the same detection method, and within the same process segment in the same project, or the allowable deviation specified by the project methodology. Under normal operating conditions, there is a fixed linear correlation between the emission factor and the material property indicator, and the change rate ratio should be stable near the benchmark value. If the measured ratio exceeds the constraint interval, it indicates that there is abnormal drift in the emission factor or material property data. When the absolute value of the change rate of the material property detection indicator is less than the preset small change threshold, the change rate ratio is not calculated, and the absolute value of the emission factor change rate is compared with the preset stability threshold for judgment. The preset small change threshold is determined based on the accuracy level of the metering equipment or the allowable fluctuation range specified by the project methodology; the preset stability threshold is determined based on the allowable deviation of the project methodology, the accuracy level of the metering equipment, or the fluctuation range of historical data without anomalies.

[0085] When the anomaly is active data, the associated production indicator is output, and the rate of change ratio is the ratio of the rate of change of active data to the rate of change of output. The constraint interval satisfies that the absolute deviation of the ratio from the theoretical ratio of active data and output determined by the material conservation equation does not exceed a preset coordination threshold. The preset coordination threshold is calculated based on the allowable deviation range of active data and output specified in the project methodology, combined with the uncertainty propagation relationship of the material conservation equation. If there is no explicit methodology, it can be preset based on the project methodology, monitoring plan, or historical fluctuation range of data without anomalies. When the absolute value of the output change rate is less than the preset small change threshold, the rate of change ratio is not calculated; instead, the absolute value of the active data change rate is compared with the preset stability threshold. The preset small change threshold is determined based on the accuracy level of the measuring equipment or the allowable fluctuation range specified in the project methodology; the preset stability threshold is determined based on the allowable deviation of the project methodology, the accuracy level of the measuring equipment, or the fluctuation range of historical data without anomalies.

[0086] Furthermore, the historical numerical sequence extraction range fully covers the set of associated data items, ensuring a high degree of consistency between the time series analysis scope and the verification scope. The system automatically filters data jumps caused by non-process factors such as equipment calibration, unit adjustments, and system upgrades. The filtering criteria include equipment verification records, unit conversion change records, and system upgrade logs, marking these situations as non-process breakpoints or removing them, ensuring the accuracy of time series segmentation and anomaly analysis. The constraint interval adopts a recursive update mechanism. For each new reporting period without anomalies, the system incorporates the current period's rate of change ratio into the mean calculation, smoothly updating the constraint interval boundary. If a new reporting period triggers a time series anomaly, the original constraint interval is retained to avoid abnormal data contaminating the baseline parameters.

[0087] This step can effectively identify hidden time-series anomalies such as single-period compliance and long-term gradual drift, making up for the technical shortcomings of traditional fixed threshold verification, which cannot adapt to process change conditions and has a high rate of missed detection.

[0088] S5. Multiply the index residual by the sensitivity coefficient of the abnormal subject in the material conservation equation and the preset threshold reduction conversion coefficient to obtain the threshold reduction amount. Reduce the review judgment threshold of the related data item set, and the reduced judgment threshold is not lower than the preset minimum threshold lower limit. Based on the reduced threshold, re-examine and reverse trace the related data item set, and update the evidence chain and abnormal event records.

[0089] Specifically, the sensitivity coefficient represents the degree of influence of a unit change in the material conservation equation on the conservation deviation of the abnormal subject. Specifically, it is the absolute value of the partial derivative of the material conservation equation with respect to the abnormal subject, used to quantify the impact of fluctuations in a single data indicator on the overall material balance relationship and the final emission reduction accounting result. The review judgment threshold is the abnormal triggering critical value of each verification rule. The threshold reduction amount is obtained by multiplying the indicator residual by the sensitivity coefficient and the preset threshold reduction conversion coefficient. The preset threshold reduction conversion coefficient is a positive value in percentage points. For each downstream verification rule, it is determined according to the difference between the original judgment threshold of the rule and the preset minimum threshold lower limit. When it is necessary to uniformly lower the threshold of all downstream verification rules of the associated data item set, the minimum value among the allowed threshold reduction amounts of each rule is taken as the uniform reduction range, which represents the tightening range of the verification standard after the deviation of the time series anomaly from the physical constraint is corrected by the conservation weight.

[0090] After the system flags a timing anomaly, it automatically retrieves the sensitivity coefficient of the corresponding anomaly in the material conservation equation, calculates the precise threshold reduction, and lowers the anomaly judgment threshold for all downstream verification rules corresponding to the associated data item set. To avoid excessive threshold tightening leading to large-scale false alarms, this step sets a minimum threshold protection mechanism: if the lowered judgment threshold is lower than the preset minimum threshold, the preset minimum threshold is used as the final verification threshold; the preset minimum threshold is determined based on the project's methodology data quality level requirements or the historical false alarm rate statistics for the same project, similar projects, or projects using the same methodology.

[0091] After the threshold is updated, the system replaces the original judgment threshold and re-executes the targeted review and reverse tracing process to complete the closed-loop in-depth review. The boundary of the second review is completely consistent with that of the first targeted review, ensuring continuous traceability of the review scope. When there are multiple time-series anomalies in the same review cycle, the system takes the largest threshold reduction as the uniform reduction range to ensure that the rigor of the review covers all time-series anomaly risks and accurately captures gradual data drift, slight correlation deviations, and abnormal operating conditions that cannot be identified by fixed thresholds.

[0092] S6. Summarize the abnormal event records within the same review cycle, and output a consistency review report after associating and aggregating them according to preset dimensions.

[0093] Specifically, the abnormal event record covers four types of abnormal data: initial compliance review abnormalities, newly added abnormalities during targeted review, time-series drift abnormalities, and abnormalities with missing evidence chains; correlation aggregation is a standardized statistical processing method that merges abnormalities from the same source and removes redundant and duplicate records according to a unified statistical dimension; and the consistency review report is a structured archived document adapted to project emission reduction verification, enterprise self-inspection, and industry supervision.

[0094] The system aggregates all abnormal event records generated throughout the current period, and completes association aggregation and deduplication based on four dimensions: abnormal subject, related data item set, missing link in the evidence chain, and time-series abnormality mark. It also completes risk stratification and classification by combining parameters such as abnormality intensity, transmission intensity, abnormality level, and number of time-series abnormalities, and finally generates a standardized structured data consistency review report.

[0095] Each step generates anomaly records with a unique event identifier, enabling precise merging of anomalies from the same source and elimination of duplicate data. All newly added anomalies during closed-loop review are included in the summary statistics. The output structured report can be directly integrated with emission reduction verification systems and industry regulatory platforms, demonstrating excellent adaptability and portability.

[0096] A carbon credit project data consistency review system, used to implement the above method, includes a data access interface, an equation library, a rule library, and a report output interface, as well as:

[0097] The initial review module is used to obtain the current multi-source raw data and perform preliminary consistency verification. When a preset verification rule is triggered, an abnormal event record is generated to determine the original data item or intermediate calculation result that triggered the rule as the abnormal subject.

[0098] The association and location module is used to determine the set of associated data items that have physical dependencies on the abnormal entity based on the material conservation equation of the project's production process.

[0099] The targeted review module is used to identify downstream verification rules in a forward manner starting from the abnormal subject, calculate the abnormality intensity value and the transmission intensity value, filter the rules that actually trigger the review, and execute the targeted review task.

[0100] The evidence chain tracing module is used to generate an evidence chain by tracing back along the data acquisition record chain. It determines the original suspicious point weight based on the interval distance of the missing links in the data acquisition record chain. When the missing link involves related data items and belongs to the input, output or loss of the material conservation equation, the suspicious point weight is corrected with an enhancement coefficient, and the abnormality level is determined by combining the abnormality intensity value and the transmission intensity value.

[0101] The timing feedback module is used to extract the historical numerical sequence of the abnormal subject and related data item set, segment it according to the time point of process change, mark the timing abnormality when the ratio of change rate in the segment exceeds the constraint interval, and calculate the index residual. Multiply the index residual by the sensitivity coefficient of the abnormal subject in the material conservation equation and the preset threshold reduction conversion coefficient to obtain the threshold reduction amount, reduce the review judgment threshold of the related data item set, and ensure that the reduced judgment threshold is not lower than the preset minimum threshold lower limit, and trigger the targeted review module and the evidence chain tracing module to re-perform targeted review and reverse tracing based on the reduced threshold;

[0102] The aggregation report module is used to summarize abnormal event records within the same review period, and output a consistency review report after being aggregated and associated according to preset dimensions.

[0103] The data access interface provides raw data to the initial review module, the equation library is called by the correlation positioning module and the time-series feedback module, the rule library is called by the initial review module and the targeted review module, and the report output interface connects to the external verification and supervision platform to output review documents.

[0104] Example 2

[0105] This embodiment strictly corresponds to the standardized technical process S1-S6 disclosed in Embodiment 1, using a domestically registered biomass direct combustion power generation greenhouse gas voluntary emission reduction project as an exemplary application scenario to demonstrate the complete execution process of the technical solution of this invention. All specific values ​​appearing in the following steps are exemplary values ​​used to facilitate the explanation of the method flow. In actual applications, each parameter depends on the project registration methodology, historical data statistical characteristics, actual monitoring data, and industry regulatory standards, and is dynamically calculated and determined by the system based on project documents and real-time data. The specific values ​​used in this embodiment do not constitute a limitation on the scope of protection of this invention.

[0106] S1. Obtain current multi-source raw data and perform preliminary consistency verification to locate abnormal entities.

[0107] Once the review period begins, the system automatically connects to six major business terminals—the project's weighbridge system, fuel inventory management system, DCS production control system (distributed control system), checkpoint power metering system, fuel testing management system, and equipment calibration ledger system—through standardized data interfaces. It then collects multi-source raw data from all dimensions for the current period in batches, covering the entire accounting chain from fuel weighing upon arrival at the plant to inventory statistics, unit consumption, power generation metering, fuel testing, and equipment calibration.

[0108] The system loads five categories of pre-set verification rules: compliance of methodology parameters, compliance of monitoring plan boundaries, compliance of equipment status, integrity of vouchers, and standardization of unit conversion. It performs preliminary consistency checks on the current multi-source raw data, checking the data value range, format specifications, unit of measurement, monitoring accuracy and collection frequency item by item, and eliminating unqualified data with abnormal format, disordered units, and invalid jumps.

[0109] After system-wide verification, an anomaly was detected at the monitoring plan data boundary during this review period. The system automatically generated a standardized anomaly event record, which included a unique event ID, anomaly timestamp, involved data identifier, measured value, anomaly type, and violation basis. The corresponding original data item or intermediate calculation result was accurately marked as the anomaly subject, serving as the sole starting point for subsequent correlation analysis and in-depth verification.

[0110] S2. Constructing a set of related data items based on the material conservation equation.

[0111] This step takes the abnormal entity located by S1 as the starting point for analysis. Based on the project filing material list, a two-level material conservation system is constructed, corresponding to the fuel inventory stage and the combustion stage respectively, to achieve multi-level correlation verification from quantity balance to element balance.

[0112] (1) Material conservation in fuel inventory

[0113] The equation describing the quantity balance of fuel from the plant and storage to the furnace is in the form of:

[0114] Total fuel input (current period input + beginning inventory) = current period fuel consumption + ending inventory + natural fuel loss

[0115] Example values ​​for each data item are shown in the table below (in this embodiment, the upper limit for fuel consumption compliance is set to 550 tons based on the project filing methodology).

[0116]

[0117] The system performs a closed-loop material balance check on the set of associated data items: Total fuel input = 580 tons + 320 tons = 900 tons; Total fuel output and loss = 562 tons + 280 tons + 6.3 tons = 848.3 tons. There is a material imbalance difference of 51.7 tons between the input and output, which is marked as "material imbalance". This material imbalance marker is used as an associated anomaly marker for evidence chain tracing; when the material balance rule is pre-configured as a preset verification rule, it can trigger a review as a preset verification rule. This, along with the set of associated data items, is then passed to step S3.

[0118] (2) Conservation of carbon during combustion

[0119] The migration equilibrium of carbon elements after combustion of fuel in the furnace is described, which is the core physical basis for verifying emission factor data. The equation is in the form of:

[0120] Total carbon content of fuel fed into the furnace = carbon emissions from flue gas + carbon content of slag + carbon content of fly ash

[0121] Example values ​​for each data item are shown in the table below.

[0122]

[0123] The system performs carbon conservation calculations on the combustion process: total output carbon = 261.8 + 6.9 + 4.7 = 273.4 tons of carbon, which deviates from the total carbon input to the furnace by 0.9 tons, a relative deviation of approximately 0.33%, within the methodological tolerance range. This level of conservation can be further correlated with detection data such as fuel carbon content and emission factors, providing a physical constraint basis for the correlation verification of emission factor anomalies.

[0124] In practical applications, the specific form of the material conservation equation, the data items involved, and the bill of materials all depend on the actual production process recorded in the project filing documents. Depending on the project type, it can be extended to material balance relationships such as water balance and carbon conservation.

[0125] S3. Targeted review of related data, reverse tracing of evidence chains, and anomaly level quantification.

[0126] (I) Targeted re-examination and conduction strength calculation

[0127] The system starts with the anomalous entity and identifies downstream verification rules along the standardized emission reduction calculation path, generating a targeted review task list. The emission reduction calculation path is a sequence of standardized calculation steps from raw monitoring data through intermediate calculation results to the final emission reduction. Downstream verification rules are those that use data items obtained directly or indirectly from the anomalous entity as core input parameters. The three thresholds exemplified in S3 are merely examples of downstream verification rules along the emission reduction calculation path; the possibility of material conservation rules triggering reviews themselves cannot be ruled out.

[0128] The system calculates the anomaly intensity value of the abnormal entity, which is the absolute value of the relative deviation between the measured value and the preset standard value. The anomaly intensity value is then multiplied by a methodological measurement error attenuation coefficient to obtain the conduction intensity value. This coefficient is set to 1 minus the maximum permissible measurement error rate specified by the applicable project methodology. The system identifies rules in the downstream verification rules whose conduction intensity value is greater than or equal to the preset anomaly judgment threshold as rules that actually trigger a review, and only performs targeted reviews on rules that actually trigger a review.

[0129] In this embodiment, the downstream verification rules are exemplarily set to three items: unit conversion verification, calorific value substitution verification, and emission range verification. In this embodiment, the original threshold for the calorific value substitution verification rule is 2.29%, used only to demonstrate the closed-loop triggering process; in actual projects, this threshold is determined by the filing methodology. The unit conversion verification threshold is exemplarily set to 5%, and the emission range verification threshold is exemplarily set to 10%. The maximum permissible measurement error rate is exemplarily set to 1.2%, corresponding to a decay coefficient of 0.988. The anomaly intensity value is the absolute value of the relative deviation between the measured value and the compliance upper limit value, calculated as |(562-550) / 550|≈2.18%, and the conduction intensity value = 2.18% × 0.988≈2.15%.

[0130] The conduction intensity value was compared one by one with the preset thresholds of each downstream rule: 2.15% was lower than the unit conversion verification threshold of 5% and the emission range verification threshold of 10%. No new rules were triggered under these two rules during the initial targeted review. The original threshold for calorific value substitution verification was 2.29%, and the conduction intensity of 2.15% during the initial targeted review was also lower than this threshold. Therefore, no new downstream rules were triggered during the initial targeted review, and the system retained the verification confirmation of the abnormal entity itself. In practical applications, the type of downstream verification rule and the corresponding threshold depend on the allowable deviation range specified in the project filing methodology and industry regulatory standards, while the measurement error rate depends on the project equipment calibration certificate and methodology specifications.

[0131] (II) Evidence chain tracing and multidimensional weighted anomaly classification

[0132] The system traces back along the data collection and recording chain, collecting the collection time, equipment number, operator identification, and original voucher number of each data item on the tracing path, marking missing links in the data collection and recording chain, and determining the suspicious point weight based on the distance between the missing link and the abnormal subject in the chain. The suspicious point weight is negatively correlated with the number of nodes.

[0133] In this embodiment, the link hierarchy is exemplarily set as L1-Abnormal Subject, L2-Daily Production Report, L3-Operation Log, and L4-Belt Scale Original Record. The interval distance is defined as the number of link levels separating the missing link from the abnormal subject (i.e., L2 corresponds to a distance of 1, and L3 corresponds to a distance of 2). The system verifies the integrity of the metadata at each level and identifies two missing links: the L2 level Daily Production Report is missing equipment number information, marked as missing link G-01, with an interval distance of 1; the L3 level Operation Log is missing the original data archiving voucher for the belt scale, marked as missing link G-02, with an interval distance of 2.

[0134] According to the rule of this invention, "the weight of suspicious points is negatively correlated with the interval distance," this embodiment exemplarily takes the mapping function as suspicious point weight = 1 / (interval distance + 1), and calculates the suspicious point weight of G-01 as 1 / (1+1) ≈ 0.5, and the suspicious point weight of G-02 as 1 / (2+1) = 0.333. In practical applications, the link hierarchy structure depends on the actual data collection process of the project, and the mapping function can be fitted based on historical project data or set based on industry experience.

[0135] The system integrates three parameters—anomaly intensity value, transmission intensity value, and corrected suspicious point weight—to quantify and classify anomaly levels. To ensure dimensional consistency, all three parameters are converted into dimensionless risk scores using a preset normalized benchmark value before synthesis.

[0136] The abnormal intensity benchmark value is set at 10%, and the abnormal intensity score = abnormal intensity value / benchmark value = 2.18% / 10% = 0.218;

[0137] The conductivity benchmark value is set at 10%, and the conductivity score = conductivity value / benchmark value = 2.15% / 10% = 0.215;

[0138] The weight of doubt is a dimensionless value in the range of 0 to 1, and is used directly as the risk score.

[0139] The three weighting coefficients are configured with the following values: anomaly intensity weight α = 0.4, transmission intensity weight β = 0.3, and suspicion weight γ = 0.3, corresponding to the three independent risk dimensions of direct deviation from anomalies, potential transmission risk, and reliability of the evidence chain, respectively. The specific values ​​of the above weighting coefficients can be preset based on the project's risk appetite or historical verification data statistics.

[0140] Taking missing link G-01 as an example, if the data items involved in this missing link do not belong to the input, output, or loss quantities of the material conservation equation, then the enhancement coefficient is taken as 1, and the corrected suspicious point weight is equal to the original suspicious point weight of 0.5. The corresponding comprehensive anomaly level (risk score) is calculated as follows:

[0141] 0.218×0.4+0.215×0.3+0.5×0.3=0.0872+0.0645+0.15=0.3017

[0142] To fully demonstrate the calculation and use of the enhancement coefficient, assume that the data item involved in the missing link G-01 is "fuel input quantity". This data item is used as an input in the material conservation equation, and its sensitivity coefficient (partial derivative) in the equation is 1.0. The preset baseline sensitivity coefficient (the arithmetic mean of the non-zero sensitivity coefficients of all related data items, ensuring it is greater than zero) is 1.0. Therefore, the enhancement coefficient = 1.0 / 1.0 = 1.0, and the corrected suspicious point weight = 0.5 × 1.0 = 0.5. Substituting this into the anomaly level calculation yields:

[0143] 0.218×0.4+0.215×0.3+0.5×0.3=0.3017

[0144] In practical applications, if the sensitivity coefficient of the data item involved in the missing link in the equation is higher, the anomaly of that data item will significantly disrupt the material balance relationship, and the corresponding weight of the suspicious point will be increased.

[0145] S4. Adaptive timing segmentation for process changes, timing anomaly identification, and index residual calculation.

[0146] The system extracts long-term historical numerical sequences from the set of related data items and adaptively segments the historical numerical sequences based on the time point of process change in project filing. The system designates the first complete reporting period after project filing without process changes and within the initial temporary constraint interval as the initial segmented normal data window. When a process change event is detected as recorded in the project design documents, monitoring plan, or confirmed process change record, a time-series segmentation breakpoint is set at the change's effective time. The first complete reporting period after a process change that does not trigger a time-series anomaly is designated as the baseline reporting period without anomalies after the change. The subsequent segmented constraint intervals are adaptively updated based on the average ratio of indicator change rates within the baseline reporting period, combined with the maximum permissible deviation of the methodology. When the absolute value of the change rate of related production indicators is less than a preset small change threshold, the change rate ratio is not calculated; instead, the absolute value of the change rate of the abnormal subject is compared with a preset stability threshold. If it exceeds this stability threshold, a time-series anomaly is marked. All change rates use the month-on-month change rate caliber between adjacent complete reporting periods. The example values ​​for the maximum permissible deviation of the methodology mentioned above are only for demonstrating the calculation process; in actual applications, this deviation is explicitly specified by the project filing methodology.

[0147] (1) Example of time-series anomaly detection for activity data

[0148] This embodiment takes activity data (fuel consumption) as an example. Based on the baseline period data, the month-on-month change rates of fuel consumption and power generation are exemplarily taken as -2.0% and -2.1% respectively, with a change rate ratio of ≈0.96. The maximum permissible deviation of the methodology is exemplarily taken as ±10%, and the constraint interval [0.96×0.9, 0.96×1.1]=[0.864, 1.056] is adaptively generated.

[0149] During the current verification period, the system calculates the ratio of change rate based on the month-on-month change rate of fuel consumption and power generation. In this embodiment, the ratio of change rate is 1.20. This ratio exceeds the upper limit of the constraint interval of 1.056, so the system marks the time series as abnormal and calculates the index residual = 1.20 - 1.056 = 0.144, providing a quantitative basis for subsequent dynamic threshold correction.

[0150] (2) Example of time series anomaly detection for emission factor class

[0151] Taking the CO2 emission factor per unit of electricity generation as an example, the associated production indicator is the carbon content of the fuel received as a basis. During the baseline period, the year-on-year change rate of the emission factor was 1.2%, and the year-on-year change rate of the fuel carbon content was 1.1%, with a change rate ratio of ≈1.09. Based on the law of carbon conservation, this ratio should be stable around 1.0 under normal operating conditions. Combining the methodological allowable deviation of ±12%, the constraint interval [0.959, 1.221] is adaptively generated.

[0152] During the current verification period, the month-on-month change rate of emission factors was 3.5%, while the month-on-month change rate of fuel carbon content was 1.0%. The ratio of these change rates was 3.5% / 1.0% = 3.5, significantly exceeding the upper limit of the constraint interval of 1.221. The system flagged this as a time-series anomaly and calculated the residual as 3.5 - 1.221 = 2.279. This residual will be used to tighten the thresholds of the corresponding downstream verification rules, enabling accurate capture of emission factor data drift.

[0153] (3) Boundary scene processing

[0154] For boundary scenarios where there is no historical data in the first phase of the project, the system directly uses the baseline parameters in the project design documents combined with the methodological allowable deviation to form a temporary constraint interval. For example, if the design unit power generation fuel consumption of this project is 0.5 tons / MWh, the temporary constraint interval is set to ±15% of the design value, i.e. [0.425, 0.575]. After accumulating data for at least two complete reporting periods, the system will switch to an adaptive segmented constraint interval.

[0155] S5. Sensitivity-weighted time-series anomaly feedback threshold reduction and closed-loop secondary review

[0156] When the S4 flag is abnormal, the system multiplies the index residual by the sensitivity coefficient of the data index in the material conservation equation and the preset threshold reduction conversion coefficient to obtain the threshold reduction amount. The system then lowers the review judgment threshold of the associated data item set and, based on the lowered threshold, conducts a new targeted review and reverse tracing of the associated data item set, updating the evidence chain and abnormal event records.

[0157] The sensitivity coefficient is the absolute value of the partial derivative of the material conservation equation with respect to the corresponding data index, and is determined by the specific form of the material conservation equation for the actual project. In this embodiment, the sensitivity coefficient for fuel consumption is 1 for example, and the preset threshold reduction conversion coefficient is 1 percentage point / ratio residual unit (this coefficient is set according to the allowable deviation range of the project methodology and the difference between the original threshold and the lower limit of the minimum threshold in the verification rules). In S4, the residual of the index obtained by calculating the time series anomaly of the activity data type is 0.144, then the threshold reduction amount = 0.144 × 1 × 1 percentage point / ratio residual unit = 0.144 percentage points.

[0158] The original threshold for the calorific value substitution verification rule is 2.29%, and the new threshold after adjustment is 2.29% - 0.144% = 2.146%. In step S3, the conduction intensity value is 2.15%. Since 2.15% > 2.146%, the system determines that the downstream verification rule has actually triggered a review. Based on this, the targeted review task and evidence chain tracing in step S3 are re-executed, and an anomaly mark is added for the "fuel calorific value substitution value", completing the closed-loop deep review. The main data in this embodiment directly demonstrates the complete closed-loop process of time-series anomaly feedback threshold reduction triggering secondary review.

[0159] To avoid large-scale false alarms caused by excessively tightening thresholds, a minimum threshold protection mechanism is implemented: The system pre-sets a minimum threshold for each verification rule based on the project's methodological data quality level requirements or historical false alarm rate statistics for the same project, similar projects, or projects using the same methodology. After each threshold update, the system automatically compares the adjusted new threshold with the minimum threshold. If the new threshold is lower than the minimum threshold, the minimum threshold is used as the final verification threshold; if the new threshold is not lower than the minimum threshold, the adjusted threshold is applied. This mechanism ensures that the threshold adjustment range is controlled, preventing large-scale invalid false alarms caused by excessively tightening thresholds due to a single abnormal deviation.

[0160] S6. Output of Multi-Dimensional Anomaly Correlation Aggregation and Standardized Review Report

[0161] The system aggregates all anomaly event records within the same review cycle, covering four categories: S1 initial compliance review anomalies, S3 targeted review new anomalies, S4 time-series drift anomalies, and evidence chain missing anomalies. The system performs correlation aggregation and deduplication based on four dimensions: anomaly subject, related data item set, missing link in the evidence chain, and time-series anomaly marker. It then uses parameters such as anomaly intensity, transmission intensity, anomaly level, and number of time-series anomalies to perform risk stratification and classification, ultimately generating a standardized structured data consistency review report. This report can be directly integrated with project emission reduction verification systems and industry regulatory platforms, demonstrating good scenario adaptability and portability.

[0162] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for verifying the consistency of carbon credit project data, applicable to voluntary greenhouse gas emission reduction projects with material input, output, or loss accounting chains, executed by computer equipment, characterized in that, Includes the following steps: Acquire current multi-source raw data and perform preliminary consistency verification. When a preset verification rule is triggered, generate an exception event record and determine the raw data item or intermediate calculation result that triggered the rule as the exception subject. Based on the material conservation equation of the project's production process, a set of associated data items that have a physical dependency relationship with the abnormal entity is determined, and a corresponding sensitivity coefficient is determined for each data item in the set of variables of the material conservation equation. The associated data item set is subject to targeted review, and an evidence chain is generated by tracing back along the data collection record chain. Extract the historical numerical sequence of the abnormal subject and associated data item set, segment the historical numerical sequence according to the process change time point, calculate the change rate of the abnormal subject and the change rate of the associated production indicator within the segment, and obtain the change rate ratio. When the change rate ratio exceeds the constraint interval corresponding to the data type of the abnormal subject, mark the time series abnormality and calculate the indicator residual. Multiply the residual of the index by the sensitivity coefficient of the abnormal subject in the material conservation equation and the preset threshold reduction conversion coefficient to obtain the threshold reduction amount, reduce the review judgment threshold of the set of related data items, and the reduced judgment threshold is not lower than the preset minimum threshold lower limit. Based on the reduced threshold, the set of related data items is re-examined and reversed, and the evidence chain and abnormal event records are updated. Summarize abnormal event records within the same review period, aggregate them according to preset dimensions, and output a consistency review report.

2. The carbon credit project data consistency review method according to claim 1, characterized in that, The preset verification rules are a set of automated data verification rules preset based on applicable project methodologies, monitoring plans and data quality requirements. The multi-source raw data includes data output from on-site metering equipment, production control systems and laboratory testing systems. The preset dimensions include abnormal subjects, sets of related data items, missing links in the evidence chain and time-series anomaly markers.

3. The carbon credit project data consistency review method according to claim 1, characterized in that, The sensitivity coefficient represents the degree of influence of a unit change in the material conservation equation on the conservation deviation of the abnormal entity, specifically the absolute value of the partial derivative of the material conservation equation with respect to the abnormal entity. The index residual is the deviation of the rate of change ratio from the nearest boundary value of the constraint interval: when the rate of change ratio is higher than the upper limit of the constraint interval, the deviation is the measured ratio minus the upper limit value; when it is lower than the lower limit, the deviation is the lower limit value minus the measured ratio. The preset threshold reduction conversion coefficient is a positive value, in percentage points, and is determined for each downstream verification rule based on the difference between the original judgment threshold of the rule and the preset minimum threshold lower limit. When it is necessary to uniformly lower the threshold for all downstream verification rules of the associated data item set, the minimum value among the allowed threshold reduction amounts of each rule is taken as the uniform reduction range.

4. The carbon credit project data consistency review method according to claim 1, characterized in that, The material conservation equation based on the project's production process determines a set of associated data items that have a physical dependency on the anomaly, including: Material input, output and loss data items for each production stage are extracted in advance from project design documents and original ledgers. Material conservation equations for each production stage are established and stored as an equation library. When the abnormal subject is an intermediate calculation result, the source variable of the intermediate calculation result is parsed, and the source variable is mapped to the input, output or loss in the material conservation equation; When the abnormal subject is a raw data item, the material conservation equation corresponding to its production process is directly located. Data items belonging to the same set of variables in the material conservation equation are identified as having physical dependencies, thus forming the set of associated data items.

5. The carbon credit project data consistency review method according to claim 1, characterized in that, Targeted review of the aforementioned set of related data items includes: Starting from the abnormal subject, the downstream verification rules are identified in a forward direction along the emission reduction calculation path corresponding to the set of associated data items, and the abnormal intensity value of the abnormal subject is calculated. The abnormal intensity value is the absolute value of the relative deviation between the measured value and the preset standard value when the preset verification rule is triggered. The abnormal intensity value is multiplied by the methodological measurement error attenuation coefficient of the data item corresponding to each downstream verification rule to obtain the transmission intensity value. When a downstream verification rule corresponds to multiple data items, the transmission intensity value of each data item is calculated separately, and the maximum transmission intensity value is taken as the transmission intensity value of the downstream verification rule. The methodological measurement error attenuation coefficient is determined as follows: when the project methodology specifies a maximum permissible measurement error rate, the attenuation coefficient is 1 minus the error rate; when not specified, it is determined according to the maximum permissible error rate in the metrological equipment calibration certificate. In the downstream verification rules, the rule whose conduction strength value is greater than or equal to the preset anomaly judgment threshold is determined as the rule that actually triggers the review, and the targeted review is only performed on the rule that actually triggers the review.

6. The carbon credit project data consistency review method according to claim 1, characterized in that, The process of tracing back along the data acquisition and recording chain to generate an evidence chain includes: Collect the collection time, device number, operator identification, and original voucher number for each data item along the traceability path; Mark the missing link in the data collection and recording chain of the evidence chain; The original suspicious point weight is determined based on the interval distance between the missing link and the abnormal subject on the data acquisition and recording chain, and the original suspicious point weight is negatively correlated with the interval distance; When the data item involved in the missing link belongs to the set of related data items and is used as an input, output or loss in the material conservation equation, the original suspicious point weight is multiplied by the enhancement coefficient to obtain the corrected suspicious point weight. Otherwise, the corrected suspicious point weight is equal to the original suspicious point weight. The enhancement coefficient is equal to the ratio of the sensitivity coefficient of the data item to the preset benchmark sensitivity coefficient. The preset benchmark sensitivity coefficient is a benchmark value greater than zero, which is the larger of the arithmetic mean, median or preset minimum benchmark value of all non-zero sensitivity coefficients of the item. The anomaly level of the abnormal event record is determined based on the anomaly intensity value, the conduction intensity value, and the corrected suspicious point weight.

7. The carbon credit project data consistency review method according to claim 1, characterized in that, The rate of change is the year-on-year change rate between adjacent complete reporting periods; the constraint interval is primarily determined based on the allowable deviation of the project methodology, and if the project methodology does not specify it, it is determined based on the material conservation equation and the statistical characteristics of historical data without anomalies; when the anomaly is an intermediate calculation result, its historical value sequence is recalculated based on the source variable of the corresponding reporting period. The time points for process changes are determined based on project design documents, monitoring plans, or confirmed process change records. The segmentation of historical value sequences based on process change time points includes: The first complete reporting period after project filing without process changes and within the initial temporary constraint range is defined as the normal data window for the initial segment; When a process change event is detected as recorded in the project design documents, monitoring plan, or a confirmed process change record, a breakpoint is set at the corresponding change time point. The first complete reporting period after the change that does not trigger a timing anomaly will be defined as the first anomaly-free reporting period after the change. Based on the mean of the ratio of the change rate during the first abnormal reporting period after the change, and the maximum permissible deviation specified by the project methodology, the constraint intervals for subsequent segments are redefined; When a project is in the initial verification stage and the historical numerical series is insufficient to construct a statistical benchmark, the initial temporary constraint interval is formed by combining the benchmark parameters in the project design documents with the allowable deviation of the project methodology. When a process change occurs, if there is no abnormality reporting period that meets the conditions after the change, the most recent effective segmented constraint interval before the change is used as a temporary reference for the transition period. If there is no segmented constraint interval before the change, the baseline parameters in the project design documents are combined with the allowable deviation of the project methodology to form a temporary constraint interval for the transition period. If the current complete reporting period after the change does not exceed the temporary constraint interval for the transition period, it is determined as the first abnormality reporting period after the change, and the adaptive segmented constraint interval based on the average value of that reporting period is switched.

8. The carbon credit project data consistency review method according to claim 7, characterized in that, When the abnormal subject is an emission factor, the associated production indicator is the property detection indicator of the corresponding material. The linkage between the emission factor and the material property detection indicator is determined based on the carbon conservation law. The change rate ratio is the ratio of the change rate of the emission factor to the change rate of the material property detection indicator. The constraint interval is determined based on the statistical value of the historical detection ratio of the same material category, the same detection method and the same process segment in the same project or the allowable deviation specified by the project methodology. When the absolute value of the change rate of the material property detection indicator is less than the preset small change threshold, the change rate ratio is not calculated, and the absolute value of the change rate of the emission factor is compared with the preset stability threshold for judgment.

9. The carbon credit project data consistency review method according to claim 7, characterized in that, When the abnormal subject is active data, the associated production indicator is output, and the change rate ratio is the ratio of the change rate of active data to the change rate of output. The constraint interval satisfies the following: the absolute deviation of the ratio from the ratio of active data to the theoretical ratio of output determined by the material conservation equation does not exceed a preset coordination threshold. When the absolute value of the change rate of output is less than a preset small change threshold, the change rate ratio is not calculated, and the absolute value of the change rate of active data is compared with a preset stability threshold for determination.

10. A carbon credit project data consistency review system, characterized in that, This includes data access interfaces, equation libraries, rule bases, and report output interfaces, as well as: The initial review module is used to obtain the current multi-source raw data and perform preliminary consistency verification. When a preset verification rule is triggered, an abnormal event record is generated to determine the original data item or intermediate calculation result that triggered the rule as the abnormal subject. The association and location module is used to determine the set of associated data items that have a physical dependency relationship with the abnormal entity based on the material conservation equation of the project production process; The targeted review module is used to identify downstream verification rules in a forward manner starting from the abnormal subject, calculate the abnormal intensity value and the transmission intensity value, filter the rules that actually trigger the review, and execute the targeted review task. The evidence chain tracing module is used to generate an evidence chain by tracing back along the data acquisition record chain. It determines the original suspicious point weight based on the interval distance of the missing links in the data acquisition record chain. When the missing link involves related data items and belongs to the input, output or loss of the material conservation equation, the suspicious point weight is corrected with an enhancement coefficient, and the abnormality level is determined by combining the abnormality intensity value and the transmission intensity value. The timing feedback module is used to extract the historical numerical sequence of the abnormal subject and the associated data item set, segment it according to the process change time point, mark the timing abnormality when the change rate ratio in the segment exceeds the constraint interval, and calculate the index residual. The index residual is multiplied by the sensitivity coefficient of the abnormal subject in the material conservation equation and the preset threshold reduction conversion coefficient to obtain the threshold reduction amount. The review judgment threshold of the associated data item set is reduced, and the reduced judgment threshold is not lower than the preset minimum threshold lower limit. The targeted review module and the evidence chain tracing module are triggered to re-perform targeted review and reverse tracing based on the reduced threshold. The aggregation report module is used to summarize abnormal event records within the same review period, and output a consistency review report after being aggregated and associated according to preset dimensions.

Images