A Big Data-Based Rating Method for Green and Smart Ports

Abstract

This invention discloses a green and smart port rating method based on big data, specifically in the field of port green rating. It addresses the lack of analysis on the impact of coordinated scheduling on carbon emissions in existing port green assessments. By collecting real-time data from multiple sources, a multimodal data pool is constructed across the entire transportation chain, enabling carbon emission tracing across transportation entities and generating a seamless carbon trajectory chain. A multi-agent strategy matrix is ​​used to simulate port-transportation coordination scenarios, extracting dynamic impact data on carbon emissions under different decision combinations, and calculating coordination efficiency indicators such as the supply chain carbon intensity variation coefficient and anchorage idling index. Furthermore, a regression model is used to establish a mapping relationship between decision-making behavior and coordination efficiency, predicting green coordination for future operational scheduling and generating a port green coordination rating. This achieves a dynamic, traceable, and predictable green evaluation mechanism, providing quantitative references for optimizing port scheduling strategies and enhancing the port's low-carbon collaborative governance capabilities.

Description

Technical Field

[0001] This invention relates to the field of green port rating technology, and more specifically, to a green and smart port rating method based on big data. Background Technology

[0002] Currently, global ports are undergoing an accelerated transformation from traditional loading and unloading hubs to smart and green hubs. However, existing port green assessments mainly remain at the static statistical level, relying on energy consumption data reported by ports themselves or carbon emission data self-declared by enterprises. They lack the ability to dynamically correlate carbon activities of multiple participating entities, making it difficult to accurately reflect the differences in carbon emissions during cross-entity collaboration processes such as maritime transport, land transport, and warehousing. Furthermore, as a supply chain hub, ports involve frequent cross-entity container transfers, resulting in highly discrete and fragmented carbon emission data. Existing systems cannot achieve precise carbon flow tracing at the container level. Simultaneously, port operation scheduling and transportation decisions directly impact carbon emission performance. For example, ship waiting at anchor, inefficient scheduling of loading and unloading equipment, and redundant transportation routes can all lead to carbon emission fluctuations, and existing assessment systems lack the ability to extrapolate the relationship between scheduling behavior and carbon emissions.

[0003] Therefore, it is necessary to propose a technical solution that can integrate multimodal data, support collaborative decision-making simulation, and provide forward-looking green ratings, so as to promote the development of port management from "emission statistics" to "carbon efficiency prediction and optimization" and provide scientific basis and quantitative support for the construction of green and smart ports. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a green smart port rating method based on big data to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A big data-based green smart port rating method includes the following steps:

[0007] S1. Based on the data interfaces of port IoT terminals and supply chain participants, collect real-time data streams to build a multimodal data pool for the entire transshipment chain;

[0008] S2. Apply dynamic traceability and anchoring processing to the multimodal data pool, and associate discrete transportation segments across entities based on container electronic tags to generate a seamless carbon trajectory chain with containers as the smallest granularity.

[0009] S3. Employ a multi-agent strategy matrix to simulate supply chain collaboration scenarios, iteratively execute container turnover strategies in a virtual environment, and record the dynamic impact of each agent's decisions on carbon emissions.

[0010] S4. Calculate the supply chain carbon intensity variation coefficient and anchorage idling index based on the dynamic impact dataset to form a set of collaborative effectiveness evaluation indicators.

[0011] S5. Using the multi-agent decision combination in the dynamic impact dataset as input features and the corresponding collaborative effectiveness evaluation index set as output labels, establish a regression model for predicting collaborative effectiveness.

[0012] S6. Predict the collaborative effectiveness of port operation scheduling decisions for future periods and output a port green collaborative rating.

[0013] In a preferred embodiment, step S1, which involves collecting real-time data streams and constructing a multimodal data pool for the entire transshipment chain based on the data interfaces of port IoT terminals and supply chain participants, specifically includes:

[0014] The port's IoT terminals collect real-time container locations and equipment operating status, while simultaneously accessing data interaction interfaces from supply chain participants to obtain the trajectory information of transportation entities.

[0015] A distributed message queue is used to receive real-time data streams of the container's real-time location, equipment operating status, and the trajectory information of the transport entity. The received heterogeneous data is parsed and mapped to generate serialized data units with a unified timestamp.

[0016] Based on container electronic tags, the parsed data units are linked and indexed in the spatiotemporal dimensions to construct a multimodal data pool that integrates records, numerical indicators and spatial coordinates for the entire transportation chain.

[0017] In a preferred embodiment, step S2, which involves applying dynamic traceability and anchoring processing to the multimodal data pool and linking discrete transport segments across entities based on container electronic tags to generate a seamless carbon trajectory chain with the container as the smallest granularity, specifically includes:

[0018] Extract discrete transport segment records with container electronic identification as the key field from the multimodal data pool. The transport segment records contain trajectory information provided by different transport entities.

[0019] Perform spatiotemporal continuity analysis on the discrete transport segment records of each container to identify logical connection points between adjacent transport segments and insert data anchoring markers at the connection points;

[0020] Based on data anchoring markers, scattered transport segments involving multiple transport entities are integrated across entities to form a continuous and complete container movement path;

[0021] The carbon emissions of each transport segment are calculated along the integrated movement path, and the carbon emission data is bound and integrated with the container movement path using a recursive accumulation method.

[0022] Generate a fixed carbon trajectory chain with containers as the tracking unit. This carbon trajectory chain contains the complete movement sequence from the starting point to the ending point and the corresponding cumulative carbon emission data.

[0023] In a preferred embodiment, step S3 employs a multi-agent policy matrix to simulate a supply chain collaboration scenario, iteratively executing container turnover strategies in a virtual environment, and recording the dynamic impact of each agent's decisions on carbon emissions. The dataset specifically includes:

[0024] A virtual collaborative environment for multiple agents is constructed based on the carbon trajectory chain, wherein the agents include port operation equipment, warehousing equipment, and transportation entities of supply chain participants;

[0025] Each type of intelligent agent is initialized and configured, including decision-making rules for transportation selection, docking planning, and loading and unloading scheduling, forming a multi-agent policy matrix;

[0026] Load historical container turnover data from the carbon trajectory chain into a virtual environment to drive multiple agents to execute transportation planning, terminal berthing arrangements, and loading and unloading equipment scheduling decisions according to the strategy matrix.

[0027] A virtual environment is run using a discrete event-driven mechanism to simulate the complete turnover process of containers between port hubs and transportation networks, and to record the decision combinations of each agent in each iteration.

[0028] Collect carbon emission data generated by agent interactions during each decision-making cycle, and construct a dynamic impact dataset that reflects the relationship between decision-making behavior and carbon emissions.

[0029] In a preferred embodiment, step S4, calculating the supply chain carbon intensity variation coefficient and anchorage idling index based on the dynamic impact dataset to form a set of collaborative effectiveness evaluation indicators, specifically includes:

[0030] Extract the total carbon emissions and corresponding freight volume data of the transport segment between each adjacent anchor point of the carbon trajectory chain within the set simulation period from the dynamic impact dataset, and calculate the carbon intensity value of each transport segment.

[0031] Based on the carbon intensity value sequence of all transportation segments, the ratio of the standard deviation to the mean of the sequence is calculated to obtain the carbon intensity variation coefficient of the supply chain.

[0032] Simultaneously extract the operational status records of vessels waiting at anchorage, calculate the proportion of the cumulative time of continuous engine idling to the total time at anchorage, and calculate the anchorage idling index.

[0033] Using the supply chain carbon intensity variation coefficient and anchorage idling index as core indicators, a set of collaborative effectiveness evaluation indicators with multiple measurement dimensions is constructed.

[0034] In a preferred embodiment, step S5, which uses the multi-agent decision combination in the dynamically impacted dataset as input features and the corresponding set of collaborative effectiveness evaluation indicators as output labels, specifically includes:

[0035] Multi-agent decision-making combination data are extracted from the dynamic impact dataset as model input features, and corresponding collaborative effectiveness evaluation index data are extracted simultaneously as output labels.

[0036] Perform feature alignment and data standardization on the input features and output labels to construct a model training sample set;

[0037] The regression model is trained based on the training sample set, and the hyperparameter combination of the model is determined by grid search to establish a nonlinear mapping relationship from decision combination to synergistic effectiveness.

[0038] In a preferred embodiment, step S6, which involves predicting the collaborative effectiveness of port operation scheduling decisions for future periods and outputting a port green collaborative rating, specifically includes:

[0039] Obtain the planned port operation scheduling decision sequence for future time periods and transform it into a standardized decision combination feature vector input into a regression model for predicting collaborative effectiveness;

[0040] The predicted values ​​of the corresponding collaborative effectiveness evaluation indicators output by the regression model are used as green collaborative rating factors.

[0041] The final port green synergy rating result is generated by weighting and comprehensively calculating based on green synergy rating factors.

[0042] The technical effects and advantages of the green smart port rating method based on big data in this invention are as follows:

[0043] By constructing a multimodal data pool covering the entire port-to-supply chain and achieving dynamic emission correlation across transportation entities based on fine-grained carbon trajectory chains for containers, port carbon emission assessment is upgraded from traditional static statistics to transparent and traceable management throughout the entire process, effectively solving the problems of fragmented carbon data and difficulty in unified integration in existing methods. By introducing a multi-agent strategy matrix, port operation scheduling and transportation strategies are simulated and extrapolated, quantifying the actual impact of different decision combinations on carbon emissions, enabling the assessment system to have forward-looking decision-making capabilities and support collaborative optimization. Simultaneously, key indicators such as the supply chain carbon intensity variation coefficient and anchorage idling index are constructed based on carbon emission topology data to achieve a comprehensive measurement of collaborative efficiency and energy utilization status, accurately reflecting the port's green collaboration level in complex logistics environments. This invention further utilizes regression models to achieve mapping prediction from scheduling behavior to collaborative effectiveness, enabling the output of green collaboration ratings in the future operation planning stage, providing a quantitative basis for port scheduling optimization and policy formulation.

[0044] Compared with existing technologies, this invention realizes a green and smart port rating mechanism that integrates precise traceability, dynamic prediction, and collaborative evaluation. It can significantly improve the accuracy of carbon emission management and the collaborative governance capabilities of the supply chain, and help ports develop towards high efficiency, low carbon, and intelligence. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of a green and smart port rating method based on big data according to the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0047] Example 1, Figure 1 This invention presents a green and smart port rating method based on big data, which includes the following steps:

[0048] S1. Based on the data interfaces of port IoT terminals and supply chain participants, collect real-time data streams to build a multimodal data pool for the entire transshipment chain;

[0049] S2. Apply dynamic traceability and anchoring processing to the multimodal data pool, and associate discrete transportation segments across entities based on container electronic tags to generate a seamless carbon trajectory chain with containers as the smallest granularity.

[0050] S3. Employ a multi-agent strategy matrix to simulate supply chain collaboration scenarios, iteratively execute container turnover strategies in a virtual environment, and record the dynamic impact of each agent's decisions on carbon emissions.

[0051] S4. Calculate the supply chain carbon intensity variation coefficient and anchorage idling index based on the dynamic impact dataset to form a set of collaborative effectiveness evaluation indicators.

[0052] S5. Using the multi-agent decision combination in the dynamic impact dataset as input features and the corresponding collaborative effectiveness evaluation index set as output labels, establish a regression model for predicting collaborative effectiveness.

[0053] S6. Predict the collaborative effectiveness of port operation scheduling decisions for future periods and output a port green collaborative rating.

[0054] In S1, a multimodal data pool for the entire transshipment chain is constructed by collecting real-time data streams based on the data interfaces of port IoT terminals and supply chain participants.

[0055] The port continuously collects dynamic attribute data of containers at each stage of port operations through IoT terminals deployed at the port. This includes real-time location information of containers in the yard, among quay cranes, trailers, and ships, specifically using positioning devices to record latitude and longitude coordinates, displacement direction, and movement speed. For quay cranes, gantry cranes, AGVs, and other operating equipment involved in loading and unloading operations, their operating status data is collected, including power output, engine speed, equipment operating time, start-stop status, and energy consumption during the specified period. Based on this, data interaction interfaces with supply chain participants such as railways, trucks, and ships are invoked to obtain the operational trajectory information of the transport entities. The trajectory information includes at least the mode of transport, the unique identifier of the transport vehicle, the coordinates of stops along the route, the transport time sequence, and the loading relationship of the transported containers. The above data is continuously input into the data acquisition module through standardized interfaces. The data acquisition module caches data with irregular arrival orders to ensure stable access capability for data from different sources.

[0056] A distributed message queue is used to receive real-time data streams of container locations, equipment operating status, and transport entity trajectory information. The message queue possesses high-concurrency write capabilities and sequential message maintenance capabilities, strictly maintaining chronological order during data access to handle sudden traffic surges caused by simultaneous data uploads from numerous devices. For heterogeneous data records from different devices and supply chain participants, format parsing is first performed to convert string data into structured record objects. Field names are uniformly mapped; for example, fields identifying different devices are standardized to the same field name, and unit differences are normalized. For instance, different coordinate systems used in location data are unified to the port's commonly used geographic coordinate system through a conversion algorithm, and energy consumption indicators in equipment status are unified to kilowatt-hours. After parsing and mapping, all records are sorted according to timestamps to form a stable temporal sequence. A data source tag is then added to each record to clearly distinguish between data collected by port equipment and data provided by external transport entities. Finally, the processed data is constructed into serialized data units, using a compact storage structure oriented towards time-series computation.

[0057] Based on the electronic identification of containers, spatiotemporal correlation processing is performed on serialized data units, with the electronic identification serving as the unique tracking field for each container in the transportation chain. Specifically, each container's electronic identification serves as the core index, rearranging the corresponding data units by timestamp to identify their relay relationship between different transportation entities. The transition times of key status nodes such as entry into the yard, loading onto the ship, unloading from the ship, and handover to transport vehicles are determined based on the spatiotemporal continuity of the movement path. During the correlation process, loading relationships, corresponding location coordinates at each time point, and equipment operating status are recorded for each trajectory segment to achieve a three-dimensional correlation between containers, transportation entities, and equipment. For missing data segments, trajectory interpolation is used to fill in the continuity, for example, inferring reasonable segments and equipment status along the intermediate path based on valid records before and after, ensuring the complete chain is uninterrupted. Ultimately, a multimodal data pool with containers as the tracking granularity is formed. Each data entry contains real-time recorded data, quantifiable numerical indicators, and precise spatial coordinate information. This data pool comprehensively covers the flow of containers within the port and surrounding transportation network.

[0058] In S2, dynamic traceability and anchoring processing is applied to the multimodal data pool, and discrete transportation segments are associated across entities based on container electronic tags to generate a seamless carbon trajectory chain with containers as the smallest granularity.

[0059] From the established multimodal data pool, using each container's electronic identifier as the unique primary key, data entries recorded for that container at different transportation stages are extracted. These entries constitute the discrete transportation segment records. Each transportation segment record includes a sequence of trajectory coordinates provided by the transportation entity, arrival and departure times, binding information with the means of transport, mode of transport identifier, and the continuous operating status of the equipment during transportation. For example, if the container is transported by trailer, the corresponding transportation segment record includes the trailer's unique identifier, the trailer's route point sequence, and the start and end times of that segment; if it is loaded onto a ship, the corresponding data record includes the ship's berth location, departure time, and voyage information. These records involve different stages such as rail, shipping, and road transport, exhibiting diverse sources and discontinuous temporal distribution. This embodiment forms a set of discrete transportation segment sequences by sorting the records by timestamp.

[0060] For each container's discrete transport segment records, a spatiotemporal continuity analysis is performed. The interval between the end time of each segment and the start time of the next segment is used to determine if a natural connection exists in actual logistics operations. If the time difference is within a set reasonable handover range (e.g., the handover interval between transport entities is less than the preset maximum vehicle turnaround time), the records are considered logically continuous. Simultaneously, location information is used to determine if the termination point of the previous transport segment and the starting point of the next transport segment are located within the same yard or port operation area. If the geographical deviation is within an acceptable range, the connection relationship is further confirmed. After confirming the connection point, data anchoring markers are inserted into the records. These markers store at least three fields: the latitude and longitude coordinates of the connection location, the exact time of the transport entity handover, and the handover role transformation information (e.g., from trailer to quay crane or from quay crane to ship). These markers serve as key nodes for cross-entity fusion in subsequent processing. With the aid of these markers, all discrete transport segments are integrated in chronological order to form a complete cross-entity movement path, clearly connecting all the container's movement trajectories from entering the port to leaving the port, without any breakpoints or logical jumps.

[0061] Carbon emissions are calculated for each transport segment along the integrated and continuous movement path, and the corresponding emissions are linked to path nodes. During the calculation, carbon emission factors corresponding to the energy consumption type are used to extrapolate emissions for different transport entities. For example, trailers are calculated based on fuel consumption, ships are estimated based on main engine operating time and power indicators, and loading / unloading equipment such as quay cranes are converted to carbon emissions based on electricity consumption. Each transport segment is used as the basic cumulative unit. To ensure the integrity of the cumulative logic of the data, the carbon emission results for each segment are recursively accumulated in the order of the path, so that each node records the total cumulative carbon emissions up to that node. The bound and integrated carbon emission data is aggregated into the path data, forming a fixed carbon trajectory chain with containers as the tracking unit. This carbon trajectory chain completely records the entire displacement sequence of the container from entering the origin point to reaching the destination, the transport mode and participating entities for each segment, information on all anchoring nodes, and the corresponding cumulative carbon emission data, achieving high-precision and fine-grained traceability of carbon emission behavior.

[0062] In S3, a multi-agent policy matrix is ​​used to simulate a supply chain collaboration scenario. The container turnover strategy is iteratively executed in a virtual environment, and a dataset of the dynamic impact of each agent's decision on carbon emissions is recorded.

[0063] Based on a constructed carbon trajectory chain with containers as the tracking granularity, a multi-agent virtual collaborative environment is established. This environment fully maps real port and supply chain scenarios, and the agents include three main categories: port operation equipment, warehousing equipment, and external transportation entities. Port operation equipment includes loading and unloading equipment such as quay cranes, gantry cranes, rail-mounted gantry cranes, and AGVs; warehousing equipment includes yard management equipment and allocation tools; and transportation entities of supply chain participants include ships, trailers, and container rail transport vehicles. For each type of agent, its behavioral decision-making capabilities are configured during the simulation initialization phase. Behavioral rules cover specific actions such as transportation selection, execution sequence, stop point planning, and loading / unloading scheduling methods. For example, a transportation entity agent can select different transportation routes according to the plan and invoke loading / unloading equipment for loading and unloading operations during node stops; equipment agents enter start / stop states and execute corresponding actions according to scheduling rules. All behavioral rules are explicitly parameterized and form targeted decision-making constraints, making them executable and derivable. Ultimately, a multi-agent policy matrix is ​​formed by the set of behavioral rules initialized by all agents. Each decision is clearly expressed in terms of execution order, constraints, and responsibility interface, thereby ensuring that after the virtual collaborative environment is established, the multi-agents have a controllable, adjustable, and simulable operational behavior foundation.

[0064] In a virtual collaborative environment, historical container turnover data from the carbon trajectory chain is loaded, using the historical real-world trajectory as the simulation input benchmark to drive various agents to execute corresponding transportation planning, terminal berthing, and loading / unloading scheduling behaviors. The simulation employs a discrete event-driven mechanism, with events triggered by actual transportation operation patterns, such as the arrival of transport entities at nodes, the completion of a loading operation by equipment, or changes in container transfer relationships. Each event has a distinct timestamp, and the entire turnover process is advanced sequentially according to time. During this process, agents execute decisions based on a strategy matrix; different decision sequences will lead to differences in loading / unloading order and waiting time, thus affecting carbon emission performance. To maintain the realism of the simulation scenario, the virtual environment strictly constrains key data in the historical records, including node berthing duration, equipment operating power range, transport vehicle speed range, and loading / unloading time, all of which are not allowed to deviate from the statistical intervals recorded in the historical data. In this way, it is ensured that the simulation process does not deviate from the real port operation patterns. Simultaneously, for each simulation iteration, the scheduling behavior sequence of all agents within the decision-making cycle is recorded, forming multi-agent decision combination data.

[0065] During the discrete event-driven simulation, node carbon emission change data generated during agent interactions within each decision cycle are collected. This change is calculated based on real-time updates of equipment operating status, transportation entity operation time, and energy consumption methods. To ensure data accuracy and validity, this embodiment establishes a comparison system based on measured data from historical carbon trajectory chains. The emission results calculated in each simulation are compared with historical data for error control. If the error exceeds a preset tolerance range, agent behavior or energy consumption parameters are recalibrated based on historical scenario rules to ensure the simulation results strictly conform to real-world business conditions. The collected carbon emission changes, together with the corresponding agent behavior decision sets, constitute a dynamic impact dataset. The structure of the dynamic impact dataset includes: decision cycle number, list of participating agents, behavior category of each agent, interaction node identifier, carbon emission change for the current cycle, and its cumulative position in the overall emission process. This dataset can quantitatively demonstrate the contribution of different decision behaviors to carbon emission growth or suppression, ensuring the simulation output has physical interpretability and business consistency.

[0066] In S4, the supply chain carbon intensity variation coefficient and anchorage idling index are calculated based on the dynamic impact dataset to form a set of collaborative effectiveness evaluation indicators.

[0067] The total carbon emissions and freight volume data corresponding to each adjacent anchor point in the transportation segment are extracted from the dynamic impact dataset. The data is strictly limited to a set simulation period to ensure data integrity and periodic consistency. To ensure the extracted data accurately reflects the actual carbon emissions of the transportation segment, this embodiment uses the transportation segment as the basic analysis unit. The extracted data for each segment includes at least the energy consumption of the transportation entity, the time range corresponding to the emissions, the actual number of containers loaded on the transport vehicle, and the freight volume completed in that segment. Freight volume data is obtained through cargo measurement records or vehicle manifest records to ensure accurate matching between load capacity and transportation links. Subsequently, based on the above total carbon emissions and freight volume data, a carbon intensity value is obtained by performing a ratio calculation between the carbon emission value and the freight volume value, which is used to measure the carbon emission level caused by a unit of transportation output. To maintain data reliability, missing or abnormally fluctuating extreme transportation segment records are deleted during the calculation process to ensure that the obtained carbon intensity sequence reflects the actual operating load and energy utilization of the transportation segment. After extraction and calculation, a carbon intensity sequence containing the carbon intensity values ​​of all transportation segments is generated, and its corresponding path location information is retained.

[0068] Based on the numerical sequence of carbon intensity values ​​for all transportation segments obtained above, a statistical dispersion analysis is further performed on this sequence. First, the mean of the sequence is calculated to obtain the average level of carbon intensity across the entire supply chain. Then, its standard deviation is calculated to measure the degree of unevenness in carbon emission efficiency among different transportation segments. The ratio between the obtained standard deviation and the mean is calculated to make this ratio dimensionless and directly reflect the degree of carbon emission fluctuation in the supply chain; this ratio is defined as the supply chain carbon intensity variation coefficient. This indicator is used to represent the magnitude of the difference in carbon emission efficiency among different transportation segments. The larger the coefficient of variation, the more significant the carbon emission differences among the transportation links within the same supply chain. To ensure the stability and validity of the evaluation results, this embodiment requires the sequence length to meet statistical significance, such as setting a minimum number of consecutive transportation segments not less than a predetermined threshold, for example, not less than 100 transportation segments, to avoid the indicator being unrepresentative due to insufficient sample size.

[0069] Simultaneously, operational status records of vessels during their waiting period at port anchorages are extracted from dynamic impact data to calculate the anchorage idling index. The anchorage waiting period refers to the continuous time a vessel enters the port anchorage but has not yet completed berthing procedures. During this period, the vessel's main engine and auxiliary engines are typically in a non-shutdown state but without propulsion load, consuming fuel and generating carbon emissions. Other port and supply chain transportation entities are not included in the idling waste statistics due to their short downtime and low unit emissions. For each anchorage stay, the cumulative duration of continuous engine idling is recorded, along with the total anchorage stay time, expressing the energy waste from engine idling and power supply activities before berthing as a time percentage. For example, if a vessel waits a total of 10 hours at anchorage, with a cumulative 4 hours of engine idling, the corresponding index value for that stay is 0.4. After multiple statistical analyses of this data, an anchorage idling index combining records from different voyages is formed. This index accurately reflects the unreasonable energy use related to scheduling patterns during vessel waiting and serves as an important supplementary quantitative indicator for the low-carbon level of supply chain collaboration.

[0070] A collaborative performance evaluation index set is jointly constructed using the supply chain carbon intensity variation coefficient and the anchorage idling index as core indicators. By standardizing the index structure, the carbon intensity distribution balance index and the anchorage energy utilization efficiency index are set up side by side in the green collaborative evaluation dimension, and necessary auxiliary index levels are added according to the port business management needs, so that the index system has multi-dimensional coverage.

[0071] In step S5, the multi-agent decision combination in the dynamic influence dataset is used as the input feature, and the corresponding collaborative effectiveness evaluation index set is used as the output label to establish a regression model for predicting collaborative effectiveness.

[0072] Decision combination data formed by multiple agents in each simulation decision-making cycle are extracted from the dynamic impact dataset and used as model input features. This decision combination data consists of the behavioral choices of multiple agents, each represented by a fixed code, such as whether a ship prioritizes berthing, whether a quay crane uses double-work loading and unloading, and whether a trailer goes directly to the handover point. The collaborative decision-making state between agent behaviors is expressed through a behavioral coding matrix. To simultaneously reflect the low-carbon performance corresponding to the decision results, the supply chain carbon intensity variation coefficient and anchorage idling index, aligned at the same time, are extracted from the collaborative effectiveness evaluation index set and combined into an output label vector. The output label is used to train the model to learn the impact of different decision combinations on green collaborative evaluation indicators. To ensure the consistency of the model training data, feature alignment is performed on the input features and output labels, accurately mapping each input record to the indicator value at the same simulation time and decision cycle, ensuring no misaligned or duplicate data matching. After alignment, a unified numerical transformation method is used to normalize and standardize all input features and output labels, making the data from different dimensions comparable. Simultaneously, an anomaly removal strategy is implemented on the dataset to remove data entries with logical anomalies or those that deviate excessively from the statistical distribution, ensuring the stability and statistical representativeness of the training data. Finally, a training sample set containing input behavioral features and corresponding collaborative performance labels is constructed, enabling the model to learn the relationship between behavior and outcome from real-world simulation data.

[0073] A regression model is trained based on a constructed training sample set. By establishing a nonlinear mapping relationship between decision combinations and collaborative effectiveness indicators, the low-carbon performance of future port operation scheduling behavior is predicted. This embodiment uses a regression model built with ensemble learning as the core algorithm architecture. During training, a grid search method is used to fine-tune hyperparameters to ensure the model has optimal learning capabilities. Specifically, the training sample set is first divided into a training set and a validation set, enabling performance supervision during parameter optimization. Then, based on the candidate hyperparameter combination table, different hyperparameter settings are loaded sequentially, such as fixed ranges for the number of decision trees, training depth, and feature splitting complexity. Each group undergoes a complete training and validation process. During training, the model updates its internal structure based on the error feedback between input features and output labels, continuously improving its fitting ability through an error minimization strategy. After training each hyperparameter group, performance on the validation set is evaluated, and the hyperparameter configuration with the best overall performance is selected based on the principle of maximizing performance indicators. After selecting hyperparameters, the model is trained using all training samples to ensure its internal structure stabilizes and converges, resulting in a regression model that accurately expresses the relationship between agent behavior and collaborative effectiveness indicators. The final model is capable of taking decision-making combination data as input and outputting predicted values ​​for the supply chain carbon intensity variation coefficient and anchorage idling index.

[0074] In S6, the collaborative effectiveness of port operation scheduling decisions in future periods is predicted, and a port green collaborative rating is output.

[0075] The port operation scheduling decision sequence planned for future time periods is obtained and used as input data for the collaborative efficiency prediction model. The scheduling sequence for future time periods is provided by the port operation plan and includes parameters such as whether ships prioritize berthing, whether quay cranes utilize dual-operation loading and unloading, and whether trailers directly reach the handover point. To ensure complete consistency between the input format and the training phase, the above decision sequence is encoded and transformed, converting the decision behavior of each agent into a feature vector. Each vector dimension corresponds to a fixed decision category. For example, ship scheduling includes three behavioral dimensions: berthing order, berth number, and loading / unloading batch arrangement; transport vehicle scheduling includes behavioral dimensions such as whether to directly load and unload, and whether to transfer across storage areas; loading / unloading equipment includes behavioral dimensions such as whether to activate dual-machine collaboration and whether to execute continuous loading / unloading mode. Each behavioral option is encoded in numerical form, for example, using 0 and 1 to indicate whether a certain strategy is activated, and using a fixed number to indicate strategy priority, forming a complete vector that ensures the decision scheme has a unique expression in the numerical space. Subsequently, the encoded vectors are normalized to ensure the data scale is consistent with the training samples. After encoding and organizing all scheduling decision sequences, a standardized decision combination feature vector is formed. This feature vector is then input into a trained collaborative performance prediction regression model, and the predicted collaborative performance index value corresponding to the decision combination is obtained through the model inference process.

[0076] The supply chain carbon intensity variation coefficient and the anchorage idling index, output from the predictive regression model, are used as green collaboration rating factors. These factors are then incorporated into a unified rating system for weighted comprehensive calculation to generate the port's green collaboration rating results. First, weight coefficients are set based on the business meaning of the indicators and their control importance in the green collaboration process. The supply chain carbon intensity variation coefficient, which focuses on evaluating the overall carbon emission balance level, has a weight of 0.6. The anchorage idling index, which mainly reflects the efficiency of berthing management and resource scheduling, has a weight of 0.4. These weights are set after statistical analysis of historical evaluation cases and are dynamically adjusted based on port size, vessel carrying capacity, and the scale of transport equipment of supply chain partners. After normalizing the two indicators, a weighted cumulative model is used to calculate the comprehensive score and convert it into a percentage score. The comprehensive score is calculated as: the supply chain carbon intensity variation score multiplied by a weight of 0.6 plus the anchorage idling index score multiplied by a weight of 0.4. The rating levels are determined based on the comprehensive index value range: an index between [0, 30) is rated as Grade A (excellent green performance and optimal scheduling and coordination); an index between [30-50) is rated as Grade B (good operation and reasonable scheduling and coordination); an index between [50, 70) is rated as Grade C (minor uneven scheduling or energy waste exists); and an index between [70, 100] is rated as Grade D (significant scheduling and coordination problems exist), requiring energy conservation supervision and structural scheduling reform.

[0077] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0078] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0079] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0080] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0081] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A green intelligent port rating method based on big data, characterized in that, It comprises the following steps: S1, based on the data interface of port Internet of Things terminal and supply chain participants, collecting real-time data flow to build a multi-modal data pool for the whole chain of transfer; S2, applying dynamic traceability anchoring processing to the multi-modal data pool, correlating discrete transportation segments across entities based on the container electronic identification, and generating a seamless carbon track chain with the container as the minimum granularity; S3, using a multi-agent strategy matrix to simulate supply chain collaboration scenarios, iteratively executing container turnaround strategies in a virtual environment, and recording the dynamic impact data set of each agent's decision on carbon emissions; S4, based on the dynamic impact data set, calculating the supply chain carbon intensity variation coefficient and the anchorage idling index to form a collaborative performance evaluation index set; S5, combining the multi-agent decision in the dynamic impact data set as input features, and the corresponding collaborative performance evaluation index set as output labels, establishing a regression model for collaborative performance prediction; S6, predicting the collaborative performance of future port operation scheduling decisions, and outputting the port green collaboration rating; In S2, the dynamic traceability anchoring processing applied to the multi-modal data pool, the discrete transportation segments are correlated across entities based on the container electronic identification, and a seamless carbon track chain with the container as the minimum granularity is generated, which specifically includes: Extracting discrete transportation segment records with container electronic identification as the key field from the multi-modal data pool, the transportation segment records contain track information provided by different transportation entities; Performing spatio-temporal continuity analysis on the discrete transportation segment records of each container to identify the logical connection points between adjacent transportation segments, and inserting data anchoring markers at the connection points; Based on the data anchoring markers, the discrete transportation segments involving multiple transportation entities are integrated across entities to form a continuous and complete container movement path; Calculate the carbon emissions of each transportation segment along the integrated movement path, and bind and fuse the carbon emission data with the container movement path in a recursive accumulation manner; Generate a fixed carbon track chain with the container as the tracking unit, which contains the complete movement sequence from the starting point to the ending point and the corresponding cumulative carbon emission data; In S4, based on the dynamic impact data set, the supply chain carbon intensity variation coefficient and the anchorage idling index are calculated to form a collaborative performance evaluation index set, which specifically includes: From the dynamic impact data set, extract the total carbon emissions and corresponding freight volume data of the transportation segment between each adjacent anchoring point of the carbon track chain within a set simulation period, and calculate the carbon intensity value of each transportation segment; Based on the sequence of carbon intensity values of all transportation segments, calculate the ratio of the standard deviation to the mean of the sequence to obtain the supply chain carbon intensity variation coefficient; Synchronously extract the running state records of the port ship during anchorage waiting period, count the proportion of cumulative time of continuous engine idling in total anchorage time, and calculate the anchorage idling index; Take the supply chain carbon intensity variation coefficient and the anchorage idling index as the core indicators, and construct a collaborative performance evaluation index set containing multiple measurement dimensions; In S5, combining the multi-agent decision in the dynamic impact data set as input features, and the corresponding collaborative performance evaluation index set as output labels, establishing a regression model for collaborative performance prediction, which specifically includes: Extract multi-agent decision combination data from dynamic impact data set as model input features, and synchronously extract corresponding collaborative efficiency evaluation index set data as output labels; Perform feature alignment and data standardization processing on the input features and output labels to construct a model training sample set; Train a regression model based on the training sample set, determine the hyperparameter combination of the model through grid search, and establish a nonlinear mapping relationship from decision combination to collaborative efficiency.

2. The green smart port rating method based on big data according to claim 1, characterized in that, In S1, based on the data interface of the port Internet of Things terminal and the supply chain participants, real-time data streams are collected to construct a multi-modal data pool of the whole chain of transshipment, which specifically includes: Collecting real-time location and equipment operating status of containers through the port Internet of Things terminal, and simultaneously calling the data interaction interface of the supply chain participants to obtain the trajectory information of the transportation entity; Using a distributed message queue to receive real-time data streams of the above-mentioned container real-time location, equipment operating status and trajectory information of the transportation entity, performing format analysis and field mapping on the received heterogeneous data, and generating serialized data units with a unified timestamp; Based on the container electronic identification, the parsed data units are associated indexed in the time and space dimensions to construct a multi-modal data pool of the whole chain of transshipment integrating records, numerical indicators and spatial coordinates.

3. The green smart port rating method based on big data according to claim 1, characterized in that, In S3, a multi-agent strategy matrix is used to simulate a supply chain collaboration scenario, and container rotation strategies are iteratively executed in a virtual environment to record a dynamic impact data set of each agent decision on carbon emissions, which specifically includes: Based on the carbon trajectory chain, a virtual collaborative environment of multi-agent is constructed, and the agents include port operation equipment, storage equipment and transportation entities of supply chain participants; Each type of agent is initialized and configured, including decision behavior rules of transportation selection, docking planning and loading and unloading scheduling, forming a multi-agent strategy matrix; In the virtual environment, load the historical rotation data of containers in the carbon trajectory chain to drive the multi-agent to execute transportation planning, port docking arrangement and loading and unloading equipment scheduling decisions according to the strategy matrix; Use a discrete event propulsion mechanism to run the virtual environment to simulate the complete rotation process of containers between port hubs and transportation networks, and record the decision combination of each agent in each iteration; Collect carbon emission data generated by agent interaction in each decision period to construct a dynamic impact data set reflecting the correlation between decision behavior and carbon emissions.

4. The green smart port rating method based on big data according to claim 1, characterized in that, In S6, the collaborative efficiency of future port operation scheduling decisions is predicted, and the port green collaboration rating is output, which specifically includes: Obtain the sequence of planned port operation scheduling decisions in the future period, and convert it into a standardized decision combination feature vector to input the regression model for collaborative efficiency prediction; The regression model outputs the corresponding collaborative efficiency evaluation index prediction value as a green collaboration rating factor; Based on the green collaboration rating factor, perform weighted comprehensive calculation to generate the final port green collaboration rating result.

Images