A city railway traffic whole-link carbon emission evaluation system and evaluation method

Abstract

The present application relates to carbon emission intelligent evaluation technical field, disclose a kind of city area railway traffic whole link carbon emission evaluation system and evaluation method, comprising the following steps: four-dimensional space-time voxelization grid covering city area railway is built and carbon polarity calibration is completed;Collecting equipment waveform, the feature is extracted by solidification base projection to calculate direct carbon emission;After being parsed into independent sequence to multiple source heterogeneous data, time delay coupling matrix is built in time domain space by calculating cross-correlation function, and is reconstructed and aligned as integrated state vector in time domain translation;Combined with direct emission, build multi-modal dynamic carbon metabolism atlas, deduce secondary cascade carbon emission increment between network nodes;Whole life cycle data is executed space-time integration clearing to output carbon emission total account bill.The present application includes physical disturbance and ecological carbon sink evolution into a unified network model, break through the technical bottleneck of data time sequence mismatch and spatial isolation, realize whole life cycle high-precision dynamic evaluation.

Description

Technical Field

[0001] This invention relates to the field of intelligent carbon emission assessment technology, specifically to a carbon emission assessment system and method for all aspects of urban railway transportation. Background Technology

[0002] Urban railway projects are characterized by large spatial span, long construction period, and multiple professional fields involved. Their carbon emissions not only cover the direct energy consumption of construction machinery, but also highly coupled with the complex indirect impacts of engineering physical disturbances on surrounding traffic flow and ecological microcirculation.

[0003] Current conventional carbon emission management often relies on static emission factor methods or periodic statistics based on macroeconomic energy consumption reports. However, this conventional assessment mechanism exposes deep-seated technical limitations when faced with the extremely complex life-cycle scenario of urban rail transit. Spatially, existing models typically treat the vast construction and environmental radiation areas as homogeneous macroeconomic statistical units, lacking refined modeling of different physical levels such as underground tunneling faces, surface road construction areas, and surrounding vegetation and water bodies. Consequently, it is difficult to accurately isolate and measure the intertwined carbon source polarity and carbon sink attributes in complex spaces.

[0004] Meanwhile, with the urgent need for refined carbon management, existing traditional physical measurement and assessment technologies have revealed the following significant shortcomings: First, purely physical metering methods cannot achieve precise decoupling of carbon emissions within the train. Suburban railway trains are essentially complex microgrids containing multiple subsystems such as traction drive, HVAC, air compressors, and auxiliary power supply. If traditional technologies are used to achieve absolute carbon emission accounting at the subsystem level, an "intrusive" approach must be adopted, involving the extensive installation of numerous secondary and tertiary physical meters within the train. This not only significantly increases the train's hardware weight, wiring costs, and communication failure points, but also often lacks the space for implementation within the compact electrical cabinets of the train, making it impossible for operators to accurately determine which specific subsystem is generating abnormal carbon emissions.

[0005] Secondly, traditional metering instruments cannot capture the "hidden carbon emissions" caused by transient waveform distortion. The urban rail system contains a massive number of high-power frequency converters and power electronic conversion devices. During the frequent starts, stops, and accelerations of trains, a large amount of high-frequency harmonics, interharmonics, and transient reactive power are injected into the power grid, resulting in severe additional heat loss (i.e., hidden carbon emissions) in the traction power supply network. Traditional smart meters, based on low-frequency sampling or steady-state power integration, lose millisecond-level or even microsecond-level transient electrical characteristic information, making it impossible to quantitatively analyze the contribution of specific transient operating conditions and high-frequency distortion to the surge in carbon emissions.

[0006] At the data processing level, the entire process of urban rail transit continuously generates massive amounts of multi-source heterogeneous data, such as high-frequency fluctuating electrical signals from engineering machinery operations and relatively low-frequency regional traffic flow and ecological monitoring sampling values. When traditional systems fuse these asynchronously sampled data streams, they often directly perform simple linear interpolation or crude numerical splicing in the time domain, completely ignoring the temporal lag coupling relationship in the cross-modal physical transmission process. This approach directly leads to phase mismatch and severe feature distortion of the underlying data on the time axis.

[0007] A more critical technological bottleneck lies in the fact that the environmental disturbances caused by large-scale transportation projects are not isolated, static events. For example, localized road construction inevitably spreads like ripples, causing traffic congestion and continuous idling emissions in the surrounding road network. In actual industrialization, to conduct "accurate accounting" of such indirect carbon emissions in a purely physical sense, it is necessary to fully access real traffic flow monitoring data from traffic management departments, which presents extremely high data acquisition barriers and lags. The traditional logic of "purely physical measurement and accounting" completely fails when faced with carbon sprawl at non-physical boundaries. Existing carbon management systems lack a graph-based deep topological simulation and transmission mechanism, and can only perform isolated point-like statistics, making it impossible to capture and predict the secondary cascade carbon emission increments caused by local physical nodes. At the same time, when a project undergoes a major physical transition from the construction phase to the operation phase, traditional static models cannot achieve adaptive carryover of ecological residual values ​​and dynamic boundary reconstruction, ultimately leading to a huge systemic bias in the carbon bill throughout the entire life cycle.

[0008] Therefore, the industry urgently needs a comprehensive dynamic assessment technology that can overcome the isolation of multidimensional spatiotemporal data and solve the frequency domain mismatch of heterogeneous data. This technology should be able to extract high-frequency features from the main line electrical signals without adding underlying sensor hardware (non-intrusive), thereby accurately decompose the carbon emissions of each subsystem inside the train. Furthermore, it should be able to use mathematical micro-simulation models to comprehensively deduce and predict the indirect carbon emissions caused by secondary cascade disturbances without relying on massive external physical monitoring data, thus achieving high-precision dynamic assessment of carbon emissions in complex transportation systems. Summary of the Invention

[0009] To address the shortcomings of existing technologies, this invention provides a carbon emission assessment system and method for the entire process of urban rail transit. It solves the technical problems that existing conventional static assessment models cannot cope with the frequency domain temporal mismatch of multi-source heterogeneous data under the complex spatiotemporal span of urban rail transit, and are difficult to accurately capture and dynamically deduce the secondary cascade carbon emission increments and the ecological carryover evolution throughout the entire life cycle caused by local physical engineering disturbances.

[0010] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of the present invention provides a method for assessing carbon emissions across the entire process of urban rail transit, comprising the following steps: constructing a four-dimensional spatiotemporal voxel grid covering the entire urban rail line and its environmental radiation zone, and defining the attributes and carbon polarity of the four-dimensional spatiotemporal voxels in the grid, including carbon sink identification; acquiring physical waveform signals of the target dynamic equipment, projecting and extracting carbon feature fingerprints using an unsupervised dimensionality reduction algorithm with a pre-trained solidified mapping substrate, and calculating the direct carbon emissions of the corresponding four-dimensional spatiotemporal voxels based on the extracted carbon feature fingerprints; and parsing the multi-source heterogeneous original data stream matrix into multiple... Independent data stream sequences are used to construct a time delay coupling matrix by calculating the cross-correlation function between each pair of data stream sequences in the time domain. The time delay coupling matrix is ​​then used to translate and reconstruct an alignment in the time domain to generate a comprehensive state vector based on a virtual evaluation timestamp. A multimodal dynamic carbon metabolism map is constructed based on the comprehensive state vector and the direct carbon emissions. Secondary cascade carbon emission increments are calculated by extrapolating among the network nodes in the multimodal dynamic carbon metabolism map. Based on the implicit carbon emissions of static entities in the multimodal dynamic carbon metabolism map, the direct carbon emissions, and the secondary cascade carbon emission increments, a comprehensive carbon emission assessment result for the entire process is output through spatiotemporal integration.

[0011] Furthermore, in order to achieve high-precision discretization of large-scale spatiotemporal scenes, the step of constructing a four-dimensional spatiotemporal voxel grid covering the entire urban railway line and its environmental radiation area specifically includes: discretizing the entire urban railway line and its environmental radiation area using a three-dimensional spatial coordinate system (longitude, latitude, and elevation) combined with a continuous time dimension; generating multiple four-dimensional spatiotemporal voxels containing independent spatial coordinates and time slices according to a preset spatial resolution and time step; splicing physical entity attribute components (such as building material type and equipment ledger) and ecological environment attribute components (such as vegetation coverage and soil type) for each four-dimensional spatiotemporal voxel to generate a multi-dimensional attribute state vector; and assigning a label polarity to the four-dimensional spatiotemporal voxel based on the project progress and the material interaction process within the four-dimensional spatiotemporal voxel, wherein the label polarity is specifically classified as a carbon source emitting greenhouse gases, a neutral carbon source balancing the budget, or a carbon sink absorbing greenhouse gases.

[0012] Furthermore, to accurately quantify the direct contribution of electromechanical equipment (such as traction transformers, construction machinery, etc.) to carbon emissions, the step of collecting the physical waveform signals of the target dynamic equipment and extracting carbon feature fingerprints using an unsupervised dimensionality reduction algorithm specifically includes: under a set preset sampling time series, using sensors deployed at the front end to collect the physical waveform signals (such as instantaneous current, voltage waveforms, or high-frequency vibration signals) of the target dynamic equipment in real time; using the unsupervised dimensionality reduction algorithm to perform feature mapping on the physical waveform signals, specifically including: using a single complete operating cycle or fixed power frequency period of the target dynamic equipment as a data extraction window to construct a high-dimensional time series input matrix; using an autoencoder network as an unsupervised dimensionality reduction architecture, utilizing package... The joint loss function, including the signal reconstruction mean square error loss and the feature non-correlation penalty term, is iteratively trained. Dimensionality reduction projection is performed through the hidden layer of the autoencoder network to filter out environmental noise background and basis features, extracting a dimensionality-reduced feature vector representing the non-stationary working state of the equipment, generating the carbon feature fingerprint containing multiple standardized feature components. Each standardized feature component is multiplied and summed with a preset physical transformation weight parameter, and a local environmental compensation constant is added to obtain the dynamic carbon polarity coefficient. The preset physical transformation weight parameter is calibrated based on a pre-constructed bench test database, which includes synchronously collected transient active power consumption increments, equipment heat dissipation, and energy carbon emission factors. The calculation process of the dynamic carbon polarity coefficient is expressed as follows:

[0013] In the formula, Indicates the dynamic carbon polarity coefficient. This represents the total number of standardized feature components. Indicates the first One standardized feature component, Indicates corresponding to the first Preset physical transformation weight parameters for each standardized feature component. The local environmental compensation constant is represented by the dynamic carbon polarity coefficient. Multiplying the dynamic carbon polarity coefficient by the operational activity level of the target dynamic device within the time period corresponding to the preset sampling time series yields the direct carbon emissions corresponding to the four-dimensional spatiotemporal voxel. The calculation process is as follows:

[0014] In the formula, This indicates direct carbon emissions. This indicates the operational activity level parameters (such as power consumption or fuel consumption) of the target dynamic equipment within the corresponding time period.

[0015] Furthermore, to address the challenges of misaligned and difficult-to-synchronize sampling periods of multi-source sensor data across systems and devices, the step of constructing a time-delay coupling matrix to reconstruct the comprehensive state vector specifically includes: parsing and splitting the multi-source heterogeneous original data stream matrix into multiple independent one-dimensional data stream sequences according to the data source; calculating the cross-correlation function between each pair of the multiple independent one-dimensional data stream sequences in the time domain; extracting the peak value of the cross-correlation function, obtaining the time delay corresponding to the peak value, and constructing the time-delay coupling matrix to characterize the time-series lag coupling relationship of physical processes (such as the time delay transmission between power grid load fluctuations and train traction energy consumption); setting a globally unified virtual evaluation timestamp, translating and aligning each one-dimensional data stream sequence in the time domain according to the time delay in the time-delay coupling matrix, and resampling and outputting the comprehensive state vector that is strictly physically aligned under the same virtual evaluation timestamp.

[0016] Furthermore, to reveal the cascading effect of carbon emissions across the entire urban railway network (i.e., the chain reaction of carbon emissions from surrounding nodes triggered by construction or operational disturbances at a certain node), the steps for calculating the incremental secondary cascading carbon emissions specifically include: mapping the four-dimensional spatiotemporal voxels to the network nodes in the multimodal dynamic carbon metabolism map, and using the direct carbon emissions as the initial capacity characteristic of the network nodes; based on the physical disturbance relationships between the four-dimensional spatiotemporal voxels (such as airflow diffusion, traffic congestion transfer, and heat island effect transmission), establishing a directed transmission edge between the source node generating the disturbance and the target node being disturbed, and calculating the transmission weight of the directed transmission edge based on the disturbance intensity of the source node and the spatial distance attenuation function (such as the Gaussian distance attenuation function or the inverse proportional attenuation function). The calculation process is expressed as follows:

[0017] In the formula, Indicates that it is from the source node Point to target node Transmission weight, Indicates the source node The initial disturbance intensity, Indicates based on source node With the target node Spatial distance between The spatial distance attenuation function; along the directed transmission edge, combined with the corresponding feature data of the target node in the comprehensive state vector, the transmission weight of the directed transmission edge is substituted into a preset nonlinear response function (such as a sigmoid activation function or an exponential response function) for mapping calculation, to obtain the secondary cascade carbon emission increment of the target node caused by the source node disturbance. The calculation process is expressed as follows:

[0018] In the formula, Represents the target node The resulting increase in secondary cascade carbon emissions, Represents a nonlinear response function. Represents the target node Its own state characteristic data.

[0019] Further, the step of outputting the comprehensive carbon emission assessment result through spatiotemporal integration includes: calculating the total system carbon emissions of the multimodal dynamic carbon metabolism map at the virtual assessment timestamp; the total system carbon emissions are calculated by summing the static entity implicit carbon emissions of all network nodes in the multimodal dynamic carbon metabolism map (the static entity implicit carbon emissions are the sum of the products of each characteristic index in the physical entity attribute component and its matching positive carbon emission conversion constant) with the direct carbon emissions, adding the sum of the secondary cascade carbon emission increments corresponding to all disturbed target nodes, and finally subtracting the dynamic carbon absorption amount obtained by the network nodes that have carbon sink attributes as determined by the carbon polarity labeling based on their ecological environment attribute components (such as vegetation photosynthetic absorption rate) in the attribute definition. The calculation process is expressed as follows:

[0020] In the formula, Indicates the virtual evaluation timestamp The total carbon emissions of the system under this condition Represents the set of all network nodes. Represents a node Direct carbon emissions, This represents the implicit carbon emissions of node u, which is equal to the sum of the products of each characteristic index in the aforementioned physical entity attribute components of that node and the corresponding positive carbon emission conversion constant. This represents the set of target nodes that are being disturbed. Indicates the disturbed target node The increase in secondary cascade carbon emissions, This represents a set of nodes that have carbon sink attributes. Nodes representing carbon sink attributes Dynamic carbon absorption.

[0021] Furthermore, considering the multi-year cyclical nature of urban railways, the steps further include: defining the time series span of the virtual assessment timestamp as the entire life cycle of the project, encompassing both the construction and operation phases; identifying the transition point from the construction phase to the operation phase (e.g., the time node for trial operation), removing the directed edges corresponding to the construction process during the construction phase, and dynamically transforming them into a directed edge topology corresponding to the traffic flow during the train operation phase; extracting the residual ecological carbon sink value (e.g., the carbon sink potential of the green belt after construction restoration) obtained by carrying over the dynamic carbon absorption amount from the four-dimensional spatiotemporal voxels labeled as having carbon sink attributes at the end of the construction phase, as the nonlinear initial boundary condition for calculating the carbon sink during the operation phase, and updating the dynamic carbon absorption amount during the operation phase accordingly; performing discrete integral summation calculations on the total carbon emissions of the system corresponding to multiple discrete virtual assessment timestamps within the entire life cycle of the project, and finally outputting a comprehensive assessment result of carbon emissions covering both temporal breadth and spatial depth, the calculation process of which is expressed as follows:

[0022] In the formula, This indicates the comprehensive assessment results of carbon emissions across all stages. This represents the total number of discrete virtual evaluation timestamps throughout the entire lifecycle of the project. Indicates the first A virtual evaluation timestamp This indicates the step size between adjacent timestamps.

[0023] This invention provides a carbon emission assessment system and method for the entire process of urban rail transit. It has the following beneficial effects: 1. This invention breaks through the limitations of spatial macroscopic fragmentation and static temporal lag in traditional carbon assessment. By constructing a four-dimensional spatiotemporal voxel grid covering the entire urban railway line and its environmental radiation zone, and by defining the attributes and carbon polarity of voxel units in a refined manner, it achieves cross-scale mapping from microscopic material interactions to macroscopic engineering progress. This not only accurately anchors the specific spatiotemporal coordinates of emission sources, but also incorporates ecological carbon sinks into the dynamic tracking system for the first time, providing a high-resolution, bottom-level spatiotemporal assessment panorama for tracing the dynamic carbon flow direction of all stages and all elements.

[0024] 2. This invention overcomes the shortcomings of traditional energy consumption metering in capturing hidden carbon emissions. Based on the law of conservation of energy, this invention rigorously anchors the underlying physical waveform distortion characteristics to the ineffective energy dissipation model of internal equipment, such as heating line losses and mechanical wear, and uses a conversion operator to accurately convert them into additional carbon emission equivalents. This mechanism provides solid physical support for interdisciplinary dimensional conversion, completely eliminates the blind spots in carbon emission calculation, and significantly improves the accuracy of the assessment.

[0025] 3. To address the challenge of asynchronous data caused by mismatched sampling cycles of cross-disciplinary and multi-level sensors in urban railway engineering, this solution abandons conventional data interpolation methods that are prone to introducing errors. Instead, it cleverly constructs a time-delay coupling matrix representing the physical temporal lag coupling relationship by introducing a cross-correlation analysis algorithm in the time domain, and reconstructs it by aligning and translating the global virtual evaluation timestamp in the time domain. This mechanism filters out sampling noise while perfectly bridging the time gaps between asynchronous data streams, ensuring the absolute rigor of the comprehensive state vector upon which subsequent carbon emission calculations rely in terms of physical causality.

[0026] 4. Addressing the blind spot where single-entity assessments easily overlook the surge in carbon emissions to the surrounding environment caused by construction or train operation (such as the large amount of exhaust fumes generated by traffic congestion due to construction site closures), this invention innovatively constructs a multimodal dynamic carbon metabolism map. It establishes directed transmission edges by leveraging the physical perturbation relationships between nodes and spatial distance attenuation functions, and uses nonlinear response functions to accurately quantify the secondary cascade carbon emission increments resulting from the perturbation waves. This design successfully expands isolated engineering bills into a realistic system-level ecological bill, restoring a comprehensive dynamic assessment of the impact of transportation infrastructure on urban microclimate and surrounding carbon flow disturbances.

[0027] 5. Considering the dramatic evolution of carbon source and sink structures at different stages of the long lifespan of urban railways, this solution achieves adaptive connection and calculation across engineering cycles. By accurately identifying the transition point from the construction phase to the operation phase, the transmission topology of the network graph is decoupled and reconstructed in a timely manner. At the same time, the residual ecological carbon sink value at the end of the construction phase is carried forward as the initial boundary condition for the new cycle. This time-integral-driven settlement model completely eliminates data gaps during the handover of engineering sections or the transition to operation, ensuring that the holographic carbon dynamic assessment from groundbreaking to long-term operation achieves lossless traceability and coherent deduction. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram of the construction of the four-dimensional spatiotemporal voxelized mesh and the carbon polarity calibration of the present invention; Figure 3 This is a schematic diagram illustrating the principle of time-domain cross-correlation analysis and time-delay shift alignment of multi-source heterogeneous raw data streams in this invention. Figure 4 This is a schematic diagram of the multimodal dynamic carbon metabolism map and secondary cascade deduction based on four-dimensional spatiotemporal voxels according to the present invention. Figure 5 This is the logic diagram for the carbon sequestration carry-over and spatiotemporal integral settlement of the entire life cycle of the project according to the present invention; Figure 6 This is a modular architecture diagram of the urban rail transit carbon emission assessment system of the present invention; Figure 7 This is a schematic diagram of carbon emission mapping for a real physical disturbance scenario of urban railways according to the present invention. Detailed Implementation

[0029] The technical solutions in 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.

[0030] Please see the appendix Figure 1 -Appendix Figure 7 This invention provides a carbon emission assessment system and method for the entire process of urban rail transit, including constructing a four-dimensional spatiotemporal voxel grid covering the entire urban rail line and its environmental radiation zone; defining attributes and carbon polarity calibration (including carbon sink identification) for the four-dimensional spatiotemporal voxels in the grid; acquiring physical waveform signals of target dynamic equipment; extracting carbon feature fingerprints using an unsupervised dimensionality reduction algorithm; calculating the direct carbon emissions of the corresponding four-dimensional spatiotemporal voxels based on the extracted carbon feature fingerprints; and acquiring a multi-source heterogeneous raw data stream matrix with different sampling periods, and parsing the multi-source heterogeneous raw data stream matrix into multiple independent data stream sequences. The process involves calculating the cross-correlation function between each pair of data stream sequences in the time domain to construct a time delay coupling matrix. This time delay coupling matrix is ​​then used to translate and reconstruct a comprehensive state vector based on a virtual evaluation timestamp in the time domain. A multimodal dynamic carbon metabolism map is constructed based on the comprehensive state vector and the direct carbon emissions. Secondary cascade carbon emission increments are then calculated and extrapolated among the network nodes in the multimodal dynamic carbon metabolism map. Finally, based on the implicit carbon emissions of static entities in the multimodal dynamic carbon metabolism map, the direct carbon emissions, and the secondary cascade carbon emission increments, a comprehensive assessment result of carbon emissions across all stages is output through spatiotemporal integration.

[0031] During the construction of a four-dimensional spatiotemporal voxel mesh covering the entire urban railway line and its surrounding environmental radiation zone, the entire urban railway line and its surrounding environmental radiation zone are discretized and modeled using a three-dimensional spatial coordinate system combined with a continuous time dimension. The three-axis spatial resolutions of the three-dimensional spatial coordinate system are set as follows: , as well as And set the time slice step size for the continuous time dimension to be... .

[0032] Through the aforementioned discretization rules, the continuous physical space of the engineering environment is divided into multiple four-dimensional spatiotemporal voxels, each containing independent spatial coordinates and time slices. Each four-dimensional spatiotemporal voxel is uniquely represented by an index. .

[0033] In this definition, , , These represent the index numbers of the voxels on the horizontal, vertical, and discrete coordinate axes in the three-dimensional coordinate system, respectively. This represents the time slice sequence number of the current voxel. The four-dimensional grid partitioning mechanism constitutes the underlying spatiotemporal foundation for carbon emission tracing across all stages, establishing the spatial absolute positioning boundary and time slice synchronization of massive data across the entire urban railway line.

[0034] After generating multiple four-dimensional spatiotemporal voxels, physical entity attribute components and ecological environment attribute components are spliced ​​together for each four-dimensional spatiotemporal voxel to generate a multi-dimensional attribute state vector.

[0035] Physical entity attribute components are used to characterize the objective existence state of man-made engineering entities or devices within a voxel. Physical entity attribute components are defined as column vectors. ,in The total number of feature dimensions representing physical attributes, with each dimension corresponding to the volume of building materials for a specific type of project, the installed power of the target dynamic equipment, and the quantitative value of the activity density of construction workers.

[0036] Ecological and environmental attribute components are used to characterize the carbon absorption potential of the natural ecological chassis within a voxel. These components are defined as column vectors. ,in The total number of characteristic dimensions representing ecological attributes, with each dimension corresponding to the vegetation leaf area index, soil organic carbon density, and surface water coverage of the voxel in the space.

[0037] Orthogonally concatenate the two column vectors mentioned above to generate a vector with a total dimension of [value missing]. Multidimensional attribute state vector The algebraic expression for this multidimensional attribute state vector is:

[0038] In the formula, the symbol This represents the concatenation operation of vectors. The multi-dimensional attribute state vector comprehensively and objectively quantifies the baseline of engineering disturbances and the existing ecological baseline within a single spatiotemporal grid cell.

[0039] Based on the project progress and the material interaction process within the four-dimensional spatiotemporal voxels, a polarity is assigned to each voxel. This polarity is specifically categorized as carbon source, neutral, or carbon sink.

[0040] To achieve dynamic quantization and automatic calibration of polarity, based on the multidimensional attribute state vector... Calculate the net instantaneous carbon flux within a voxel. The calculation process for the net instantaneous carbon flux is expressed as follows:

[0041] In the formula, This indicates that the four-dimensional spacetime voxel is in a time slice. Net instantaneous carbon flux; Represents the first of the physical entity attribute components Feature indicators; Indicates the relationship with the first The positive carbon emission conversion constant matched with the characteristic indicators; The first component representing the ecological environment attribute is... Feature indicators; Indicates the relationship with the first The negative carbon absorption curing constant matched with the characteristic index.

[0042] Based on the calculated net instantaneous carbon flux Using polarity determination function Perform the final carbon polarity calibration. The mapping rule for the polarity determination function is expressed as follows:

[0043] In the formula, This represents the preset polarity determination tolerance threshold, used to filter out background noise disturbances during multidimensional data acquisition. When the output value is 1, the four-dimensional spacetime voxel is assigned carbon source polarity; when the output value is 0, the four-dimensional spacetime voxel is assigned neutral polarity; when the output value is -1, the four-dimensional spacetime voxel is assigned carbon sink polarity.

[0044] The carbon polarity calibration mechanism transforms the entire voxel grid from a static data container into a set of active nodes with definite physical emission topological properties, providing clear initial boundary conditions for subsequent cascaded carbon emission algorithms in multimodal dynamic carbon metabolism maps.

[0045] In this embodiment, in order to accurately quantify the actual direct carbon emissions of the target dynamic equipment under complex operating conditions, and to break through the limitation that traditional static emission factors cannot reflect the transient energy consumption degradation of the equipment, the fundamental relationship between the micro-operational characteristics of the equipment and macro-carbon emissions is deeply analyzed.

[0046] In this invention, physical waveform signals of the target dynamic equipment are acquired in real time under a preset sampling time sequence. Specifically, a high-frequency sensor array is deployed in bypass mode at key energy-consuming nodes of the target dynamic equipment to obtain continuous transient physical signals. The target dynamic equipment includes core equipment that generates large-scale energy consumption during the construction and operation phases of urban railways, such as the main drive motor of tunnel boring machines, large earthmoving machinery, and transformer components of traction substations along the line.

[0047] The acquired physical waveform signal is defined as a time-domain time series matrix. This matrix records with high fidelity the instantaneous current distortion characteristics or high-frequency mechanical vibration envelope characteristics of the equipment under different load abrupt changes, providing a raw physical foundation for subsequent exploration of carbon emission surge factors under non-standard working conditions.

[0048] Subsequently, the unsupervised dimensionality reduction algorithm is used to perform feature mapping on the physical waveform signal. In this embodiment, an autoencoder network is preferably used as the core architecture for unsupervised dimensionality reduction. Specifically, a single complete operating cycle or a fixed power frequency cycle of the target dynamic device is first used as the data extraction window to analyze the time-domain time series matrix. The system first segments the data to construct a high-dimensional temporal input matrix. Then, during offline training, it uses a joint loss function, including a mean square error loss for signal reconstruction and a penalty term for feature irrelevance, to perform unsupervised iterative training on the autoencoder network, thereby constructing a baseline feature space with fixed physical correspondences. During real-time computation, the system projects the acquired high-dimensional temporal input matrix onto the hidden layers of the autoencoder network for dimensionality reduction, performs nonlinear feature space compression, filters out environmental noise background and normal baseline features, and extracts dimensionality-reduced feature vectors representing the non-stationary working state of the equipment.

[0049] Through the above projection dimensionality reduction process, the system can automatically remove the conventional power frequency interference and environmental white noise that accompany the physical waveform signal, and extract the energy consumption distortion eigenvector hidden deep in the complex waveform based on a fixed feature basis.

[0050] The extracted energy consumption distortion eigenvector is used to generate the carbon feature fingerprint containing multiple standardized feature components, denoted as the feature vector. .in, This represents the total dimension of the standardized feature components. It is necessary to specifically disclose the cross-scale dimension conversion relationship and calibration logic from the standardized feature components (dimensionless) to the dynamic carbon polarity coefficient (corresponding to the carbon emission mass dimension) in this embodiment: Although the additional energy consumption caused by equipment degradation is macroscopically reflected in the active power at the input end, conventional electricity meters, limited by low-frequency sampling rates and steady-state metering principles, cannot accurately isolate transient ineffective dissipation caused by high-frequency waveform distortion (such as cable heating caused by high-order current harmonics and negative work done by high-frequency mechanical vibration). The feature component fifi extracted by this system is not a black box data without physical meaning, but rather, through joint calibration by bench experiments, a rigorous mapping with a specific physical heat / dissipation model has been established.

[0051] The specific calibration logic is as follows: In offline bench experiments, a high-frequency power analyzer and thermal sensing equipment are used to synchronously acquire the physical waveform signals of the equipment under different distortion conditions, as well as the corresponding transient additional ineffective power dissipation (such as harmonic heating loss power and frictional heat generation power) and energy carbon emission factor. The aforementioned dimensionality reduction algorithm is then used to extract the characteristic components of the waveform. Then, multiple regression or system identification algorithms are used to identify each dimensionless feature component. By strictly anchoring to a specific physical dissipation model collected above, the corresponding physical transformation weight parameters are jointly solved. .

[0052] Therefore, the preset physical transformation weight parameter It is not an empirical constant; it has definite physical dimensions. Its essence is the 'carbon emission equivalent conversion rate of the ineffective dissipation power in a specific frequency band corresponding to a surge in a unit characteristic component'. Its dimensional unit is essentially [mass / dimensionless characteristic unit] (e.g., ...). Furthermore, by each of the standardized feature components... Its corresponding physical transformation weight parameters By performing product summation, the system strictly adheres to the law of conservation of energy at its mathematical level. It accurately restores and reconstructs the dimensionless distortion features implicit in high-frequency waveforms into additional carbon emissions with real physical dimensions caused by heating or wear. Adding a local environmental compensation constant, it ultimately obtains the dynamic carbon polarity coefficient. This mechanism completely eliminates the black-box nature of feature mapping, providing solid thermodynamic and electrophysical support for cross-dimensional carbon emission calculations.

[0053] The calculation process for the dynamic carbon polarity coefficient is expressed as follows:

[0054] In the formula, This represents the calculated dynamic carbon polarity coefficient; Represents the total number of standardized feature components; Indicates carbon feature fingerprint After normalization, the first One standardized feature component; Indicates corresponding to the first Preset physical transformation weight parameters for each standardized feature component.

[0055] Preset physical transformation weight parameters Calibration is performed based on a pre-established equipment energy consumption bench test database. It should be noted that this embodiment breaks down the technical barriers of interdisciplinary dimension conversion; the parameters... It is not simply a mathematical fitting coefficient, but a "feature-energy consumption-carbon emission" conversion operator with definite physical dimensions. Its physical mapping logic lies in the fact that waveform distortion of equipment (such as current harmonics and high-frequency mechanical vibration) essentially corresponds to energy dissipation that cannot perform work (such as heat loss and negative work from mechanical wear). This ineffective energy dissipation inevitably leads to additional intake from the upstream power grid or fossil fuels. During the offline calibration phase, this system synchronously collects waveform data, transient active power consumption increments, heat dissipation, and energy carbon emission factors of the equipment through bench tests. Through joint calibration, each standardized feature component extracted in the dimensionality reduction process is... Strictly anchored to a specific energy dissipation model, and then calculate the corresponding weight parameters. .therefore The actual physical dimension is "the equivalent of an additional carbon emission surge per unit feature surge". This is because the baseline feature space of the unsupervised dimensionality reduction algorithm and the transformation weight parameter... All calibrations originate from the same "bench test database" and are performed offline. Therefore, the first real-time projection extracted each time... Standardized feature components Physical semantics and It is a one-to-one mapping and strictly aligned. Within this logical framework, the formula not only lacks randomness, but also strictly follows the laws of energy conservation and dissipation in its mathematical form. Its essence is to precisely map the abstract mathematical feature components in a fixed feature space to the carbon emission surge coefficient dimension with standard engineering dimensions.

[0056] In the formula This represents a local environmental compensation constant, used to compensate for errors in external baseline factors that affect the thermodynamic efficiency of the target dynamic device, such as ambient temperature, humidity, or atmospheric pressure within the four-dimensional spatiotemporal voxel where the device is currently located.

[0057] Finally, the dynamic carbon polarity coefficient is multiplied by the operational activity level of the target dynamic device within the time period corresponding to the preset sampling time series to obtain the direct carbon emissions for the corresponding four-dimensional spatiotemporal voxel. The calculation process for direct carbon emissions is expressed as follows:

[0058] In the formula, This represents the direct carbon emissions contributed by the target dynamic device within the current spatiotemporal voxel unit; This indicates the operational activity level parameters of the target dynamic equipment within a corresponding time period. These operational activity level parameters are derived from macroscopic, real-time consumption statistics obtained from the equipment control host computer or on-site intelligent metering instruments, such as the cumulative energy consumption in kilowatt-hours or fuel volume within a single operating cycle.

[0059] By introducing a feature analysis mechanism for the underlying physical waveform, this scheme achieves a deep penetration from macroscopic energy consumption records to microscopic transient physical work states. In practical engineering, conventional smart meters and other metering instruments can only record the active power consumption of terminal equipment, failing to capture the additional line losses and carbon emissions from the upstream power grid caused by waveform distortions (such as high-order harmonics and reactive power surges) resulting from equipment aging and overload. Simultaneously, the high-frequency distortion of physical waveforms accompanying non-stationary and severe operating conditions often signifies a significant increase in "non-electricity-related ancillary carbon footprints," such as a surge in coolant and lubricant consumption and premature scrapping of mechanical wear parts. Therefore, the dynamic carbon polarity coefficient... It replaces the traditional, unchanging empirical carbon emission factor. It is not used to replace the active reading of the meter, but as a key implicit carbon emission compensation factor. This allows the final assessment result of direct carbon emissions to keenly capture and reflect the additional real carbon emission anomalies caused by hidden adverse factors such as equipment wear and tear, and unstable overload operation, which cannot be measured by conventional electricity meters. It completely eliminates the blind spots in carbon emission calculation caused by a single energy consumption metering method.

[0060] In this embodiment, in response to the technical barrier of severe mismatch in sampling cycles of massive monitoring equipment in the cross-professional system of urban railways, the synchronization and alignment mechanism of multi-source heterogeneous data at the underlying physical logic is described in detail.

[0061] In this invention, a multi-source heterogeneous raw data stream matrix with different sampling periods is acquired. The underlying acquisition network within the system synchronously accesses waveform sampling data from high-frequency dynamic devices and ecological flow data from low-frequency environmental monitoring stations. The multi-source heterogeneous raw data stream matrix acquired in the time domain is defined as follows: .

[0062] in, Represents the total number of channels for heterogeneous data streams. Representing the The discrete time series of each physical sampling channel in this time domain. Due to the different working mechanisms of the physical sensors in each channel, there are significant differences in sampling rate and data discontinuity in their time series.

[0063] In order to accurately extract the time delay caused by causal relationship between different physical sampling channels (such as the physical time delay characteristics between power grid load fluctuation and downstream train traction energy consumption transmission), this embodiment abandons the frequency domain transformation that is prone to causing non-stationary signal distortion, and instead directly performs cross-correlation analysis on each data stream sequence in the multi-source heterogeneous original data stream matrix in the time domain space.

[0064] For any two heterogeneous data streams in the time domain and Within the set time observation window Internal calculation of the discrete cross-correlation function of the two. The calculation process is expressed as follows:

[0065] In the formula, This represents the time-shifted variable (lag). It is calculated by... The cross-correlation function value under the given value can measure the similarity between two signals at different time misalignments.

[0066] Subsequently, the cross-correlation function was extracted using an optimization algorithm. The global peak value, and the time-slip variable corresponding to this peak value is the actual physical time delay between data stream a and data stream b. Its calculation formula is expressed as:

[0067] By traversing all data stream channels in the multi-source heterogeneous original data stream matrix, a global time delay coupling matrix is ​​constructed using the obtained series of time delay values. This matrix quantitatively establishes the temporal sequence of material and energy transfer and carbon emission impact in complex engineering construction and operation networks within the time domain.

[0068] Set a globally unified virtual evaluation timestamp Combined with the time delay coupling matrix Using a certain baseline data stream as a time anchor point, a forced translation and sliding alignment is performed on each heterogeneous data stream in the time domain (i.e., After time-domain shift compensation, the data streams in each path are... Resampling is performed at various times to output the comprehensive state vector that is strictly physically aligned under the virtual evaluation timestamp, denoted as... In this vector, the originally misaligned physical and ecological components are forcibly converged and aligned on the absolute time axis. time.

[0069] The aforementioned time-domain cross-correlation alignment mechanism for multi-source heterogeneous data strictly follows the objective evolution law of non-periodic engineering signals, avoids artificially created time axis distortion, and provides a rigorous guarantee for the absolute physical synchronization of the multi-dimensional boundary conditions on which the subsequent carbon metabolism map relies for cascade extrapolation.

[0070] In this embodiment, in order to address the systemic blind spots that are easily overlooked in assessments of a single entity, such as the surge in carbon emissions in the surrounding environment caused by engineering construction or train operation, a global transmission network spanning physical entities and the ecological environment was constructed, realizing the dynamic extrapolation and precise quantification of cascaded carbon emissions.

[0071] In this invention, a multimodal dynamic carbon metabolism map is constructed based on the integrated state vector and the direct carbon emissions. This multimodal dynamic carbon metabolism map is essentially a complex directed graph network with spatial geographic constraints. The system maps the previously generated four-dimensional spatiotemporal voxels one by one to the network nodes in the multimodal dynamic carbon metabolism map.

[0072] After node mapping is completed, the direct carbon emissions are used as the initial capacity feature of the network nodes. That is, for any node in the graph... The direct carbon emissions obtained by analyzing physical waveform features under the current virtual assessment timestamp. Assign it an absolute baseline emission level.

[0073] Based on the physical perturbation relationships between the four-dimensional spatiotemporal voxels, a directed propagation edge is established between the source node generating the perturbation and the target node being perturbed. During the construction or operation of urban railway projects, physical perturbations inevitably radiate to the surrounding area. For example, construction encroachment leads to a redistribution of traffic flow in the surrounding road network and causes congestion, or large-scale underground excavation causes changes in the direction of the groundwater system, thus affecting the vegetation carbon sequestration rate. This cross-voxel causal relationship is visualized in the graph as a causal relationship starting from the source node... Point to target node The directed edge of the conduction.

[0074] Following the aforementioned topology, the propagation weights of the directed edges are calculated based on the disturbance intensity of the source node and the spatial distance attenuation function. The propagation of physical disturbances exhibits objective energy dissipation patterns in real geographic space. The calculation process for the propagation weights is as follows:

[0075] In the formula, Indicates that it is from the source node Point to target node Transmission weight; Indicates the source node The initial intensity of the disturbance released outward from itself; Indicates based on source node With the target node Spatial distance between Construct a spatial distance decay function.

[0076] Specifically, the spatial distance decay function uses a Gaussian distance decay model to characterize the nonlinear dissipation characteristics of the physical field, and its expanded expression is as follows:

[0077] In the formula, This parameter represents the spatial influence radius corresponding to a specific type of physical disturbance. It controls the boundary range affected by the disturbance, ensuring that redundant nodes that are too far away do not participate in cascading operations, thus greatly reducing the computational cost of full graph traversal while maintaining physical rigor.

[0078] Along the directed edge of the transmission, and combining the corresponding feature data of the target node in the comprehensive state vector, the transmission weight of the directed edge of the transmission is substituted into a preset nonlinear response function for mapping calculation. Due to the objective physical capacity limitations of the disturbed node (such as the saturation limit of the maximum vehicle traffic capacity of the road), the carbon emission feedback of the target node to external disturbances does not increase linearly indefinitely.

[0079] The secondary cascade carbon emission increment at the target node caused by the source node disturbance is obtained. The calculation process for this secondary cascade carbon emission increment is expressed as follows:

[0080] In the formula, Represents the target node The passively generated secondary cascade carbon emission increments; This represents a preset nonlinear response function; Represents the target node The basic feature data extracted from the integrated state vector after alignment and reconstruction are used to reflect the current vulnerability and disturbance resistance of the node.

[0081] In this embodiment, the nonlinear response function An improved logistic saturation activation function is used, and its specific mapping formula is expressed as follows:

[0082] In the formula, Represents the target node The maximum potential carbon emission capacity limit under the current state data constraints; Indicates the disturbance sensitivity amplification factor; Indicates from feature data The target node state susceptibility parameters are extracted from the data.

[0083] Specifically, the target node state susceptibility parameter It is through multidimensional state feature data The specific analytical mapping formula obtained from the preset feature capacity saturation mapping model is expressed as follows:

[0084] In the formula, Representing feature data The total dimension of effective sensitive features participating in cascaded simulations; Representing multidimensional vectors The first in The current real-time value of each feature component (e.g., node) (Current road traffic volume or exposed area of ​​the foundation pit); Indicates the relationship with the first The theoretical capacity benchmark value of the physical environment corresponding to each characteristic component (e.g., the maximum design traffic capacity of the road or the total area of ​​the plot). Indicates that for the first Each feature component has a preset sensitivity weight constant, and satisfies .

[0085] Through this mapping formula, the system can map high-dimensional node state vectors with different physical dimensions. The dimensionality reduction transforms it into a dimensionless scalar with values ​​between [0,1] (or higher). This parameter mathematically quantifies the current physical or ecological saturation of the target node: the higher the node's saturation (i.e., the higher the saturation level), the more saturated the node becomes (i.e., the higher the saturation level). The closer ),That The larger the value, the more likely it is to be substituted into the aforementioned... After the function, the node responds to external physical disturbances ( This will exhibit a more dramatic nonlinear surge in secondary carbon emissions.

[0086] By constructing a multimodal dynamic carbon metabolism map and performing network cascade simulations, this scheme breaks through the technical bottleneck of traditional carbon footprint tracking, which only focuses on a single source. It scientifically and accurately quantifies the "ripple" secondary ecological costs emanating from the core project, and restores the full picture of the real disturbance of urban transportation infrastructure to carbon flow in urban microsystems.

[0087] In this embodiment, considering the multi-year cyclical characteristics of urban railways and the dynamic evolution of ecological carbon balance, a spatiotemporal integral settlement mechanism and a nonlinear carbon sink carry-over model that spans the entire life cycle of the project are constructed.

[0088] In this invention, based on the direct carbon emissions and the secondary cascade carbon emission increments in the multimodal dynamic carbon metabolism map, a comprehensive assessment result of carbon emissions across all stages is output through spatiotemporal integration. First, the total system carbon emissions of the multimodal dynamic carbon metabolism map at the virtual assessment timestamp are calculated.

[0089] The total carbon emissions of the system are calculated as follows: the static entity implicit carbon emissions of all network nodes in the multimodal dynamic carbon metabolism map are obtained by multiplying and summing the characteristic indicators in the physical entity attribute components with their matching positive carbon emission conversion constants, and then summing the sum with the direct carbon emissions. This is then added to the sum of the secondary cascade carbon emission increments corresponding to all the disturbed target nodes, and finally subtracted from the dynamic carbon absorption amount obtained by the network nodes that are identified as having carbon sink attributes by the carbon polarity labeling, based on their ecological environment attribute components in the attribute definition.

[0090] The above calculation process for the total carbon emissions of the system, which integrates physical carbon source emissions and ecological carbon sink absorption, is expressed as follows:

[0091] In the formula, Indicates the time stamp of the given virtual evaluation. The total carbon emissions of the entire system in the multimodal dynamic carbon metabolism map; This represents the set of all network nodes in the graph that possess carbon source attributes; Represents the nodes in the set Direct carbon emissions; Represents a node The implicit carbon emissions of the static entity are equal to the characteristic indicators of each physical entity attribute component of the node when constructing the multidimensional attribute state vector. Matching positive carbon emission conversion constant The product summation term (which essentially corresponds to the one mentioned earlier) part); This represents the set of all target nodes affected by external physical disturbances; Indicates the disturbed target node The passively generated secondary cascade carbon emission increments; This represents a set of ecological nodes that are identified as having carbon sink properties based on carbon polarity. Nodes representing carbon sink attributes The dynamic carbon uptake is obtained by converting its ecological environment attribute components (such as the carbon sequestration rate of vegetation photosynthesis) at the current timestamp.

[0092] The time series span of the virtual assessment timestamp is defined as the entire life cycle of the project, including the construction and operation phases. The carbon emission characteristics of urban railways exhibit fundamental physical differences at different life cycle stages. The construction phase is mainly characterized by high-intensity civil engineering energy consumption and mechanical work, while the operation phase is mainly characterized by train traction energy consumption and traffic flow radiation.

[0093] The system identifies the transition point from the construction phase to the operation phase throughout the project's entire lifecycle. When the system determines that this transition point has been reached, the directed edges corresponding to the construction phase are removed. Specifically, this involves disconnecting the physical disturbance edges radiating from phased project nodes such as large temporary construction equipment and foundation pit excavation faces in the multimodal dynamic carbon metabolism map, and converting them into directed edges corresponding to the operation phase. The converted network topology is then re-established based strictly on the track network planning, train departure frequency, and normalized passenger flow at stations during the operation phase. This adaptive reconstruction mechanism ensures absolute consistency between the carbon metabolism evolution logic and the actual project progress.

[0094] The residual ecological carbon sink value is obtained by extracting the four-dimensional spatiotemporal voxels, which are calibrated as having carbon sink attributes at the end of the construction period, based on the carryover of the dynamic carbon absorption. This residual ecological carbon sink value characterizes the absolute carbon sequestration potential base established after the original surface vegetation was damaged during the construction period and restored through subsequent engineering revegetation operations.

[0095] The residual value of the ecological carbon sink is used as the nonlinear initial boundary condition for calculating the carbon sink during the operation period, and the dynamic carbon uptake during the operation period is updated accordingly. Since ecological restoration and plant community succession are not a simple linear continuation of the pre-construction state, this nonlinear initial boundary condition re-anchors the calculation benchmark for dynamic carbon uptake under each time slice of the operation period by introducing the vegetation restoration growth coefficient and soil carbon pool remodeling parameter.

[0096] Finally, the total carbon emissions of the system corresponding to multiple discrete virtual assessment timestamps throughout the entire lifecycle of the project are calculated by discrete integral summation over time, and the comprehensive carbon emission assessment result for the entire process is output. The process of integral summation over the entire lifecycle is expressed as follows:

[0097] In the formula, It represents the comprehensive assessment results of carbon emissions across all stages, encompassing both temporal breadth and spatial depth; This represents the total number of discrete virtual evaluation timestamps throughout the entire lifecycle of the project; Indicates the first A virtual evaluation timestamp; This indicates the time span between adjacent virtual evaluation timestamps; Indicates the first Total system carbon emissions calculated based on a multimodal dynamic carbon metabolism map under a virtual assessment timestamp.

[0098] By introducing a time-series-based discrete integral calculation model and a carbon sink carry-over mechanism, this scheme completely eliminates the data gaps caused by the physical handover of engineering sections or the conversion of construction and operation identities, and realizes the continuous flow and non-destructive traceability of the holographic carbon ledger from groundbreaking to long-term service.

[0099] In this embodiment, in order to intuitively demonstrate the execution logic and multi-dimensional data calculation process of the above-mentioned urban railway transportation carbon emission assessment method in a real and complex engineering environment, a typical construction and surrounding environment physical interaction scenario is set up for complete simulation.

[0100] In this invention, the construction phase of the starting shaft of an underground shield tunnel section of a municipal railway and the traffic management phase of the surrounding main roads are selected as a specific observation window. During this phase, the system first constructs a four-dimensional spatiotemporal voxel grid under a set three-dimensional space and continuous time slice, and then discretizes and delimits the high-frequency physical disturbance region within the space.

[0101] The system identifies and locks three core spatial interaction entities, and maps them as network nodes: the underground construction area containing the tunnel boring machine's main drive system and the excavated soil removal equipment is defined as the source node voxel. The adjacent urban arterial roads that experience traffic congestion due to road construction are defined as target node voxels. The green belts along the river that are affected by fluctuations in groundwater levels and dust are defined as ecological node voxels. .

[0102] In the grid attribute definition phase, based on the calculation rules of the instantaneous net carbon flux, the source node voxels... Due to its extremely high energy consumption, the physical entity component is labeled as a carbon source polarity; target node voxel Because it carries fuel-powered traffic flow, it is also categorized as a carbon source polarity; ecological node voxels Therefore, due to its ecological environment attribute of vegetation photosynthesis, it is labeled as a carbon sink polarity.

[0103] In the direct carbon emissions assessment phase, the system's front-end sensors collect data in real time from voxels deployed at the source nodes. The high-frequency transient current waveform signal of the cutterhead motor of the internal shield tunneling machine. The waveform signal is subjected to unsupervised dimensionality reduction mapping using an autoencoder network. By combining the loss function and the hidden layer projection, the fundamental frequency interference of the normal power grid is accurately removed, and the standardized feature components that characterize the overload distortion of the motor when the cutterhead cuts hard rock (corresponding to internal heating and mechanical loss) are extracted.

[0104] The extracted standardized feature components and preset physical transformation weight parameters are substituted into the product summation formula to calculate the dynamic carbon polarity coefficient for this specific overload condition. Subsequently, The actual power consumption parameters of the tunnel boring machine during the corresponding time period Multiply to obtain the source node voxel direct carbon emissions This direct calculation process breaks away from empirical formulas and truly reflects the surge in carbon emissions caused by the reduced efficiency of mechanical work under harsh geological conditions.

[0105] Due to significant order-of-magnitude barriers in the millisecond-level current sampling period of the tunnel boring machine (TBM), the minute-level monitoring period of traffic flow on main roads, and the hour-level assessment period of carbon absorption in green belts, the system acquires a multi-source heterogeneous raw data stream matrix. To overcome the technical bottleneck of being unable to directly perform frequency domain coupling analysis on heterogeneous data across different order-of-magnitude scales due to the lack of overlapping frequency bands, the system parses the millisecond-level high-frequency physical data stream (such as current waveforms) and the minute-level traffic flow data into independent time series. Following the maximum time step of the virtual assessment timestamp, the high-frequency data stream sequences are low-pass filtered and resampled using feature envelopes. After ensuring that the multiple independent one-dimensional data stream sequences have the same time resolution, the system directly calculates the cross-correlation function between the TBM transient current waveform sequence and the surrounding road network traffic flow fluctuation sequence in the time domain. By extracting the peak value of the cross-correlation function, the system accurately captures the actual time delay caused by the redistribution of surrounding traffic flow due to TBM construction and constructs a time delay coupling matrix accordingly.

[0106] Based on the set global virtual evaluation timestamp The system, in conjunction with the aforementioned time delay coupling matrix, performs translation compensation and resampling alignment on the misaligned data streams in the time domain, reconstructing and generating a comprehensive state vector. This time delay alignment mechanism completely eliminates asynchronous misalignment of cross-disciplinary monitoring data, ensuring causal synchronization between construction operations and traffic congestion on the absolute time axis.

[0107] In the multimodal dynamic carbon metabolism mapping calculation stage, the system uses source node voxels. Starting from the target node voxel Establish a directed edge for transmission. This directed edge physically maps the process of traffic spillover and idling congestion caused by construction road closures.

[0108] Based on source node voxels The extent and intensity of the obstruction were determined, and the voxel pointing towards the target node was calculated using the Gaussian distance decay function. Transmission weight The transmission weight is then correlated with the target node voxel. By substituting its own basic traffic flow saturation characteristic data into the improved logistic nonlinear response function, the secondary cascade carbon emission increment caused by the extra idling work of fuel vehicles due to construction congestion can be accurately derived. .

[0109] In the spatiotemporal integration evaluation phase, this specific virtual evaluation timestamp is used. The system summarizes dynamic carbon emission and carbon sink data within the graph. The total carbon emission calculation model for the microsystem at this moment is expressed as:

[0110] In the formula, Indicates the virtual evaluation timestamp Total carbon emissions of the system within the next observation window; Represents the source node voxel The static carbon emissions from the engineering building materials, such as tunnel segments, included within the tunnel; This represents the source node voxel obtained from waveform fingerprint parsing. Direct carbon emissions; This represents the target node voxel derived from the network graph. The increase in secondary cascade carbon emissions; Representing ecological node voxels The dynamic carbon uptake is calculated based on the current leaf area index of its vegetation.

[0111] When the project's progress transitions from the civil construction phase of that section to the full-line operation phase, the system triggers a full lifecycle state topology update. At this time, the system automatically removes the source node voxels from the multimodal dynamic carbon metabolism map. With the target node voxel The directed edges representing the road construction are transformed into new operational directed edges that radiate passenger flow from the station to surrounding bus connection points.

[0112] At the same time, the system extracts the voxels of ecological nodes at the end of the construction period. The residual value of the ecological carbon sink established through the revegetation operation is seamlessly transferred into the nonlinear initial boundary conditions of the calculus equations for the operation period. Based on the continuous iteration of this micro-sand table calculation logic, the system performs time-dimensional integral summation on discrete time points, and finally outputs a comprehensive assessment result of carbon emissions throughout the entire process, supported by solid physical data and rigorous time causal extrapolation.

[0113] In this embodiment, based on the same inventive concept, the present invention also provides a carbon emission assessment system for all aspects of urban rail transit. This system, as a physical entity combining hardware and software, is configured to implement the carbon emission assessment method for all aspects of urban rail transit described in any of the above embodiments. The various functional modules deployed within the system work collaboratively to support the underlying logic of carbon emission assessment in cross-scale spatiotemporal data processing.

[0114] In this invention, the urban rail transit carbon emission assessment system specifically includes a grid construction and calibration module, a waveform feature extraction module, a data entanglement and alignment module, a graph calculation module, and a full-cycle comprehensive assessment module. Each module achieves secure interaction and state quantification transmission of multi-dimensional data through an internal high-speed bus and shared memory mechanism.

[0115] The mesh construction and calibration module is used to construct a four-dimensional spatiotemporal voxel grid covering the entire urban railway line and its environmental radiation area, and to define the attributes of the four-dimensional spatiotemporal voxels in the grid and calibrate their carbon polarity, including carbon sink identification. This module relies on a pre-set three-dimensional geographic information engine and a clock divider to deconstruct the complex macroscopic physical world into structured multi-dimensional attribute state vectors.

[0116] The waveform feature extraction module is used to acquire the physical waveform signals of the target dynamic device, extract carbon feature fingerprints through an unsupervised dimensionality reduction algorithm, and calculate the direct carbon emissions of the corresponding four-dimensional spatiotemporal voxels based on the extracted carbon feature fingerprints. This module directly interfaces with the field sensor array at the hardware level, utilizing the internally encapsulated autoencoder hidden layer projection function and dissipative mapping operator to complete the dimensionality reduction transformation from physical waveforms to dynamic carbon polarity coefficients.

[0117] The direct carbon emission calculation engine built into the waveform feature extraction module follows the following evaluation logic:

[0118] In the formula, This represents the direct carbon emissions of the target dynamic device within the corresponding four-dimensional spatiotemporal voxel; This represents the dynamic carbon polarity coefficient obtained based on the mapping between the carbon feature fingerprint and the physical transformation weight parameters; This represents the operational activity level parameters of the target dynamic equipment, which are transmitted back in real time by field metering instruments.

[0119] The data entanglement and alignment module is used to acquire multi-source heterogeneous raw data stream matrices with different sampling periods. It then parses these matrices into multiple independent data stream sequences, calculates the cross-correlation functions between each pair of these sequences in the time domain to construct a time delay coupling matrix, and uses this time delay coupling matrix to reconstruct and align the data streams in the time domain, generating a comprehensive state vector based on a virtual evaluation timestamp. This module undertakes the crucial task of "time stitching" between heterogeneous systems, eliminating timing misalignment errors during multi-stream data interaction through the mathematical core of time-domain cross-correlation analysis.

[0120] The graph computation module is used to construct a multimodal dynamic carbon metabolism graph based on the comprehensive state vector and the direct carbon emissions, and to extrapolate and calculate the secondary cascade carbon emission increments between network nodes in the multimodal dynamic carbon metabolism graph. This module dynamically maintains a directed graph data structure in memory and continuously updates the directed edges and physical transmission weights between nodes.

[0121] The graph calculation module performs cascade operations by calling a preset nonlinear response function to obtain the secondary cascade carbon emission increment of the target node caused by the perturbation of the source node. Its built-in calculation rules are expressed as follows:

[0122] In the formula, Represents the target node The increase in secondary cascade carbon emissions; Represents a nonlinear response function; This represents the weight propagated from the source node to the target node; Represents the target node The corresponding feature data in the comprehensive state vector.

[0123] The full-cycle comprehensive assessment module is used to calculate and output the comprehensive assessment result of carbon emissions throughout the entire process based on the direct carbon emissions and the secondary cascade carbon emission increments in the multimodal dynamic carbon metabolism map through spatiotemporal integration. This module has both state monitoring and full life cycle settlement functions, and is responsible for identifying the switching moments of the engineering phases and performing the carry-over and update of the nonlinear initial boundary conditions of carbon sinks.

[0124] The time-dimensional discrete integral summation calculation logic ultimately invoked by the full-cycle comprehensive evaluation module is expressed as follows:

[0125] In the formula, This indicates the comprehensive assessment results of carbon emissions across all stages; This represents the total number of discrete virtual evaluation timestamps; Indicates the first A virtual evaluation timestamp; Indicates the time slice step size; This represents the total carbon emissions of the system calculated at the corresponding timestamp.

[0126] Furthermore, embodiments of the present invention also provide a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it implements all the steps of the aforementioned method for assessing carbon emissions across the entire process of urban rail transit. The computer-readable storage medium includes non-volatile memory and volatile memory, such as a solid-state drive, read-only memory, or random access memory.

[0127] This invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to drive the logical operation of the aforementioned system modules, completing time-domain alignment, spatial map mapping, and full-cycle dynamic simulation of massive multi-source heterogeneous time-series data, ultimately outputting a comprehensive carbon emission assessment result reflecting the characteristics of real engineering physical disturbances.

[0128] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for assessing carbon emissions across all stages of urban rail transit, characterized in that, Includes the following steps: A four-dimensional spatiotemporal voxel grid covering the entire urban railway line and its environmental radiation area is constructed, and the attributes of the four-dimensional spatiotemporal voxels in the four-dimensional spatiotemporal voxel grid are defined and carbon polarity is calibrated, including carbon sink identification. The physical waveform signals of the target dynamic device are collected, including instantaneous current, voltage waveforms, or high-frequency vibration signals. Carbon feature fingerprints are extracted by projection using an unsupervised dimensionality reduction algorithm based on a pre-trained solidified mapping substrate. The direct carbon emissions of the corresponding four-dimensional spatiotemporal voxels are calculated by combining the extracted carbon feature fingerprints. The solidified mapping substrate is a reference feature space with fixed physical correspondences constructed by unsupervised iterative training of an autoencoder network. A multi-source heterogeneous raw data stream matrix with different sampling periods is obtained. The multi-source heterogeneous raw data stream matrix is ​​parsed into multiple independent data stream sequences. The cross-correlation function between each pair of the data stream sequences is calculated in the time domain to construct a time delay coupling matrix. The time delay coupling matrix is ​​used to translate and reconstruct alignment in the time domain to generate a comprehensive state vector based on a virtual evaluation timestamp. The multi-source heterogeneous raw data stream consists of waveform sampling data from high-frequency dynamic equipment and ecological flow data from low-frequency environmental monitoring stations. The four-dimensional spatiotemporal voxels are mapped as network nodes in a multimodal dynamic carbon metabolism map, and the direct carbon emissions are used as the initial capacity feature of the network nodes. Based on the physical perturbation relationship between the four-dimensional spatiotemporal voxels, a directed edge is established between the source node that generates the perturbation and the target node that is perturbed, and the transmission weight of the directed edge is calculated based on the perturbation intensity of the source node and the spatial distance attenuation function. Along the directed edge of the transmission, combined with the corresponding feature data of the target node in the comprehensive state vector, the transmission weight of the directed edge of the transmission is substituted into a preset nonlinear response function for mapping calculation to obtain the secondary cascade carbon emission increment of the target node caused by the disturbance of the source node. Based on the implicit carbon emissions of static entities in the multimodal dynamic carbon metabolism map, the direct carbon emissions, and the incremental secondary cascade carbon emissions, a comprehensive assessment result of carbon emissions across all stages is output through spatiotemporal integration. The steps of constructing a four-dimensional spatiotemporal voxel grid covering the entire urban railway line and its environmental radiation area, and defining the attributes of the four-dimensional spatiotemporal voxels in the grid and including carbon polarity labeling for carbon sink identification, specifically include: The entire urban railway line and its environmental radiation zone are discretized and modeled using a three-dimensional spatial coordinate system combined with a continuous time dimension, generating multiple four-dimensional spatiotemporal voxels containing independent spatial coordinates and time slices; For each of the four-dimensional spatiotemporal voxels, physical entity attribute components and ecological environment attribute components are spliced ​​together to generate a multi-dimensional attribute state vector; the physical entity attribute components include building material type and equipment ledger; the ecological environment attribute components include vegetation coverage and soil type. Based on the project progress and the material interaction process within the four-dimensional spatiotemporal voxel, a label polarity is assigned to the four-dimensional spatiotemporal voxel, and the label polarity is specifically classified as carbon source, neutral or carbon sink.

2. The method for assessing carbon emissions across all stages of urban rail transit according to claim 1, characterized in that, The steps of acquiring the physical waveform signal of the target dynamic device, extracting carbon feature fingerprints using an unsupervised dimensionality reduction algorithm, and calculating the direct carbon emissions of the corresponding four-dimensional spatiotemporal voxels based on the extracted carbon feature fingerprints specifically include: The physical waveform signal of the target dynamic device is acquired in real time under a preset sampling time sequence; Using a single complete operation cycle or a fixed power frequency cycle of the target dynamic equipment as the data capture window, the physical waveform signal is segmented to construct a high-dimensional time-series input matrix; An autoencoder network is used as the core architecture of the unsupervised dimensionality reduction algorithm. Iterative training is performed using a joint loss function that includes signal reconstruction mean square error loss and feature non-correlation penalty term. Dimensionality reduction projection is performed through the hidden layer of the autoencoder network to filter out environmental noise background and basis features, extract the dimensionality reduction feature vector that represents the non-stationary working state of the device, and generate the carbon feature fingerprint containing multiple standardized feature components. Each standardized feature component is multiplied and summed with a preset physical conversion weight parameter, and a local environmental compensation constant is added to obtain the dynamic carbon polarity coefficient; wherein, the preset physical conversion weight parameter is calibrated based on a pre-constructed bench test database, which includes synchronously collected transient active power consumption increments, equipment heat dissipation and energy carbon emission factors. The direct carbon emissions of the corresponding four-dimensional spatiotemporal voxel are obtained by multiplying the dynamic carbon polarity coefficient by the operational activity level of the target dynamic device during the time period corresponding to the preset sampling time series.

3. The method for assessing carbon emissions across all stages of urban rail transit according to claim 1, characterized in that, The steps of obtaining a multi-source heterogeneous raw data stream matrix with different sampling periods, parsing the multi-source heterogeneous raw data stream matrix into multiple independent data stream sequences, calculating the cross-correlation function between each pair of the data stream sequences in the time domain to construct a time delay coupling matrix, and using the time delay coupling matrix to translate and reconstruct alignment in the time domain to generate a comprehensive state vector based on a virtual evaluation timestamp specifically include: The multi-source heterogeneous original data stream matrix is ​​parsed and split into multiple independent one-dimensional data stream sequences according to the data source; In the time domain, calculate the cross-correlation function between each pair of the multiple independent one-dimensional data stream sequences; Extract the peak value of the cross-correlation function, obtain the time delay corresponding to the peak value, and construct a time delay coupling matrix to characterize the time lag coupling relationship of the physical process. A globally unified virtual evaluation timestamp is set, and each of the one-dimensional data stream sequences is shifted and aligned in the time domain according to the time delay amount in the time delay coupling matrix, and the comprehensive state vector is resampled and output.

4. The method for assessing carbon emissions across all stages of urban rail transit according to claim 1, characterized in that, The steps of constructing a multimodal dynamic carbon metabolism map based on the integrated state vector and the direct carbon emissions, and extrapolating and calculating the secondary cascade carbon emission increments among the network nodes in the multimodal dynamic carbon metabolism map, specifically include: The four-dimensional spatiotemporal voxels are mapped to the network nodes in the multimodal dynamic carbon metabolism map, and the direct carbon emissions are used as the initial capacity feature of the network nodes. Based on the physical perturbation relationship between the four-dimensional spatiotemporal voxels, a directed edge is established between the source node that generates the perturbation and the target node that is perturbed, and the transmission weight of the directed edge is calculated based on the perturbation intensity of the source node and the spatial distance attenuation function. Along the directed transmission edge, combined with the corresponding feature data of the target node in the comprehensive state vector, the transmission weight of the directed transmission edge is substituted into a preset nonlinear response function for mapping calculation to obtain the secondary cascade carbon emission increment of the target node caused by the disturbance of the source node.

5. The method for assessing carbon emissions across all stages of urban rail transit according to claim 4, characterized in that, The step of outputting a comprehensive assessment result of carbon emissions across all stages through spatiotemporal integration based on the implicit carbon emissions of static entities in the multimodal dynamic carbon metabolism map, the direct carbon emissions, and the incremental secondary cascade carbon emissions includes: Calculate the total system carbon emissions of the multimodal dynamic carbon metabolism map at the virtual evaluation timestamp; The total carbon emissions of the system are calculated as follows: the static entity implicit carbon emissions of all network nodes in the multimodal dynamic carbon metabolism map are obtained by multiplying and summing the characteristic indicators in the physical entity attribute components with their matching positive carbon emission conversion constants, and then summing the sum with the direct carbon emissions. This is then added to the sum of the secondary cascade carbon emission increments corresponding to all the disturbed target nodes. Finally, the dynamic carbon absorption amount obtained by the network nodes that are identified as having carbon sink attributes by the carbon polarity labeling is based on their ecological environment attribute components in the attribute definition.

6. The method for assessing carbon emissions across all stages of urban rail transit according to claim 5, characterized in that, The step of outputting a comprehensive assessment result of carbon emissions across all stages through spatiotemporal integration based on the implicit carbon emissions of static entities in the multimodal dynamic carbon metabolism map, the direct carbon emissions, and the incremental secondary cascade carbon emissions further includes: The time series span of the virtual evaluation timestamp is defined as the entire life cycle of the project, including the construction and operation periods; Identify the transition point from the construction phase to the operation phase throughout the entire lifecycle of the project, remove the directed edges corresponding to the construction phase, and transform them into directed edges corresponding to the operation phase; The residual value of ecological carbon sink obtained by carrying over the dynamic carbon absorption amount based on the four-dimensional spatiotemporal voxel that is calibrated as having carbon sink attributes at the end of the construction period is used as the nonlinear initial boundary condition for carbon sink calculation during the operation period, and the dynamic carbon absorption amount during the operation period is updated accordingly. The total carbon emissions of the system corresponding to multiple discrete virtual evaluation timestamps throughout the entire life cycle of the project are calculated by discrete integral summation over time, and the comprehensive carbon emission evaluation result of the entire process is output.

7. A carbon emission assessment system for the entire process of urban rail transit, characterized in that, The system for implementing the carbon emission assessment method for all aspects of urban rail transit as described in any one of claims 1 to 6 includes: The grid construction and calibration module is used to construct a four-dimensional spatiotemporal voxel grid covering the entire urban railway line and the environmental radiation area, and to define the attributes of the four-dimensional spatiotemporal voxels in the four-dimensional spatiotemporal voxel grid and calibrate the carbon polarity including carbon sink identification. The waveform feature extraction module is used to collect the physical waveform signal of the target dynamic device, and extract carbon feature fingerprints by projecting the unsupervised dimensionality reduction algorithm of the pre-trained solidified mapping substrate. The direct carbon emissions of the corresponding four-dimensional spatiotemporal voxel are calculated by combining the extracted carbon feature fingerprints. The data entanglement and alignment module is used to obtain a multi-source heterogeneous original data stream matrix with different sampling periods, parse the multi-source heterogeneous original data stream matrix into multiple independent data stream sequences, calculate the cross-correlation function between each pair of the data stream sequences in the time domain to construct a time delay coupling matrix, and use the time delay coupling matrix to translate and reconstruct alignment in the time domain to generate a comprehensive state vector based on a virtual evaluation timestamp. The graph calculation module is used to construct a multimodal dynamic carbon metabolism graph based on the integrated state vector and the direct carbon emissions, and to deduce and calculate the secondary cascade carbon emission increments among the network nodes in the multimodal dynamic carbon metabolism graph. The full-cycle comprehensive assessment module is used to output a comprehensive assessment result of carbon emissions throughout the entire process by spatiotemporal integration based on the implicit carbon emissions of static entities in the multimodal dynamic carbon metabolism map, the direct carbon emissions, and the incremental secondary cascade carbon emissions.

8. A carbon emission assessment system for all aspects of urban rail transit according to claim 7, characterized in that, The graph calculation module is specifically used for: The four-dimensional spatiotemporal voxels are mapped to network nodes in the multimodal dynamic carbon metabolism map; a directed transmission edge is established between the source node that generates the disturbance and the target node that is disturbed, and the transmission weight is calculated; along the directed transmission edge, combined with the corresponding feature data of the target node in the comprehensive state vector, the transmission weight of the directed transmission edge is substituted into a preset nonlinear response function for mapping calculation to obtain the secondary cascade carbon emission increment of the target node caused by the disturbance of the source node.

9. A carbon emission assessment system for all aspects of urban rail transit according to claim 8, characterized in that, The full-cycle comprehensive evaluation module is specifically used for: The total carbon emissions of the system in the multimodal dynamic carbon metabolism map at the virtual evaluation timestamp are calculated as follows: the static entity implicit carbon emissions of all network nodes in the multimodal dynamic carbon metabolism map are obtained by multiplying and summing the characteristic indicators in the physical entity attribute components with their matching positive carbon emission conversion constants, and summing them with the direct carbon emissions, then adding the sum of the secondary cascade carbon emission increments corresponding to all the disturbed target nodes, and finally subtracting the dynamic carbon absorption of the network nodes that are identified as having carbon sink attributes by the carbon polarity labeling based on their ecological environment attribute components in the attribute definition. The time series span of the virtual evaluation timestamp is defined as the entire life cycle of the project, including the construction period and the operation period; the switching moment from the construction period to the operation period of the entire life cycle of the project is identified, the directed transmission edge corresponding to the construction period is removed and transformed into a directed transmission edge corresponding to the operation period; the ecological carbon sink residual value obtained by the four-dimensional spatiotemporal voxel with carbon sink attributes at the end of the construction period is extracted based on the carry-over of the dynamic carbon absorption as the nonlinear initial boundary condition for the carbon sink calculation in the operation period, and the dynamic carbon absorption in the operation period is updated accordingly; The total carbon emissions of the system corresponding to multiple discrete virtual evaluation timestamps throughout the entire life cycle of the project are calculated by discrete integral summation over time, and the comprehensive carbon emission evaluation result of the entire process is output.

Images