Machine learning diagnosis method for rural carbon source abnormity

Abstract

The invention relates to the technical field of rural carbon source abnormity diagnosis, and discloses a machine learning diagnosis method for rural carbon source abnormity. The method comprises the following steps: step 1, setting inspection points, acquiring management data of all carbon sources of a country, actual measurement data of all the inspection points and management data of fluctuation factors, and classifying to form a data set; 2, the distribution rationality of rural carbon source inspection points is evaluated, a coverage index is generated, different rural carbon source structures are adapted, and the comprehensive adaptation precision is high; 3, evaluating the linear correlation degree of the greenhouse gas total CO2 equivalent of all inspection points and the linear distance between the inspection points, generating a correlation coefficient, and accurately judging the risk of carbon source leakage; 4, a fixed threshold value is set, corresponding management measures are triggered, the problems of insufficient routing inspection coverage, carbon source leakage and the like are rapidly solved, and the prevention and control timeliness is improved; and 5, monitoring and evaluating the abnormal degree of the rural carbon source, generating an abnormal score, and triggering management measures, so that the abnormal response is more timely.

Description

Technical Field

[0001] This invention relates to the field of rural carbon source anomaly diagnosis technology, specifically a machine learning diagnosis method for rural carbon source anomalies. Background Technology

[0002] Rural carbon source anomalies refer to the phenomenon in rural ecosystems where the emission intensity, structure, or spatiotemporal distribution of carbon sources deviates from the regional natural level and the expected benchmark for socio-economic development. Rural carbon sources themselves have complex and diverse characteristics, exhibiting a mixed distribution of point sources (such as enclosed spaces like livestock sheds and biogas digesters), non-point sources (such as open spaces like farmland and rice paddies), and line sources (such as rural roads). Furthermore, emission intensity shows significant spatiotemporal heterogeneity, with distinct differences in carbon emissions across different regions and time periods.

[0003] Currently, in terms of spatiotemporal coverage, traditional methods mostly employ a discrete inspection point monitoring model, setting only a fixed number of inspection points with an extremely low detection frequency, typically only once a month. This makes it difficult to comprehensively cover the complex distribution range of rural carbon sources, resulting in a serious lack of spatiotemporal representativeness of the monitoring data and an inability to reflect the overall trend of carbon source emissions. Regarding data accuracy, traditional methods do not fully consider various interference factors and are easily affected by key fluctuations such as air pressure fluctuations in enclosed spaces, average wind speeds in open spaces, inspection point positioning errors, and the frequency of temporary carbon source emissions. Furthermore, they do not employ differentiated accounting logic for the different emission characteristics of enclosed and open spaces, leading to significant errors in the monitoring data and making it difficult to accurately capture dynamic changes in greenhouse gas concentrations, such as N2O after farmland fertilization. Key emission points such as short-term peaks and nighttime CH4 emission peaks in livestock sheds are often overlooked. Currently, in terms of anomaly identification, traditional methods lack a scientific quantitative assessment system and a timely response mechanism, relying heavily on subjective judgment. This makes it difficult to quickly and accurately identify carbon source anomalies, resulting in delayed anomaly identification and missed opportunities for optimal prevention and control. Consequently, traditional diagnostic methods are ill-suited to the complex characteristics of rural carbon sources and cannot provide accurate and reliable decision-making basis for carbon source management. This seriously hinders the achievement of carbon emission reduction targets and fails to meet the practical needs of rural ecological environmental protection and sustainable development. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a machine learning diagnostic method for rural carbon source anomalies, which has the advantages of comprehensive adaptability, high accuracy, and more timely anomaly response. It solves the problems of traditional rural carbon source anomaly diagnostic methods being unable to adapt to carbon source mixing characteristics and having poor anomaly response effects.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a machine learning diagnostic method for rural carbon source anomalies, comprising the following steps: Step 1: Set up carbon source protection systems in rural areas. Each inspection point is connected to a database and gas detection equipment to obtain management data of all carbon sources in the village, measured data of all inspection points, and management data of fluctuation factors. These data are then classified and compiled into carbon source datasets, inspection datasets, and fluctuation datasets. Step 2: Based on the carbon source dataset, assess the rationality of the distribution of rural carbon source inspection points and generate the corresponding coverage index. ; Step 3: Based on the inspection dataset, assess the linear correlation between the total CO2 equivalent of greenhouse gases at all inspection points and the straight-line distance between inspection points, and generate the corresponding correlation coefficients. ; Step 4: Set a coverage threshold with a fixed value. Combined with the coverage index and correlation coefficient This triggers corresponding management measures; Step 5: Set a fixed monitoring period Combined with the coverage index Correlation coefficient Using a fluctuating dataset, assess the degree of anomaly in rural carbon sources and generate corresponding anomaly scores. This will trigger corresponding management measures.

[0006] Preferably, in step one, the carbon source dataset includes the functional type of each carbon source in the village, the number of inspection points, the effective coverage area of ​​a single inspection point, the total agricultural production area, the number of devices covered by a single inspection point, the number of energy consumption devices, the number of links covered by a single inspection point, the number of waste treatment links, the number of sewage treatment links, the number of functions covered by a single inspection point, and the number of building land functions. The functional types include agricultural production, energy consumption, waste treatment, sewage treatment, and building land.

[0007] Preferably, in step one, the inspection dataset includes the spatial type, CO2 mass concentration, CH4 mass concentration, N2O mass concentration, fluorine-containing gas mass concentration, spatial volume, effective sampling volume, and straight-line distance for each inspection point. The spatial type includes enclosed spaces and open spaces.

[0008] Preferably, in step one, the fluctuation dataset includes the average air pressure fluctuation in enclosed spaces, the average wind speed in open spaces, the average location error of inspection points, and the number of temporary carbon source emissions.

[0009] Preferably, in step two, the coverage index The calculation process is as follows: S11. Based on the carbon source dataset, extract the management data of the current timestamp of all carbon sources; S12. For agricultural production carbon sources, the number of inspection points for agricultural production carbon sources is recorded as follows: The effective coverage area of ​​a single inspection point is denoted as The total area of ​​agricultural production is recorded as Then calculate the area coverage of agricultural production carbon sources. ; S13. For energy-consuming carbon sources, the number of inspection points for these sources is recorded as follows: The number of devices covered by a single inspection point is recorded as The number of energy consumption devices is recorded as Then calculate the equipment coverage of energy consumption carbon sources. ; S14. For waste treatment-type carbon sources, the number of inspection points for these carbon sources is recorded as follows: The number of links covered by a single inspection point is recorded as The number of waste disposal steps is recorded as Then calculate the stage coverage of waste treatment carbon sources. ; S15. For wastewater treatment-type carbon sources, the number of inspection points for wastewater treatment-type carbon sources is recorded as follows: The number of links covered by a single inspection point is recorded as The number of wastewater treatment stages is recorded as Then calculate the coverage rate of wastewater treatment carbon source sites. ; S16. For carbon sources located on building sites, the number of inspection points for these carbon sources is recorded as follows: The number of functions covered by a single inspection point is denoted as The number of building land functions is recorded as Then calculate the functional coverage of carbon source sites on building land. ; S17. Based on S11-S16, calculate the coverage index of rural carbon source inspection points using a weighted method. .

[0010] Preferably, in step three, the correlation coefficient The calculation process is as follows: S21. Based on the inspection dataset, extract the first... The measured data of the current timestamp of each inspection point. and will the The CO2 mass concentration at the current timestamp of each inspection point is recorded as follows: , will the The CH4 mass concentration at the current timestamp of each inspection point is recorded as follows: , will the The N2O mass concentration at the current timestamp of each inspection point is recorded as follows: , will the The fluorine gas concentration at the current timestamp of each inspection point is recorded as follows: ; S22. Calculate the first according to the space type. Total CO2 equivalent of greenhouse gases at the current timestamp of each inspection point ; S23. Based on S21-S22, calculate the total CO2 equivalent of greenhouse gases at the current timestamp of each inspection point. , recorded as Then calculate the average total CO2 equivalent of greenhouse gases at the current timestamp of all inspection points. ; S24. Based on the inspection dataset, the straight-line distance between each inspection point is denoted as... , This represents the number of pairwise combinations of all inspection points, and then the average straight-line distance between all pairwise combinations of inspection points is calculated. ; S25. Based on S21-S24, calculate the correlation coefficient between the total CO2 equivalent of greenhouse gases at all inspection points and the straight-line distance between inspection points. .

[0011] Preferably, in step four, the coverage index ≤Coverage threshold When this occurs, it indicates that the distribution of rural carbon source inspection points does not meet the standards. Triggering measures include increasing the number of inspection points, expanding the coverage of individual inspection points, increasing the frequency of inspections, and extending the inspection period.

[0012] Preferably, in step four, the correlation coefficient When the value is less than 0, it indicates that the total CO2 equivalent of greenhouse gases at all inspection points is negatively correlated with the linear distance between inspection points. Triggering measures include tracing all paired inspection points, locating closed carbon sources with leakage risks, repairing sealing defects of closed carbon sources and implementing isolation and control measures to prevent the spread of leaked gases to open areas.

[0013] Preferably, in step five, the anomaly scoring... The evaluation process is as follows: S31. The monitoring cycle is arranged in chronological order from morning to evening. Recorded as , Indicates the monitoring period The total number of timestamps within the period, and the statistical monitoring period in chronological order from earliest to latest. Coverage index within and correlation coefficient ; S32. Based on the fluctuation dataset, the monitoring period... Within the space, the average air pressure fluctuation at each time point is denoted as... The monitoring cycle Within, the average wind speed in the open space at each time stamp is recorded as follows: The monitoring cycle Within this timeframe, the average positioning error of the inspection point for each timestamp is recorded as follows: The monitoring cycle Within this time, the number of temporary carbon source emissions for each timestamp is recorded as follows: ; S33. Based on S31-32, assess the degree of anomalousness of rural carbon sources and assign a rural carbon source anomalousness score. The initial value is set to 0 points; If three consecutive timestamps cover the index ≤Coverage threshold This indicates changes in the rural carbon source structure and assigns rural carbon source anomaly scores. Add 1 point; If there is a correlation coefficient for three consecutive timestamps A score less than 0 indicates that the risk of leakage from the closed carbon source persists and is not under control, thus the rural carbon source anomaly score will be lowered. Add 1 point; During the monitoring period Within a confined space, if the average pressure fluctuation at any given time point is greater than +5 Pa or less than -5 Pa, it indicates that the pressure fluctuation amplitude is causing increased errors in the gas monitoring equipment, thus affecting the rural carbon source anomaly scoring. Add 1 point; During the monitoring period Within the open space, if the average wind speed at any time point is greater than 3 m / s, it indicates that the wind speed accelerates gas dilution, thereby increasing the error of the gas detection equipment and affecting the rural carbon source anomaly scoring. Add 1 point; During the monitoring period If the average positioning error of any inspection point at any time point is greater than 1m, it indicates that the positioning error has increased the error of the gas monitoring equipment, and the rural carbon source anomaly score will be adjusted. Add 1 point; During the monitoring period If, within any given time point, the number of temporary carbon source emissions is ≥1, it indicates that the temporary carbon source emissions have increased the error of the gas monitoring equipment, and the rural carbon source anomaly score will be adjusted accordingly. Add 1 point.

[0014] Preferably, in step five, anomaly scoring... A score greater than 0 indicates that the abnormality of rural carbon sources has exceeded the standard, and the triggering measures include re-examining all carbon sources and suspending temporary carbon source emissions.

[0015] Compared with existing technologies, this invention provides a machine learning diagnostic method for rural carbon source anomalies, which has the following beneficial effects: 1. This invention, through setting By collecting management data on all carbon sources in the village, measured data from all inspection points, and management data on fluctuation factors, the data are categorized into carbon source datasets, inspection datasets, and fluctuation datasets. This provides comprehensive data support for subsequent accurate diagnosis. Based on the carbon source datasets, the rationality of the distribution of rural carbon source inspection points is assessed, and a coverage index is generated. It adapts to different rural carbon source structures, with comprehensive and highly accurate adaptation.

[0016] 2. This invention generates a correlation coefficient by evaluating the linear correlation between the total CO2 equivalent of greenhouse gases at all inspection points and the straight-line distance between inspection points. Accurately assess the risk of carbon source leakage, quickly resolve issues such as insufficient inspection coverage and carbon source leakage, improve the timeliness of prevention and control, and set a fixed monitoring cycle. Combined with the coverage index Correlation coefficient Using a fluctuating dataset, assess the degree of anomaly in rural carbon sources and generate corresponding anomaly scores. This triggers corresponding management measures, resulting in more timely responses to anomalies. Attached Figure Description

[0017] Figure 1 This is a diagram illustrating the steps of the method of the present invention. Detailed Implementation

[0018] 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.

[0019] Example Please see Figure 1 Based on the experimental data of coverage index (Table 1), correlation coefficient (Table 2), and anomaly scoring (Table 3), this invention provides a machine learning diagnostic method for rural carbon source anomalies, comprising the following steps: Step 1: Set up carbon source protection systems in rural areas. Each inspection point is connected to a database and gas detection equipment to obtain management data of all carbon sources in the village, measured data of all inspection points, and management data of fluctuation factors. These data are then classified and compiled into carbon source datasets, inspection datasets, and fluctuation datasets. The carbon source dataset includes the functional type of each carbon source in the village, the number of inspection points, the effective coverage area of ​​a single inspection point, the total agricultural production area, the number of devices covered by a single inspection point, the number of energy consumption devices, the number of links covered by a single inspection point, the number of waste treatment links, the number of sewage treatment links, the number of functions covered by a single inspection point, and the number of building land functions. Among them, the functional types include agricultural production type, energy consumption type, waste treatment type, sewage treatment type, and building land type. The inspection dataset includes the spatial type, CO2 mass concentration, CH4 mass concentration, N2O mass concentration, fluorine-containing gas mass concentration, spatial volume, effective sampling volume, and straight-line distance for each inspection point. The spatial type includes enclosed space and open space. Specifically, the main sources of CO2 emissions in rural areas are biomass burning, soil respiration, and energy consumption; the main sources of CH4 emissions are paddy fields, livestock and poultry farming, and straw fermentation; the main sources of N2O emissions are soil nitrification / denitrification and livestock and poultry manure; and fluorinated greenhouse gases include HFC-134a, CF4, and SF6. The main source of HFC-134a emissions in rural areas is leakage from rural refrigerators and air conditioners; the main source of CF4 emissions in rural areas is degradation of fluorinated agricultural films and application of fluorinated fertilizers; and the main source of SF6 emissions in rural areas is high-voltage switchgear in rural power grids. The fluctuation dataset includes the mean air pressure fluctuation in enclosed spaces, the average wind speed in open spaces, the mean location error of inspection points, and the number of temporary carbon source emissions. Specifically, rural carbon emissions are characterized by a mixed distribution of point sources (closed spaces such as livestock sheds), area sources (open spaces such as farmland), and line sources (rural roads). Furthermore, emission intensity exhibits strong spatiotemporal heterogeneity. The current discrete inspection point monitoring model suffers from low detection frequency (e.g., once a month) and is susceptible to interference from key fluctuation factors such as the average air pressure fluctuation in closed spaces, the average wind speed in open spaces, the average location error of inspection points, and the frequency of temporary carbon source emissions. This not only fails to accurately capture the dynamic changes in greenhouse gas concentrations (e.g., the short-term peak of N2O after farmland fertilization and the nighttime CH4 emission peak in livestock sheds), but also suffers from insufficient spatiotemporal representativeness. In contrast, constructing a multi-scale monitoring network combining vehicle-mounted mobile inspections and drone inspections can not only save on the high operating costs of fixed monitoring stations but also provide real-time dynamic correction based on environmental fluctuation factors to meet the different monitoring needs of closed and open spaces. This enables continuous and refined monitoring of greenhouse gases, effectively compensating for the shortcomings of traditional discrete inspection points in terms of spatiotemporal coverage and data accuracy. Step 2: Based on the carbon source dataset, assess the rationality of the distribution of rural carbon source inspection points and generate the corresponding coverage index. The calculation process is as follows: S11. Based on the carbon source dataset, extract the management data of the current timestamp of all carbon sources; S12. For agricultural production carbon sources, the number of inspection points for agricultural production carbon sources is recorded as follows: The effective coverage area of ​​a single inspection point is denoted as The total area of ​​agricultural production is recorded as Agricultural production includes arable land, orchards, and aquaculture land, and then the area coverage rate of agricultural production carbon sources is calculated. Its expression is as follows: In the formula, This indicates the effective coverage area of ​​all inspection points for agricultural production carbon sources. S13. For energy-consuming carbon sources, the number of inspection points for these sources is recorded as follows: The number of devices covered by a single inspection point is recorded as The number of energy consumption devices is recorded as Energy consumption equipment includes both production and living equipment, and then the equipment coverage rate of energy consumption carbon sources is calculated. Its expression is as follows: In the formula, This indicates the number of effective coverage devices at all inspection points for energy-consuming carbon sources. S14. For waste treatment-type carbon sources, the number of inspection points for these carbon sources is recorded as follows: The number of links covered by a single inspection point is recorded as The number of waste disposal steps is recorded as The waste treatment process includes stages such as "collection → transportation → transfer → treatment → leachate disposal," and then the stage coverage rate of waste treatment-related carbon sources is calculated. Its expression is as follows: In the formula, This indicates the number of effective coverage points for all inspection points of waste-to-energy carbon sources. S15. For wastewater treatment-type carbon sources, the number of inspection points for wastewater treatment-type carbon sources is recorded as follows: The number of links covered by a single inspection point is recorded as The number of wastewater treatment stages is recorded as The wastewater treatment process includes stages such as "bar screen → equalization tank → biological treatment → advanced treatment → effluent discharge," and then the stage coverage rate of the wastewater treatment-type carbon source site is calculated. Its expression is as follows: In the formula, This indicates the number of effective coverage points for all inspection points of wastewater treatment carbon sources. S16. For carbon sources located on building sites, the number of inspection points for these carbon sources is recorded as follows: The number of functions covered by a single inspection point is denoted as The diverse uses of building land are matched with the functional characteristics of the land use. Different functions have significantly different carbon emissions, with commercial buildings consuming more energy than residential buildings. Therefore, the number of building land functions is denoted as... The functional coverage of carbon source sites in the building land area is calculated based on the building land use functions such as residential, commercial, office, medical, and educational. This reflects the coverage of differentiated management and control, and its expression is as follows: In the formula, This indicates the number of effective coverage points for all inspection points of carbon sources located on building sites. S17. Based on S11-S16, calculate the coverage index of rural carbon source inspection points using a weighted method. Its expression is as follows: In the formula, The weight representing the area coverage of agricultural production carbon sources. The weight representing the coverage of energy-consuming carbon source devices. The weight representing the coverage of carbon source processes in waste treatment. The weight representing the coverage rate of carbon source processes in wastewater treatment. The weight representing the functional coverage of carbon sources on built-up land. , , , and All are constants, and ; Specifically, the carbon source structure varies significantly among different villages. Taking a single village as an example, the proportion of carbon sources is positively correlated with the weighting. The proportions of the five types of carbon sources—agricultural production, energy consumption, waste disposal, sewage treatment, and building land use—are different, and the corresponding weighting needs to be adjusted accordingly. This flexible weighting design can adapt to the carbon source structure characteristics of different villages, which can not only scientifically assess the coverage integrity of rural carbon source inspections, but also facilitate horizontal comparison of the coverage levels of different villages, thereby providing a quantitative basis for carbon source management and monitoring optimization. The following is the experimental data for the coverage index, as shown in Table 1: Table 1: Experimental Data for the Coverage Index Table 1 shows the experimental data for the coverage index. Village A was selected as the experimental target, with weights... , , , , ; Step four sets a fixed coverage threshold value. This is used to quickly determine whether the distribution of rural carbon source monitoring points meets the standards. The value directly affects the comprehensiveness and accuracy of rural carbon source emission monitoring. An excessively high coverage threshold... This can lead to the system becoming overly sensitive to minor gaps in inspection coverage, frequently triggering inspection optimization measures and increasing the cost of manpower, materials, and resources for inspections. Conversely, it can reduce the ability to identify blind spots in inspections, making it difficult to detect carbon source monitoring loopholes in a timely manner, hindering a comprehensive understanding of the true situation of rural carbon source emissions, affecting the scientific nature of carbon emission reduction and control decisions, and creating potential risks of carbon accounting bias. Therefore, the optimal value of this parameter needs to be obtained through the following system calibration experiments, coverage threshold. The calibration method is as follows: A simulation scenario of rural carbon source distribution was built using simulation tools. Different carbon source densities, terrain complexity, and accessibility conditions were set, and multiple sets of experiments were conducted. In each set of experiments, the actual carbon source coverage index in the rural area was recorded. The change curves and the system's judgment on the rationality of the distribution were used to simulate the coverage effect under single inspection mode and mixed inspection mode (such as fixed inspection + mobile inspection). The layout of inspection points, coverage radius and inspection time period were adjusted, and multiple sets of experiments were carried out to record the coverage index. The fluctuations and whether the system triggers optimization measures. For each candidate threshold, the collected coverage index will be... The change curve is used as input to count the number of times optimization measures were not triggered in actual inspection blind spots due to improper threshold settings, which are counted as missed detections; and the number of times optimization measures were mistakenly triggered under normal coverage conditions, which are counted as false detections. Finally, the value that minimizes both the missed detection rate and the false detection rate is selected as the coverage threshold. The preferred value; In Table 1, the coverage threshold is used in the experimental data of the coverage index. The preferred value is set to 1.2. Based on the assessment, the coverage index of village A is... <Coverage threshold This indicates that the distribution of carbon source inspection points in rural area A is not reasonable enough. The triggering measures include increasing the number of inspection points, expanding the coverage of a single inspection point, increasing the inspection frequency and extending the inspection period. For weak links with a coverage rate of less than 1, the inspection frequency of agricultural production carbon sources will be increased from once a month to 2-3 times a month. Temporary inspections will be added during key periods (such as the period of concentrated straw and garbage after the busy farming season and the peak garbage period during holidays). Step 3: Based on the inspection dataset, assess the linear correlation between the total CO2 equivalent of greenhouse gases at all inspection points and the straight-line distance between inspection points, and generate the corresponding correlation coefficients. The calculation process is as follows: S21. Based on the inspection dataset, extract the first... The measured data of the current timestamp of each inspection point. and will the The CO2 mass concentration at the current timestamp of each inspection point is recorded as follows: , will the The CH4 mass concentration at the current timestamp of each inspection point is recorded as follows: , will the The N2O mass concentration at the current timestamp of each inspection point is recorded as follows: , will the The fluorine gas concentration at the current timestamp of each inspection point is recorded as follows: ; S22. Calculate the first according to the space type. Total CO2 equivalent of greenhouse gases at the current timestamp of each inspection point Its expression is as follows: If the first The space type of the first inspection point is an enclosed space, which will be the first The spatial volume of each inspection point is denoted as . , In the formula, Indicates the first The mass concentration of the gas Indicates the first The GWP value of the gas on a 100-year timescale. The mapping relationship is When corresponding to CO2, , When corresponding to CH4, , When dealing with N2O, , When dealing with fluorine-containing gases, , Indicates the first The total CO2 equivalent of greenhouse gases at the current timestamp of each closed inspection point; If the first The space type of the first inspection point is an open space, which will be the first The effective sampling volume of the gas detection equipment at each inspection point is denoted as . , In the formula, Indicates the first The first inspection point, under conditions of zero greenhouse gas emissions, The mass concentration of the gas The mapping relationship is the same as above. Indicates the first The quality of the newly added emissions of this gas, Indicates the first The total CO2 equivalent of greenhouse gases at the current timestamp of each open inspection point; Specifically, for enclosed spaces (including greenhouses, livestock sheds, biogas digesters, etc.) in rural areas, the emissions are more accurately quantified by directly measuring the volume of the space because gas emissions tend to accumulate. However, for open spaces (including paddy fields, farmland, ponds, etc.), the background concentration needs to be deducted to eliminate its interference with the calculation of carbon source emissions because the gas diffusion rate is fast. The background concentration under conditions without greenhouse gas emissions is the blank sample measured on site. S23. Based on S21-S22, calculate the total CO2 equivalent of greenhouse gases at the current timestamp of each inspection point. , recorded as Then calculate the average total CO2 equivalent of greenhouse gases at the current timestamp of all inspection points. Its expression is as follows: S24. Based on the inspection dataset, the straight-line distance between each inspection point is denoted as... , This represents the number of pairwise combinations of all inspection points, and then the average straight-line distance between all pairwise combinations of inspection points is calculated. Its expression is as follows: In the formula, Indicates the first The first inspection point and the first The straight-line distance between each inspection point , ; S25. Based on S21-S24, calculate the correlation coefficient between the total CO2 equivalent of greenhouse gases at all inspection points and the straight-line distance between inspection points. Its expression is as follows: In the formula, This represents the sum of the products of the total CO2 equivalent of greenhouse gases and the straight-line distance from the mean. This represents the sum of squares of the mean deviations of the total CO2 equivalent of greenhouse gases. This represents the sum of squared deviations from the mean of the straight-line distance; Specifically, correlation coefficient The range of values ​​for is [ [1,1] is used to determine whether there is a linear correlation and the strength of the correlation between the emission intensity of an inspection point and the spatial distance between inspection points. That is, whether there is a linear correlation and the strength of the correlation between the total greenhouse gas CO2 equivalent of a single inspection point and the other inspection point that is closest to it in a straight line. When the coefficient is close to -1, it indicates that the two are strongly negatively correlated, that is, the closer the straight line distance between the two inspection points, the higher their total greenhouse gas CO2 equivalent, and the higher the probability of abnormal carbon source leakage from a closed inspection point to an open inspection point. When the coefficient is close to 0, it indicates that the two are almost not linearly correlated, that is, there is no obvious linear correspondence between the emission intensity of the inspection point and the spatial distance, and the probability of abnormal carbon source leakage is low. When the coefficient is close to 1, it indicates that the two are strongly positively correlated, that is, the farther the straight line distance, the higher the total greenhouse gas CO2 equivalent of the inspection point, which contradicts the leakage logic and there is no possibility of leakage. The following are the experimental data for the correlation coefficient, as shown in Table 2: Table 2: Experimental Data for Correlation Coefficient In the correlation coefficient experimental data in Table 2, rural village B was selected as the experimental target. The agricultural production carbon sources in rural village B include greenhouses, livestock sheds, biogas digesters, fermentation chambers, paddy fields, and orchards. Inspection points were set up sequentially for each carbon source. Greenhouse No. 1 (60, 120), livestock shed No. 2 (180, 240), biogas digester No. 3 (300, 360), and fermentation chamber No. 4 (240, 180) are all closed inspection points. Paddy field No. 5 (120, 180) and orchard No. 6 (270, 300) are both open inspection points. The GWP value of CO2 gas on a 100-year timescale is 1, that of CH4 gas on a 100-year timescale is 28, that of N2O gas on a 100-year timescale is 265, and that of fluorine-containing gas on a 100-year timescale is 14800. The combination and pairing methods are as follows: The straight-line distance between numbers 1 and 2 is 169.7056m, between numbers 1 and 3 is 339.4113m, between numbers 1 and 4 is 189.7367m, between numbers 1 and 5 is 84.8528m, and between numbers 1 and 6 is 276.5863m. The straight-line distance between points 2 and 6 is 108.1665m; the straight-line distance between points 3 and 4 is 189.7367m; the straight-line distance between points 3 and 5 is 254.5584m; the straight-line distance between points 3 and 6 is 67.0820m; the straight-line distance between points 4 and 5 is 120.0000m; the straight-line distance between points 4 and 6 is 123.6932m; and the straight-line distance between points 5 and 6 is 192.0937m. In Table 2, based on the experimental data of correlation coefficients, it was determined that the correlation coefficient of village B was... When the value is less than 0, it indicates that the total CO2 equivalent of greenhouse gases at all inspection points is negatively correlated with the straight-line distance between inspection points. The triggering measures include tracing the source of all paired inspection points, locating the closed carbon source with leakage risk, and the total CO2 equivalent of greenhouse gases in Greenhouse No. 1 and Livestock House No. 2 being significantly higher than other inspection points, indicating that Greenhouse No. 1 and Livestock House No. 2 have leakage risk. The measures include repairing the sealing defects of the closed carbon source and implementing isolation and control to prevent the leakage gas from spreading to open areas. Step 4: Set a coverage threshold with a fixed value. Combined with the coverage index and correlation coefficient This triggers corresponding management measures; Step 5: Set a fixed monitoring period Combined with the coverage index Correlation coefficient Using a fluctuating dataset, assess the degree of anomaly in rural carbon sources and generate corresponding anomaly scores. This will trigger corresponding management measures. Abnormal scoring The evaluation process is as follows: S31. The monitoring cycle is arranged in chronological order from morning to evening. Recorded as , Indicates the monitoring period The total number of timestamps within the period, and the statistical monitoring period in chronological order from earliest to latest. Coverage index within and correlation coefficient ; S32. Based on the fluctuation dataset, the monitoring period... Within the space, the average air pressure fluctuation at each time point is denoted as... The monitoring cycle Within, the average wind speed in the open space at each time stamp is recorded as follows: The monitoring cycle Within this timeframe, the average positioning error of the inspection point for each timestamp is recorded as follows: The monitoring cycle Within this time, the number of temporary carbon source emissions for each timestamp is recorded as follows: ; S33. Based on S31-32, assess the degree of anomalousness of rural carbon sources and assign a rural carbon source anomalousness score. The initial value is set to 0 points; If three consecutive timestamps cover the index ≤Coverage threshold This indicates changes in the rural carbon source structure, requiring adjustments. , , , and The weighting of rural carbon source anomaly scores Add 1 point; If there is a correlation coefficient for three consecutive timestamps A value less than 0 indicates that the risk of leakage from the closed carbon source persists and is not under control, requiring a reinvestigation of the leaking carbon source and an anomaly score for the rural carbon source. Add 1 point; During the monitoring period If the average air pressure fluctuation in the enclosed space at any time point is greater than +5 Pa or less than -5 Pa, it indicates that the air pressure fluctuation amplitude has increased the error of the gas detection equipment. The error needs to be corrected before re-measurement, and the rural carbon source anomaly score should be adjusted accordingly. Add 1 point; During the monitoring period If the average wind speed in the open space at any time point is greater than 3 m / s, it indicates that the wind speed accelerates gas dilution, thus increasing the error of the gas detection equipment. A windproof protective cover needs to be installed, and the measurement re-tested. The rural carbon source anomaly score should be adjusted accordingly. Add 1 point; During the monitoring period If the average positioning error of any inspection point at any time point is greater than 1m, it indicates that the positioning error has increased the error of the gas monitoring equipment. The error needs to be corrected before re-measurement, and the rural carbon source anomaly score should be adjusted accordingly. Add 1 point; During the monitoring period If the number of temporary carbon source emissions at any time point is ≥1, it indicates that the temporary carbon source emissions have increased the error of the gas monitoring equipment. Therefore, it is necessary to expand the calculation scope and re-measure, and then score the rural carbon source anomalies. Add 1 point; The following is the experimental data for anomaly scoring, as shown in Table 3: Table 3: Experimental Data for Anomaly Scoring In Table 3, five consecutive monitoring timestamps were selected as the monitoring period for the anomaly scoring experimental data. Rural area B was used as the experimental target, with each timestamp spaced 1 hour apart, covering the threshold. Set to 0.75. Based on the assessment, the monitoring period Within, the abnormal rating of Village B A score greater than 0 indicates that the carbon source anomaly in Village B has exceeded the standard. Triggering measures include re-examining all carbon sources, suspending temporary carbon source emissions, and ensuring the effectiveness of carbon source monitoring and control. In this embodiment, the machine learning process for diagnosing rural carbon source anomalies is as follows: 1. Data Acquisition Phase: By connecting the database and gas detection equipment, the distribution data of historical inspection points, management data of all carbon sources in the village, measured data of all inspection points, and management data of fluctuation factors are obtained. These data are then classified into carbon source datasets, inspection datasets, and fluctuation datasets. The carbon source datasets cover key management information such as the functional type of each carbon source in the village and the number of inspection points. The inspection datasets contain measured data such as the spatial type of each inspection point and the mass concentration of various greenhouse gases. The fluctuation datasets involve data on fluctuation factors such as the average air pressure fluctuation in enclosed spaces and the average wind speed in open spaces. 2. Data processing stage: Based on the monitoring cycle The dataset is split and divided into training set (for model training), validation set (for adjusting model parameters), and real-time dataset (for inference evaluation) in chronological order from early to late. This ensures that the different datasets are continuous in time and cover various carbon emission scenarios, providing comprehensive data support for model learning. 3. Model selection stage: A weighted fusion machine learning model, specifically gradient boosting tree (XGBoost / LightGBM), is used to learn the nonlinear relationships between multi-dimensional data such as rural carbon source cover, the correlation between greenhouse gas emissions and spatial distance, and the impact of fluctuation factors. 4. Model training phase: Input training set: historical monitoring periods Previous carbon source dataset, inspection dataset, fluctuation dataset, coverage threshold Coverage Index Correlation coefficient and abnormal scoring and historical monitoring cycles The carbon source dataset, inspection dataset, and fluctuation dataset within the system; Tag verification set: historical monitoring period Within, coverage threshold Coverage Index Correlation coefficient and abnormal scoring The true value; Training objective: To reduce computational load based on actual application scenarios, pairing each open inspection point with the nearest closed inspection point in a straight line and calculating the correlation coefficient. To minimize the mean square error between the model's predicted threshold and the actual threshold, as well as the mean square error between the model's predicted index and coefficient and the actual index and coefficient, ensuring the accuracy of the model's prediction of carbon source anomaly-related indicators. Hyperparameter tuning: Set the learning rate (0.01~0.1) and tree depth (3~8), and adjust the hyperparameters through multiple iterations based on the characteristic distribution of rural carbon source data to ensure the model's prediction accuracy; 5. Model Evaluation and Optimization Phase: The evaluation metrics are the mean squared error between the predicted and actual values ​​of the index and coefficients, where the coverage threshold is also considered. The mean square error between the predicted and actual values ​​is ≤5%, and the coverage index is [not specified]. The mean square error between the predicted and actual values ​​is ≤5%, and the correlation coefficient is... The mean square error between the predicted and actual values ​​is ≤5%, and anomaly scoring is applied. The mean square error between the predicted and actual values ​​is ≤5%; Model optimization: If the mean square error between any predicted value and the true value is greater than 5%, the historical monitoring period is supplemented again. Incremental training was performed using the carbon source dataset, inspection dataset, and fluctuation dataset within the model, and the weight coefficients were autonomously adjusted through model iteration. , , , and The optimal model is obtained when the mean square error between all predicted values ​​and the true values ​​is ≤5%. 6. Implementation and Deployment Phase: Input real-time dataset (current monitoring period) (Includes carbon source datasets, inspection datasets, and fluctuation datasets). Using the current best model, the coverage threshold is automatically output. Coverage Index Correlation coefficient and abnormal scoring This triggers targeted management measures.

[0020] The threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by those skilled in the art for each set of sample data; as long as it does not affect the ratio between the parameter and the quantized value, it is acceptable.

[0021] The above formulas are all derived from software simulation using a large amount of data and are selected to be close to the actual values. The coefficients in the formulas are set by those skilled in the art according to the actual situation. The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any equivalent substitutions or changes made by those skilled in the art within the technical scope disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the protection scope of the present invention.

Claims

1. A machine learning-based diagnostic method for abnormal carbon sources in rural areas, characterized in that, Includes the following steps: Step 1: Setting up carbon sources in rural areas Each inspection point is connected to a database and gas detection equipment to obtain management data of all carbon sources in the village, measured data of all inspection points, and management data of fluctuation factors. These data are then classified and compiled into carbon source datasets, inspection datasets, and fluctuation datasets. Step 2: Based on the carbon source dataset, assess the rationality of the distribution of rural carbon source inspection points and generate the corresponding coverage index. ; Step 3: Based on the inspection dataset, assess the linear correlation between the total CO2 equivalent of greenhouse gases at all inspection points and the straight-line distance between inspection points, and generate the corresponding correlation coefficients. ; Step 4: Set a coverage threshold with a fixed value. Combined with the coverage index and correlation coefficient This triggers corresponding management measures; Step 5: Set a fixed monitoring period Combined with the coverage index Correlation coefficient Using a fluctuating dataset, assess the degree of anomaly in rural carbon sources and generate corresponding anomaly scores. This will trigger corresponding management measures.

2. The machine learning diagnostic method for rural carbon source anomalies according to claim 1, characterized in that: In step one, the carbon source dataset includes the functional type of each carbon source in the village, the number of inspection points, the effective coverage area of ​​a single inspection point, the total agricultural production area, the number of devices covered by a single inspection point, the number of energy consumption devices, the number of links covered by a single inspection point, the number of waste treatment links, the number of sewage treatment links, the number of functions covered by a single inspection point, and the number of building land functions. The functional types include agricultural production, energy consumption, waste treatment, sewage treatment, and building land.

3. The machine learning diagnostic method for rural carbon source anomalies according to claim 2, characterized in that: In step one, the inspection dataset includes the spatial type, CO2 mass concentration, CH4 mass concentration, N2O mass concentration, fluorine-containing gas mass concentration, spatial volume, effective sampling volume, and straight-line distance for each inspection point. The spatial type includes enclosed space and open space.

4. The machine learning diagnostic method for rural carbon source anomalies according to claim 3, characterized in that: In step one, the fluctuation dataset includes the average air pressure fluctuation in enclosed spaces, the average wind speed in open spaces, the average location error of inspection points, and the number of temporary carbon source emissions.

5. The machine learning diagnostic method for rural carbon source anomalies according to claim 4, characterized in that: In step two, the coverage index The calculation process is as follows: S11. Based on the carbon source dataset, extract the management data of the current timestamp of all carbon sources; S12. For agricultural production carbon sources, the number of inspection points for agricultural production carbon sources is recorded as follows: The effective coverage area of ​​a single inspection point is denoted as The total area of ​​agricultural production is recorded as Then calculate the area coverage of agricultural production carbon sources. ; S13. For energy-consuming carbon sources, the number of inspection points for these sources is recorded as follows: The number of devices covered by a single inspection point is recorded as The number of energy consumption devices is recorded as Then calculate the equipment coverage of energy consumption carbon sources. ; S14. For waste treatment-type carbon sources, the number of inspection points for these carbon sources is recorded as follows: The number of links covered by a single inspection point is recorded as The number of waste disposal steps is recorded as Then calculate the stage coverage of waste treatment carbon sources. ; S15. For wastewater treatment-type carbon sources, the number of inspection points for wastewater treatment-type carbon sources is recorded as follows: The number of links covered by a single inspection point is recorded as The number of wastewater treatment stages is recorded as Then calculate the coverage rate of wastewater treatment carbon source sites. ; S16. For carbon sources located on building sites, the number of inspection points for these carbon sources is recorded as follows: The number of functions covered by a single inspection point is denoted as The number of building land functions is recorded as Then calculate the functional coverage of carbon source sites on building land. ; S17. Based on S11-S16, calculate the coverage index of rural carbon source inspection points using a weighted method. .

6. The machine learning diagnostic method for rural carbon source anomalies according to claim 5, characterized in that: In step three, the correlation coefficient The calculation process is as follows: S21. Based on the inspection dataset, extract the first... The measured data of the current timestamp of each inspection point. and the first The CO2 mass concentration at the current timestamp of each inspection point is recorded as follows: , will the The CH4 mass concentration at the current timestamp of each inspection point is recorded as follows: , will the The N2O mass concentration at the current timestamp of each inspection point is recorded as follows: , will the The fluorine gas concentration at the current timestamp of each inspection point is recorded as follows: ; S22. Calculate the first according to the space type. Total CO2 equivalent of greenhouse gases at the current timestamp of each inspection point ; S23. Based on S21-S22, calculate the total CO2 equivalent of greenhouse gases at the current timestamp of each inspection point. , recorded as Then calculate the average total CO2 equivalent of greenhouse gases at the current timestamp of all inspection points. ; S24. Based on the inspection dataset, the straight-line distance between each inspection point is denoted as... , This represents the number of pairwise combinations of all inspection points, and then the average straight-line distance between all pairwise combinations of inspection points is calculated. ; S25. Based on S21-S24, calculate the correlation coefficient between the total CO2 equivalent of greenhouse gases at all inspection points and the straight-line distance between inspection points. .

7. The machine learning diagnostic method for rural carbon source anomalies according to claim 6, characterized in that: Step four, coverage index ≤Coverage threshold When this occurs, it indicates that the distribution of rural carbon source inspection points does not meet the standards. Triggering measures include increasing the number of inspection points, expanding the coverage of individual inspection points, increasing the frequency of inspections, and extending the inspection period.

8. The machine learning diagnostic method for rural carbon source anomalies according to claim 7, characterized in that: Step four, correlation coefficient When the value is less than 0, it indicates that the total CO2 equivalent of greenhouse gases at all inspection points is negatively correlated with the linear distance between inspection points. Triggering measures include tracing all paired inspection points, locating closed carbon sources with leakage risks, repairing sealing defects of closed carbon sources and implementing isolation and control measures to prevent the spread of leaked gases to open areas.

9. The machine learning diagnostic method for rural carbon source anomalies according to claim 8, characterized in that: In step five, anomaly scoring The evaluation process is as follows: S31. The monitoring cycle is arranged in chronological order from morning to evening. Recorded as , Indicates the monitoring period The total number of timestamps within the period, and the statistical monitoring period in chronological order from earliest to latest. Coverage index within and correlation coefficient ; S32. Based on the fluctuation dataset, the monitoring period... Within the space, the average air pressure fluctuation at each time point is denoted as... The monitoring cycle Within, the average wind speed in the open space at each time stamp is recorded as follows: The monitoring cycle Within this timeframe, the average positioning error of the inspection point for each timestamp is recorded as follows: The monitoring cycle Within this time, the number of temporary carbon source emissions for each timestamp is recorded as follows: ; S33. Based on S31-32, assess the degree of anomalousness of rural carbon sources and assign a rural carbon source anomalousness score. The initial value is set to 0 points; If three consecutive timestamps cover the index ≤Coverage threshold This indicates changes in the rural carbon source structure and assigns rural carbon source anomaly scores. Add 1 point; If there is a correlation coefficient for three consecutive timestamps A score less than 0 indicates that the risk of leakage from the closed carbon source persists and is not under control, thus the rural carbon source anomaly score will be lowered. Add 1 point; During the monitoring period Within a confined space, if the average pressure fluctuation at any given time point is greater than +5 Pa or less than -5 Pa, it indicates that the pressure fluctuation amplitude is causing increased errors in the gas monitoring equipment, thus affecting the rural carbon source anomaly scoring. Add 1 point; During the monitoring period Within the open space, if the average wind speed at any time point is greater than 3 m / s, it indicates that the wind speed accelerates gas dilution, thereby increasing the error of the gas detection equipment and affecting the rural carbon source anomaly scoring. Add 1 point; During the monitoring period If the average positioning error of any inspection point at any time point is greater than 1m, it indicates that the positioning error has increased the error of the gas monitoring equipment, and the rural carbon source anomaly score will be adjusted. Add 1 point; During the monitoring period If, within any given time point, the number of temporary carbon source emissions is ≥1, it indicates that the temporary carbon source emissions have increased the error of the gas monitoring equipment, and the rural carbon source anomaly score will be adjusted accordingly. Add 1 point.

10. The machine learning diagnostic method for rural carbon source anomalies according to claim 9, characterized in that: Step five, anomaly scoring A score greater than 0 indicates that the abnormality of rural carbon sources has exceeded the standard, and the triggering measures include re-examining all carbon sources and suspending temporary carbon source emissions.

Images