A dynamic tracking system and method for railway transportation carbon footprint based on blockchain and edge computing

Abstract

This invention relates to the field of dynamic carbon footprint technology, specifically to a dynamic carbon footprint tracking system and method for railway transportation based on blockchain and edge computing. It includes a data acquisition unit for collecting train operating status, environmental parameters, and traction energy consumption data; an edge computing unit for fusing the multi-source data collected by the data acquisition unit using adaptive Kalman filtering to obtain standardized data, converting the train operating status into carbon emissions based on an instantaneous power model, and calculating cumulative carbon emissions by introducing a dynamic emission factor; and a blockchain network unit for compressing the cumulative carbon emission data and generating a Merkle root hash for on-chain storage, storing the carbon emission data on the blockchain. This achieves incremental on-chain data storage and rapid verification, improves on-chain data processing efficiency and query traceability capabilities, and solves the problems of high storage pressure and low verification efficiency in traditional on-chain evidence storage methods under high-frequency rail data writing scenarios.

Description

Technical Field

[0001] This invention relates to the field of dynamic carbon footprint technology, and more specifically, to a dynamic carbon footprint tracking system and method for railway transportation based on blockchain and edge computing. Background Technology

[0002] Emissions data statistics often rely on periodic aggregation and macro-level estimation, resulting in low data granularity and poor timeliness. Dynamic tracking can monitor emissions at the second or minute level, enabling operators to understand the emission characteristics of different lines and operating conditions in a timely manner, facilitating targeted optimization and adjustments. However, traditional static rules are untimely in response to sudden high-carbon emission events and have a high false alarm rate; moreover, traditional on-chain evidence storage methods face significant storage pressure in high-frequency rail data writing scenarios. Therefore, this paper proposes a dynamic carbon footprint tracking system and method for railway transportation based on blockchain and edge computing. Summary of the Invention

[0003] The purpose of this invention is to provide a dynamic tracking system and method for railway transportation carbon footprint based on blockchain and edge computing, in order to solve the problems mentioned in the background art, such as untimely response and high false alarm rate to sudden high carbon emission events under traditional static rules; and the high storage pressure of traditional on-chain evidence storage methods in high-frequency track data writing scenarios.

[0004] To achieve the above objectives, on the one hand, the present invention aims to provide a dynamic tracking system for the carbon footprint of railway transportation based on blockchain and edge computing, comprising:

[0005] A data acquisition unit is used to collect train operating status, environmental parameters, and traction energy consumption data.

[0006] The edge computing unit fuses multi-source data collected by the data acquisition unit using adaptive Kalman filtering to obtain standardized data, and converts the train's operating status into carbon emissions based on an instantaneous power model, while also introducing a dynamic emission factor. Calculation of cumulative carbon emissions ;

[0007] A blockchain network unit is used to compress the cumulative carbon emission data and generate a Merkle tree root hash for on-chain storage, thereby storing the carbon emission data on the blockchain.

[0008] The dynamic tracking unit, based on the LSTM prediction model and correcting on-chain data and real-time edge computing data, employs the NSGA-II multi-objective genetic algorithm with an elitist retention strategy for global optimization to generate a Pareto solution set. and Pareto solution set The data is then distributed to the edge computing unit for closed-loop optimization.

[0009] As a further improvement to this technical solution, the data acquisition unit includes an on-board sensor module, a track environment monitoring module, and an energy consumption metering module;

[0010] Among them, the on-board sensor module is used to monitor train operating energy consumption, load, speed and location data;

[0011] The track environment monitoring module is used to collect track gradient, weather conditions, and signal system status parameters;

[0012] The energy consumption metering module is used to collect energy consumption data from the traction power supply system and stations.

[0013] As a further improvement to this technical solution, the edge computing unit includes a local preprocessing node module, a distributed computing module, and an edge prediction module;

[0014] The local preprocessing node module uses adaptive Kalman filtering to fuse the multi-source data collected by the data acquisition unit to obtain preprocessed standardized data.

[0015] The distributed computing module uses preprocessed standardized data and an instantaneous power model to convert the train's operating status into carbon emissions.

[0016] Among them, the train's time is calculated based on the instantaneous power model. Instantaneous total power and improve traction system efficiency Introducing instantaneous total power Obtain cumulative energy consumption Meanwhile, based on real-time emission factors Cumulative energy consumption This translates into cumulative carbon emissions.

[0017] Furthermore, the instantaneous power model is composed of air resistance power, slope resistance power, and inertial power coupled together;

[0018] The edge prediction module predicts carbon emissions based on a carbon emission prediction model. It also outputs Boolean anomaly flags by introducing an anomaly detection auxiliary branch. .

[0019] As a further improvement to this technical solution, the specific steps involved in the edge prediction module are as follows:

[0020] A carbon emission prediction model was constructed using a prediction model based on long short-term memory networks and attention mechanisms to predict carbon emission levels in future periods.

[0021] The preprocessed, standardized data is used as input to the carbon emission prediction model, and the output is the predicted carbon emission value. ;

[0022] An anomaly detection auxiliary branch, constructed using binary classification logic, is introduced into the carbon emission prediction model, outputting a Boolean anomaly flag. and prediction confidence value ;

[0023] Finally, the carbon emission prediction model packages the regression and classification results into a triple. .

[0024] As a further improvement to this technical solution, the blockchain network unit includes a blockchain storage module and a blockchain early warning module;

[0025] The blockchain storage module is used to hash and compress carbon emission data from edge computing units according to time windows, construct dynamic sharded Merkle trees, and write them into the blockchain to ensure data tamper-proof and traceability.

[0026] The Merkle tree uses a dynamic sharding structure, and the sharding logic is based on the division of train operating sections.

[0027] The blockchain early warning module uses on-chain carbon emission data and a threshold judgment algorithm to determine thresholds and trigger early warning events, providing timely responses for operation and maintenance and scheduling.

[0028] As a further improvement to this technical solution, the dynamic segmentation rule for the train operating section is as follows:

[0029] The railway network is composed of It consists of several stations, and the tracks between adjacent stations form a basic operating section. :

[0030] ;

[0031] in, Indicates the starting station of the interval; Indicates the terminal station of the section; Represents the smallest geographic coordinate boundary of the interval; Indicates the maximum geographic coordinate boundary of the interval;

[0032] Each interval Corresponding to a dynamic partition :

[0033] ;

[0034] In the formula, Indicates the slice key, derived from the running interval Location and time window Sure; This represents the carbon emissions dataset stored within the partition; Indicates the initial capacity threshold for fragmentation; Indicates the segmentation threshold; Indicates the threshold for merging fragments;

[0035] Based on train location and time window Determine the sharding key ;

[0036] As the train enters the new section At that time, continuously check whether its real-time position and timestamp meet the new shard creation logic. If they do, trigger the creation of a new shard. create;

[0037] When fragmentation The number of data records stored in the middle exceeds the split threshold. When this happens, a split operation is triggered;

[0038] When adjacent fragments and If both low load and physical adjacency are met, a merge operation is triggered, and adjacent shards are merged. and merged into .

[0039] As a further improvement to this technical solution, the specific steps involved in constructing the Merkle tree are as follows:

[0040] For each fragment Carbon emission data blocks in Calculate the hash of the leaf node ;

[0041] Hash all leaf nodes within the shard. Sort by time, pair them up to generate parent node hashes, and repeat the above steps until a unique shard root hash is generated. ;

[0042] hash all shard roots As leaf nodes, construct the global Merkle tree.

[0043] As a further improvement to this technical solution, the specific steps involved in the threshold judgment algorithm determining the threshold and triggering the warning event in the blockchain early warning module are as follows:

[0044] The blockchain early warning module monitors newly added shard root hashes on the chain. ;

[0045] The carbon emission dataset within the fragment was located using on-chain pointer indexes and the Merkle path. ;

[0046] Traversing the carbon emission dataset For each data point, if the actual carbon emissions... Exceeding the set threshold is recorded as an abnormal event. ;

[0047] At the same time, when new anomalies are to be evaluated, the most recent ones are traced back. On the chain at that moment sequence;

[0048] Anomalies are determined based on a threshold-based algorithm combined with historical on-chain data.

[0049]

[0050] In the formula, Indicates the first Detected carbon emission anomalies in real time; Indicates the first The actual carbon emissions detected at any given time; Indicates the current train operating section Historical average carbon emissions; Indicates at time The probability coefficient of continuous anomalies on the chain is a weighting index of continuous anomalies in historical periods, reflecting whether the current anomaly is in a period of continuous anomalies (a value close to 1 is more likely to be a continuous anomaly). Indicates in The isolated anomaly penalty coefficient at any given time is used to suppress the effects of random or sporadic mutation events; This represents the weighting coefficient for the historical deviation term, used to measure the strength of the impact of the difference between the current emission value and the historical average. This represents the weighting coefficient for continuous anomaly detection; This represents the isolated outlier discrimination weight coefficient, used to control the degree of influence of sudden outliers on the judgment result; and ;

[0051] like Exceeding the preset alarm threshold If the exception is entered into the summary data structure, then the exception will be included in the summary data structure. ;

[0052] Among them, after the abnormal event is confirmed, it is encapsulated as an off-chain early warning record into a summary data structure. ;

[0053] Abstract data structure Once uploaded to the blockchain, the data is synchronously sent to the scheduling center, triggering an operations and maintenance response.

[0054] As a further improvement to this technical solution, the dynamic tracking unit includes a collaborative correction module, a multi-objective optimization decision module, a full life cycle carbon integration module, and a closed-loop strategy distribution module.

[0055] Among them, the collaborative correction module is based on the summary data structure. Triples Based on historical carbon emission sequences, an LSTM prediction model was used to assess the efficiency of the traction system. With real-time emission factors To conduct short-term forecasting and correction;

[0056] The multi-objective optimization decision module uses edge computing latency, energy consumption, and on-chain overhead as objective functions, comprehensively considering the sampling interval. Edge computing quota and on-chain frequency The NSGA-II multi-objective genetic algorithm with an elitist retention strategy is used for global optimization to generate the Pareto solution set. ;

[0057] The life-cycle carbon integration module calculates the comprehensive carbon emissions based on the carbon footprint extension model. If comprehensive carbon emissions are detected On-chain carbon trading is triggered when the threshold is exceeded. The carbon footprint extension model is based on cumulative carbon emissions. It was constructed by incorporating indirect carbon emission factors generated during the use of infrastructure;

[0058] The closed-loop strategy distribution module is used to resolve Pareto data. Distribute to edge computing units.

[0059] On the other hand, the present invention provides a method for dynamic tracking of railway traffic carbon footprint based on blockchain and edge computing, used in any of the above-mentioned dynamic tracking system for railway traffic carbon footprint based on blockchain and edge computing, comprising the following steps:

[0060] S10.1 The edge computing unit fuses the multi-source data collected by the data acquisition unit, and calculates the real-time energy consumption of the train by combining the instantaneous power model, introducing the dynamic emission factor, and converting it into cumulative carbon emissions.

[0061] S10.2 The blockchain network unit compresses and encodes data from edge computing, constructs the hash structure of carbon emission data blocks through a dynamic sharded Merkle tree, and uploads them to the chain to achieve traceable evidence storage; when a train crosses an operating section or time window, it automatically triggers the creation, splitting, or merging of new shards.

[0062] Meanwhile, by monitoring newly added shard hashes on the chain, abnormal events are identified based on threshold judgment algorithms combined with historical data, and the abnormal summary data structure is uploaded to the chain and pushed to the scheduling center to trigger operation and maintenance response;

[0063] S10.3. The on-chain data and edge prediction results are collaboratively corrected by the dynamic tracking unit. The Pareto solution set is generated by the NSGA-II multi-objective optimization algorithm. After comprehensively considering latency, energy consumption and on-chain overhead, the optimal strategy is fed back to the edge computing unit by the closed-loop strategy distribution module to achieve full-process optimization.

[0064] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0065] 1. In a dynamic tracking system and method for railway traffic carbon footprint based on blockchain and edge computing, a dynamic sharded Merkle tree is constructed based on blockchain, and carbon emission data for different operating intervals are sharded and hashed for storage, so as to realize cross-interval traceability and historical auditing.

[0066] Meanwhile, the system constructs shards based on train operating sections and time windows, supporting splitting, merging, and archiving. Based on the dynamic sharding mechanism of train operating sections and time windows, and combined with the Merkle tree structure, carbon emission data is sharded and hashed for storage, enabling incremental data uploading and rapid verification. This improves on-chain data processing efficiency and query traceability capabilities, solving the problems of high storage pressure and low verification efficiency in traditional on-chain storage methods under high-frequency rail data writing scenarios.

[0067] 2. In a dynamic tracking system and method for railway traffic carbon footprint based on blockchain and edge computing, a carbon emission prediction model based on LSTM and attention mechanism is constructed, historical operation trajectory is integrated and an auxiliary binary classification detection branch is introduced, and dynamic threshold is controlled by prediction confidence to improve the sensitivity and accuracy of abnormal event identification. This solves the problems of untimely response and high false alarm rate of sudden high carbon emission events under traditional static rules.

[0068] 3. In a dynamic tracking system and method for railway traffic carbon footprint based on blockchain and edge computing, the NSGA-II multi-objective optimization algorithm with an elite retention strategy is used to perform global optimization among sampling interval, edge computing latency and on-chain overhead, generate Pareto solution set and combine it with resource status to dynamically issue scheduling strategy, which significantly enhances the system's adaptability to resource fluctuations and network congestion, and solves the problems of resource allocation imbalance and high response latency under the single objective scheduling mechanism. Attached Figure Description

[0069] Figure 1 This is a flowchart illustrating the overall process of the present invention.

[0070] The meanings of the labels in the diagram are as follows:

[0071] 1. Data acquisition unit;

[0072] 2. Edge computing unit; 21. Local preprocessing node module; 22. Distributed computing module; 23. Edge prediction module;

[0073] 3. Blockchain network unit; 31. Blockchain storage module; 32. Blockchain early warning module;

[0074] 4. Dynamic tracking unit; 41. Collaborative correction module; 42. Multi-objective optimization decision-making module; 43. Full life cycle carbon integration module; 44. Closed-loop strategy distribution module. Detailed Implementation

[0075] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0076] Example 1: Please refer to Figure 1 As shown, a dynamic tracking system for railway traffic carbon footprint based on blockchain and edge computing is provided, including a data acquisition unit 1, which is used to collect train operating status, environmental parameters and traction energy consumption data.

[0077] The data acquisition unit 1 includes an on-board sensor module, a track environment monitoring module, and an energy consumption metering module.

[0078] Among them, the on-board sensor module is used to monitor train operating energy consumption, load, speed and location data.

[0079] The track environment monitoring module is used to collect track gradient, weather conditions, and signal system status parameters.

[0080] The energy consumption metering module is used to collect energy consumption data from the traction power supply system and stations.

[0081] In this embodiment, the railway traffic carbon footprint dynamic tracking system based on blockchain and edge computing also includes an edge computing unit 2. The edge computing unit 2 fuses the multi-source data collected by the data acquisition unit 1 through adaptive Kalman filtering to obtain standardized data, and converts the train operation status into carbon emissions based on the instantaneous power model, and introduces a dynamic emission factor. Calculation of cumulative carbon emissions .

[0082] The dynamic emission factor model, based on the LSTM time series prediction algorithm, involves the following construction process:

[0083] The historical multi-source standardized data obtained from preprocessing in the edge computing unit is organized into a fixed-length time series window and used as the model input sequence.

[0084] An LSTM layer is followed by a fully connected regression output layer to predict the emission factor at the next time step.

[0085] The network loss function uses a weighted average absolute error to improve prediction stability near outliers;

[0086] In the prediction phase, the model uses the latest... The feature sequence of each time step is taken as input, and the predicted emission factor at the current time is output.

[0087] Subsequently, the predicted values ​​are corrected using carbon emission observation data and energy consumption data within the historical sliding window to obtain a calibration emission factor. This dynamic emission factor model can continuously predict changes in grid carbon emission intensity during train operation and calculate dynamic carbon emissions by combining traction energy consumption. It is suitable for scenarios such as carbon footprint calculation, early warning judgment, and carbon trading response.

[0088] Specifically, the LSTM model is used to predict real-time emission factors, and a historical sliding window is introduced for error backtracking and bias self-calibration. Combined with the sliding variance analysis of traction efficiency, regional carbon intensity adaptation is achieved, which effectively suppresses the bias caused by regional power grid fluctuations or data drift in carbon emission estimation and solves the problem that fixed emission factors are difficult to adapt to different lines and time periods.

[0089] In this embodiment, the edge computing unit 2 includes a local preprocessing node module 21, a distributed computing module 22, and an edge prediction module 23.

[0090] The local preprocessing node module 21 uses adaptive Kalman filtering to fuse the multi-source data collected by the data acquisition unit 1 to obtain preprocessed standardized data.

[0091] The local preprocessing node module 21 is deployed on the station's edge server to perform data cleaning, anomaly detection, and preliminary carbon emission calculation.

[0092] All parameters are aligned to a unified timestamp (GNSS clock) with millisecond-level accuracy.

[0093] The distributed computing module 22 uses the preprocessed standardized data and an instantaneous power model to convert the train's operating status into carbon emissions.

[0094] Among them, the train's time is calculated based on the instantaneous power model. Instantaneous total power and improve traction system efficiency Introducing instantaneous total power Obtain cumulative energy consumption Meanwhile, based on real-time emission factors Cumulative energy consumption This translates into cumulative carbon emissions.

[0095] Furthermore, the instantaneous power model is formed by coupling air resistance power, slope resistance power, and inertial power.

[0096] Wherein, the air resistance power ;

[0097] The slope resistance power ;

[0098] The inertial power ;

[0099] but .

[0100] In the formula, Indicates the static mass of the train; This represents the gravitational acceleration constant.

[0101] Operating parameters: speed acceleration Load capacity and slope angle ;

[0102] Environmental parameters: air density air drag coefficient and windward area ;

[0103] Energy efficiency parameters: Traction system efficiency .

[0104] Further:

[0105]

[0106] in, As the load on the traction motor changes, the traction system data is fed back in real time from the on-board control system. This indicates the train's traction power (unit: kW). It represents an infinitesimal time increment and is used as the time differential unit in the integration process; Indicates the start time of the tracking; This represents the time variable for integration.

[0107]

[0108] In the formula, Indicates time Instantaneous carbon emission factor (unit: kg CO2 / kWh).

[0109] The edge prediction module 23 predicts carbon emission values ​​based on a carbon emission prediction model. It also outputs Boolean anomaly flags by introducing an anomaly detection auxiliary branch. .

[0110] Specifically, the edge prediction module 23 involves the following steps:

[0111] A carbon emission prediction model is constructed using a prediction model based on a long short-term memory network (LSTM) and an attention mechanism to dynamically predict carbon emission levels for future periods.

[0112] The preprocessed, standardized data is used as input to the carbon emission prediction model, and the output is the predicted carbon emission value. ;

[0113] An anomaly detection auxiliary branch, constructed using binary classification logic, is introduced into the carbon emission prediction model, outputting a Boolean anomaly flag. and prediction confidence value ; Generate prediction confidence values ​​based on the discrimination criteria automatically learned during model training. ;

[0114] Deep modeling is performed on the operational status and carbon emission trajectory in historical time series data, and auxiliary binary classification results are output, namely, a prediction marker of whether there is an abnormal carbon emission at the current moment. This auxiliary output result, together with the timestamp, geographical location, and carbon emission value, constitutes a data structure that can be uploaded to the blockchain.

[0115] Meanwhile, to support the closed-loop design of edge-side decision-making and on-chain collaboration, the computational logic of the carbon emission prediction model is decentralized to the edge computing node and deployed locally together with the power modeling and carbon emission calculation modules, making full use of the computing power of edge resources, reducing response latency and improving system real-time performance and resilience.

[0116] Finally, the carbon emission prediction model packages the regression and classification results into a triple. .

[0117] In the formula, This represents the predicted carbon emissions of edge nodes for future periods based on their current orbital state. This indicates whether the carbon emission prediction model determines that abnormal carbon emission behavior will occur during this period. This represents the confidence level of the predicted abnormal event, used to determine credibility or alarm triggering threshold.

[0118] Edge nodes generate the above structure based on standardized multi-source data and synchronize the triples through an off-chain data synchronization mechanism. The data is uploaded to the blockchain system in a summary data structure and used as an input parameter for the blockchain early warning module 32 in the judgment process.

[0119] In this embodiment:

[0120] The preprocessed standardized data in edge computing unit 2 is organized into a fixed-length time series window and fed into the carbon emission prediction model.

[0121] In Long Short-Term Memory (LSTM) networks, for the first... Features of each moment Among them, features This represents the data vector at each moment in the input sequence (a multi-dimensional feature vector composed of preprocessed standardized data, including speed, energy consumption, slope, carbon emissions, etc.).

[0122] First, the forget gate is used to determine which historical information needs to be retained at the current moment;

[0123] Updates to new information are generated through input gates and candidate memories;

[0124] The results of the forget gate and the input gate are merged to update the internal memory state;

[0125] Finally, the hidden state at that moment is generated through the output gate. ;

[0126] Repeat the above steps until all time points have been traversed, obtaining the hidden state sequence of the entire window. .

[0127] For the hidden state at each time step Calculate attention score This is used to characterize the importance of that moment to the final prediction;

[0128] All attention scores Normalization yields a set of attention weights. ;

[0129] We use attention weights to perform a weighted summation of the hidden state sequence to obtain a comprehensive context vector. ;

[0130] context vector By inputting a linear regression layer, we can obtain the predicted carbon emissions for future times. .

[0131] In the same context vector An additional small neural network is connected to it for binary classification prediction:

[0132] First, perform a non-linear mapping (e.g., a fully connected layer activated by ReLU) to extract discriminative features;

[0133] Then, the Sigmoid function outputs a probability value between 0 and 1. This is used to indicate the confidence level of an abnormal carbon emission trend in the current period;

[0134] The probability A Boolean anomaly flag is obtained by comparing it with the preset threshold (0.7-0.8). :

[0135] If the probability value If the value is not lower than the threshold, it is marked as abnormal; otherwise, it is marked as normal.

[0136] Finally, the carbon emission prediction model packages the regression and classification results into a triple:

[0137] ;

[0138] In the formula, This indicates the prediction phase of a carbon emission prediction model, for future moments. Abnormal predictions; This represents the confidence level of the anomaly probability in the carbon emission prediction model.

[0139] This ternary structure includes accurate predictions of future carbon emissions, anomaly detection at future moments, and confidence values ​​for the probability of anomalies, providing effective data support for downstream on-chain notarization and early warning logic.

[0140] In this embodiment, the railway transportation carbon footprint dynamic tracking system based on blockchain and edge computing also includes a blockchain network unit 3. The blockchain network unit 3 is used to compress the cumulative carbon emission data and generate a Merkle tree root hash for on-chain storage, thereby storing the carbon emission data on the blockchain.

[0141] Among them, the blockchain network unit 3 includes a blockchain storage module 31 and a blockchain early warning module 32;

[0142] The blockchain storage module 31 is used to hash and compress carbon emission data from edge computing unit 2 according to time windows, construct dynamic sharded Merkle trees and write them into the blockchain to ensure the data is tamper-proof and traceable.

[0143] The Merkle tree employs a dynamic sharding structure, with sharding logic based on the division of train operating sections.

[0144] The blockchain early warning module 32 uses on-chain carbon emission data to determine thresholds and trigger early warning events based on a threshold judgment algorithm, providing timely responses for operation and maintenance and scheduling.

[0145] In this embodiment, when a high-speed train departs from section A, passes through section B, and finally arrives at section C, the system needs to dynamically track the carbon emission data of the train in the three operating sections and realize the segmented storage of evidence through blockchain.

[0146] Each operating range (such as range A) corresponds to an independent data shard, which stores the raw carbon emission data for that range.

[0147] When a train crosses a section boundary (such as entering the track under the jurisdiction of section B), a new segment is automatically created.

[0148] If the amount of data in a certain interval is too large (e.g., more than 10,000 records), it is necessary to further divide it into sub-shards.

[0149] Specifically, the dynamic segmentation rules for the runtime interval are as follows:

[0150] The railway network is composed of It consists of several stations, and the tracks between adjacent stations form a basic operating section. :

[0151] ;

[0152] in, Indicates the starting station of the interval; Indicates the terminal station of the section; Represents the smallest geographic coordinate boundary (longitude and latitude) of the interval. ; Represents the maximum geographic coordinate boundary (longitude and latitude) of the interval. ; Indicates the current operating range Minimum longitude of a geographical boundary; Indicates the current operating range The maximum longitude of the geographical boundary; Indicates the current operating range The minimum latitude of a geographical boundary; Indicates the current operating range The maximum latitude of the geographical boundary.

[0153] Each interval Corresponding to a dynamic partition :

[0154] ;

[0155] In the formula, Indicates the slice key, derived from the running interval Location and time window Sure; This represents the carbon emission dataset stored within a partition, satisfying... Its location And timestamp ; This indicates the initial capacity threshold for the partition (e.g., 1000 records). Indicates the threshold for fragmentation (e.g., 10,000 records); This indicates the threshold for merging fragments (e.g., 200 records).

[0156] Based on train location and time window Determine the sharding key Ensure that the data fragments are aligned with the operating range and data timing:

[0157] ;

[0158] in, Indicates the longitude of the train at the time the event occurred. Indicates the latitude of the train at the time the event occurred; The granularity of spatial segmentation along the longitude direction is used to divide the entire longitude range into several small grids. Indicates the granularity of spatial segmentation along the latitudinal direction; The timestamp represents the current moment; This indicates the length of the time partitioning window (e.g., 300 seconds or 5 minutes), used for time partitioning. This is the XOR operator.

[0159] As the train enters the new section At the same time, continuously check whether its real-time location (latitude and longitude coordinates) and timestamp meet the new fragment creation logic. If they do, trigger the creation of a new fragment. create.

[0160] New Segmentation The creation logic is as follows:

[0161] The train's current location's latitude and longitude are outside the geographical coverage of any existing segment, meaning it does not belong to any current segment. of and When the interval is , where, Indicates the first One valid shard instance;

[0162] The current time exceeds the time window of the current slice. (e.g., 5 minutes) Even if the train is still within the same geographical area, a new segment needs to be created;

[0163] Create new shard Initial capacity is (e.g., 1000 records), set the status to Active;

[0164] Original fragment (Such as the previous interval's fragment) is converted to Frozen (read-only);

[0165] When fragmentation The number of data records stored in the middle exceeds the split threshold. When this happens, a split operation is triggered.

[0166] The splitting operation is as follows:

[0167] Will The data is divided into two segments based on timestamps:

[0168] First half Store sub-shards ;

[0169] Second half Store sub-shards ;

[0170] Sub-partitions inherit the geographical scope of the original partition. and time window Only time boundary updates;

[0171] In the formula, Indicates a certain dynamic partition The start timestamp, which is the time when the segment begins recording carbon emission data; Indicates a certain dynamic partition The timestamp of the end of the record, that is, the time of the last piece of data before the fragment was frozen or archived; This indicates the intermediate time point used in the sharding and splitting operation, i.e., the time boundary for splitting the sharded data.

[0172] When adjacent fragments and If both low load and physical adjacency are met, a merge operation is triggered, and adjacent shards are merged. and merged into (Indicates a new partition), whose geographical scope covers the original two partitions, and its capacity is reset to [value missing]. Original fragment and Marked as Archived.

[0173] Low load conditions specifically refer to situations where the data volume of both shards is below the merging threshold. (e.g., 200 items), that is and .

[0174] Physical adjacency specifically means that the running intervals corresponding to two slices are physically adjacent, i.e. The final stop yes The starting station .

[0175] In this embodiment, for each fragment The lifecycle states are as follows:

[0176] Active: The only writable data segment, corresponding to the current operating section of the train;

[0177] Frozen (Read-only): After the train leaves the section, it becomes read-only, retaining complete data for querying;

[0178] Archived: Fragments are marked as archived after being split or merged, and only historical tracing is supported.

[0179] Furthermore, the specific steps involved in constructing the Merkle tree are as follows:

[0180] For each fragment Carbon emission data blocks in (Carbon emission data block) Includes timestamps, location, carbon emissions, and supplementary markers. Among them, the actual carbon emissions are stored by the blockchain storage module 31. (Actual carbon emissions) (Calculated by distributed computing module 22) and auxiliary flags Save as data block At the same time, historical anomaly markers are generated. If the model flag detects an anomaly or the threshold judgment detects an anomaly, then Equals 1; otherwise, An anomalous historical marker is generated when the value equals 0. (The sequence is permanently recorded on the chain), calculate the hash of the leaf nodes. :

[0181]

[0182] in, Indicates the train is at Longitude of each data point; Indicates the train is at The dimensions of each data point; Indicates the first Instantaneous carbon emissions at each data point.

[0183] Hash all leaf nodes within the shard. Sort by time and pair them to generate parent node hashes. If the number of leaf nodes is odd, the last node copies itself and participates in the calculation. Repeat the above steps until a unique shard root hash is generated. :

[0184] ;

[0185] .

[0186] hash all shard roots As leaf nodes, construct the global Merkle tree:

[0187] .

[0188] In the formula, Indicates the first Each carbon emission data block contains a timestamp, location, and carbon emission amount; Represents data block The timestamp is synchronized with the GNSS clock; This indicates a 256-bit secure hash algorithm used to generate unique hash values ​​of fixed length. This represents the hash value of the left child node of the current node; This represents the hash value of the right child node of the current node; This represents the hash value of the parent node generated by concatenating the hashes of the left and right child nodes; Indicates fragmentation Merkle tree root hash; This indicates that during the construction of the Merkle tree, for each carbon emission data block... The calculated hash value; Indicates fragmentation The last carbon emissions data block The hash value of the leaf node; The root hash of the global Merkle tree is constructed from the root hashes of all the shards; Indicates the total number of fragments; Indicates the first Merkle root hash of each slice; This represents the Merkle tree construction function, which generates the root node through layer-by-layer hash calculation. The input is an ordered set of hashes of all leaf nodes within the shard, and the output is the root hash value.

[0189] Root hash of each shard Root hash of shard It participates in the global Merkle tree construction as an independent unit, supporting the dynamic creation, merging, or splitting of segments (such as generating new segments when a train enters a new section).

[0190] When a new shard is created or an old shard is modified, the global Merkle tree only needs to update the affected branches.

[0191] In this embodiment:

[0192] The specific steps involved in data sharding are as follows:

[0193] Interval A: Generate fragments It stores data from 14:00 to 14:30, totaling 800 records;

[0194] Section B: The train enters a new section, creating... Initial capacity: 1000 items;

[0195] Interval C: Due to a surge in data volume (1500 records), a partitioning was triggered. , .

[0196] The specific steps involved in hash-based evidence storage are as follows:

[0197] Fragmented root hash:

[0198]

[0199] Global root hash:

[0200]

[0201] In the formula, Indicates fragmentation The root hash; Indicates the first The hash value of each (e.g., the 1st, 2nd, 800th) carbon emission data block; This represents the root hash calculated independently for each shard (e.g., A, B, C1, C2), which is a digest of all data records in that shard.

[0202] Extracting interval B data:

[0203] Provide data blocks Merkle path:

[0204] ;

[0205] In the formula, Representation and data block Sibling nodes at the same level (i.e., another child node in the tree structure used to calculate its parent node); This represents the set of nodes that, when merged with the current branch, form an intermediate hash path. The root hash of shard B is the hash containing the data block. The intermediate goal; This represents the global root hash, which must ultimately be matched with the hash value recorded on the chain to confirm that the data has not been tampered with.

[0206] Calculate the hash layer by layer to confirm. Consistent with on-chain records.

[0207] In this embodiment, the specific steps involved in the threshold judgment algorithm determining the threshold and triggering the warning event in the blockchain early warning module 32 are as follows:

[0208] Based on the blockchain early warning module 32, monitor the newly added shard root hash on the chain. ;

[0209] The carbon emission dataset within the fragment was located using on-chain pointer indexes and the Merkle path. ;

[0210] Decompress carbon emission data blocks ;

[0211] The above This indicates the flag used during the on-chain evidence storage phase, representing an index of a specific data point at which the data is actually received. (For example, when a new carbon emission record is written to or detected within a fragment) (At that time), the determination of whether the record is abnormal is written onto the chain, which represents the evidence storage stage, for the actual time. Anomaly markers; Boolean anomaly markers in carbon emission prediction model outputs After being verified on-chain or through a threshold, it becomes on-chain. (i.e., save as) ).

[0212] Based on vehicle operating range The threshold for dynamically adjusting the difficulty level (such as slope and weather) is adjusted.

[0213] Traversing the carbon emission dataset For each data point, if the actual carbon emissions... Exceeding the set threshold is recorded as an abnormal event. ;

[0214]

[0215] Meanwhile, when new anomalous events need to be evaluated (e.g., in the carbon emissions dataset within a shard), Encountering new data blocks ), back to the most recent On the chain at that moment sequence:

[0216]

[0217]

[0218] Each Includes data from edge models result, and The output of the auxiliary branch in the implicit inheritance model.

[0219] Determining whether an anomaly (continuous anomaly or isolated anomaly) is based on a threshold-based algorithm combined with historical on-chain data:

[0220]

[0221] In the formula, Indicates the first Detected carbon emission anomalies in real time; Indicates the first The actual carbon emissions detected at any given time; Indicates the current train operating section Historical average carbon emissions; Indicates at time The probability coefficient of continuous anomalies on the chain is a weighting index of continuous anomalies in historical periods, reflecting whether the current anomaly is in a period of continuous anomalies (a value close to 1 is more likely to be a continuous anomaly). Indicates at time The isolated anomaly penalty coefficient is used to suppress the effects of random or sporadic mutation events; This represents the weighting coefficient for the historical deviation term, used to measure the strength of the impact of the difference between the current emission value and the historical average. This represents the weighting coefficient for continuous anomaly detection; This represents the isolated outlier discrimination weight coefficient, used to control the degree of influence of sudden outliers on the judgment result; and .

[0222] Furthermore, Indicates the length of the time window; Indicates time Does anything unusual happen at any given time? (1 indicates an anomaly, 0 indicates no anomaly); Indicates the first The timestamp of each data point, i.e., the time when the abnormal carbon emission event occurred, is synchronized with the GNSS global clock; This indicates the reason for triggering the abnormal event or the basis for judgment, and provides decision support for subsequent early warning or scheduling response (including but not limited to the abnormality exceeding the set threshold, the Boolean abnormality flag output by the edge model, and the Boolean abnormality flag output by the edge model). Its value range is , among which, if A value close to 1 indicates that anomalies have occurred frequently recently, and the current anomaly is persistent. A value close to 0 indicates that there have been almost no recent anomalies, and the current anomaly is an isolated anomaly. This helps to determine the background of the current anomaly and prevents accidental sudden points from being misjudged. This represents the penalty weight coefficient (0.5-0.8), used to control the weight of isolated anomalies; Indicates the train's longitude at the time the event occurred; Indicates the latitude of the train at the time the event occurred.

[0223] like Exceeding the preset alarm threshold If the exception is entered into the summary data structure, then the exception will be included in the summary data structure. .

[0224] Among them, after the abnormal event is confirmed, it is encapsulated as an off-chain early warning record into a summary data structure. ;

[0225] Abstract data structure Once the data is uploaded to the blockchain (or stored in an off-chain database and its fingerprint uploaded to the blockchain), it is simultaneously sent to the scheduling center to trigger an operational response (such as slowdown, temporary shutdown, load adjustment, etc.).

[0226] In this embodiment, the blockchain-based and edge computing-based dynamic tracking system for railway transportation carbon footprint also includes a dynamic tracking unit 4. The dynamic tracking unit 4 corrects on-chain data and real-time edge computing data based on an LSTM prediction model, and uses an NSGA-II multi-objective genetic algorithm with an elitist retention strategy for global optimization to generate a Pareto solution set. and Pareto solution set The data is then sent to edge computing unit 2 for closed-loop optimization.

[0227] Specifically, the dynamic tracking unit 4 includes a collaborative correction module 41, a multi-objective optimization decision-making module 42, a full life cycle carbon integration module 43, and a closed-loop strategy distribution module 44.

[0228] Among them, the collaborative correction module 41 is based on the summary data structure Triples Based on historical carbon emission sequences, an LSTM prediction model was used to assess the efficiency of the traction system. With real-time emission factors It performs short-term forecasting and correction to adjust marginal parameters.

[0229] In this embodiment, the input to the LSTM prediction model includes:

[0230] Summary Data Structure Triples and historical carbon emission sequences ;

[0231] Real-time emission factors based on LSTM predictions calibrated using on-chain historical data :

[0232]

[0233] In the formula, This represents the weighting coefficient, used to balance the influence of LSTM model predictions and on-chain historical data, with a value range of [0,1]. Indicates the calibration window length (default) (seconds); the larger the window, the more stable the calibration but the slower the response. This indicates that the LSTM model is at the current time step. The predicted instantaneous emission factor has not been corrected by on-chain data; Indicates a point in history The actual instantaneous carbon emissions; Indicates a point in time Instantaneous energy consumption; This indicates the calibrated emission factor.

[0234] Traction system efficiency Compensation mechanism:

[0235] When the efficiency prediction reported by the edge node deviates from the measured value by more than 5%, initiate a sliding window analysis of variance (window = 10 seconds):

[0236]

[0237] In the formula, Indicates the attenuation factor (default) ); This represents the traction system efficiency obtained from edge prediction at the current moment; This indicates the efficiency deviation between the predicted and measured values; This represents the dynamically corrected traction system efficiency, which serves as the input value for subsequent carbon emission calculations and power modeling.

[0238] If a persistent positive deviation is detected, the prediction efficiency value is adjusted downwards exponentially.

[0239] If a persistent negative bias is detected, the prediction efficiency value is adjusted upwards exponentially.

[0240] The multi-objective optimization decision module 42 uses edge computing latency, energy consumption, and on-chain overhead as objective functions, and comprehensively considers the sampling interval. Edge computing quota and on-chain frequency The NSGA-II multi-objective genetic algorithm with an elitist retention strategy is used for global optimization to generate the Pareto solution set. This provides the system with multiple optimal operating schemes that balance performance, among which... Indicates the first The sampling time interval of the solution; Indicates the first The proportion of edge computing resources used to solve the problem; Indicates the first The frequency of the solution going up the chain; This represents the number of Pareto solutions (the number of optimal combinations that satisfy different performance preferences). Indicates an index variable.

[0241] In this embodiment, the sampling interval Second;

[0242] Edge computing quota CPU;

[0243] On-chain frequency Hz.

[0244] The NSGA-II multi-objective genetic algorithm with an elitist retention strategy is used for global optimization to generate the Pareto solution set. The specific steps involved are as follows:

[0245] Initialize the population (population size) Individual coding format is The triplet;

[0246] Generate the Pareto front solution set using non-dominated sorting;

[0247] The crowding distance operator is used to maintain the diversity of the solution set;

[0248] Output the optimal solution set Optimal solution set Include Groups of parameters to satisfy different performance preferences Based on the real-time network status and the remaining resources of edge nodes, from the optimal solution set Choose the operation strategy that best suits the current scenario.

[0249] The life-cycle carbon integration module 43 calculates the comprehensive carbon emissions based on the carbon footprint extension model. If comprehensive carbon emissions are detected On-chain carbon trading is triggered when the threshold is exceeded. The carbon footprint extension model is based on cumulative carbon emissions. It was constructed by incorporating indirect carbon emission factors generated during the use of infrastructure.

[0250] In this embodiment, the carbon footprint extended model is based on cumulative carbon emissions. By incorporating and weighting indirect carbon emission factors generated during infrastructure use, a full life-cycle carbon emission model is constructed:

[0251]

[0252] In the formula, Indicates time The total carbon emissions throughout the entire life cycle (unit: kg CO2); This indicates the total number of facility types, specifically the total number of infrastructures participating in carbon sharing (such as rail, signaling systems, power stations, etc.). Indicates the first Carbon emissions per unit life cycle of such facilities (unit: kg CO2). Indicates the first Carbon weight coefficient of facilities (based on train load on facilities) (Determined by the proportion of time occupied).

[0253] in, In the formula, Indicates the impact of train operation tasks on facilities Actual occupied time (in hours); Indication facilities Design life cycle (in hours).

[0254] Cumulative carbon emissions These parameters are used to characterize the direct carbon emissions generated by trains due to energy consumption during operation and are the basic parameters for constructing a train lifecycle carbon footprint model.

[0255] A carbon footprint extension model is constructed to dynamically estimate the total indirect carbon emissions based on the infrastructure lifecycle and train occupancy ratio. When the threshold is exceeded, an on-chain carbon trading mechanism is triggered to achieve the integrated fusion of multi-source carbon factors and the carbon efficiency assessment of operational strategies. This solves the problem of isolated emission assessment during the operation phase and difficulty in connecting with the carbon market. Specifically, the full lifecycle carbon integration module 43 is used to further integrate indirect carbon emission factors at the infrastructure level on the basis of completing the dynamic calculation of carbon emissions during train operation, to construct a full lifecycle carbon emission model, and improve the system's ability to perceive and manage the carbon footprint of the entire railway transportation process.

[0256] The closed-loop strategy distribution module 44 is used to resolve Pareto data. The data is sent to edge computing unit 2.

[0257] Example 2: The difference between Example 2 and Example 1 is that this example introduces a method for dynamic tracking of railway traffic carbon footprint based on blockchain and edge computing, used in a dynamic tracking system for railway traffic carbon footprint based on blockchain and edge computing.

[0258] A method for dynamic tracking of railway traffic carbon footprint based on blockchain and edge computing, used in any of the above-mentioned dynamic tracking systems for railway traffic carbon footprint based on blockchain and edge computing, includes the following steps:

[0259] S10.1 The edge computing unit 2 fuses the multi-source data collected by the data acquisition unit 1, and calculates the real-time energy consumption of the train by combining the instantaneous power model, and introduces the dynamic emission factor to convert it into cumulative carbon emissions.

[0260] S10.2 and Blockchain Network Unit 3 compress and encode data from edge computing, construct the hash structure of carbon emission data blocks through dynamic sharding Merkle trees and put them on the chain to achieve traceable evidence storage; when the train crosses the operating section or time window, it automatically triggers the creation, splitting or merging of new shards;

[0261] Meanwhile, by monitoring newly added shard hashes on the chain, abnormal events are identified based on a threshold judgment algorithm combined with historical data. The abnormal summary data structure is then uploaded to the chain and pushed to the scheduling center to trigger an operation and maintenance response.

[0262] S10.3. The on-chain data and edge prediction results are collaboratively corrected by the dynamic tracking unit 4. The Pareto solution set is generated by the NSGA-II multi-objective optimization algorithm. After comprehensively considering latency, energy consumption and on-chain overhead, the optimal strategy is fed back to the edge computing unit by the closed-loop strategy distribution module to achieve full-process optimization.

[0263] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A dynamic tracking system for the carbon footprint of railway transportation based on blockchain and edge computing, characterized in that, include: Data acquisition unit (1), the data acquisition unit (1) is used to collect train operating status, environmental parameters and traction energy consumption data; The edge computing unit (2) fuses the multi-source data collected by the data acquisition unit (1) through adaptive Kalman filtering to obtain standardized data, and converts the train operation status into carbon emissions based on the instantaneous power model, and introduces a dynamic emission factor. Calculation of cumulative carbon emissions ; Blockchain network unit (3), the blockchain network unit (3) is used to compress the cumulative carbon emission data and generate a Merkle tree root hash to put it on the chain, and store the carbon emission data on the blockchain; The dynamic tracking unit (4) is based on the LSTM prediction model to correct on-chain data and real-time edge computing data. It uses the NSGA-II multi-objective genetic algorithm with an elite retention strategy to perform global optimization and generate Pareto solution set. and Pareto solution set The data is sent to the edge computing unit (2) for closed-loop optimization.

2. The dynamic tracking system for railway transportation carbon footprint based on blockchain and edge computing according to claim 1, characterized in that, The data acquisition unit (1) includes an on-board sensor module, a track environment monitoring module, and an energy consumption metering module; Among them, the on-board sensor module is used to monitor train operating energy consumption, load, speed and location data; The track environment monitoring module is used to collect track gradient, weather conditions, and signal system status parameters; The energy consumption metering module is used to collect energy consumption data from the traction power supply system and stations.

3. The dynamic tracking system for railway transportation carbon footprint based on blockchain and edge computing according to claim 1, characterized in that, The edge computing unit (2) includes a local preprocessing node module (21), a distributed computing module (22), and an edge prediction module (23). The local preprocessing node module (21) uses adaptive Kalman filtering to fuse the multi-source data collected by the data acquisition unit (1) to obtain preprocessed standardized data; The distributed computing module (22) uses the instantaneous power model to convert the train's operating status into carbon emissions based on the preprocessed standardized data. Among them, the train's time is calculated based on the instantaneous power model. Instantaneous total power and will improve the efficiency of the traction system. Introducing instantaneous total power Obtain cumulative energy consumption Meanwhile, based on real-time emission factors Cumulative energy consumption This translates into cumulative carbon emissions; Furthermore, the instantaneous power model is composed of air resistance power, slope resistance power, and inertial power coupled together; The edge prediction module (23) predicts carbon emission values ​​based on the carbon emission prediction model. It also outputs Boolean anomaly flags by introducing an anomaly detection auxiliary branch. .

4. The dynamic tracking system for railway transportation carbon footprint based on blockchain and edge computing according to claim 3, characterized in that, The specific steps involved in the edge prediction module (23) are as follows: A carbon emission prediction model is constructed using a prediction model based on long short-term memory networks and attention mechanisms to dynamically predict carbon emission levels in future periods. The preprocessed, standardized data is used as input to the carbon emission prediction model, and the output is the predicted carbon emission value. ; An anomaly detection auxiliary branch, constructed using binary classification logic, is introduced into the carbon emission prediction model, outputting a Boolean anomaly flag. and prediction confidence value ; Finally, the carbon emission prediction model packages the regression and classification results into a triple. .

5. The dynamic tracking system for railway transportation carbon footprint based on blockchain and edge computing according to claim 1, characterized in that, The blockchain network unit (3) includes a blockchain storage module (31) and a blockchain early warning module (32). The blockchain storage module (31) is used to perform hash compression on carbon emission data from the edge computing unit (2) according to time windows, construct a dynamic sharded Merkle tree and write it into the blockchain; The Merkle tree uses a dynamic sharding structure, and the sharding logic is based on the division of train operating sections. The blockchain early warning module (32) determines the threshold and triggers an early warning event based on the carbon emission data stored on the chain and the threshold judgment algorithm.

6. The dynamic tracking system for railway transportation carbon footprint based on blockchain and edge computing according to claim 5, characterized in that, The dynamic segmentation rules for the train operating section are as follows: The railway network is composed of It consists of several stations, and the tracks between adjacent stations form a basic operating section. ; Each interval Corresponding to a dynamic partition ; Based on train location and time window Determine the sharding key ; As the train enters the new section At that time, continuously check whether its real-time position and timestamp meet the new shard creation logic. If they do, trigger the creation of a new shard. create; When fragmentation The number of data records stored in the middle exceeds the split threshold. When this happens, a split operation is triggered; When adjacent fragments and If both low load and physical adjacency are met, a merge operation is triggered, and adjacent shards are merged. and merged into .

7. The dynamic tracking system for railway transportation carbon footprint based on blockchain and edge computing according to claim 6, characterized in that, The specific steps involved in constructing the Merkle tree are as follows: For each fragment Carbon emission data blocks in Calculate the hash of the leaf node ; Hash all leaf nodes within the shard. Sort by time, pair them up to generate parent node hashes, and repeat the above steps until a unique shard root hash is generated. ; hash all shard roots As leaf nodes, construct the global Merkle tree.

8. The dynamic tracking system for railway transportation carbon footprint based on blockchain and edge computing according to claim 5, characterized in that, In the blockchain early warning module (32), the specific steps involved in the threshold judgment algorithm to judge the threshold and trigger the early warning event are as follows: Based on the blockchain early warning module (32), monitor the newly added shard root hash on the chain. ; The carbon emission dataset within the fragment was located using on-chain pointer indexes and the Merkle path. ; Traversing the carbon emission dataset For each data point, if the actual carbon emissions... Exceeding the set threshold is recorded as an abnormal event. ; At the same time, when new anomalies are to be evaluated, the most recent ones are traced back. On the chain at that moment sequence; Anomalies are determined based on a threshold-based algorithm combined with historical on-chain data. In the formula, Indicates the first Detected carbon emission anomalies in real time; Indicates the first The actual carbon emissions detected at any given time; Indicates the current train operating section Historical average carbon emissions; Indicates at time The probability coefficient of continuous anomalies on the chain; Indicates at time The penalty coefficient for isolated anomalies; This represents the weighting coefficient for the historical deviation term; This represents the weighting coefficient for continuous anomaly detection; This represents the weighting coefficient for identifying isolated anomalies; like Exceeding the preset alarm threshold If the exception is entered into the summary data structure, then the exception will be included in the summary data structure. ; Among them, after the abnormal event is confirmed, it is encapsulated as an off-chain early warning record into a summary data structure. ; Abstract data structure Once uploaded to the blockchain, the data is synchronously sent to the scheduling center, triggering an operations and maintenance response.

9. The dynamic tracking system for railway transportation carbon footprint based on blockchain and edge computing according to claim 1, characterized in that, The dynamic tracking unit (4) includes a collaborative correction module (41), a multi-objective optimization decision module (42), a full life cycle carbon integration module (43), and a closed-loop strategy distribution module (44). Among them, the collaborative correction module (41) is based on the summary data structure Triples Based on historical carbon emission sequences, an LSTM prediction model was used to assess the efficiency of the traction system. With real-time emission factors Short-term forecasting and correction are performed to adjust marginal parameters. The multi-objective optimization decision module (42) takes edge computing latency, energy consumption, and on-chain overhead as objective functions, and comprehensively considers the sampling interval. Edge computing quota and on-chain frequency The NSGA-II multi-objective genetic algorithm with an elitist retention strategy is used for global optimization to generate the Pareto solution set. ; The life-cycle carbon integration module (43) calculates the comprehensive carbon emissions based on the carbon footprint extension model. If comprehensive carbon emissions are detected On-chain carbon trading is triggered when the threshold is exceeded. The carbon footprint extension model is based on cumulative carbon emissions. It was constructed by incorporating indirect carbon emission factors generated during the use of infrastructure; The closed-loop strategy distribution module (44) is used to resolve Pareto data. The data is distributed to the edge computing unit (2).

10. A method for dynamic tracking of railway traffic carbon footprint based on blockchain and edge computing, used in a dynamic tracking system for railway traffic carbon footprint based on blockchain and edge computing as described in any one of claims 1-9, characterized in that, Includes the following steps: S10.1 The edge computing unit (2) fuses the multi-source data collected by the data acquisition unit (1), and calculates the real-time energy consumption of the train by combining the instantaneous power model, introducing the dynamic emission factor, and converting it into the cumulative carbon emissions. S10.2, Blockchain network unit (3) compresses and encodes data from edge computing, constructs the hash structure of carbon emission data blocks through dynamic sharding Merkle tree and puts them on the chain to achieve traceable evidence storage; when the train crosses the operating section or time window, it automatically triggers the creation, splitting or merging of new shards; Meanwhile, by monitoring newly added shard hashes on the chain, abnormal events are identified based on threshold judgment algorithms combined with historical data, and the abnormal summary data structure is uploaded to the chain and pushed to the scheduling center to trigger operation and maintenance response; S10.

3. The on-chain data and edge prediction results are collaboratively corrected by the dynamic tracking unit (4). The Pareto solution set is generated by the NSGA-II multi-objective optimization algorithm. After comprehensively considering the delay, energy consumption and on-chain overhead, the optimal strategy is fed back to the edge computing unit by the closed-loop strategy distribution module to achieve full-process optimization.

Images