Ecological environment multi-source heterogeneous data intelligent auditing method and system based on agent technology
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-05-21
- Publication Date
- 2026-08-11
AI Technical Summary
[0002]随着物联网、大数据及人工智能技术的融合，生态环境监测进入以数据驱动为核心的智能化时代，环境数据是进行污染溯源、生态评估和制定科学治理策略的依据，对于保障可持续发展具有重要的意义；在此背景下，如来自固定监测站、移动无人机、遥感卫星等多源异构数据，使得难以高效、准确地对这些数据进行审核与校验
1、本发明通过校验模型对次数据执行初步校验，针对未通过初步校验的待校验数据，基于正常数据计算得到修正校验评分，并以此为基准参考值，结合增量差值形成动态评分值集合；再将待校验数据的数据校验评分与评分值集合一同输入判断模型，输出决策结果，并建基于正常数据动态变化的动态判断基准，取代固定阈值判断方式，能够区分真实异常与数据在复杂环境下的正常波动，提高对生态环境多源异构数据审核的审核准确性与可靠性。
Smart Images
Figure CN122548563A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of environmental monitoring technology, specifically relating to an intelligent auditing method and system for multi-source heterogeneous ecological and environmental data based on agent technology. Background Technology
[0002] With the integration of IoT, big data and AI technologies, ecological and environmental monitoring has entered an intelligent era driven by data. Environmental data serves as the basis for pollution source tracing, ecological assessment and the formulation of scientific governance strategies, and is of great significance for ensuring sustainable development. In this context, heterogeneous data from multiple sources, such as fixed monitoring stations, mobile drones and remote sensing satellites, makes it difficult to efficiently and accurately review and verify these data.
[0003] Existing multi-source heterogeneous data auditing technologies, including some preliminary attempts based on agent technology, in practical applications, traditional methods use verification modes based on preset thresholds or isolated rules to independently evaluate a single data stream, lacking in-depth mining of the inherent correlation between different data sources; they cannot effectively identify complex anomalies caused by the coupling of multiple factors, such as misjudging the real data fluctuations at a water quality monitoring point caused by upstream rainfall and sewage discharge as sensor failure, thus leading to false alarms and missed alarms, affecting the accuracy of environmental event early warning; In addition, the system lacks adaptive and intelligent capabilities, making it difficult to dynamically adapt to seasonal changes in environmental background values, natural degradation of sensor performance, or the emergence of new pollution events. Its effectiveness continues to decline over time, requiring a large amount of manpower for maintenance and calibration.
[0004] In view of this, the present invention proposes an intelligent auditing method and system for multi-source heterogeneous data of the ecological environment based on agent technology. Summary of the Invention
[0005] The purpose of this invention is to provide an intelligent auditing method for multi-source heterogeneous ecological and environmental data based on agent technology. This method can standardize the processing of ecological and environmental big data from different sources and in different formats, and support mutual verification and cross-checking between data, thereby avoiding duplicate judgments and reducing the probability of misjudgment.
[0006] The technical solution adopted in this invention is as follows: An intelligent verification method for multi-source heterogeneous ecological and environmental data based on agent technology. This method is applied to multi-source heterogeneous ecological and environmental data, including air quality, noise, rainfall, water quality, and meteorological data. The method includes: when a secondary data point in the ecological and environmental data fails a preset preliminary verification, the secondary data point is identified as data to be verified, and the following steps are executed: The data to be verified is assessed for its credibility to determine whether it is reliable data. Furthermore, based on trusted data and the offset data identified during the data verification process, data correction processing is performed to construct a time-continuous error zone; The process of judging the credibility of the data to be verified to determine it as credible data includes: data verification scoring based on the data to be verified, and a set of score values formed based on multiple normal data. The data to be verified and the set of score values are input into the judgment model together to output a decision result that determines the data to be verified as credible data.
[0007] Preferably, the set of scores formed based on multiple normal data includes: a data verification score based on multiple normal data, forming a corrected verification score, and setting the corrected verification score as a benchmark reference value; The incremental difference is calculated based on the data verification scores of multiple normal data sets; and a set of score values is formed based on the benchmark reference value and the incremental difference.
[0008] Preferably, the data correction process includes: calculating a reference time period based on the acquisition time of reliable data and the fault tolerance deviation of the offset data; The process involves identifying offset time points within a reference time period and excluding these offset time points from the reference time period to determine the time segment requiring correction; extracting the data to be corrected within the time segment requiring correction; and constructing a time continuity error zone based on the error information between the data to be corrected and the reference set data.
[0009] Preferably, the method further includes performing a data verification process before performing the credibility judgment. The data verification process includes: extracting sample data from the secondary data as a reference set data; forming a verification time range based on the reference set data, and extracting data falling within the verification time range from the secondary data as data to be verified; and inputting the data to be verified into the verification model for processing, so as to determine the data to be verified as normal data or offset data.
[0010] Preferably, inputting the data to be verified into the verification model for processing to determine the data to be verified as offset data includes: comparing the data to be verified with the reference set data to output the difference; if the difference is greater than a preset tolerance standard, the data to be verified is determined as offset data; and inputting the offset data into the fault tolerance model for processing to obtain the fault tolerance deviation used to calculate the reference time period.
[0011] Preferably, the method further includes: performing adaptive update processing; If the data to be verified is marked as slightly anomalous, the adaptive update process includes extracting neighboring data to update the baseline data for subsequent verification; and if the data to be verified is marked as moderately anomalous, the adaptive update process includes updating the judgment rules of the verification model using the corresponding verification results. Among them, slightly abnormal data refers to abnormal data whose deviation value is less than or equal to the first preset threshold, and moderately abnormal data refers to abnormal data whose deviation value is greater than the first preset threshold and less than or equal to the second preset threshold.
[0012] The intelligent auditing system for multi-source heterogeneous data of ecological environment based on agent technology includes: a data verification module, which is used to monitor secondary data in ecological environment data and identify secondary data as data to be verified when the secondary data fails the preset preliminary verification; The data verification module is used to perform verification on the sub-data to classify it as normal data or offset data. The credibility judgment module is used to respond to the identification of the data to be verified by the data verification module, and to judge the credibility of the data to be verified so as to determine that it is credible data. The data correction module is used to perform data correction processing based on the credible data determined by the credibility judgment module and the offset data identified by the data verification module, so as to construct a time continuous error zone; The adaptive update module is used to perform adaptive update processing based on the anomaly level of the data to be verified. The adaptive update processing includes updating the benchmark data used for subsequent verification, or updating the judgment rules of the verification model adopted by the data verification module.
[0013] Preferably, when the data verification module inputs the data to be verified into the verification model for processing to determine that the data to be verified is offset data, it is configured to: compare the data to be verified with the reference set data to output the difference; if the difference is greater than the preset tolerance standard, then the data to be verified is determined to be offset data; input the offset data into the fault tolerance model for processing to obtain the fault tolerance deviation used to calculate the reference time period.
[0014] Preferably, when the credibility judgment module performs credibility judgment on the data to be verified to determine it as credible data, it is configured to: input the data to be verified and the set of scores formed based on multiple normal data into the judgment model to output a decision result that determines the data to be verified as credible data.
[0015] Preferably, the data correction module is configured to: calculate a reference time period based on the acquisition time of reliable data and the fault tolerance deviation of the offset data; The process involves excluding offset time points from the reference time period to determine the time segment requiring correction; extracting the data to be corrected within the time segment requiring correction; and constructing a time continuity error zone based on the error information between the data to be corrected and the reference set data.
[0016] Beneficial effects 1. This invention performs preliminary verification on the data through a verification model. For the data to be verified that fails the preliminary verification, a corrected verification score is calculated based on normal data. This score is used as a benchmark reference value and combined with incremental differences to form a dynamic score value set. The data verification score and the score value set of the data to be verified are then input into the judgment model to output the decision result. A dynamic judgment benchmark based on the dynamic changes of normal data is established, replacing the fixed threshold judgment method. This can distinguish between real anomalies and normal fluctuations of data in complex environments, and improve the accuracy and reliability of the review of multi-source heterogeneous data of the ecological environment.
[0017] 2. After completing the credibility judgment, the present invention calculates the reference time period based on the acquisition time of the credible data and the fault tolerance deviation of the offset data, and locates the time segment to be corrected by identifying and excluding the offset time points in the time period; then extracts the data to be corrected in the segment and calculates the error information between it and the reference set data; by constructing a time continuous error zone, the abnormal data is located and quantitatively analyzed, avoiding the general processing or direct discarding of large-scale data.
[0018] 3. This invention establishes a closed-loop adaptive optimization mechanism to mark the data to be verified as slightly anomalous or moderately anomalous: for slightly anomalous data, the system extracts its adjacent data as a new benchmark set to dynamically adjust the reference benchmark for subsequent verification; for moderately anomalous data, the verification results are used to update the judgment rules of the verification model itself; through the system's self-learning and iterative optimization capabilities, it can automatically adapt to long-term changes in data characteristics or sensor drift. Attached Figure Description
[0019] Figure 1 This is a flowchart of the method for extracting data to be corrected provided by the present invention; Figure 2 This is a flowchart of the method for outputting the final review and correction results provided by the present invention; Figure 3 This is a system module diagram provided by the present invention. Detailed Implementation
[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0021] Example 1 like Figure 1-2As shown, this embodiment provides an intelligent verification method for multi-source heterogeneous ecological and environmental data based on distributed processing technology. By constructing a data processing and verification process, it achieves quality control of multi-source heterogeneous ecological and environmental data, identifies and classifies potentially failed verification data or offset data, and improves the credibility and application accuracy of multi-source heterogeneous data. Specifically, it includes the following steps: The data acquisition module obtains source data and secondary data from multiple data sources, including meteorological data such as air quality, noise, and rainfall. Source data is usually obtained by direct sampling from authoritative or standard measuring equipment, while secondary data may come from edge devices, mobile devices, or third-party platforms. The acquired data is uniformly formatted and transformed according to a set of preset data structure mapping rules and parsing libraries to make the heterogeneous data comparable and processable in terms of data structure, naming conventions, and timestamp consistency.
[0022] Furthermore, the data is submitted to the data quality assessment and processing unit for processing according to a preset set of verification rules. Based on statistical analysis of a large amount of historical data, the data covers multiple dimensions of validity characteristics, including data format, normal range of values, reasonable slope of trend, and completeness of data items. By quantitatively assessing the degree of conformity between the data points and the set of verification rules, a data verification score using a preset unified scoring unit is output to ensure that data with different physical dimensions can be uniformly evaluated and compared in subsequent processes. For example, all scores are normalized to dimensionless scores from 0 to 100, where 100 represents the highest data quality and 0 represents completely invalid data. If the data fully conforms to the verification rule set, it is determined to be verified data, and the data verification score corresponding to the verified data is output. If the data fails to fully comply with the verification rule set, such as missing data, excessive deviation, or inconsistent units, the data will be identified as data to be verified and submitted to the next step for error tolerance processing.
[0023] Furthermore, all data to be verified is submitted to a fault-tolerant processing mechanism for correction attempts, including: constructing trend lines based on data from adjacent time periods for fitting, filling missing values using smooth interpolation methods, or retrieving neighboring data based on historical similar conditions for substitution, and correcting offset values or minor errors. After fault-tolerant processing, the difference between the data verification score of the data to be verified and the benchmark reference value generated in subsequent steps is calculated to obtain the deviation, enabling refined management of anomalies of different degrees. Based on preset processing standards, the data to be verified is marked as slightly anomalous or moderately anomalous according to the multiple relationship between the deviation value and the standard deviation of historical data. If the data to be verified is marked as slightly anomalous, it indicates that there is a small instantaneous fluctuation in the data. Then, based on the collection time of the slightly anomalous data, a data sampling period is set, for example, 5 minutes before and after the time point. Adjacent data is extracted within the data sampling period and determined as a new benchmark data set for subsequent verification. Based on the strategy of dynamically adjusting the benchmark, normal short-term fluctuations of environmental factors are dealt with to avoid misjudging non-critical data jitter as serious errors.
[0024] If the data to be verified is marked as moderately anomalous, it indicates that the sensor has drifted or that the local environment has undergone a continuous change. The verification results corresponding to the moderately anomalous data, such as the type, magnitude, and duration of the anomaly, are used to update the verification rule set within the data quality assessment processing unit. Based on the feedback mechanism, this processing unit can learn new anomaly patterns and continuously improve its accuracy in identifying similar anomalies in the future, thereby achieving adaptive optimization of the verification standard.
[0025] Further, sample data is extracted from the secondary data set and designated as the baseline data set. Based on the collection time, a maximum query time is set on the time series to form a verification time range. Then, data falling within this verification time range is extracted from the secondary data set and designated as the data to be verified. This data is then submitted to the data quality assessment and processing unit for processing, and the verification results are output. Finally, the data to be verified is compared item by item with the baseline data set, and the difference is output. Based on the relationship between this difference and the preset tolerance standard, the following classification results are output: If the verification result is unsuccessful, the data to be verified will be judged as verification failed data; if the verification result is successful and the difference is greater than the preset tolerance standard, the system will judge it as offset data; if the verification result is successful and the data to be verified is within the preset allowable range, it will be judged as normal data; if the difference is less than or equal to the tolerance standard and the data to be verified is higher than the data verification score of the benchmark data, it indicates that the newly acquired data is of higher quality while meeting the consistency requirements. At this point, output the optimized results of data accuracy and data validation score, such as suggesting that this high-quality data point be updated as the new benchmark to improve the benchmark level of the entire dataset; The tolerance standard is a threshold jointly set based on the reasonable numerical range corresponding to the data type of the monitoring indicator, such as concentration, decibel value, and water level, and the statistical fluctuation range of the indicator in historical data of the same period, such as the standard deviation. Specifically, the process of setting the tolerance standard includes: obtaining historical data of the monitoring indicator under normal environmental conditions and calculating its mean and standard deviation; setting an absolute lower limit or upper limit based on the physical meaning of the indicator, such as PM2.5 concentration not being negative; combining the absolute limit with the statistical threshold, such as mean ± 3 times the standard deviation, and taking the more stringent one as the tolerance standard to define the preset allowable range.
[0026] Further, in the data credibility judgment stage, the multiple data verification scores output by the data quality assessment and processing unit are comprehensively calculated. For example, the multiple data verification scores output by the data quality assessment and processing unit are weighted and summed. The weight of each score is pre-set according to the historical credibility or data timeliness of its corresponding data source to obtain the total score. Normal data is submitted to the data quality assessment and processing unit again to obtain the deviation value corresponding to the normal data, so as to establish a dynamic evaluation benchmark that can reflect the recent data quality trend. Based on the deviation value, the data validation score of the normal data is recalculated. For example, based on the deviation value, the data validation score of the normal data is recalculated using the following formula: Corrected score = Original score × (1 - Deviation influence factor), where the deviation influence factor is positively correlated with the magnitude of the deviation value and its value range is [0, 0.5]. The recalculated data validation score is combined with the total score to form a corrected validation score, and the corrected validation score is set as the benchmark reference value. Based on the data validation scores of multiple normal data, the incremental difference is calculated to construct a dynamic threshold for decision-making. The incremental difference is obtained by calculating the numerical difference between the data verification scores of two adjacent normal data in the time series, reflecting the short-term rate of change of data quality. Based on the incremental difference, a set of score values is formed. Specifically, the score values in the set of score values are obtained by superimposing the benchmark reference value and the incremental difference, thereby forming a reliable interval centered on the benchmark reference value and dynamically adjusted according to the normal fluctuation range of recent data quality. Before submitting the data verification score of the data to be verified to the decision logic unit, multiple data to be verified are organized into a data sequence based on the collection time of the data to be verified; the data sequence is arranged in order from the oldest to the most recent collection time; then the data verification score of the data to be verified and the set of score values are submitted to the decision logic unit together. The core judgment logic of the decision logic unit is: to determine whether the data verification score of the data to be verified falls within the confidence interval defined by the score value set, and output the decision result accordingly; if the decision result is to accept, the data to be verified is determined to be reliable data; if the decision result is to reject, it is determined to be unreliable data and recorded in the database for later use.
[0027] Furthermore, based on the acquisition time of reliable data and the fault tolerance deviation of the offset data, a baseline time period is calculated; the offset data and slightly abnormal data identified in the aforementioned steps are recorded as correction data to ensure the accuracy of the correction process. When determining the time interval to be corrected, the correction data is excluded to prevent known outliers from interfering with the correction benchmark; the offset time points within the benchmark time period are identified and excluded from the benchmark time period to determine the time interval to be corrected, and the data to be corrected is extracted; then the error information between the data to be corrected and the benchmark data is calculated to form multiple data nodes; then a time continuous error zone is constructed, which represents the overall offset trend and range of the data within a specific time period in a visual or data-driven form; Based on the acquisition time of the calibration data and the data to be calibrated, the time offset difference is calculated to further define the boundary of the calibration operation, and the time offset difference is compared with the preset offset standard. If the time offset difference meets the offset standard, such as the continuous offset time exceeding the set minimum response period, the correction time result is output; and the end point of the correction time result is marked. This end point will be used as the basis for the time period of subsequent data correction to ensure that the scope of the correction operation is clear and controllable.
[0028] Furthermore, the system will output the final identified reliable data, unreliable data, corrected data, and time-continuous error zones in a unified manner for use by upper-layer application systems; and provide a visual interface to display information such as data fluctuation trends, reliability changes, and abnormal band locations to facilitate user understanding and decision-making. In practical implementation, the following auxiliary logic can be configured: historical trend comparison function, which compares the trend of historical data for the same period based on seasonal, time or climate characteristics to improve the accuracy of judgment; Priority scheduling logic sets data processing priorities for key areas and important monitoring indicators to improve the response rate of critical data. The anomaly attribution module performs external causal retrieval for large-scale data anomalies, such as associating them with abnormal weather or temporary construction. In addition, a manual verification channel will be introduced when necessary to improve the accuracy of judgments at key points and the robustness of the system. This embodiment constructs a complete closed loop from data collection, unified processing, and hierarchical verification to data verification scoring, data correction processing, and output of the time continuous error zone. The entire system emphasizes unified data standards, fault-tolerant processing, dynamic backtracking, and proactive intervention. It is suitable for the complex multi-source data structures commonly found in ecological and environmental monitoring scenarios, and solves the defects of traditional methods such as easy omission, misjudgment, or delayed response. Practical verification has improved the stability of the environmental multi-source data acquisition system and the reliability of the decision-making system, providing solid data support and methodological guarantee for intelligent and automated ecological and environmental quality assessment.
[0029] Example 2 like Figure 3 As shown, this embodiment provides an intelligent auditing system for multi-source heterogeneous ecological and environmental data. It audits and corrects multi-source heterogeneous ecological and environmental data, including air quality, noise, water quality, and meteorological data, identifies and processes anomalies and offsets in the data stream, and constructs a time-continuous error zone, thereby improving the overall quality of ecological and environmental monitoring data.
[0030] In its implementation, this system can be deployed on servers, cloud computing platforms, or dedicated data processing centers. It interacts with various environmental monitoring sensors deployed in different geographical locations via network interfaces, such as PM2.5 monitors, noise sensors, rain gauges, water quality analyzers, and weather stations. The system specifically includes the following modules: The data verification module performs rapid preliminary verification to continuously monitor secondary data in the ecological and environmental data received from various data sources. For each piece of secondary data, a preset set of verification rules is applied, including verification of the physical range of data values, data format verification, timestamp continuity verification, or data transmission integrity verification. When a data point fails to pass any of the aforementioned preset preliminary checks, the data is identified and marked as data to be checked, and then submitted to the subsequent credibility judgment module and data verification module for further analysis.
[0031] The data verification module performs in-depth verification of the data stream, including: extracting a representative sample of data from the received sub-data stream, especially the data that has passed the initial verification; and constructing a baseline set of data that represents the "normal" state of the data under the current environment. Then, based on the timestamp range of the benchmark data, a verification time zone is formed, and all data falling within this verification time zone are extracted from the original sub-data as data to be verified. Each piece of data to be verified is then input into the internal verification model for processing. The difference is output by comparing the data to be verified with the benchmark data. If the difference is greater than the preset tolerance standard, the data to be verified is determined to be offset data; otherwise, it is determined to be normal data. For cases where the data is determined to be offset, the offset data is further input into the fault tolerance model for processing to obtain the fault tolerance deviation that quantifies the degree of offset. This fault tolerance deviation will be provided to the data correction module for use.
[0032] The credibility judgment module processes the data to be verified identified by the data verification module and determines whether it has value for correction rather than discarding it directly. When the data to be verified is received, a data verification score is calculated for the data to be verified. At the same time, multiple data that have been judged as normal data are obtained from the data verification module, and a set of score values is formed based on the data verification scores of these normal data through a specific algorithm. The process of forming this set of scores may include: calculating the corrected values of multiple normal data verification scores as benchmark reference values, and calculating the incremental difference between each normal data score and the benchmark reference value; then constructing a set of scores representing the distribution of normal data scores based on the benchmark reference value and the incremental difference; then inputting the data verification score of the data to be verified and the set of scores together into the judgment model; the judgment model outputs a decision result, and if the decision result is acceptable, the data to be verified is finally determined to be reliable data, indicating that although the data has preliminary anomalies, its intrinsic information still has reference value and can be used for subsequent correction processes.
[0033] The data correction module is the final execution unit of the entire review process, responsible for generating the final correction result; it receives credible data determined by the credibility judgment module and offset data identified by the data verification module; its data correction process calculates the baseline time period based on the collection time of credible data and the error tolerance deviation of offset data. Within this baseline period, all known offset time points are identified and excluded, such as timestamps corresponding to offset data, to determine the time intervals that need correction. All data to be corrected is then extracted within these time intervals. Next, by combining the error information between the baseline data provided by the data verification module, a time-continuous error zone is constructed through interpolation, fitting, or other time-series data processing techniques. This error zone corrects abnormal data points while also filling in data gaps.
[0034] The adaptive update module performs adaptive update processing based on the anomaly level of the data to be verified. Specifically, if the data to be verified is marked as slightly abnormal by the system's internal rules, such as only slightly exceeding the normal range, the module will trigger the baseline update mechanism. Extract the normal data adjacent to the data point to update the baseline data used in subsequent verification processes, so that the system can adapt to the slow drift of data. If the data to be verified is marked as moderately anomalous, such as a new, unseen anomalous pattern, the module will use the verification result to update the judgment rules of the verification model used by the data verification module, thereby adjusting the model parameters, adding new classification rules, or using the sample to incrementally train the model to improve the system's future accuracy in identifying similar anomalies.
[0035] Through the collaborative work of the above modules, this embodiment can identify and correct various anomalies and deviations in ecological and environmental monitoring data, and continuously optimize its own auditing capabilities through continuous learning, ensuring stable and high-precision data support for application scenarios such as smart environmental protection, environmental early warning, and scientific research analysis.
Claims
1. A method for intelligent auditing of multi-source heterogeneous data on the ecological environment based on agent technology, characterized in that: The method is applied to multi-source heterogeneous ecological and environmental data, including air quality, noise, rainfall, water quality and meteorological data. It includes: when a sub-data in the ecological and environmental data fails the preset preliminary verification, the sub-data is identified as data to be verified, and the following is performed: the credibility of the data to be verified is judged to determine it as credible data. Furthermore, based on trusted data and the offset data identified during the data verification process, data correction processing is performed to construct a time-continuous error zone; The process of judging the credibility of the data to be verified to determine it as credible data includes: data verification scoring based on the data to be verified, and a set of score values formed based on multiple normal data. The data to be verified and the set of score values are input into the judgment model together to output a decision result that determines the data to be verified as credible data.
2. The intelligent auditing method for multi-source heterogeneous ecological environment data based on agent technology according to claim 1, characterized in that, The set of scores formed based on multiple normal data includes: a data verification score based on multiple normal data, a corrected verification score, and a corrected verification score set as a benchmark reference value; The incremental difference is calculated based on the data verification scores of multiple normal data sets; and a set of score values is formed based on the benchmark reference value and the incremental difference.
3. The intelligent auditing method for multi-source heterogeneous ecological environment data based on agent technology according to claim 1, characterized in that, Data correction processing includes: calculating a baseline time period based on the acquisition time of reliable data and the fault tolerance deviation of the offset data; The process involves identifying offset time points within a reference time period and excluding these offset time points from the reference time period to determine the time segment requiring correction; extracting the data to be corrected within the time segment requiring correction; and constructing a time continuity error zone based on the error information between the data to be corrected and the reference set data.
4. The intelligent auditing method for multi-source heterogeneous ecological environment data based on agent technology according to claim 3, characterized in that, The method also includes performing a data verification process before performing the credibility judgment. The data verification process includes: extracting sample data from the secondary data as a baseline data set. Based on the benchmark data, a verification time zone is formed, and data falling within the verification time zone is extracted from this data as data to be verified; and the data to be verified is input into the verification model for processing, so as to determine whether the data to be verified is normal data or offset data.
5. The intelligent auditing method for multi-source heterogeneous ecological environment data based on agent technology according to claim 4, characterized in that, The process of inputting the data to be verified into the verification model for processing to determine the data to be verified as offset data includes: comparing the data to be verified with the reference set data to output the difference; If the difference is greater than the preset tolerance standard, the data to be verified will be judged as offset data. The offset data is input into the fault-tolerant model for processing to obtain the fault-tolerant deviation used to calculate the baseline period.
6. The intelligent auditing method for multi-source heterogeneous ecological environment data based on agent technology according to claim 1, characterized in that, The method also includes: performing adaptive update processing; If the data to be verified is marked as slightly anomalous, the adaptive update process includes extracting neighboring data to update the baseline data for subsequent verification; and if the data to be verified is marked as moderately anomalous, the adaptive update process includes updating the judgment rules of the verification model using the corresponding verification results. Among them, slightly abnormal data refers to abnormal data whose deviation value is less than or equal to the first preset threshold, and moderately abnormal data refers to abnormal data whose deviation value is greater than the first preset threshold and less than or equal to the second preset threshold.
7. An intelligent auditing system for multi-source heterogeneous data of the ecological environment based on agent technology, characterized in that: include: The data verification module is used to monitor secondary data in ecological and environmental data, and to identify secondary data as data to be verified when it fails the preset preliminary verification. The data verification module is used to perform verification on the sub-data to classify it as normal data or offset data. The credibility judgment module is used to respond to the identification of the data to be verified by the data verification module, and to judge the credibility of the data to be verified so as to determine that it is credible data. The data correction module is used to perform data correction processing based on the credible data determined by the credibility judgment module and the offset data identified by the data verification module, so as to construct a time continuous error zone; The adaptive update module is used to perform adaptive update processing based on the anomaly level of the data to be verified. The adaptive update processing includes updating the benchmark data used for subsequent verification, or updating the judgment rules of the verification model adopted by the data verification module. 8.The agent technology-based ecological environment multi-source heterogeneous data intelligent auditing system according to claim 7, characterized in that, When the data verification module inputs the data to be verified into the verification model for processing, and determines that the data to be verified is offset data, it is configured as follows: Compare the data to be verified with the benchmark set data to output the difference; If the difference is greater than the preset tolerance standard, the data to be verified will be judged as offset data. The offset data is input into the fault-tolerant model for processing to obtain the fault-tolerant deviation used to calculate the baseline period. 9.The agent technology-based ecological environment multi-source heterogeneous data intelligent auditing system according to claim 7, characterized in that, When the credibility judgment module judges the credibility of the data to be verified and determines it as credible data, it is configured to: input the data to be verified and the set of scores formed by multiple normal data into the judgment model together, and output the decision result that determines the data to be verified as credible data. 10.The agent technology based ecological environment multi-source heterogeneous data intelligent auditing system according to claim 7, characterized in that, The data correction module is configured to calculate a baseline time period based on the acquisition time of reliable data and the fault tolerance deviation of the offset data; Exclude offset time points from the baseline time period to determine the time segment that needs correction; Furthermore, it extracts the data to be corrected within the time interval to be corrected, and constructs a time-continuous error region based on the error information between the data to be corrected and the reference set data.