Rail transit carbon emission management method and system

Abstract

The application relates to a rail transit carbon emission management method and system, which comprises the following steps: if the carbon emission abnormality risk level of this time is a low risk level, a visual interface displays the current overall carbon emission value, the deviation amplitude from the predicted value, and a trend arrow indicating the trend of carbon emission in a preset time range in the future; if the carbon emission abnormality risk level of this time is a medium risk level, the visual interface displays the detailed prediction deviation curve of each component and the dynamic change process of the deviation; if the carbon emission abnormality risk level of this time is a high risk level, the visual interface displays the detailed prediction deviation curve of each component and the dynamic change process of the deviation, and simultaneously, the high risk level and the abnormality reason are warned. The application has the following effects: accurate carbon emission prediction, early warning and reason analysis are realized, the risk level is visualized, and efficient management is facilitated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of information management, in particular to a rail transit carbon emission management method and system. BACKGROUND

[0002] At present, with the increasing environmental awareness and the global consensus on sustainable development, as the core part of urban transportation system, the carbon emission problem of rail transit has a significant impact on the environment. Efficient carbon emission management not only helps to reduce the negative effects of rail transit on the environment, but also helps the green transformation and long-term development of the industry.

[0003] At present, the carbon emission management of rail transit mainly focuses on data-driven analysis methods. In the data processing stage, the focus is on data collection and preliminary processing. Through various sensors, multi-source and heterogeneous carbon emission related data such as train operation and station equipment are collected, and then simple data cleaning work is performed.

[0004] The prior art has a major deficiency in the core carbon emission early warning analysis mechanism. Due to the inability to accurately compare real-time data with predicted values, it is difficult to accurately judge the abnormal risk level of carbon emissions, and it is even more difficult to determine the root cause of the abnormality. Moreover, the visualization display link fails to effectively link with the carbon emission early warning level, and lacks targeted design according to different early warning levels. This makes it difficult for management departments to intuitively and quickly grasp the real situation of carbon emissions when facing a large amount of carbon emission data, and unable to plan and develop effective countermeasures in advance, which seriously hinders the progress of rail transit carbon emission management towards a scientific and efficient direction. SUMMARY

[0005] In order to realize accurate carbon emission prediction, early warning and cause analysis, and to assist efficient management through risk level visualization, the present application provides a rail transit carbon emission management method and system.

[0006] In a first aspect, the present application provides a rail transit carbon emission management method, which adopts the following technical solution:

[0007] A rail transit carbon emission management method, comprising:

[0008] Obtaining multi-source and heterogeneous data related to carbon emissions and performing data preprocessing;

[0009] Extracting carbon emission time series data from the multi-source and heterogeneous data related to carbon emissions after completing data preprocessing, and using a preset seasonal decomposition method to strip the influence of periodic factors, and decomposing the time series into trend items, seasonal items and residual items;

[0010] Based on the decomposed trend items, seasonal items and residual item time series information, extract the preset deep features;

[0011] inputting the extracted deep features into a pre-constructed ARIMA model, and predicting a numerical range of a trend item, a seasonal item and a residual item of carbon emission in a future preset time range as a prediction value of the ARIMA model;

[0012] comparing the prediction value of the ARIMA model with real-time carbon emission related data, and calculating a prediction deviation of each component including the trend item, the seasonal item and the residual item of carbon emission;

[0013] inputting the prediction deviation of each component into a pre-constructed carbon emission early warning analysis model, and outputting a carbon emission abnormal risk level and an abnormal reason of this time, wherein the carbon emission abnormal risk level includes a low risk level, a medium risk level and a high risk level;

[0014] if the carbon emission abnormal risk level of this time is the low risk level, a current overall carbon emission value, a deviation amplitude between the current overall carbon emission value and a prediction value, and a trend arrow indicating a trend of carbon emission in a future preset time range are displayed on a visual interface;

[0015] if the carbon emission abnormal risk level of this time is the medium risk level, a detailed prediction deviation curve and a dynamic change process of the deviation of each component are displayed on the visual interface;

[0016] if the carbon emission abnormal risk level of this time is the high risk level, the detailed prediction deviation curve and the dynamic change process of the deviation of each component are displayed on the visual interface, and the high risk level and the abnormal reason are simultaneously warned.

[0017] By using the above technical solution, the carbon emission trend can be accurately predicted through multi-source data processing and analysis. By comparing the prediction value with the real-time data, the prediction deviation can be found in time. With the help of the early warning model, the abnormal risk level and the reason can be determined. Different risk levels correspond to different visual displays, which is convenient and intuitive for understanding the carbon emission situation, and is beneficial for relevant departments to take measures in advance to realize effective management of carbon emission of rail transit.

[0018] Optionally, if the carbon emission abnormal risk level of this time is the medium risk level or the high risk level, the method further comprises:

[0019] acquiring environmental monitoring satellite data, wherein the environmental monitoring satellite data includes regional meteorological dynamics and geographic information;

[0020] extracting a feature vector related to rail transit carbon emission from the environmental monitoring satellite data, inputting the feature vector related to rail transit carbon emission as an input object into a pre-constructed comprehensive model for simulating changes in carbon emission under different emergency scenarios, and outputting a simulation result;

[0021] According to the output simulation result, a curve of rail transit carbon emission change over time in a future preset time range is drawn, and emergency scenario related information is marked.

[0022] By adopting the technical solution, the environment monitoring satellite data is acquired, and regional meteorological and geographical information can be comprehensively understood. The related feature vectors are extracted from the environment monitoring satellite data and input into the comprehensive model, so that the carbon emission change under different emergency scenarios can be simulated. According to the simulation result, the change curve is drawn and the information is marked, so that the future carbon emission trend can be intuitively presented, and the manager can clearly grasp the carbon emission dynamics under different emergency scenarios, thereby providing a strong basis for formulating scientific and effective response strategies.

[0023] Optionally, the feature vectors related to rail transit carbon emission are extracted from the environment monitoring satellite data, including:

[0024] The carbon emission abnormality risk level of this time is acquired;

[0025] The real-time satellite data feature fluctuation information is analyzed and determined according to the environment monitoring satellite data, the carbon emission abnormality risk level and the real-time satellite data feature fluctuation information are input into the pre-trained feature selection standard analysis model, and the feature selection standard is output.

[0026] The feature vectors related to rail transit carbon emission are extracted from the environment monitoring satellite data according to the feature selection standard.

[0027] By adopting the technical solution, the risk level and the satellite data fluctuation information are input into the model to obtain the selection standard, which can accurately adapt to the current carbon emission abnormality. According to this standard, the feature vectors are extracted, so that the extracted information is highly targeted and closely related to rail transit carbon emission. This not only improves the data screening efficiency, but also provides more valuable data support for simulating the carbon emission change under different emergency scenarios, thereby assisting accurate decision-making.

[0028] Optionally, the feature vectors related to rail transit carbon emission are taken as input objects and input into the pre-constructed comprehensive model for simulating the carbon emission change under different emergency scenarios, and the simulation result includes:

[0029] Whether three-dimensional terrain model information scanned by a laser radar device deployed on the top of a train or at key points along the line and environmental parameters detected by a dust sensor network deployed along the rail transit line and stations are acquired is analyzed;

[0030] If not, the feature vectors related to rail transit carbon emission are taken as input objects and input into the pre-constructed comprehensive model for simulating the carbon emission change under different emergency scenarios, and the simulation result is output.

[0031] If yes, extract a laser radar feature vector from the acquired three-dimensional terrain model information, and calibrate and integrate the acquired environmental parameters and extract an intelligent micro-dust sensor network feature vector;

[0032] Fuse the laser radar feature vector, the intelligent micro-dust sensor network feature vector, and the feature vector extracted based on the environmental monitoring satellite data to form a fusion vector;

[0033] Adjust the architecture of the comprehensive model according to the feature vectors contained in the fusion vector and a preset comprehensive model architecture quantitative adjustment method;

[0034] Input the fusion vector into the comprehensive model simulating carbon emission changes under different emergency scenarios after the architecture adjustment, and output the simulation results.

[0035] By using the above technical solution, the accuracy of the simulation results can be significantly improved through the judgment and use of additional equipment data. When there is laser radar and micro-dust sensor data, corresponding feature vectors are extracted and fused, and the architecture of the comprehensive model is also adjusted to make the model better adapt to complex data. This enables the model to comprehensively reflect the influence of terrain and environmental factors on carbon emissions and output more realistic simulation results, providing more reliable decision-making basis for rail transit carbon emission management.

[0036] Optionally, the high-risk level and the abnormal reason are also warned, including:

[0037] Analyze whether to acquire abnormal sound information captured by an acoustic sensor array deployed on a train operation track and key equipment;

[0038] If no, maintain the original settings;

[0039] If yes, extract acoustic features from the abnormal sound information, and extract core sound features from the acoustic features through a dimension reduction method of principal component analysis;

[0040] Input the core sound features and the prediction deviation of each component into an abnormal sound diagnosis model to output the abnormal reason;

[0041] Update the abnormal reason of the warning according to the output abnormal reason, and display the high-risk level together.

[0042] By using the above technical solution, the clues behind the abnormal sound can be deeply mined by using the acoustic sensor array information. By extracting core features from the abnormal sound and combining the prediction deviation to accurately output the abnormal reason through the diagnosis model, the warning content can be updated in real time. This not only makes the abnormal reason of the warning more accurate, but also helps staff quickly locate the root cause of the problem and take effective measures in a timely manner to ensure the safe and low-carbon operation of the rail transit system.

[0043] Optionally, the step of synchronously updating the abnormal reason of the warning according to the output abnormal reason and displaying the high risk level together is performed, and the step is specifically as follows:

[0044] Analyze the object and the abnormal position associated with the abnormal reason, wherein the object includes a train running track and a key device;

[0045] If the abnormal reason is associated with the train running track, obtain the three-dimensional terrain model information and the environmental parameters of the location area where the abnormal position is located, and according to the combination of the train running track associated with the abnormal reason, the three-dimensional terrain model information of the location area where the abnormal position is located, and the environmental parameters, match the track fault influence range from the pre-constructed track fault influence model, and obtain the fault influence range analysis diagram;

[0046] If the abnormal reason is associated with the key device, obtain the three-dimensional terrain model information and the environmental parameters of the location area where the abnormal position is located, and according to the combination of the key device associated with the abnormal reason, the three-dimensional terrain model information of the location area where the abnormal position is located, and the environmental parameters, match the key device fault influence range from the pre-constructed fault propagation model, and obtain the fault influence range analysis diagram;

[0047] Obtain the passenger flow distribution data of the location area where the abnormal position is located, and use dynamic visualization technology to superimpose and display the personnel flow in the fault influence range analysis diagram in real time.

[0048] By using the above technical solution, the terrain and environmental information can be obtained by analyzing the object and the position associated with the abnormal reason, and the fault influence range can be accurately determined by using the model. At the same time, the relationship between the personnel flow and the fault area can be intuitively presented by superimposing the passenger flow distribution data and dynamically displaying it. This helps the relevant departments to quickly evaluate the potential risks, develop scientific evacuation and response strategies, ensure personnel safety, and reduce accident losses.

[0049] Optionally, the step of superimposing and displaying the personnel flow in the fault influence range analysis diagram in real time is specifically as follows:

[0050] Obtain the personnel behavior state data and environmental factor data of the location area where the abnormal position is located, wherein the personnel behavior state data includes action posture, stay time and gathering mode, and the environmental factors include temperature, humidity, air flow speed and carbon dioxide concentration;

[0051] Synchronously integrate the personnel behavior state data and the environmental factor data according to the time and space dimensions;

[0052] According to the identified different personnel action postures, adapt individualized carbon emission coefficients for each individual;

[0053] The environmental factor data is input into a carbon emission coefficient dynamic correction factor analysis model, and a dynamic correction factor is output to adjust the carbon emission coefficient;

[0054] The personalized carbon emission coefficient of all personnel in the location area is multiplied by the corresponding residence time, and then summed to obtain the real-time personnel flow carbon emission total of the location area;

[0055] On the fault influence range analysis diagram, the personnel flow carbon emission data is displayed in a preset visual manner.

[0056] By adopting the above technical solution, the personnel flow carbon emission of the abnormal area can be accurately quantified. The personnel behavior and environmental data are obtained and integrated, the carbon emission coefficient is adapted for individuals, and the coefficient is more realistic after dynamic correction. The real-time personnel flow carbon emission total is calculated and visualized, which can intuitively present the influence of personnel activities on carbon emission. It is helpful to understand the details of carbon emission in the fault area, provide a basis for formulating energy-saving and emission-reducing and optimizing environmental measures, and help low-carbon operation and environmental management.

[0057] In a second aspect, the present application provides a rail transit carbon emission management system, which adopts the following technical solution:

[0058] A rail transit carbon emission management system includes a memory, a processor, and a program stored on the memory and executable on the processor. The program can be loaded and executed by the processor to implement the rail transit carbon emission management method as described in the first aspect. BRIEF DESCRIPTION OF DRAWINGS

[0059] Fig. 1 is a flowchart of a rail transit carbon emission management method according to an embodiment of the present application.

[0060] Fig. 2 is a flowchart of another embodiment of the present application when the carbon emission abnormality risk level of this time is a medium risk level or a high risk level.

[0061] Fig. 3 is a flowchart of another embodiment of the present application for extracting a feature vector related to rail transit carbon emission from environmental monitoring satellite data. DETAILED DESCRIPTION

[0062] The present application will be further described in detail below with reference to the accompanying drawings.

[0063] REFERENCE Fig. 1 A rail transit carbon emission management method disclosed by the present application includes:

[0064] Step S100, obtaining multi-source heterogeneous data related to carbon emission and performing data preprocessing.

[0065] “Multi-source” means that the data comes from multiple different channels. For example, on one hand, it comes from inside the rail transit system, such as the power consumption data recorded by various energy consumption monitoring devices on the train, which reflects the energy use during the train operation and is directly related to carbon emissions; on the other hand, it also comes from external relevant departments, such as meteorological data such as air temperature, wind speed, and wind direction provided by the meteorological department along the line, because the meteorological conditions will affect the energy consumption of the train operation and thus affect the carbon emissions. “Heterogeneous” means that the structure and format of these data are different, such as the energy consumption data may be structured data recorded in numerical form at certain time intervals, while the meteorological report provided by the meteorological department may be semi-structured or unstructured data containing various forms such as text description and chart.

[0066] Data preprocessing refers to a series of operations to process the collected raw data to make it more suitable for subsequent analysis and modeling requirements. Specifically, it covers data cleaning, which removes error values, duplicate values, and missing values, and other abnormal situations in the data.

[0067] Step S200, from the carbon emission related multi-source heterogeneous data after data preprocessing, extract the carbon emission time series data, and use the preset seasonal decomposition method to strip the periodic factor influence, and decompose the time series into trend item, seasonal item and residual item.

[0068] Carbon emission time series data: within the scope of rail transit carbon emission management, it refers to carbon emission related data records arranged in a certain time sequence. The way to extract carbon emission time series data is as follows: from the multi-source heterogeneous data after step S100 data preprocessing, filter out the data directly related to carbon emissions and arranged in chronological order. For example, from the integrated train energy consumption data, station equipment energy consumption data and corresponding timestamp information, extract the carbon emission total value at each time point according to the predetermined time interval (such as every hour, every day, etc.), and form the time series data.

[0069] Preset seasonal decomposition method: this is a mathematical method specially used to analyze the characteristics of time series data, the purpose of which is to separate different components contained in the time series. In this application, it is to distinguish the part affected by seasonal, periodic and other factors from other components. Common seasonal decomposition methods include additive model and multiplicative model, etc. For example, for rail transit, there may be seasonal factors, such as in summer, due to high temperature, the air conditioning and refrigeration equipment in the station is used frequently, the energy consumption increases, leading to a certain seasonal variation of carbon emissions, and through this method, the part related to such rules (i.e. seasonal component) can be separated.

[0070] Trend component: represents the overall trend of the time series data over a longer period of time, reflecting the general trend of carbon emissions over time. It is a relatively stable and slowly changing component. Seasonal component: as mentioned earlier, it is related to the seasonal regularity in the time series. Residual component: it is the remaining part after removing the trend component and the seasonal component, containing random fluctuation factors in the time series that cannot be explained by the trend and seasonal regularity. For example, a sudden equipment failure causes temporary increase in energy consumption of trains, resulting in abnormal fluctuations in carbon emissions on that day. This abnormal fluctuation is reflected in the residual component after removing the trend and seasonal factors.

[0071] Step S300, based on the decomposed trend component, seasonal component and residual component time series information, extract the preset deep features.

[0072] Deep features refer to features with representative and key information extracted from original data after a series of complex and deep processing and analysis.

[0073] Trend component deep feature extraction can involve the following: 1. Long-term trend slope and curvature: calculate the slope of the trend component over a longer period of time, which represents the average rate of change of carbon emissions over time. For example, by fitting the trend component curve using linear regression or other methods, the slope is obtained. A positive slope indicates an overall upward trend in carbon emissions, while a negative slope indicates a downward trend. At the same time, the curvature of the curve is calculated, which reflects the acceleration or deceleration of the trend change. If the curvature gradually increases, it means that the speed of carbon emissions rising or falling is accelerating.

[0074] Seasonal component deep feature extraction can involve the following: 1. Seasonal peak and valley: identify the peak and valley of carbon emissions in each season of the seasonal component. 2. Seasonal fluctuation amplitude: calculate the fluctuation amplitude of carbon emissions in each season, i.e. the difference between the peak and the valley. By comparing the fluctuation amplitudes of different seasons, we can understand which seasons have more dramatic changes in carbon emissions. 3. Seasonal phase and delay: analyze the phase of the seasonal component, i.e. the starting position of seasonal change on the time axis. In addition, consider whether there is a delay phenomenon in the seasonal change in different years or time periods.

[0075] Residual term depth feature extraction can involve the following: 1. Residual standard deviation and variance: Calculate the standard deviation and variance of the residual term, which reflect the degree of dispersion of the residual. A larger standard deviation or variance means that the residual fluctuates more, i.e. there are more random fluctuations that cannot be explained by trend and seasonal factors. Outlier detection and frequency: By setting a certain threshold, detect outliers in the residual term. Outliers may be due to equipment failure or other reasons causing abnormal fluctuations in carbon emissions. The frequency and distribution of statistical outliers can help identify potential problems in a timely manner.

[0076] Step S400, input the extracted depth features into the pre-constructed ARIMA model, and predict the numerical range of the carbon emission trend term, seasonal term and residual term in the future preset time range as the ARIMA model prediction value.

[0077] ARIMA model: ARIMA, which stands for Autoregressive Integrated Moving Average Model, is a commonly used time series prediction model. In rail transit carbon emission management, this model can be used to predict future carbon emissions based on historical carbon emission-related data features (i.e. the depth features extracted earlier). The future preset time range refers to the time period for which carbon emissions are to be predicted, which can be one month, one quarter or one year, etc. according to actual needs.

[0078] Step S500, compare the ARIMA model prediction value with real-time carbon emission-related data, and calculate the prediction bias of each component, including the carbon emission trend term, seasonal term and residual term.

[0079] The ARIMA model prediction value is the numerical range of the carbon emission trend term, seasonal term and residual term for the future preset time range predicted by the ARIMA model based on the previously extracted carbon emission depth features in step S400. For example, the trend term may have a specific numerical range in the next quarter, the seasonal term will show a corresponding periodic fluctuation range, and the residual term will also have a corresponding expected fluctuation range. These predicted results constitute the ARIMA model prediction value, which is a kind of estimation of future carbon emissions.

[0080] The prediction bias is the difference between the ARIMA model prediction value and the real-time carbon emission-related data, which is obtained through certain calculation methods. It can directly reflect the accuracy of the model prediction and the degree of deviation between the actual carbon emissions and the expected value. For each component (trend term, seasonal term, residual term), its prediction bias can be calculated separately to understand the prediction accuracy in different aspects in detail.

[0081] Step S600, input the predicted deviation of each component into the pre-constructed carbon emission early warning analysis model, output the current carbon emission abnormal risk level and the abnormal reason, the carbon emission abnormal risk level includes low risk level, medium risk level and high risk level.

[0082] The carbon emission early warning analysis model is a model specially constructed for rail transit carbon emission situation, which is based on machine learning, data analysis and other related technologies and algorithms, and through comprehensive analysis and processing of the input component (carbon emission trend item, seasonal item and residual item) predicted deviation data, the current carbon emission abnormal risk level is judged, and the corresponding abnormal reason is given.

[0083] Model analysis and judgment mechanism: the carbon emission early warning analysis model will comprehensively consider the input component predicted deviation data according to the pre-set algorithm and rules. It will refer to the corresponding relationship between different deviation combinations and risk levels, abnormal reasons learned during previous training. For example, if the trend item deviation continues to increase and at the same time the residual item appears more abnormal values, the model may determine that this situation conforms to the characteristics of high risk level in the past, and then evaluate towards the direction of high risk level.

[0084] Step S700, if the current carbon emission abnormal risk level is low risk level, the current overall carbon emission value, the deviation amplitude of the current overall carbon emission value and the predicted value are displayed on the visual interface, and the trend arrow is used to show the trend of carbon emission in the future preset time range.

[0085] Visual interface: an interactive interface that displays data in the form of intuitive graphics, charts and other forms, which can convert complex carbon emission data into visual information easy to understand, so that staff can quickly grasp the overall situation.

[0086] Current overall carbon emission value display: the current overall carbon emission value of the rail transit system is presented in the form of numbers or column charts in a prominent position on the visual interface.

[0087] Trend arrow display: according to the carbon emission trend in the future preset time range predicted by ARIMA model, draw a trend arrow on the visual interface. If the predicted carbon emission shows an upward trend, draw an upward arrow; if it shows a downward trend, draw a downward arrow; if it is basically stable, draw a horizontal arrow. The predicted approximate change range can be marked beside the arrow, such as "future one week carbon emission is expected to rise / fall [X]%", which helps users understand the future carbon emission trend.

[0088] Step S800, if the carbon emission anomaly risk level of this time is a medium risk level, the visualization interface displays the detailed prediction deviation curve of each component and the dynamic change process of the deviation.

[0089] Detailed prediction deviation curve of each component: for the carbon emission trend term, seasonal term and residual term, respectively, the deviation between the actual value and the ARIMA model prediction value is displayed in the form of a curve. These curves can clearly reflect the deviation size and change trend of each component at different time points.

[0090] Dynamic change process of deviation: in a dynamic demonstration way, the continuous change of the prediction deviation of each component over time is presented, so that users can intuitively see how the deviation develops and evolves.

[0091] Step S900, if the carbon emission anomaly risk level of this time is a high risk level, the visualization interface displays the detailed prediction deviation curve of each component and the dynamic change process of the deviation, and at the same time, the high risk level and the abnormal reason are warned.

[0092] The warning involves the following:

[0093] Highlight the high risk level: use a prominent color (such as red) to identify the "high risk level" word, or place it in a prominent position on the visualization interface, such as the top center of the interface, in a larger font, to ensure that staff can see at a glance that the current carbon emission is in a high risk state.

[0094] Show the abnormal reason: in the area immediately below or next to the high risk level identification, clearly and concisely list the abnormal reasons for this time being judged as a high risk level. For example, write in text form "abnormal reason: a key equipment failure, resulting in a significant increase in energy consumption, which in turn causes carbon emissions to exceed expectations, and the failure has lasted [X] days", so that staff can quickly know the root cause of the high risk, providing key evidence for quickly developing and implementing appropriate solutions.

[0095] Referring to Fig. 2 , if the carbon emission anomaly risk level of this time is a medium risk level or a high risk level, the following steps are also included:

[0096] Step SA00, obtain environmental monitoring satellite data, wherein the environmental monitoring satellite data includes regional meteorological dynamics and geographic information.

[0097] Environmental monitoring satellite data: data collected and transmitted back to the ground by a dedicated environmental monitoring satellite.

[0098] Regional meteorological dynamics: refers to the changes of meteorological elements in a specific region over time, including but not limited to temperature, pressure, humidity, wind speed, wind direction, precipitation, etc.

[0099] Geographical Information: This encompasses various information related to the geographical features of the Earth's surface, such as topography (mountains, plains, rivers, etc.), elevation, land use types (cities, forests, farmland, etc.).

[0100] Data filtering and extraction involves the following: 1. Determine the regional scope: in combination with the actual orientation and coverage area of the rail transit line, the corresponding geographical scope is delineated in satellite data. 2. Time filtering: according to the time requirements of rail transit carbon emission analysis, satellite data within a specific time period is selected.

[0101] Step SB00, extract the feature vector related to rail transit carbon emissions from environmental monitoring satellite data, input the feature vector related to rail transit carbon emissions as the input object into the pre-constructed comprehensive model simulating carbon emission changes under different emergency scenarios, and output the simulation results.

[0102] Feature vector: In this context, it refers to a set of data combinations extracted from environmental monitoring satellite data that can reflect the characteristics closely related to rail transit carbon emissions.

[0103] Comprehensive model simulating carbon emission changes under different emergency scenarios: This is a specially constructed model designed to simulate how rail transit carbon emissions will change under various emergency scenarios such as extreme weather, natural disasters, sudden equipment failure, etc. by inputting relevant feature data. The model takes into account multiple factors and their interrelationships, and is constructed based on mathematical algorithms, physical principles, and past actual case data, etc. It can output the estimated changes in carbon emissions under corresponding scenarios, providing a reference for formulating countermeasures in advance.

[0104] Step SC00, according to the output simulation results, draw the curve of rail transit carbon emissions over time within the future preset time range, and mark the emergency scenario related information.

[0105] Future preset time range: According to actual needs and analysis purposes, a time period is set in advance, such as the next three days, a week, or a month, etc.

[0106] The specific operation is as follows: 1. Construct the coordinate system: in the visual drawing interface (such as professional drawing software or customized data analysis platform interface), first determine the coordinate system. Usually take time as the horizontal axis, divide according to equal interval, the size of the interval depends on the length of the preset time range and the requirement of display accuracy. For example, if the preset time range is one week in the future, it can be divided by day; if the preset time range is one month in the future, it can be divided by week. Set the vertical axis as the carbon emission value, and reasonably determine the scale range of the vertical axis according to the approximate range of the carbon emission value in the simulation result, to ensure that the change of carbon emission can be completely and clearly displayed. 2. Draw the change curve:

[0107] According to the carbon emission values corresponding to different time points in the simulation result, the corresponding coordinate points are found out in the coordinate system, and then these points are connected in turn by a smooth curve. If multiple emergency scenarios are simulated, a corresponding curve needs to be drawn for each scenario, and different colors and line styles (such as solid lines, dashed lines, etc.) can be used to distinguish them, making it easy to view and compare the differences in carbon emission changes under different scenarios. 3. Label the emergency scenario related information as follows: 3.1, scenario name label: label the name of the emergency scenario represented by each curve in a prominent position above or beside the curve, such as "heavy rain emergency scenario", "earthquake emergency scenario", etc., so that the viewer can immediately understand which specific situation this curve corresponds to. 3.2, key influencing factor labeling: briefly label and explain the key factors that have a significant impact on carbon emissions in the emergency scenario. For example, label "railway water accumulation leads to increased friction and energy consumption" next to the "heavy rain emergency scenario" curve; label "partial line interruption, increased travel distance for train detours, increased energy consumption and carbon emissions" next to the "earthquake emergency scenario" curve, etc. This helps the viewer better understand the reasons behind the carbon emission changes reflected by the curve. 3.3, special time point labeling: for key time points in the emergency scenario that may have significant changes in carbon emissions, such as turning points where carbon emissions suddenly increase or decrease, label the corresponding position on the curve and briefly explain the reason. For example, label "here, due to the completion of equipment repair, carbon emissions begin to decline" at a certain time point on a curve, so that the viewer can better understand the details of the carbon emission changes.

[0108] Reference Fig. 3 From the environmental monitoring satellite data, the feature vectors related to rail transit carbon emissions include:

[0109] Step SB10, obtain the carbon emission abnormal risk level of this time.

[0110] Step SB20, according to the environmental monitoring satellite data analysis to determine the real-time satellite data feature fluctuation information, the carbon emission abnormal risk level and the real-time satellite data feature fluctuation information are input into the pre-trained feature selection standard analysis model, and the feature selection standard is output.

[0111] Real-time satellite data feature fluctuation information: a set of information reflecting the fluctuation and change of various environmental related data (such as meteorological elements, geographical features, etc.) at the current time or in the near future, which is analyzed from environmental monitoring satellite data.

[0112] The analysis and determination of real-time satellite data feature fluctuation information specifically involves the following contents: 1. Meteorological data fluctuation analysis: for the temperature data obtained from environmental monitoring satellite data, calculate the range (difference between maximum and minimum) and standard deviation of statistical quantities within a certain time period (such as every hour, every day, etc.), to measure the fluctuation degree of temperature. For example, if the temperature range of a certain area within a day reaches 10℃, it means that the temperature fluctuation is large, which may have a significant impact on the temperature control energy consumption and carbon emissions of rail transit. 2. For wind speed and direction data, analyze their change frequency and amplitude in different time periods. For example, by observing whether the wind direction frequently changes within a few hours, whether the wind speed increases or decreases sharply, etc., to determine the dynamic impact on train operation resistance and energy consumption, and to organize these analysis results into feature fluctuation information of meteorological data. 3. Geographical data fluctuation analysis: if the geographical information related to topography and geomorphology data (such as elevation, slope, etc.) is involved, check whether these geographical features in the area along the line have changed, such as whether there is a slight local elevation change due to geological activity, or new construction changes the local terrain slope, etc. By comparing geographical data at different times, determine the fluctuation of geographical features. Although geographical data changes relatively slowly compared to meteorological data, some special cases can also affect rail transit energy consumption and carbon emissions, which also need to be organized into corresponding fluctuation information. 4. Integration of fluctuation information: integrate the fluctuation information obtained from meteorological data and geographical data, etc. to form a complete set of real-time satellite data feature fluctuation information, which can be used as model input in the future. For example, the integrated information may be in the form of "temperature daily range 10℃, wind speed changes 3 times per hour, elevation of a certain section increases by 2 meters compared with last month, etc."

[0113] Inputting data into the feature selection standard analysis model and obtaining output involves the following: 1. Data input: The carbon emission abnormal risk level obtained in step SB10 (such as "low risk level", "medium risk level" or "high risk level" and the like) and the real-time satellite data feature fluctuation information determined by analysis are arranged according to the input format required by the feature selection standard analysis model. For example, the risk level is in the form of a string, and the fluctuation information is in a specific data structure (such as a dictionary, a list, etc. according to the model setting) respectively input into the corresponding input interface of the model to ensure that the model can accurately identify and receive these data. 2. Model operation and output: The feature selection standard analysis model will analyze the input carbon emission abnormal risk level and real-time satellite data feature fluctuation information comprehensively according to the pre-learned knowledge and algorithm rules. For example, if the input is a medium risk level and the information that the temperature fluctuation is large and the terrain of a certain section changes, the model will refer to the experience of similar situations in the past and weigh the importance of different environmental factors on carbon emission.

[0114] Step SB30, according to the feature selection standard, the feature vector related to rail transit carbon emission is extracted from the environmental monitoring satellite data.

[0115] The process of extracting and integrating data to form a feature vector involves the following: 1. Data extraction: According to the analyzed feature selection standard, each type of data is extracted one by one. 2. Feature vector integration: The extracted data is arranged and integrated in a certain order to form a complete feature vector.

[0116] The feature vector related to rail transit carbon emission is input into the pre-constructed comprehensive model simulating the change of carbon emission under different emergency scenarios, and the simulation results include:

[0117] Step SBA0, analyze whether to obtain three-dimensional terrain model information scanned by laser radar devices deployed on the top of the train or at key points along the line, and environmental parameters detected by the dust sensor network deployed along the rail transit line and station. If not, execute step SBB0; if yes, execute step SBC0.

[0118] Laser radar device: It is an active sensor device that uses laser beams to detect the position, speed, shape and other information of the target.

[0119] Three-dimensional terrain model information: The information contained in the three-dimensional model presenting the spatial terrain conditions along the rail transit line constructed by the data generated by the laser radar device after scanning, including the coordinate position, height value and terrain fluctuation of each point, etc.

[0120] Micro-dust sensor network: a network system formed by deploying multiple micro-dust sensors in a certain layout and connection manner at various places along the rail transit line and station. These micro-dust sensors can detect various environmental parameters in the surrounding environment in real time, such as particulate matter concentration in the air, humidity, temperature, etc. Through network transmission, the data detected at each point is collected to provide a basis for comprehensively grasping the environmental conditions around the rail transit, and then analyze its impact on carbon emissions.

[0121] Environmental parameters: refers to the various specific index values detected by the micro-dust sensor network, reflecting the physical properties of the environment along the rail transit line and around the station, such as the above-mentioned air particulate matter concentration, humidity, temperature, etc. Different environmental parameters will affect the running state and energy consumption of train equipment to varying degrees, thereby being associated with carbon emissions.

[0122] Step SBB0: input the feature vector related to rail transit carbon emissions as an input object into the comprehensive model pre-constructed to simulate the changes in carbon emissions under different emergency scenarios, and output the simulation results.

[0123] Feature vector related to rail transit carbon emissions: this is a data set that reflects various key factors affecting rail transit carbon emissions, which is screened and extracted through a series of previous steps. It covers feature information closely related to carbon emissions such as meteorology, geography, etc. Combined in a specific order and format, it serves as the basis for input content for subsequent model analysis, such as containing relevant data elements such as wind speed, wind direction, air temperature, altitude, etc.

[0124] Comprehensive model simulating changes in carbon emissions under different emergency scenarios: a mathematical model specially constructed for the rail transit field, which is based on a large amount of historical data, physical principles and related professional knowledge, and comprehensively considers various possible emergency scenarios (such as extreme weather, equipment failure, natural disasters, etc.) and the interaction of various factors on carbon emissions under these scenarios. By inputting the corresponding feature data, it can simulate the changes in rail transit carbon emissions over time under different emergency conditions, providing a reference basis for decision-making.

[0125] Step SBC0: extract the lidar feature vector from the acquired three-dimensional terrain model information, and simultaneously calibrate and integrate the acquired environmental parameters, and extract the intelligent micro-dust sensor network feature vector.

[0126] The extraction of the lidar feature vector from the three-dimensional terrain model information involves the following: 1. Identifying key terrain elements: a comprehensive and in-depth analysis of the three-dimensional terrain model information is conducted to accurately identify terrain elements closely related to rail transit carbon emissions. 2. Quantification and construction of feature vectors: The identified key terrain elements are quantified, with slope expressed in specific angle values or percentages, terrain relief quantified by changes in height difference per unit distance, and elevation difference directly recorded as the height difference of the corresponding section. These quantified terrain element data are organized in a logical order to construct the lidar feature vector.

[0127] Calibration and integration of environmental parameters and extraction of intelligent micro dust sensor network feature vectors involve the following processes:

[0128] Environmental parameter calibration: 1. Temperature calibration: Due to differences in environmental microclimate at different micro dust sensor locations and sensor precision deviations, temperature data detected by each sensor may have errors. Therefore, a standard thermometer or a reference source with known temperature is used to conduct comparative measurements near the installation location of the sensor, and the temperature measurements of each sensor are corrected based on the comparison results to accurately reflect the true environmental temperature. 2. Humidity calibration: Similarly, for humidity sensors, a professional high-precision humidity calibration device or a standard environmental chamber with stable humidity is used to calibrate the humidity values detected by each sensor, eliminating measurement deviations caused by sensor aging, installation conditions, and other factors, ensuring that all humidity data are under a unified and accurate standard. 3. Calibration of other parameters: For air pressure, particulate matter concentration, harmful gas concentration, and other environmental parameters, corresponding professional calibration methods are used, such as comparing and calibrating air pressure values with a standard barometer, and calibrating particulate matter concentration with high-precision particulate matter monitoring instruments, to ensure the accuracy and reliability of each environmental parameter data, laying a foundation for subsequent integration and analysis.

[0129] Integration and extraction of feature vectors: From the calibrated environmental parameters, key indicators with strong correlation to rail transit carbon emissions are selected. For example, temperature and humidity affect the energy consumption of the train's temperature control system, high particulate matter concentration may increase the energy consumption of the ventilation system, and changes in air pressure affect the aerodynamic performance of the train, thereby affecting energy consumption. These key indicators are selected as the core elements of the feature vector. These selected key environmental parameter data are arranged in a reasonable order to form the intelligent micro dust sensor network feature vector.

[0130] Step SBD0, fuse the lidar feature vector, intelligent micro dust sensor network feature vector, and feature vector extracted based on environmental monitoring satellite data to form a fused vector.

[0131] Fusion vector: a new vector formed by organically integrating the three feature vectors of different sources, which comprehensively integrates topography, environmental parameters, and macro-environment, and other key information related to rail transit carbon emissions.

[0132] The specific fusion operation involves the following processes:

[0133] Fusion operation: 1. Sequential fusion by element order: A common fusion method is to sequentially combine the corresponding elements according to the order of the elements of each feature vector to form a fusion vector. For example, assuming that the laser radar feature vector is [slope value 1, elevation difference value 1, …], the intelligent dust sensor network feature vector is [temperature value 1, humidity value 1, …], and the feature vector extracted based on environmental monitoring satellite data is [wind speed value 1, wind direction value 1, …], then the fusion vector can be constructed in order as [slope value 1, elevation difference value 1, temperature value 1, humidity value 1, wind speed value 1, wind direction value 1, …], so that the fused vector completely contains key feature information from different aspects. 2. Consider weight fusion (if applicable): In some cases, according to the importance of different features on rail transit carbon emissions, different weights may be assigned to the elements in each feature vector before fusion. For example, it is found through analysis that the weight of terrain factors (represented by the laser radar feature vector) on carbon emissions in a specific scenario is 0.4, the weight of environmental parameters (represented by the intelligent dust sensor network feature vector) is 0.3, and the weight of macro-environment (represented by the feature vector extracted based on environmental monitoring satellite data) is 0.3. Then, during fusion, for each corresponding element, multiply it by the weight corresponding to the vector it belongs to, and then combine it to reflect the relative importance of each factor in the overall carbon emission impact, so that the fusion vector more scientifically and reasonably reflects the actual situation.

[0134] Step SBE0, according to the feature vectors contained in the fusion vector and the preset comprehensive model architecture quantization adjustment method, adjust the architecture of the comprehensive model.

[0135] Pre-set comprehensive model architecture quantization adjustment method: This is a set of rules and calculation methods that are pre-set when building a comprehensive model to simulate carbon emission changes under different emergency scenarios, which is used to adjust the architecture of the model according to the characteristics of the input data (here, the features contained in the fusion vector), such as changing the number of layers, the number of neurons (for neural network models), or adjusting the weights of parameters in the equation, etc., the purpose is to make the model better adapt to new input data, improve the accuracy of the model in simulating carbon emission changes under different emergency scenarios.

[0136] Architecture of the integrated model: refers to the internal structure of the integrated model that simulates the changes in carbon emissions under different emergency scenarios, including the types of mathematical models used (such as linear regression models, neural network models, etc.), the connection between each component, parameter settings, etc. It determines how the model processes input data and outputs corresponding simulation results. A reasonable architecture is the key to ensuring the performance and simulation effect of the model.

[0137] The specific operation involves the following processes:

[0138] Analysis of fusion vector features: 1. Feature importance evaluation: carefully study each feature vector and its specific elements contained in the fusion vector, and evaluate the importance of different features in simulating the changes in rail transit carbon emissions. For example, analyze the weight of the terrain slope feature (from the laser radar feature vector) in the fusion vector on train energy consumption and carbon emission changes under specific rail transit lines and operating environments, compared with environmental parameter features such as temperature and humidity (from the intelligent micro-dust sensor network feature vector) and macro-environmental features such as wind speed and direction (from the feature vector extracted based on environmental monitoring satellite data), to determine which feature is more critical in the current scenario, or which features have strong interaction relationships. 2. Feature correlation analysis: calculate the correlation coefficient between each feature in the fusion vector to determine the degree of association between different features. For example, check if there is a significant correlation between the altitude feature and the temperature feature, because altitude changes may affect air temperature, which in turn affects train temperature control energy consumption and carbon emissions. Through correlation analysis, the synergistic effect or mutual restraint relationship of each feature in the process of influencing carbon emissions can be better understood, providing a basis for subsequent model architecture adjustment.

[0139] Adjusting the model architecture according to the quantization adjustment method: 1. Adjust the model complexity (take the neural network model as an example): 1.1. Adjust the number of layers: If the fusion vector contains more and complex features, according to the preset comprehensive model architecture quantization adjustment method, it may be necessary to increase the number of layers of the model to better fit the nonlinear relationship between these features and carbon emissions. For example, originally a three-layer neural network model, after analysis found that the features in the fusion vector need a deeper network structure to mine their internal relationship, then according to the adjustment method, increase the model layer to four or even more, to enhance the expression ability of the model, so that it can accurately capture the influence of different feature combinations on carbon emission changes. 1.2. Adjust the number of neurons: At the same time, the number of neurons in each layer is also adjusted accordingly. For those layers corresponding to the parts that are closely related to carbon emissions and have rich feature changes, appropriately increase the number of neurons to more finely process and analyze these key feature information. For example, between the input layer and the hidden layer, if the environmental parameter related features in the input fusion vector are more and the changes are complex, increase the number of neurons in the hidden layer, so that the model can fully learn the influence mode of these environmental factors on carbon emissions, and improve the simulation accuracy. 1.3. Adjust the parameter weight: For the existing parameters in the model (such as the coefficients in the linear regression model, the connection weights in the neural network model, etc.), according to the importance and correlation analysis results of the features in the fusion vector, the weights are redistributed according to the preset quantization adjustment method.

[0140] Step SBF0, input the fusion vector into the comprehensive model for simulating carbon emission changes under different emergency scenarios after adjusting the architecture, and output the simulation results.

[0141] At the same time, high-risk levels and abnormal reasons are warned, including:

[0142] Step S910, analyze whether to obtain abnormal sound information captured by the acoustic sensor array deployed on the train running track and key equipment. If not, step S920 is executed; if yes, step S930 is executed.

[0143] Among them, the acoustic sensor array: a system composed of multiple acoustic sensors arranged in a specific combination, deployed around the train running track and key equipment, used for omnidirectional and high-precision collection of sound signals.

[0144] In the rail transit scene, the train running track and key equipment will produce specific frequency and characteristic sounds when running normally. When the track has cracks, parts are loose, or the key equipment (such as train traction motor, braking system, etc.) fails, the sound emitted will change. The acoustic sensor array can capture these sound changes to provide raw data for judging the running state of the equipment and the track condition.

[0145] Abnormal sound information: refers to the sound data that has significant differences in frequency, amplitude, tone, and other acoustic characteristics from the sound emitted by the normal operation of the train running track and key equipment. These abnormal sounds often serve as early signals of equipment failure or track problems. Role: Through the analysis of abnormal sound information, potential fault sources can be quickly located, and timely measures can be taken to repair them, avoiding further deterioration of the fault and ensuring the safe operation of rail transit. At the same time, it also helps to analyze the reasons for abnormal carbon emissions, as equipment failure or track abnormalities may lead to increased energy consumption, which in turn affects carbon emissions.

[0146] Step S920, maintain the original settings.

[0147] Step S930, extract acoustic features from abnormal sound information, and extract core sound features from acoustic features through the dimension reduction method of principal component analysis.

[0148] Principal component analysis principle and calculation: 1. Principle: PCA aims to convert the original high-dimensional feature data into a set of new, mutually orthogonal low-dimensional eigenvectors, i.e. principal components, through linear transformation. 2. Calculation steps: Data standardization: standardize each column of data in the feature matrix to have a mean of 0 and a variance of 1. This is to eliminate the differences in dimension and numerical range between different features, ensuring that each feature has equal importance in subsequent calculations. For example, the numerical range of the short-time energy feature may be larger, while the numerical range of some frequency features is smaller. Through standardization, they can be treated fairly in principal component analysis. Calculate the covariance matrix: calculate the covariance matrix of the standardized data, which reflects the linear correlation between different features. For example, if two acoustic features (such as short-time energy and a certain frequency component) have a strong positive correlation, the corresponding element value in the covariance matrix will be larger. Calculate the eigenvalues and eigenvectors: solve the eigenvalues and eigenvectors of the covariance matrix, the eigenvalues represent the variance contained in each principal component, and the eigenvectors determine the direction of the principal component. Sort the eigenvalues from large to small, select the eigenvectors corresponding to the first few large eigenvalues, and these eigenvectors form the new coordinate system of the reduced principal component space. For example, by calculating 10 eigenvalues, select the eigenvectors corresponding to the first 3 largest eigenvalues, and project the original 10-dimensional acoustic features into a 3-dimensional principal component space to achieve dimension reduction.

[0149] Core sound feature extraction: project the original acoustic feature data onto the selected principal component space to obtain the reduced core sound features. These core sound features not only retain most of the important information in the original acoustic features, but also remove redundant and weakly correlated information, making subsequent analysis more efficient and accurate.

[0150] Step S940: Input the core sound features and the prediction deviations of each component into the abnormal sound diagnosis model, and output the cause of the abnormality.

[0151] Abnormal sound diagnostic models are built upon extensive historical data, equipment operating principles, and acoustic analysis knowledge. Common model types include neural network models (such as multilayer perceptrons and convolutional neural networks), decision tree models, and support vector machine models. The most suitable model is selected based on the characteristics of the rail transit field and the available data.

[0152] After receiving input data, the abnormal sound diagnosis model performs complex calculations based on its internal algorithms and structure. Taking a neural network model as an example, the data first passes through the input layer, and then undergoes multiple nonlinear transformations and feature extractions in the hidden layers. Neurons in the hidden layers perform weighted summations of the input data according to pre-trained weights and biases, and then perform nonlinear transformations using activation functions (such as ReLU, Sigmoid, etc.) to gradually uncover potential patterns and features in the data. Different models employ different computational methods, but all aim to find the correlation between the input data and known causes of anomalies.

[0153] After analysis and computation by the model, the final output is a diagnostic result for the cause of the abnormal sound. This result may be presented in various forms, such as a text description explicitly stating the cause as "train traction motor bearing wear" or "loose track fasteners"; it may also be output as classification labels corresponding to different anomaly types, such as "equipment failure type 1" or "track problem type 3". The model output may also include relevant confidence information, indicating the model's degree of certainty regarding the diagnostic result. For example, an output result of "train traction motor bearing wear, confidence level 90%" provides a clear direction and reference for subsequent fault investigation and handling of abnormal carbon emissions.

[0154] Step S950: Update the alarm's cause of error based on the output error cause, and display the high-risk level as well.

[0155] Choose an appropriate display platform to present high-risk levels and updated causes of anomalies. Common display platforms include large-screen monitoring systems in rail transit control centers, handheld terminal devices (such as tablets and smartphones) used by maintenance personnel, and dedicated carbon emission management software interfaces. Configure the platform according to its characteristics to ensure that the information is displayed clearly and intuitively.

[0156] Display high-risk levels and abnormal reasons in an intuitive and visual manner. For high-risk levels, use prominent colors (such as red) and larger fonts to highlight, such as displaying "High Risk Level" on the screen with a red background fill. For abnormal reasons, present them in simple and clear language, avoiding overly complex terms and sentence structures. For example, display "Abnormal Reason: Train braking system brake pad excessive wear" below the high-risk level.

[0157] The steps of updating the abnormal reason of the warning according to the output abnormal reason and synchronously displaying the high-risk level are as follows:

[0158] Step Sa00, analyze the objects and abnormal positions associated with the abnormal reason, wherein the objects include train running tracks and key equipment.

[0159] Extract keywords related to objects and positions from the abnormal reason description. For example, the keywords for "Train traction motor brush wear is serious" are "train traction motor" (associated object) and "brush" (further specifying the specific part of the associated object). For "Track circuit choke transformer has inter-turn short circuit", the keywords are "track circuit" (associated object) and "choke transformer" (specifying the specific equipment of the associated object). By extracting these keywords, a foundation is provided for subsequent determination of the objects and positions associated with the abnormality.

[0160] The associated object is determined as follows: 1. Determine the object category: According to the extracted keywords, determine whether the object associated with the abnormal reason belongs to the train running track or the key equipment. For example, if the abnormal reason involves "loose track fastener", it is clear that the associated object is the train running track; while for "train air conditioning compressor failure", the associated object is the key equipment of the train. Establishing clear object category determination criteria helps to quickly classify and handle different types of abnormal situations. For example, establish the rule: any description related to track structure, track bed, track accessories, etc. is determined as a train running track associated object; any description related to train-mounted equipment, power system, control system, etc. is determined as a key equipment associated object. 2. Refine object information: For objects associated with train running tracks, further refine the information. If the keyword is "track fastener", it needs to be clear which segment of the track the fastener is on, whether it is a straight segment, a curve segment, or a turnout area fastener. It can be accurately located through track number, mileage mark, etc. For example, in the track transportation line map, the track number and mileage information are clearly marked, and when an abnormality is found, the specific track section can be quickly located. For key equipment associated objects, determine the specific location of the equipment on the train, such as the train traction motor, which is usually installed on the bogie, and determine which bogie of which car the traction motor is on. By consulting the train equipment layout diagram and related technical documents, these information can be accurately obtained.

[0161] The abnormal position determination can be based on the location lookup of the object and the positioning in combination with auxiliary information.

[0162] If the associated object is a train operation track, according to the track number, mileage mark and detailed record of track facilities, the specific track position area where the abnormal position is located is determined. For example, it is known that the abnormal reason is “track weld cracking at K10+200”, through the mileage measurement system of the track transportation line and the track facility account, the position of the weld in the actual track line can be accurately found. For key equipment, the specific installation position of the equipment is determined by using the train equipment layout diagram and the equipment number system. For example, the brake system of the train is distributed in each carriage, and through the carriage number and brake equipment number, the carriage and the specific installation position where the faulty brake component is located can be quickly located.

[0163] Step Sb00, if the abnormal reason is associated with the train operation track, the three-dimensional terrain model information and environmental parameters of the position area where the abnormal position is located are obtained, and the track fault influence range is matched from the pre-constructed track fault influence model according to the combination of the three of the abnormal reason associated with the train operation track, the three-dimensional terrain model information of the position area where the abnormal position is located, and the environmental parameters, and the fault influence range analysis diagram is obtained.

[0164] Three-dimensional terrain model information acquisition: The three-dimensional terrain model information of the position area where the abnormal position is located is found from the geographic information system (GIS) or a special track terrain database. These information may be stored in a special file format, such as a digital elevation model (DEM) file, which contains detailed terrain data of the area where the track is located. For example, the corresponding DEM file can be opened using ArcGIS software, the terrain information of the area is extracted by locating the coordinates of the abnormal position, including the altitude, slope and other information of different positions.

[0165] Environmental parameter acquisition: The environmental parameters of the position area where the abnormal position is located are obtained from the environmental monitoring system. The environmental monitoring system may include multiple sensors distributed along the track, which will collect environmental data in real time or periodically and store the data in the database. The temperature, humidity, wind speed and other environmental parameters of the position area where the abnormal position is located can be extracted from the database through data query statements.

[0166] The abnormal reason, the three-dimensional terrain model information of the position area where the abnormal position is located, and the environmental parameters are sorted to meet the input requirements of the track fault influence model.

[0167] The processed data is input into a pre-constructed track failure impact model. The model considers the influence of terrain on failure propagation (e.g., on steep slopes, track failure may cause trains to slide down, affecting a larger range), environmental factors on track structure (e.g., high temperatures may cause track expansion, potentially expanding the failure range), and the characteristics of the anomaly itself (e.g., track cracks may expand at different speeds and directions under different terrain and environmental conditions), and calculates the impact range of the failure.

[0168] Regarding the generation of the failure impact range analysis chart, the type of chart is selected: according to the needs of the analysis, the appropriate chart is selected to represent the impact range of the failure. It can be a two-dimensional planar impact range chart, showing the impact range of the failure on the track plane, marking the areas that may be affected based on the track line; it can also be a three-dimensional impact range chart, more intuitively showing the impact range on the terrain, such as showing the impact range at different altitudes and planar positions.

[0169] Drawing tool usage: professional drawing software (such as AutoCAD, MATLAB, etc.) is used to generate the failure impact range analysis chart. In the chart, different colors or patterns are used to distinguish different degrees of impact range, such as red area representing severe impact area, yellow area representing moderate impact area, and green area representing slight impact area. Key information is marked, such as marking the anomaly location, terrain contour, environmental factors (such as wind direction arrow indicating the direction of wind influence), and the possible expansion direction of the failure, to help analysts intuitively understand the impact range of the failure.

[0170] Step Sc00, if the anomaly cause is associated with a key device, obtain the three-dimensional terrain model information and environmental parameters of the location area where the anomaly is located, and according to the combination of the anomaly cause associated key device, three-dimensional terrain model information of the location area where the anomaly is located, environmental parameters, match and obtain the key device failure impact range from the pre-constructed failure propagation model, and get the failure impact range analysis chart.

[0171] Integrate the obtained three-dimensional terrain model information, environmental parameters, and key device information associated with the anomaly cause. Convert the three-dimensional terrain model data into a format suitable for model input, such as organizing the terrain elevation, slope, and other data into a matrix form; organize the environmental parameters into an array according to the category, such as [temperature value, humidity value, air pressure value, wind speed value, wind direction value]; at the same time, associate the type, model, failure mode, and other information of the key device with the above data in text or specific coding form, forming a complete input data set.

[0172] Input fault propagation model: The integrated dataset is accurately input into the model according to the input interface and format required by the pre-constructed fault propagation model. For example, if the model is a machine learning model developed based on Python, the input data set is passed into the model as a parameter by calling the prediction function of the model, triggering the calculation and analysis process of the model.

[0173] Model calculation: After receiving the input data, the fault propagation model first analyzes the fault mode of the key equipment and determines the possible propagation path of the fault in the equipment according to the physical structure and working principle of the equipment. For example, if the key equipment is the braking system of the train, when the brake cylinder fails, the model will analyze how the fault affects the action of the brake shoe and the transmission of the braking force according to the pipeline connection, mechanical transmission relationship, etc. of the braking system.

[0174] Then, the model will consider the influence of three-dimensional terrain model information and environmental parameters on fault propagation. In complex terrain areas, equipment failure may cause changes in train operation state, which in turn affects other related equipment. For example, in the climbing section, the failure of the traction motor may cause the train speed to drop, causing the braking system to act frequently and expanding the scope of the fault. In terms of environmental parameters, high temperature may accelerate the damage of equipment, and high humidity may cause short circuit of electrical equipment. The model will comprehensively consider these factors for complex calculation and simulation.

[0175] Result acquisition: After the calculation of the model, the result data of the influence range of the key equipment fault is obtained. These data may be presented in different forms, such as the list of affected equipment, the geographic coordinate range of the fault affected area, and the hierarchical structure of the affected system. For example, the result data shows that the failure of the braking system of the train will affect the power supply system, signal system of the same car, and part of the control circuit of the adjacent car, and gives the specific location coordinates of these affected equipment and systems.

[0176] Step Sd00, obtain the flow distribution data of the location area where the abnormal position is located, and use dynamic visualization technology to real-time superimposed display the personnel flow in the fault influence range analysis diagram.

[0177] The flow distribution data can be obtained based on sensor data collection or video monitoring data extraction.

[0178] Based on sensor-based data collection, the specific way is as follows: 1. Infrared sensor: Infrared sensors are deployed at key locations such as entrances and exits, corridors, and platforms in rail transit stations. These sensors identify the entry and movement of personnel by detecting the infrared radiation emitted by the human body. For example, a row of infrared sensors is set up at the edge of the platform. When a passenger passes through, the sensor will produce an electrical signal change according to the change in infrared intensity. The system counts and analyzes these signals to count the number of people passing through the area. At the same time, combined with the position information of the sensor, it can preliminarily judge the approximate distribution of personnel on the platform. 2. Pressure sensor: Pressure sensors are installed in areas where personnel frequently pass through, such as stairs and escalators. When passengers step on these devices, pressure sensors will sense the change in pressure and convert it into an electrical signal to transmit to the data collection system. Through analysis and processing of pressure signals, not only can the number of people passing through be counted, but also the walking speed and density of personnel can be inferred according to the change in pressure distribution.

[0179] Based on video monitoring data extraction, the specific way is as follows: 1. Video analysis algorithm: Advanced video analysis algorithms are used to process real-time monitoring camera videos in rail transit areas. These algorithms can identify human silhouettes in videos through image recognition technology and track and analyze them. For example, by using background difference method, the current video frame is compared with the pre-set background image to extract moving human targets. Then, using target tracking algorithm, each identified human target is continuously tracked to record its movement trajectory and location. Through analysis of a large number of video frames, people flow data in different time periods and different locations is counted. 2. Deep learning model: Use deep learning-based target detection and tracking models such as YOLO (You Only Look Once) series models and Deepsort. These models are trained on a large number of human image data and can accurately identify humans in videos and efficiently track them. For example, the video captured by the monitoring camera is input into the trained YOLO model in real time, which quickly detects human targets in the video and assigns each target a unique identifier. Then, the Deepsort model tracks the target based on its movement characteristics and appearance characteristics, obtaining the flow trajectory and real-time location information of the personnel. Through the fusion analysis of multiple monitoring camera video data, the people flow distribution data in the location area where the abnormal position is located can be obtained.

[0180] The principle of layer superposition in the superimposed display is that in the visualization tool, the personnel flow data is superimposed on the base map of the fault impact range analysis diagram as a new layer. Each layer has its own attributes and data, but when displayed in the visualization, they are superimposed in a certain order. For example, in ArcGIS, the fault impact range analysis diagram is set as the base layer and displayed at the bottom; the personnel flow data is created in the form of a point layer or a line layer and displayed at the top. By adjusting the transparency and display order of the layers, the personnel flow situation can be clearly displayed on the fault impact range analysis diagram, while the key information of the base map is not blocked.

[0181] The display effect optimization can be as follows:

[0182] Color and symbol design: In order to make the personnel flow situation more intuitive and easy to understand, the color and symbol design of the superimposed personnel flow data is carried out. For example, different colors are used to represent different people flow densities, red for high-density areas, yellow for medium-density areas, and green for areas with relatively sparse personnel. For the direction of personnel flow, arrow symbols can be used to represent the direction of personnel flow, and the length or thickness of the arrow can represent the size of the flow speed. Through reasonable color and symbol design, users can quickly understand the distribution and flow trend of personnel in the fault area.

[0183] Interactive function addition: Add interactive functions in dynamic visualization display to improve user experience and data analysis efficiency. For example, in the Web visualization interface, add a mouse hover prompt function, which displays detailed people flow information such as specific number of people, time, etc. when the user hovers the mouse over a personnel flow data point or area. At the same time, zoom and pan functions can be added to allow users to zoom in or out of the fault impact range analysis diagram as needed to view the personnel flow situation in different areas; through the pan operation, the overall view of the entire fault area can be viewed. Through the addition of these interactive functions, users can more flexibly analyze and understand the relationship between personnel flow data and fault impact range.

[0184] The steps of real-time superimposed display of personnel flow situation in the fault impact range analysis diagram are as follows:

[0185] Step Se00, obtain personnel behavior state data and environmental factor data in the location area where the abnormal position is located, wherein the personnel behavior state data includes action posture, stay time and aggregation mode, and the environmental factors include temperature, humidity, air flow speed and carbon dioxide concentration.

[0186] Action posture recognition: relying on advanced computer vision technology, real-time analysis of video captured by surveillance cameras in abnormal position areas. Using a convolutional neural network (CNN) algorithm based on deep learning, which can accurately extract human motion features through learning from a large number of human motion samples. For example, for walking motion, the algorithm captures features such as the motion trajectory of human joints and limb swing amplitude. When analyzing new video streams, it can quickly determine whether the person is in a walking state and further calculate the walking speed. Taking the monitoring video analysis of a subway station platform as an example, by tracking the coordinate changes of the hip joint, knee joint, and ankle joint in the video, combined with the video frame rate, the displacement per second of the person is calculated, and the accurate walking speed value is obtained.

[0187] Residence time monitoring: with the help of video tracking technology, the system continuously tracks each personnel target. Set a time threshold, such as 30 seconds, when the person's stay in a certain position exceeds the threshold, the system automatically records the residence time. For example, at the entrance of a store in the station hall, the video monitoring system monitors a passenger for 5 minutes, and this data will be accurately recorded. To ensure the accuracy of residence time calculation, the system will judge the person's small movements, if the moving distance is less than a certain threshold (such as 0.5 meters), it is still considered as a stay state, to avoid false calculation of residence time due to the person's slight adjustment action.

[0188] Aggregation pattern analysis: set a specific area range, when a certain number (such as more than 5 people) of people appear in this area at the same time, it is determined as personnel aggregation. By analyzing the time when personnel enter and leave the area, as well as the relative position relationship between personnel, further judge the closeness and duration of aggregation. For example, in a narrow section of the transfer channel, the system monitors that 8 people gather in a short time, and the distance between the people is small, the duration is 2 minutes, these data will be recorded in detail for subsequent analysis of the impact of personnel aggregation on the environment and carbon emissions.

[0189] Step Sf00, synchronize and integrate personnel behavior state data and environmental factor data according to time and space dimensions.

[0190] After obtaining personnel behavior and environmental data, they need to be integrated according to time and space dimensions, laying a solid foundation for subsequent carbon emission analysis.

[0191] The time dimension integration is specifically as follows: 1. Time stamp calibration: check the time stamp format of personnel behavior and environmental data, and unify it into a standard format such as “YYYY-MM-DDHH:MM:SS”. For time stamps of data from different sources, round to the nearest second to eliminate millisecond-level differences and ensure consistency of data time points. 2. Data completion: due to different data collection frequencies, data may be missing. For missing values, interpolation is used to complete them. For example, for environmental temperature data, linear interpolation is used based on adjacent time temperature; for personnel gathering data, weighted average calculation is used to complete the data based on the time points before and after. 3. Construct sequence: sort the two types of data after time stamp unification and data completion by time to construct a time series data set. With time as the index, combine the personnel behavior (such as the number of people gathered, the time of stay) and environmental data (such as temperature, humidity, and carbon dioxide concentration) at each time to facilitate analysis of data correlation at different times.

[0192] The spatial dimension integration is specifically as follows: 1. Position matching: determine the geographical position of personnel behavior and environmental data. Personnel behavior data is determined by monitoring equipment and Bluetooth positioning; environmental data is determined by sensor installation point positioning. Accurately match the two types of data in the same area, such as personnel monitoring and temperature and humidity sensor data in platform A area. 2. Range division: for data that cannot be accurately positioned, divide the space range. For example, for a large station hall, divide it into sub-areas, combine monitoring pictures and sensor coverage range, and correspond personnel behavior and environmental data to the corresponding sub-area. 3. Fusion structure: create a fusion space data structure, use GIS feature class concepts, and organize personnel behavior and environmental data according to spatial position. Set corresponding personnel behavior and environmental attribute fields for each spatial position (point or polygon) to realize synchronous integration in the spatial dimension.

[0193] Step Sg00, according to the identified different personnel action postures, adapt individualized carbon emission coefficients for each individual.

[0194] Action posture and carbon emission correlation analysis is specifically as follows: 1. Detailed action classification: according to personnel behavior data, subdivide action postures, such as standing, subdivided into static standing, standing with luggage, etc.; walking is divided into slow, normal, fast walking, and pushing luggage cart walking, etc. according to speed and whether to push a cart. Use image recognition and motion analysis algorithms such as OpenPose algorithm to accurately identify postures based on human joint motion. 2. Explore carbon emission correlation: through experimental measurement, theoretical derivation, and reference to research results, determine the energy consumption and carbon emission of each posture. For example, simulate different walking speeds in the laboratory to measure carbon dioxide emissions. At the same time, consider the influence of air flow, temperature and humidity in rail transit on energy consumption, and establish a relationship model between environmental factors and carbon emission correction coefficients.

[0195] Adapt the personalized carbon emission coefficient, as follows: 1. Individual action recognition and tracking: use multi-camera fusion and human recognition algorithms to monitor individual actions in abnormal areas. Give each individual a unique identifier and build an action information library to record their posture, position, and duration at different times. 2. Use adaptive algorithms: based on the results of action classification and carbon emission correlation, design an algorithm. This algorithm matches the corresponding coefficient from the carbon emission coefficient database according to the individual's real-time actions, and adjusts the coefficient considering the individual's physical characteristics.

[0196] Step Sh00, input environmental factor data into the carbon emission coefficient dynamic correction factor analysis model, output dynamic correction factor, adjust carbon emission coefficient.

[0197] Carbon emission coefficient dynamic correction factor analysis model operation involves the following: 1. Model operation: input the sorted environmental data into the carbon emission coefficient dynamic correction factor analysis model. This model is based on a large amount of historical data and research results on the relationship between environment and carbon emissions, and operates internally through complex algorithms and logic. For example, the model may use machine learning algorithms to analyze the correlation between input environmental data characteristics and carbon emission coefficient adjustment. 2. Output correction factor: after model operation, output one or more dynamic correction factors. These factors reflect the adjustment range of the basic carbon emission coefficient under current environmental factors. For example, in a high-temperature and low-airflow-speed environment, the model's output correction factor may increase the basic carbon emission coefficient by 10%, indicating that carbon emissions from personnel activities in this environment are relatively increased.

[0198] Carbon emission coefficient adjustment involves the following processes: 1. Coefficient update: apply the dynamic correction factor output by the model to the personalized carbon emission coefficient adapted for each individual in step Sg00. Adjust through multiplication or other established operation rules. For example, if an individual's original carbon emission coefficient is 0.5 and the correction factor is 1.1, the adjusted carbon emission coefficient becomes 0.5 x 1.1 = 0.55. 2. Real-time adjustment mechanism: as environmental factors change in real time, continuously repeat the above steps to ensure that the carbon emission coefficient can be updated in a timely manner according to the latest environmental conditions, providing support for subsequent accurate calculation of total personnel flow carbon emissions.

[0199] Step Si00, multiply the personalized carbon emission coefficient of all personnel in the location area by their corresponding residence time, and then accumulate and sum to obtain the real-time total personnel flow carbon emissions in the location area.

[0200] Data Preparation: 1. Confirm individual information and coefficients: Sort out all monitored individual information in the abnormal position area, including the unique identification assigned to each individual and the personalized and dynamically adjusted carbon emission coefficient determined in steps Sg00 and Sh00. For example, in a certain subway station, the system records the information of 100 individuals, each with a corresponding unique ID, such as P001, P002, etc., and their latest carbon emission coefficients, such as individual P001's carbon emission coefficient is 0.6 and P002's carbon emission coefficient is 0.55.

[0201] Extracting stay time data: From the personnel behavior state data, accurately extract the stay time of each individual in the position area. These stay time data record the activity duration of individuals in the area, which is indispensable for calculating the total carbon emission. For example, individual P001 stays in the area for 10 minutes and P002 stays for 8 minutes. Ensure the accuracy and completeness of the stay time data to avoid calculation bias due to data loss or errors.

[0202] The calculation process is as follows: 1. Individual carbon emission calculation: For each individual, multiply its personalized carbon emission coefficient by the corresponding stay time to get the carbon emission of that individual in that position area. Use the formula: Individual carbon emission = carbon emission coefficient x stay time. For example, individual P001's carbon emission coefficient is 0.6 and stay time is 10 minutes, so its carbon emission is 0.6 x 10 = 6 (unit: kilogram of carbon dioxide equivalent, assuming the unit of carbon emission coefficient is kilogram of carbon dioxide equivalent / minute, and the following is the same); individual P002's carbon emission coefficient is 0.55 and stay time is 8 minutes, so its carbon emission is 0.55 x 8 = 4.4.

[0203] 2. Total accumulation sum: Add up the carbon emissions of all individuals in the position area to get the real-time personnel flow carbon emission total of the position area. Use the formula: Personnel flow carbon emission total = Σ(individual carbon emission coefficient x individual stay time). For example, add up the carbon emissions of the above 100 individuals one by one, assuming that after calculation, the total carbon emission of the first 99 individuals is 300, plus the carbon emission of individual P100, which is 5, the final real-time personnel flow carbon emission total of the position area is 300 + 5 = 305 kilogram of carbon dioxide equivalent.

[0204] Step Sj00, on the fault influence range analysis map, use the preset visualization method to display the personnel flow carbon emission data.

[0205] Based on the format and characteristics of the fault impact range analysis map, select an appropriate visualization tool. If the analysis map is generated based on Geographic Information System (GIS) software, the software's visualization functions can be used, such as ArcGIS's symbolization and annotation tools. If the analysis map is displayed in web page format, JavaScript visualization libraries such as D3.js and Echarts can be used.

[0206] Determine the visualization techniques to be used, such as using heat maps to display carbon emission intensity in different regions, using bar charts to present changes in total carbon emissions over different time periods, and using scatter plots to show the relationship between per capita carbon emissions and population density. Based on the characteristics of the data and the purpose of the presentation, select the visualization technique that best conveys the information.

[0207] Furthermore, considering that when the total carbon emissions from population movement in a certain area increase sharply in a short period of time and exceed the preset high carbon emission threshold, triggering personnel relocation, carbon emission control also needs to be considered during personnel relocation, specifically as follows:

[0208] Step Sk00: Analyze whether the total carbon emissions from personnel movement in the location area within a preset time period exceed the preset total carbon emissions.

[0209] Step SL00: If not, maintain the original settings.

[0210] If the total carbon emissions from personnel movement in the location area do not exceed the preset total carbon emissions within the preset time period, it indicates that the current carbon emissions are within a relatively controllable range. At this time, the original personnel scheduling and resource allocation settings should be maintained, and routine monitoring of carbon emissions should continue.

[0211] Step Sm00, if yes, then by using high-precision personnel flow sensors and positioning systems, determine the number of people in different areas, personnel flow trends, and the urgency of personnel's safety protection (such as lighting and ventilation needs), and by using carbon emission monitoring equipment and energy consumption monitoring devices, obtain existing carbon emission data for each area and energy consumption data for various emergency resources (lighting equipment, ventilation equipment, etc.).

[0212] Step Sn00: Construct a matrix with regions as rows and emergency resources as columns. Each element in the matrix represents the benefit value generated by allocating a certain type of resource to a specific region.

[0213] Each row of this matrix represents a specific area, and each column represents an emergency resource, such as lighting equipment, ventilation equipment, etc.

[0214] Benefit value quantification: Each element in the matrix represents the benefit value generated by allocating a certain type of resource to a specific area. When quantifying benefit value, three key factors are considered comprehensively: the degree to which personnel needs are met, the increase in carbon emissions, and energy consumption.

[0215] The quantification benefit values are as follows: for the personnel demand satisfaction degree, the proportion of the number of personnel in the region to the total number of personnel, the demand urgency, and other factors can be quantified; for the carbon emission increase, the carbon emission data during equipment operation is directly referenced; the energy consumption can also be quantified according to the equipment energy consumption monitoring data. By setting different weights, the benefit values of each element are calculated comprehensively. For example, the personnel demand satisfaction degree weight is set to 0.5, the carbon emission increase weight is set to 0.3, and the energy consumption weight is set to 0.2. The benefit value is calculated by the formula (benefit value = 0.5 x personnel demand satisfaction degree score + 0.3 x carbon emission increase score + 0.2 x energy consumption score).

[0216] Step So00, input the constructed resource allocation matrix into the Hungarian algorithm model. The Hungarian algorithm determines the optimal resource allocation scheme by constantly searching for independent zero elements in the matrix (i.e., a combination of one zero element in each row and column).

[0217] In this process, the algorithm performs a series of transformation operations on the matrix, such as row transformation and column transformation, to increase the number of zero elements and find the optimal match.

[0218] After the algorithm runs, an optimal resource allocation scheme is obtained. This scheme clearly shows which types of emergency resources are allocated to which regions to meet personnel safety needs while maximizing carbon emission control. For example, the results may show that several low-energy, low-carbon emission ventilation devices are allocated to the waiting area where personnel are relatively concentrated and carbon emission is high, which can meet the personnel's demand for ventilation and effectively control carbon emission; high-brightness, medium-energy lighting devices are allocated to key evacuation channels to ensure personnel safety while minimizing unnecessary carbon emission.

[0219] Step Sp00, according to the results obtained by the Hungarian algorithm, send the deployment task information to relevant personnel.

[0220] These information details the types and quantities of emergency resources that need to be deployed, the purpose of the deployment area, and the time requirements for deployment, etc. Then, these deployment task information is sent to relevant personnel, such as equipment maintenance personnel, dispatchers, etc.

[0221] After receiving the deployment task information, the relevant personnel immediately organize the action. They quickly take out the corresponding emergency resources from the equipment storage area and transport the resources to the designated area according to the specified route and time requirements. Upon arrival at the destination, they follow the installation requirements and debugging standards of the equipment to accurately install and debug, ensuring that the equipment can operate normally, thereby achieving effective control of personnel safety and carbon emission.

[0222] Based on the same inventive concept, the embodiment of the present application provides a rail transit carbon emission management system, which comprises a memory and a processor, the memory is stored with a program capable of being run on the processor to realize the method as Figs. 1 to 3 The program of any method.

[0223] The embodiments of the specific implementation are the preferred embodiments of the present application, but do not limit the protection scope of the present application, so that: any equivalent changes made according to the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. A rail transportation carbon emission management method, characterized by, The method comprises the following steps: acquiring carbon emission related multi-source heterogeneous data, and performing data preprocessing, wherein the multi-source heterogeneous data comprises internal data of a rail transit system and external related department data; extracting carbon emission time series data from the carbon emission related multi-source heterogeneous data after data preprocessing, and using a preset seasonal decomposition method to strip periodic factor influence, and decomposing the time series into a trend item, a seasonal item and a residual item; based on the trend item, the seasonal item and the residual item time series information after decomposition, extracting preset deep features, wherein the deep features extracted from the trend item are the slope and curvature of the trend item in a long time period, the slope is obtained by fitting the trend item curve through a linear regression method, a positive slope indicates that the overall carbon emission presents an upward trend, and a negative slope indicates that the overall presents a downward trend; the curvature reflects the acceleration or deceleration of the trend change, and an increase in the curvature indicates that the carbon emission rising or falling speed is accelerated; the deep features extracted from the seasonal item are the peak and valley of the seasonal item; the deep features extracted from the residual item are the standard deviation and variance of the residual item; inputting the extracted deep features into a pre-constructed ARIMA model to predict the numerical range of the trend item, the seasonal item and the residual item of the carbon emission in a future preset time range as the ARIMA model prediction value; comparing the ARIMA model prediction value with real-time carbon emission related data to calculate the prediction deviation of each component, including the carbon emission trend item, the seasonal item and the residual item; inputting the prediction deviation of each component into a pre-constructed carbon emission early warning analysis model to output the carbon emission abnormal risk level and the abnormal reason of this time, and the carbon emission abnormal risk level includes a low risk level, a medium risk level and a high risk level; if the carbon emission abnormal risk level of this time is a low risk level, the current overall carbon emission value, the deviation amplitude of the current overall carbon emission value and the prediction value are displayed on a visual interface, and a trend arrow is used to indicate the trend of the carbon emission in a future preset time range; if the carbon emission abnormal risk level of this time is a medium risk level, the visual interface displays the detailed prediction deviation curve and the dynamic change process of the deviation of each component; if the carbon emission abnormal risk level of this time is a high risk level, the visual interface displays the detailed prediction deviation curve and the dynamic change process of the deviation of each component, and simultaneously warns the high risk level and the abnormal reason.

2. The rail transit carbon emission management method according to claim 1, wherein if the carbon emission abnormal risk level of this time is a medium risk level or a high risk level, the method further comprises the following steps: acquiring environmental monitoring satellite data, wherein the environmental monitoring satellite data comprises regional meteorological dynamics and geographic information; from the environmental monitoring satellite data, extracting a feature vector related to rail transit carbon emission, taking the feature vector related to rail transit carbon emission as an input object, inputting into a pre-constructed comprehensive model for simulating carbon emission changes under different emergency scenarios, and outputting simulation results; according to the output simulation results, drawing a rail transit carbon emission change curve over time in a future preset time range, and marking emergency scenario related information. ​ 3. The rail transit carbon emission management method of claim 2, wherein, The feature vector related to the carbon emission of the rail transit is extracted from the environmental monitoring satellite data according to the feature selection standard. Obtain the carbon emission abnormal risk level this time; According to the real-time satellite data feature fluctuation information determined by analyzing the environmental monitoring satellite data, input the carbon emission abnormal risk level and the real-time satellite data feature fluctuation information into the pre-trained feature selection standard analysis model, and output the feature selection standard. According to the feature selection standard, the feature vector related to the carbon emission of the rail transit is extracted from the environmental monitoring satellite data.

4. The rail transit carbon emission management method of claim 2, wherein, The feature vector related to the carbon emission of the rail transit is input into the pre-constructed comprehensive model for simulating the change of carbon emission under different emergency scenarios as an input object, and the simulation result is output, including: Analyze whether to obtain three-dimensional terrain model information scanned by laser radar devices deployed on the top of the train or at key points along the line, and environmental parameters detected by a micro dust sensor network deployed along the rail transit line and station; If not, the feature vector related to the carbon emission of the rail transit is input into the pre-constructed comprehensive model for simulating the change of carbon emission under different emergency scenarios as an input object, and the simulation result is output; If yes, the laser radar feature vector is extracted from the obtained three-dimensional terrain model information, and the obtained environmental parameters are calibrated and integrated to extract the intelligent micro dust sensor network feature vector, wherein the intelligent micro dust sensor network feature vector refers to the key indicators with strong correlation with the carbon emission of the rail transit in the rail transit scene, which are screened from the calibrated environmental parameters, arranged in a reasonable order, and formed into a vector; The laser radar feature vector, the intelligent micro dust sensor network feature vector and the feature vector extracted based on the environmental monitoring satellite data are fused to form a fusion vector; According to the feature vector contained in the fusion vector and the preset comprehensive model architecture quantitative adjustment method, the architecture of the comprehensive model is adjusted, and the architecture of the comprehensive model refers to the structure of the comprehensive model for simulating the change of carbon emission under different emergency scenarios, including the type of mathematical model used, the connection relationship between each component and the parameter setting. The adjustment of the comprehensive model architecture is specifically adjusting the number of layers, the number of neurons and the parameter weight of the model according to the feature vector contained in the fusion vector by using the preset comprehensive model architecture quantitative adjustment method, including adjusting the number of layers, the number of neurons and the parameter weight of the model. The adjustment method of the comprehensive model architecture is as follows: when the feature of the fusion vector is complex, the number of layers is increased, the number of neurons is appropriately increased for the layer corresponding to the part closely related to the carbon emission and rich in feature change, and the weight of the existing parameters of the model is redistributed according to the importance and correlation analysis result of the feature of the fusion vector, so as to adjust the architecture of the comprehensive model. The fusion vector is input into the comprehensive model for simulating the change of carbon emission under different emergency scenarios with the adjusted architecture, and the simulation result is output.

5. The rail transit carbon emission management method of claim 4, wherein, Meanwhile, the high-risk level and the abnormal reason are warned, including: Analyze whether to obtain abnormal sound information captured by an acoustic sensor array deployed on the rail track and key equipment, including train traction motor and brake system; If no, the original setting is maintained; If yes, acoustic features are extracted from the abnormal sound information, and core sound features are extracted from the acoustic features through a dimension reduction method of principal component analysis; The core sound features and the prediction bias of each component are input into the abnormal sound diagnosis model to output the abnormal reason; The abnormal reason of the warning is updated according to the output, and the high risk level is also displayed.

6. The rail transit carbon emission management method of claim 5, wherein, While performing the step of updating the abnormal reason of the warning according to the output and displaying the high risk level, the following steps are also performed: Analyze the objects and abnormal positions associated with the abnormal reason, wherein the objects include train running tracks and key equipment; If the abnormal reason is associated with the train running track, obtain the three-dimensional terrain model information and environmental parameters of the location area where the abnormal position is located, and according to the combination of the train running track associated with the abnormal reason, the three-dimensional terrain model information of the location area where the abnormal position is located, and the environmental parameters, match the track fault influence range from the pre-constructed track fault influence model, and obtain the fault influence range analysis diagram; If the abnormal reason is associated with key equipment, obtain the three-dimensional terrain model information and environmental parameters of the location area where the abnormal position is located, and according to the combination of the key equipment associated with the abnormal reason, the three-dimensional terrain model information of the location area where the abnormal position is located, and the environmental parameters, match the key equipment fault influence range from the pre-constructed fault propagation model, and obtain the fault influence range analysis diagram; The environmental parameters refer to the various specific index values reflecting the physical properties of the environment along the rail transit line and the station detected by the dust sensor network deployed along the rail transit line and the station, including particulate matter concentration in the air, humidity in the air, and temperature in the air; Obtain the passenger flow distribution data of the location area where the abnormal position is located, and use dynamic visualization technology to real-time superimposed display the personnel flow in the fault influence range analysis diagram.

7. The rail transit carbon emission management method of claim 6, wherein, The step of real-time superimposed display the personnel flow in the fault influence range analysis diagram is as follows: Obtain the personnel behavior state data and environmental factor data of the location area where the abnormal position is located, wherein the personnel behavior state data includes action posture, stay time and aggregation mode, and the environmental factors include temperature, humidity, air flow speed and carbon dioxide concentration; Synchronously integrate the personnel behavior state data and environmental factor data according to time and space dimensions; According to the identified different personnel action postures, adapt personalized carbon emission coefficients for each individual, wherein the carbon emission coefficient is a quantitative index for measuring the carbon emission amount of individual personnel activities per unit time in the rail transit scene, considering personnel action postures, individual physical characteristics and environmental factors; Input the environmental factor data into the carbon emission coefficient dynamic correction factor analysis model to output the dynamic correction factor and adjust the carbon emission coefficient; Multiply the personalized carbon emission coefficients of all personnel in the location area by their corresponding stay time, and then accumulate and sum to obtain the real-time personnel flow carbon emission total amount of the location area; On the fault influence range analysis diagram, the personnel flow carbon emission data is displayed in a preset visual manner.

8. A rail transportation carbon emission management system, characterized by, The computer readable storage medium stores a program, and the program can be loaded and executed by the processor to implement the rail transit carbon emission management method according to any one of claims 1 to 7.

Images