Steel mill electric locomotive transportation and charging cooperative scheduling method and related device

Abstract

The invention provides a steel mill electric locomotive transportation and charging cooperative scheduling method and a related device, and relates to the technical field of software, and the method comprises the steps: collecting multi-modal data of a to-be-scheduled task, the multi-modal data comprising transportation associated data and charging scheduling associated data; performing multi-task learning on the transportation associated data to predict and obtain a transportation plan scheme of the to-be-scheduled task; and processing the transportation plan scheme and the charging scheduling associated data based on an NSGA-II optimization algorithm to output a vehicle assignment scheme, a charging plan scheme and a path planning scheme of the to-be-scheduled task. According to the method, a double-layer collaborative framework of upper-layer transportation plan prediction and lower-layer transportation and charging scheduling optimization can be constructed, transportation and charging collaborative scheduling is realized, and the molten steel transfer efficiency and the scheduling intelligent level are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of software, in particular to a steel plant electric locomotive transportation and charging collaborative scheduling method and related device. BACKGROUND

[0002] High-power pure electric locomotives, with their zero-emission, low-noise characteristics and excellent traction performance, are gradually becoming the preferred solution for steel plant green transformation.

[0003] However, large-scale deployment of these electric locomotives in the complex rail transportation network of the steel plant faces the contradiction between charging demand and production continuity, as well as the dynamic scheduling problem in complex environments. SUMMARY

[0004] In view of the above problems, the present application provides a steel plant electric locomotive transportation and charging collaborative scheduling method and related device to realize the purpose of collaborative scheduling of steel plant electric locomotive transportation and charging. The specific scheme is as follows:

[0005] The first aspect of the present application provides a steel plant electric locomotive transportation and charging collaborative scheduling method, which comprises:

[0006] Collecting multi-modal data of a to-be-scheduled task, the multi-modal data including transportation-related data and charging scheduling-related data;

[0007] Multi-task learning is performed on the transportation-related data to predict a transportation plan scheme of the to-be-scheduled task;

[0008] The transportation plan scheme and the charging scheduling-related data are processed based on an NSGA-II optimization algorithm to output a vehicle assignment scheme, a charging plan scheme and a path planning scheme of the to-be-scheduled task.

[0009] In one possible implementation, the transportation-related data includes MES plan data, a steel plant physical transportation network and transportation task data, and the multi-task learning performed on the transportation-related data to predict a transportation plan scheme of the to-be-scheduled task comprises:

[0010] A graph structure corresponding to the steel plant physical transportation network is constructed, the graph structure including a node set and an edge set, and each edge in the edge set being used to describe the track connection relationship between the two connected nodes;

[0011] A space-time feature vector of each node in the node set is constructed based on the MES plan data and the transportation task data, and a weight of each edge in the edge set is determined based on the transportation task data, the weight being used to represent the physical distance and transportation strength of the edge;

[0012] inputting the spatio-temporal feature vector of each node in the node set and the weight of each edge in the edge set into a transportation plan prediction model to obtain the transportation plan scheme output by the transportation plan prediction model, the transportation plan prediction model being obtained by pre-training using a multi-task learning architecture.

[0013] In a possible implementation, the transportation plan prediction model comprises a spatial feature extraction layer, a temporal feature extraction layer, a feature fusion layer and a multi-task prediction output layer, and the inputting the spatio-temporal feature vector of each node in the node set and the weight of each edge in the edge set into a transportation plan prediction model to obtain the transportation plan scheme output by the transportation plan prediction model comprises:

[0014] extracting a corresponding spatial feature vector from the input spatio-temporal feature vector and the input weight by the spatial feature extraction layer;

[0015] extracting a corresponding temporal feature vector from the input spatio-temporal feature vector by the temporal feature extraction layer;

[0016] concatenating the spatial feature vector and the temporal feature vector to obtain a spatio-temporal feature vector by the feature fusion layer;

[0017] predicting the transportation task type, the transportation starting work site, the transportation destination work site and the transportation start time corresponding to the spatio-temporal feature vector by the multi-task prediction output layer.

[0018] In a possible implementation, the steel plant electric locomotive transportation and charging collaborative scheduling method further comprises:

[0019] issuing a scheduling instruction according to the vehicle assignment scheme, the charging plan scheme and the path planning scheme, and optimizing the transportation plan prediction model according to the execution of the scheduling instruction.

[0020] In a possible implementation, the charging scheduling related data comprises vehicle data, charging station data, power grid data and a steel plant physical transportation network, and the processing of the transportation plan scheme and the charging scheduling related data based on the NSGA-II optimization algorithm to output the vehicle assignment scheme, the charging plan scheme and the path planning scheme of the to-be-scheduled task comprises:

[0021] obtaining configuration parameters, the configuration parameters comprising decision variables and objective functions, the decision variables comprising vehicle assignment variables and charging decision variables, and the objective functions comprising a transportation and charging time function and a power grid load peak-valley difference function;

[0022] According to the decision variable, the transportation plan scheme, the vehicle data, the charging station data, the power grid data and the steel plant physical transportation network are initialized to obtain a chromosome population, each chromosome in the chromosome population being composed of a vehicle assignment section, a charging plan section and a path planning section;

[0023] According to the transportation and charging time function and the power grid load peak-valley difference function, a transportation and charging time value and a power grid load peak-valley difference value of each chromosome in the chromosome population are calculated;

[0024] The number of updates of the chromosome population is obtained;

[0025] If the number of updates of the chromosome population does not reach a corresponding threshold value, based on the transportation and charging time value and the power grid load peak-valley difference value of each chromosome in the chromosome population, part of the chromosomes in the chromosome population are selected for genetic operation to update the chromosome population, and the step of calculating the transportation and charging time value and the power grid load peak-valley difference value of each chromosome in the chromosome population according to the transportation and charging time function and the power grid load peak-valley difference function is returned to be executed;

[0026] If the number of updates of the chromosome population reaches the corresponding threshold value, according to the transportation and charging time value and the power grid load peak-valley difference value of each chromosome in the chromosome population, a global performance index of each chromosome in the chromosome population is determined;

[0027] A target chromosome with the optimal global performance index is selected from the chromosome population, and the vehicle assignment scheme, the charging plan scheme and the path planning scheme are respectively output according to the vehicle assignment section, the charging plan section and the path planning section of the target chromosome.

[0028] In a possible implementation, the configuration parameters further include constraint conditions, the constraint conditions including a locomotive exclusivity constraint condition, a charging pile exclusivity constraint condition and a path connectivity constraint condition, and the processing of the transportation plan scheme and the charging scheduling associated data based on the NSGA-II optimization algorithm to output the vehicle assignment scheme, the charging plan scheme and the path planning scheme of the to-be-scheduled task further includes:

[0029] Before updating the chromosome population, based on the locomotive exclusivity constraint condition, the vehicle assignment section of the part of the chromosomes that have been subjected to genetic operation is subjected to anomaly detection, based on the charging pile exclusivity constraint condition, the charging plan section of the part of the chromosomes that have been subjected to genetic operation is subjected to anomaly detection, and based on the path connectivity constraint condition, the path planning section of the part of the chromosomes that have been subjected to genetic operation is subjected to anomaly detection.

[0030] A second aspect of this application provides a steel plant electric locomotive transportation and charging coordinated dispatching device, the steel plant electric locomotive transportation and charging coordinated dispatching device comprising:

[0031] The data acquisition module is used to collect multimodal data of the task to be scheduled, including transportation-related data and charging scheduling-related data.

[0032] The transportation planning prediction module is used to perform multi-task learning on the transportation-related data in order to predict and obtain the transportation planning scheme for the task to be scheduled.

[0033] The transportation and charging scheduling module is used to process the transportation plan and the charging scheduling associated data based on the NSGA-II optimization algorithm, so as to output the vehicle assignment plan, charging plan and route planning plan for the task to be scheduled.

[0034] A third aspect of this application provides a computer program product, including computer-readable instructions, which, when executed on an electronic device, cause the electronic device to implement the steel plant electric locomotive transportation and charging coordinated scheduling method described in the first aspect or any implementation thereof.

[0035] A fourth aspect of this application provides an electronic device, including at least one processor and a memory connected to the processor, wherein:

[0036] The memory is used to store computer programs;

[0037] The processor is used to execute the computer program so that the electronic device can implement the steel plant electric locomotive transportation and charging coordinated scheduling method described in the first aspect or any implementation thereof.

[0038] The fifth aspect of this application provides a computer storage medium carrying one or more computer programs, which, when executed by an electronic device, enable the electronic device to implement the steel plant electric locomotive transportation and charging coordinated scheduling method described in the first aspect or any implementation thereof.

[0039] By employing the above technical solution, this application provides a method and related apparatus for coordinated scheduling of electric locomotive transportation and charging in steel plants. The method includes: collecting multimodal data of the task to be scheduled, including transportation-related data and charging scheduling-related data; performing multi-task learning on the transportation-related data to predict the transportation plan scheme for the task to be scheduled; and processing the transportation plan scheme and charging scheduling-related data based on the NSGA-II optimization algorithm to output the vehicle assignment scheme, charging plan scheme, and route planning scheme for the task to be scheduled. This application can collect multimodal data of the task to be scheduled, divide it into transportation-related data and charging scheduling-related data, then perform multi-task learning on the transportation-related data to predict the transportation plan scheme, and further process the transportation plan scheme and charging scheduling-related data based on the NSGA-II optimization algorithm to obtain the vehicle assignment scheme, charging plan scheme, and route planning scheme. In other words, this application can construct a two-layer collaborative framework of upper-level transportation plan prediction and lower-level transportation and charging scheduling optimization, realizing coordinated scheduling of transportation and charging, and improving the efficiency of molten steel transfer and the level of intelligent scheduling. Attached Figure Description

[0040] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0041] Figure 1 A flowchart illustrating a method for coordinated scheduling of electric locomotive transportation and charging in a steel plant, provided as an embodiment of this application;

[0042] Figure 2 This application provides a partial flowchart illustrating a method for coordinated scheduling of electric locomotive transportation and charging in steel plants.

[0043] Figure 3 This is another schematic flowchart of a method for coordinated scheduling of electric locomotive transportation and charging in steel plants, provided in an embodiment of this application.

[0044] Figure 4 This is another schematic flowchart of a method for coordinated scheduling of electric locomotive transportation and charging in steel plants, provided in an embodiment of this application.

[0045] Figure 5 A schematic diagram of the structure of a steel plant electric locomotive transportation and charging coordinated dispatching device provided in this application embodiment;

[0046] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0047] The embodiments of this application are described below with reference to the accompanying drawings. The terminology used in the implementation section of this application is for explaining specific embodiments only and is not intended to limit the scope of this application.

[0048] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.

[0049] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of elements is not necessarily limited to those elements, but may include other elements not explicitly listed or inherent to those processes, methods, products, or apparatuses.

[0050] In the steel manufacturing process, the high-temperature transfer of molten steel is a crucial link between steelmaking and continuous casting. Its timeliness and temperature stability directly determine production efficiency and billet quality. Traditional transportation methods mainly rely on diesel locomotives or fixed-track ladle cars. These methods not only generate high emissions and noise but also exhibit significant inflexibility when facing complex production scheduling. With the deepening implementation of the "dual-carbon" strategic goal, high-power pure electric locomotives, with their zero-emission, low-noise characteristics and excellent traction performance, are gradually becoming the preferred solution for steel plants' green transformation.

[0051] However, deploying these electric locomotives on a large scale within the complex rail transport network of steel plants still faces core technological challenges in meeting actual production needs. Specifically, the application of electric locomotives faces the following two key issues:

[0052] 1) The contradiction between charging demand and production continuity. The charging process of electric locomotives is time-consuming, while the task of transporting molten steel is characterized by high temperature, continuous operation, and extremely high time sensitivity. If charging is arranged during peak production periods or critical heat transport, it is very easy to cause production interruption or excessive molten steel temperature, resulting in serious quality accidents and economic losses.

[0053] 2) The challenge of dynamic scheduling in complex environments. The steel plant's rail network structure is complex and ever-changing, and transportation tasks are dynamically adjusted according to the production plan. How to intelligently allocate transportation resources, optimize travel routes, and rationally arrange charging times based on the production instructions generated by the Manufacturing Execution System (MES) constitutes a highly complex time and air scheduling problem.

[0054] Current scheduling strategies mostly focus on single-objective optimization (such as finding the shortest path) and lack deep integration with production plans, making it difficult to achieve global optimization of transportation efficiency and charging demand. Therefore, developing an intelligent scheduling method for transportation and charging that can be deeply integrated with production plans and support multi-objective collaborative optimization has become an urgent need to improve the efficiency and reliability of molten steel transfer.

[0055] To address the aforementioned issues, this application provides a method for coordinated scheduling of electric locomotive transportation and charging in steel plants, relating to intelligent manufacturing scenarios. Addressing the contradiction between electric locomotive charging demand and production continuity, as well as the dynamic scheduling challenges in the complex environment of steel plants, this application constructs a two-layer collaborative framework consisting of upper-level transportation plan prediction and lower-level transportation and charging scheduling optimization. A closed-loop feedback mechanism connects the upper and lower layers for rolling optimization, forming a dynamic collaborative system of "prediction-optimization-execution-feedback." This system enables coordinated transportation and charging scheduling, replacing manually generated scheduling plans and improving steel transfer efficiency and intelligent scheduling levels. The method for coordinated scheduling of electric locomotive transportation and charging in steel plants, as described in this application, will be described in detail below with reference to the accompanying drawings.

[0056] See Figure 1 , Figure 1 This is a flowchart illustrating a method for coordinated scheduling of electric locomotive transportation and charging in a steel plant, provided as an embodiment of this application. Figure 1 As shown in the embodiment of this application, a method for coordinated scheduling of electric locomotive transportation and charging in a steel plant may include steps S101 to S103, which are described in detail below.

[0057] S101 collects multimodal data of the tasks to be scheduled, including transportation-related data and charging scheduling-related data.

[0058] In this embodiment of the application, multimodal data related to the task to be scheduled can be collected. The task to be scheduled is generated by real-time MES plan data. The multimodal data includes MES plan data (including historical MES plan data and current MES plan data), vehicle data, charging station data, power grid data, steel plant physical transportation network and transportation task data (associated with historical MES plan data), etc. All multimodal data are obtained in real time through API interface and are accompanied by timestamps to form a unified spatiotemporal data system. For example, the MES (Manufacturing Execution System) generates a production plan table with the smelting number as the unique identifier, containing process paths, key time points, and equipment allocation. This results in two types of transportation tasks: transporting heavy ladle (steelmaking → continuous casting) and transporting empty ladle (continuous casting → steelmaking). Vehicle data is obtained through the vehicle dispatching system, including the number of locomotives, locomotive number, location, speed, battery status (SOC), and task status (idle, transport, charging). Charging station data is obtained through the charging station system, including the total number of charging piles, charging pile number, location, number of idle charging piles, charging pile power, queuing status, and waiting time. Grid data is obtained through the energy management system, including electricity load curves. The steel plant's physical transportation network is obtained through the steel plant's track topology system, including each node ID, node type, and node coordinates. Transportation task data is obtained through the transportation management system, representing completed transportation tasks, including task ID, transportation task type, transportation start point, transportation destination point, transportation start time, transportation end time, vehicle number, and path nodes.

[0059] Furthermore, after obtaining multimodal data, to address the issue of multi-source heterogeneity, preprocessing can be performed on the multimodal data, including data cleaning, data transformation and standardization, and data integration and fusion. Data cleaning includes missing value imputation, outlier removal or correction, and duplicate record deletion. Data transformation and standardization include unifying units and formats and normalizing timestamps; text is numerated (e.g., One-Hot encoding), and numerical features are standardized using Z-scores. Data integration and fusion uses attributes with strong correlations, such as smelting numbers and timestamps, to link MES plan data, vehicle data, charging station data, power grid data, steel plant physical transportation network, and transportation task data, forming a unified wide table containing complete spatiotemporal information, with each row representing a feature snapshot at a given moment.

[0060] After performing the above preprocessing on the multimodal data, it can be divided into transportation-related data and charging scheduling-related data.

[0061] S102, perform multi-task learning on transportation-related data to predict and obtain transportation plans for tasks to be scheduled.

[0062] In this embodiment, the transportation-related data is learned through multi-task learning based on the Graph-ConvMixer architecture. Spatiotemporal features are extracted from the data to perform multi-task prediction and obtain the transportation plan scheme for the task to be scheduled, including the transportation task type, transportation starting point, transportation destination point, and transportation start time.

[0063] In one possible implementation, the transportation-related data includes MES planning data, the steel plant's physical transportation network, and transportation task data. This is achieved by constructing a graph structure for the steel plant's physical transportation network, performing feature engineering on different nodes within the graph structure based on the MES planning data and transportation task data, and calculating the weights of different edges in the graph structure based on the transportation task data. This enables multi-task learning to accurately predict transportation plans and ensure the efficiency of coordinated optimization between transportation and charging. See also... Figure 2 , Figure 2 This is a partial flowchart illustrating a method for coordinated scheduling of electric locomotive transportation and charging in a steel plant, provided as an embodiment of this application. Figure 2 As shown in the embodiment of this application, a method for coordinated scheduling of electric locomotive transportation and charging in a steel plant is provided. Step S102, "performing multi-task learning on transportation-related data to predict and obtain transportation plans for tasks to be scheduled," may include steps S201 to S203. These steps are described in detail below.

[0064] S201. Construct the graph structure corresponding to the physical transportation network of the steel plant. The graph structure contains a set of nodes and a set of edges. Each edge in the edge set is used to describe the track connection relationship between the two connected nodes.

[0065] In this embodiment of the application, the physical transportation network of the steel plant is a passable transportation network that connects various work sites, charging stations, and turnout nodes. The physical transportation network of the steel plant is abstracted as a graph structure G=(V, E), where V is the set of nodes and E is the set of edges.

[0066] It should be noted that the node set includes work site nodes (representing the starting and ending points of transportation), charging station nodes (corresponding to the locomotive charging position), turnout nodes (representing the intersection or branching position of tracks), etc. Each node has a unique ID and basic attributes (such as type and location).

[0067] It should also be noted that the edge set is used to describe the orbital connections between nodes. For example, if there is a direct orbit between node a and node b, then an edge is created. .

[0068] S202, construct the spatiotemporal feature vector of each node in the node set based on MES plan data and transportation task data; and determine the weight of each edge in the edge set based on transportation task data, the weight is used to characterize the physical distance and transportation intensity of the edge.

[0069] In this embodiment, on the one hand, based on MES plan data and transportation task data, a feature vector reflecting the spatiotemporal state of each node in the node set can be constructed, namely, a spatiotemporal feature vector. The spatiotemporal feature vector is obtained by fusing the following time-series features, node state features, and transportation task features:

[0070] 1) Time series characteristics, including absolute time (time, day of the week, whether it is a working day, using sin / cos encoding), relative time (task interval, deviation between plan and actual), sliding window statistical characteristics (number of tasks, average interval), etc.

[0071] 2) Node status characteristics, including work point (type, real-time status, historical task volume), turnout (connectivity, traffic frequency), etc.

[0072] 3) Transportation task characteristics, including task type, origin and destination work sites, start time, etc.

[0073] On the other hand, based on the transportation task data, each edge in the edge set is assigned a corresponding weight to form an adjacency matrix describing the connection relationships and interaction strength between nodes. For example, the weight between node a and node b. Taking into account both static physical distance and dynamic transport intensity, the following formula (1) can be used for calculation:

[0074] (1)

[0075] in, This represents the preset static-dynamic balance coefficient, with a value ranging from 0.5 to 0.8; This represents a static adjacency matrix, indicating whether there is a physical connection between nodes, and expressing the physical distance between nodes. Cost factors are incorporated and can be calculated according to the following formula (2); The static adjacency matrix is ​​calculated from the transportation task data and can be calculated according to the following formula (3):

[0076] (2)

[0077] in, , Represents the physical coordinates of node a. This represents the physical coordinates of node b. It can be seen that the closer the physical distance between node a and node b, the better. The larger the value, the greater the physical distance between node a and node b. The smaller.

[0078] (3)

[0079] in, This represents the task time series of node a; This represents the task time series of node b; This represents the covariance of the time series of tasks at node a and node b. This represents the standard deviation of node a; This represents the standard deviation of node b; This represents the number of transport tasks from node a to node b, and is directional. ); This represents the maximum number of computational tasks performed on all nodes in the node set V.

[0080] S203, input the spatiotemporal feature vector of each node in the node set and the weight of each edge in the edge set into the transportation planning prediction model to obtain the transportation planning scheme output by the transportation planning prediction model. The transportation planning prediction model is pre-trained using a multi-task learning architecture.

[0081] In this embodiment, a transportation planning prediction model can be obtained based on deep learning training. This model adopts a multi-task learning Graph-ConvMixer architecture to collaboratively predict the transportation task type (classification task), the transportation start point (classification task), the transportation destination point (classification task), and the transportation start time (regression task). Advanced optimization strategies are introduced to ensure model convergence, generalization ability, and final performance. The transportation planning prediction model integrates the spatial modeling capabilities of graph neural networks and the temporal modeling capabilities of convolutional neural networks to achieve joint modeling of the spatiotemporal characteristics of steel plant transportation tasks.

[0082] The loss function for the transportation planning prediction model can be constructed during the training process as shown in the following formula (4):

[0083] (4)

[0084] in, Indicates overall loss; The predicted loss is represented by the type of transportation task, using a binary cross-loss method. The predicted loss for the initial transportation work site is represented by multi-class cross-entropy loss. The loss for predicting work sites at transportation destinations is represented by multi-class cross-entropy loss. The loss is represented by the predicted transportation start time, using the mean squared error loss. The weights of the prediction losses for transportation task type, transportation start point, transportation destination, and transportation start time are respectively represented. The weights are automatically adjusted using an uncertainty weighting method to avoid the subjectivity of manual setting and enable the model to adaptively balance the learning progress of each task.

[0085] During model training, the AdamW optimizer was employed, along with the CosineAnnealingWarmRestarts learning rate scheduling strategy. Regularization techniques included Dropout and weight decay. Early stopping was implemented during training. Furthermore, for model evaluation, root mean square error and timeliness were used for regression tasks, while accuracy and F1 score were used for classification tasks.

[0086] After training and obtaining the transportation calculation prediction model, the spatiotemporal feature vector of each node in the node set obtained in step S202 above, and the weight of each edge in the edge set, can be input into the transportation planning prediction model to obtain the transportation plan output by the model. This includes the transportation task type (e.g., "loaded tank" or "empty tank"), the transportation starting work point (e.g., KG_trans_1), the transportation destination work point (e.g., ZG_trans_2), and the transportation start time (e.g., 2025 / 6 / 26 15:18:32). The transportation planning prediction model can also simultaneously output a confidence score for the predicted transportation plan to assess its reliability. Transportation plans with high confidence scores can be used for lower-level scheduling, while those with low confidence scores trigger the transportation planning prediction model to recalculate.

[0087] In one possible implementation, the transportation planning prediction model includes a spatial feature extraction layer, a temporal feature extraction layer, a feature fusion layer, and a multi-task prediction output layer. See also Figure 3 , Figure 3 This is another schematic flowchart illustrating a method for coordinated scheduling of electric locomotive transportation and charging in a steel plant, provided as an embodiment of this application. Figure 3 As shown in the embodiment of this application, a method for coordinated scheduling of electric locomotive transportation and charging in a steel plant is provided. In step S203, "inputting the spatiotemporal feature vector of each node in the node set and the weight of each edge in the edge set into the transportation plan prediction model to obtain the transportation plan scheme output by the transportation plan prediction model", it may include steps S301 to S304. These steps are described in detail below.

[0088] S301 extracts the corresponding spatial feature vector from the input spatiotemporal feature vector and the input weight through the spatial feature extraction layer.

[0089] In this embodiment, the spatial feature extraction layer uses a graph attention network (GAT), which dynamically assigns weights to different neighbor nodes through an attention mechanism to learn more expressive spatial features. The spatiotemporal feature vector at the current time and the input weights are used as the inputs to the spatial feature extraction layer, and the output is a feature embedding vector rich in spatial context information for each node, i.e., a spatial feature vector.

[0090] S302 extracts the corresponding time feature vector from the input spatiotemporal feature vector through the time feature extraction layer.

[0091] In this embodiment, the temporal feature extraction layer employs the ConvMixer module to process the feature sequences of each node that change over time, capturing their temporal dependencies. The ConvMixer module consists of depthwise convolution and pointwise convolution; the depthwise convolution uses a 9x1 kernel to independently process the temporal sequences of each feature channel, effectively capturing local patterns and long-term trends; the pointwise convolution uses a 1x1 convolution to achieve interactive integration between feature channels. The spatiotemporal feature vector with a time sequence is used as the input to the temporal feature extraction layer, and its corresponding temporal feature embedding vector, i.e., the temporal feature vector, is output.

[0092] S303 obtains a spatiotemporal feature vector by splicing spatial feature vectors and temporal feature vectors through a feature fusion layer.

[0093] In this embodiment of the application, a unified spatiotemporal feature vector is obtained by concatenating the spatial feature vector and the temporal feature vector through a feature fusion layer.

[0094] S304 predicts the transportation task type, transportation start point, transportation destination point, and transportation start time corresponding to the spatiotemporal feature vector through the multi-task prediction output layer.

[0095] In this embodiment, the multi-task prediction output layer jointly connects the spatiotemporal feature vectors to multiple parallel task heads, each task head using a different activation function based on its task nature. Specifically, the transportation task type prediction uses a fully connected layer + Softmax to output the probability distribution of "loaded tank" or "empty tank"; the transportation start point prediction uses a fully connected layer + Softmax to output the probability distribution of the transportation start point; the transportation destination point prediction uses a fully connected layer + Softmax to output the probability distribution of the transportation destination point; and the transportation start time uses a fully connected layer + linear activation to output the absolute value of the transportation start time.

[0096] S103 processes the transportation plan and charging scheduling data based on the NSGA-II optimization algorithm to output the vehicle assignment scheme, charging plan scheme and route planning scheme for the task to be scheduled.

[0097] In this embodiment, based on the transportation plan scheme output from the upper layer, and combined with the charging scheduling associated data, a multi-combination and multi-constraint optimization model of lower-level transportation and charging scheduling is constructed. The NSGA-II optimization algorithm (i.e., non-dominated sorting genetic algorithm) is used to solve the model, and the output is a scheduling strategy that includes vehicle assignment scheme, charging plan scheme and route planning scheme.

[0098] In one possible implementation, by configuring decision variables, objective functions, and constraints, path, charging, and resource allocation are coordinated and optimized while satisfying all physical and operational constraints. See also Figure 4 , Figure 4 This is another schematic flowchart illustrating a method for coordinated scheduling of electric locomotive transportation and charging in a steel plant, provided as an embodiment of this application. Figure 4 As shown in the embodiment of this application, a method for coordinated scheduling of electric locomotive transportation and charging in a steel plant is provided. The charging scheduling associated data includes vehicle data, charging station data, power grid data, and the physical transportation network of the steel plant. Step S103, "processing the transportation plan and charging scheduling associated data based on the NSGA-II optimization algorithm to output the vehicle assignment scheme, charging plan scheme, and route planning scheme for the task to be scheduled," may include steps S401 to S407. These steps are described in detail below.

[0099] S401, Obtain configuration parameters. Configuration parameters include decision variables and objective functions. Decision variables include vehicle assignment variables and charging decision variables. Objective functions include transportation and charging time functions and grid load peak-valley difference functions.

[0100] In this embodiment of the application, configuration parameters are obtained, which include decision variables and objective functions.

[0101] Among them, the decision variables include vehicle assignment variables. and charging decision variables Vehicle assignment variables It is a binary variable, representing the time step. Get off the vehicle Is a task being performed? Charging decision variables It is a binary variable, representing the time step. off the locomotive Is it at a charging station? Charging. Among them, , , , , Represents a set of transportation tasks, Indicates a collection of locomotives. This represents a collection of charging stations. To represent a day, for example, a day can be divided into 144 time steps, each time step... It lasts for 10 minutes.

[0102] The objective functions include the transportation and charging time functions and the grid load peak-valley difference function. Among them, the transportation and charging time functions... The total time taken for all vehicles to complete the transportation and charging tasks can be expressed by the following formula (5); Grid load peak-valley difference function To punish charging during peak hours, promote grid load balancing, and achieve "peak shaving and valley filling," the following formula (6) can be used:

[0103] (5)

[0104] (6)

[0105] in, Indicates at time step The maximum value of the sum of the power load and the charging load. Indicates at time step The minimum value of the sum of the power load and the charging load.

[0106] S402, based on decision variables, performs population initialization on transportation plan scheme, vehicle data, charging station data, power grid data and steel plant physical transportation network to obtain chromosome population. Each chromosome in the chromosome population consists of vehicle assignment segment, charging plan segment and route planning segment.

[0107] In this embodiment, using decision variables as the objective, a random population initialization is performed on the obtained transportation plan, vehicle data, charging station data, power grid data, and steel plant physical transportation network to obtain a chromosome population composed of multiple chromosomes. Each chromosome uses a segmented encoding method to represent a complete transportation and charging coordinated scheduling scheme, consisting of a vehicle assignment segment, a charging plan segment, and a route planning segment. The vehicle assignment segment indicates which locomotive will execute each task, using integer encoding. This segment has... Each gene segment contains 1 gene bit, where the sequence number of each gene bit represents the task number, and the value of each gene bit represents the vehicle number performing the task. The charging plan segment represents the charging schedule for each vehicle (when, where, and for how long), using integer encoding. This segment has 1000 bytes. There are 12 gene positions, each with a corresponding vehicle number. The value of each gene position is a tuple containing the charging start time, charging station number, and charging duration. The path planning segment represents the specific driving path selected for each task-vehicle combination, encoded using an integer sequence. This segment has 12 gene positions. There are 10 gene positions, each with a sequence number corresponding to a task number, and the value of each gene position is a sequence of node numbers.

[0108] S403 calculates the transport and charging time value and the grid load peak-valley difference value for each chromosome in the chromosome population based on the transport and charging time function and the grid load peak-valley difference function.

[0109] In this embodiment of the application, for each chromosome in the chromosome population, the corresponding transport and charging time value is calculated according to the transport and charging time function, and the grid load peak-valley difference value is calculated according to the grid load peak-valley difference function.

[0110] S404 retrieves the number of times the chromosome population has been updated.

[0111] In this embodiment of the application, the number of times the chromosome population is updated is obtained, that is, the number of times genetic operations are performed.

[0112] S405, if the number of times the chromosome population is updated has not reached the corresponding threshold, based on the transport and charging time value of each chromosome in the chromosome population and the peak-valley difference of the power grid load, select some chromosomes in the chromosome population for genetic operation to update the chromosome population, and return to step 403.

[0113] In this embodiment, if the number of updates to the chromosome population does not reach the corresponding threshold, a binary tournament selection is adopted based on the transportation and charging time values ​​of each chromosome in the chromosome population and the peak-valley difference of the power grid load. Chromosomes with high non-dominated levels and large congestion distances are preferentially selected to perform genetic operations, including crossover (simulated binary crossover for vehicle assignment segments, uniform crossover for charging plan segments and path planning segments) and mutation (randomly redistributing tasks for vehicle assignment segments, making minor perturbations to charging plan segments, and inserting, deleting, or replacing nodes in path planning segments). A new chromosome population is formed by merging chromosomes that did not perform this genetic operation. Further, the process returns to step S403 to enter the next genetic iteration, continuing until the number of updates to the chromosome population reaches the corresponding threshold.

[0114] S406, when the number of updates to the chromosome population reaches the corresponding threshold, the global performance index of each chromosome in the chromosome population is determined based on the transport and charging time value of each chromosome in the chromosome population and the peak-valley difference of the power grid load.

[0115] In this embodiment of the application, if the number of times the chromosome population is updated reaches the corresponding threshold, the global performance index of the chromosome is determined based on the transport and charging time value of each chromosome in the chromosome population and the peak-valley difference of the grid load. The global performance index is negatively correlated with the charging time value and negatively correlated with the peak-valley difference of the grid load.

[0116] S407 selects the target chromosome with the best global performance index from the chromosome population, and outputs the vehicle assignment scheme, charging plan scheme and path planning scheme respectively based on the vehicle assignment segment, charging plan segment and path planning segment of the target chromosome.

[0117] In this embodiment, a target chromosome with the best global performance index is selected from the chromosome population, and a vehicle assignment scheme is output based on the vehicle assignment segment of the target chromosome, a vehicle assignment scheme is output based on the charging plan segment of the target chromosome, and a path planning scheme is output based on the path planning segment of the target chromosome.

[0118] In one possible implementation, after genetic operations are performed on a portion of the chromosomes, anomaly detection can be performed on the vehicle assignment segment, charging plan segment, and path planning segment of these chromosomes to ensure the rationality of the updated chromosome population. In this regard, an embodiment of this application provides a method for coordinated scheduling of electric locomotive transportation and charging in a steel plant. The configuration parameters further include constraints, including locomotive exclusivity constraints, charging pile exclusivity constraints, and path connectivity constraints. Step S103, "Processing the transportation plan scheme and charging scheduling association data based on the NSGA-II optimization algorithm to output the vehicle assignment scheme, charging plan scheme, and path planning scheme for the task to be scheduled," further includes the following steps:

[0119] Before updating the chromosome population, anomaly detection is performed on the vehicle assignment segments of some chromosomes that have undergone genetic operations based on locomotive exclusivity constraints, as well as on the charging plan segments of some chromosomes that have undergone genetic operations based on charging pile exclusivity constraints, and on the path planning segments of some chromosomes that have undergone genetic operations based on path connectivity constraints.

[0120] In this embodiment, the constraints include locomotive exclusivity constraints, charging pile exclusivity constraints, and path connectivity constraints. The locomotive exclusivity constraint ensures that a vehicle can only perform one task (transportation or charging) at a time, and can be expressed using formula (7). The charging pile exclusivity constraint ensures that a charging pile can only charge one vehicle at a time, and can be expressed using formula (8). The path connectivity constraint ensures that the assigned path must be physically connected to the steel plant's physical transportation network, and can be expressed using formula (9).

[0121] (7)

[0122] (8)

[0123] (9)

[0124] in, Indicates vehicle Execute the task The selected path.

[0125] For chromosomes that have undergone genetic manipulation, anomaly detection is performed on the vehicle assignment segment, charging plan segment, and route planning segment of each chromosome. Specifically, based on locomotive exclusivity constraints, the coding of the vehicle assignment segment is checked to see if it violates the locomotive exclusivity constraint. If it does, the positions of duplicate vehicle numbers are recorded from left to right, and the positions violating the constraint are inserted into non-duplicate vehicle numbers. Based on charging pile exclusivity constraints, the coding of the charging plan segment is checked to see if it violates the charging pile exclusivity constraint. If it does, the charging start time, charging pile number, and charging duration corresponding to the tuples of each gene position are traversed from left to right. The charging end time is calculated using the charging start time and charging duration. Gene positions with the same time period and the same charging pile number are recorded, and the numbers of unused charging piles are inserted into the duplicate positions. If all charging pile numbers are occupied, the charging start time or charging duration is adjusted to eliminate time overlap and charging pile number duplication. Based on path connectivity constraints, the coding of the path planning segment is checked to see if it violates the path connectivity constraint. If it does, the node number sequence of each gene position is traversed from left to right. The connectivity of the sequence is checked according to the track network topology. If there are disconnected nodes in the node sequence, the disconnected nodes are replaced to make them connected.

[0126] In one possible implementation, vehicle assignment schemes, charging plans, and route planning schemes can be converted into specific scheduling instructions, issued to the corresponding vehicles and charging piles for execution, and the execution status of the scheduling instructions can be monitored in real time. If a significant deviation occurs (such as equipment failure or emergency task insertion), the abnormal information can be returned as feedback to the upper-level transportation plan prediction model, triggering the model's rolling re-optimization, thereby forming a closed-loop collaborative mechanism of "prediction-optimization-execution-feedback" to ensure continuous adaptation to the dynamic environment. In this regard, the steel plant electric locomotive transportation and charging collaborative scheduling method provided in this application embodiment further includes the following steps:

[0127] Dispatch instructions are issued based on vehicle assignment schemes, charging plans, and route planning schemes, and the transportation plan prediction model is optimized based on the execution of the dispatch instructions.

[0128] Based on the above description, the steel plant electric locomotive transportation and charging collaborative scheduling method provided in this application constructs a two-layer collaborative framework of upper-level transportation plan prediction and lower-level transportation and charging scheduling optimization, which can realize the transportation-charging collaborative scheduling optimization of the molten steel transportation process. The upper layer constructs a transportation plan prediction model based on multimodal data and deep learning, and synchronously outputs the transportation task type, transportation start point, transportation destination point, and transportation start time through multi-task learning. The lower layer outputs vehicle assignment scheme, charging plan scheme, and route planning scheme based on the NSGA-II optimization algorithm. The upper and lower layers are connected through a closed-loop feedback mechanism for rolling optimization, forming a dynamic collaborative system of "prediction-optimization-execution-feedback".

[0129] The above describes a method for coordinated scheduling of electric locomotive transportation and charging in steel plants, provided by embodiments of this application. The following describes the apparatus for implementing the above-described method for coordinated scheduling of electric locomotive transportation and charging in steel plants.

[0130] See Figure 5 , Figure 5 This is a schematic diagram of a steel plant electric locomotive transportation and charging coordinated dispatching device provided as an embodiment of this application. Figure 5 As shown in the embodiment of this application, a steel plant electric locomotive transportation and charging coordinated dispatching device includes:

[0131] The data acquisition module 501 is used to collect multimodal data of the task to be scheduled, including transportation-related data and charging scheduling-related data.

[0132] The transportation planning prediction module 502 is used to perform multi-task learning on transportation-related data in order to predict and obtain transportation planning schemes for tasks to be scheduled.

[0133] The transportation and charging scheduling module 503 is used to process the transportation plan and charging scheduling related data based on the NSGA-II optimization algorithm to output the vehicle assignment plan, charging plan and route planning plan for the task to be scheduled.

[0134] In one possible implementation, the transportation-related data includes MES planning data, steel mill physical transportation network, and transportation task data. The transportation planning prediction module 502 is specifically used for:

[0135] A graph structure corresponding to the physical transportation network of the steel plant is constructed. The graph structure contains a set of nodes and a set of edges. Each edge in the edge set describes the track connection relationship between two connected nodes. Based on MES planning data and transportation task data, a spatiotemporal feature vector of each node in the node set is constructed. Based on the transportation task data, the weight of each edge in the edge set is determined. The weight is used to characterize the physical distance and transportation intensity of the edge. The spatiotemporal feature vector of each node in the node set and the weight of each edge in the edge set are input into the transportation planning prediction model to obtain the transportation planning scheme output by the transportation planning prediction model. The transportation planning prediction model is pre-trained using a multi-task learning architecture.

[0136] In one possible implementation, the transportation planning prediction model includes a spatial feature extraction layer, a temporal feature extraction layer, a feature fusion layer, and a multi-task prediction output layer. A transportation planning prediction module 502, used to input the spatiotemporal feature vector of each node in the node set and the weight of each edge in the edge set into the transportation planning prediction model to obtain the transportation planning scheme output by the model, is specifically used for:

[0137] The spatial feature extraction layer extracts the corresponding spatial feature vector from the input spatiotemporal feature vector and the input weights; the temporal feature extraction layer extracts the corresponding temporal feature vector from the input spatiotemporal feature vector; the feature fusion layer concatenates the spatial feature vector and the temporal feature vector to obtain the spatiotemporal feature vector; and the multi-task prediction output layer predicts the transportation task type, transportation start point, transportation destination point and transportation start time corresponding to the spatiotemporal feature vector.

[0138] In one possible implementation, the aforementioned steel plant electric locomotive transportation and charging coordinated dispatching device further includes:

[0139] The scheduling optimization module is used to issue scheduling instructions based on vehicle assignment schemes, charging plans, and route planning schemes, and to optimize the transportation plan prediction model based on the execution of the scheduling instructions.

[0140] In one possible implementation, the charging scheduling associated data includes vehicle data, charging station data, power grid data, and the steel plant's physical transportation network. The transportation and charging scheduling module 503 is specifically used for:

[0141] The process involves obtaining configuration parameters, including decision variables and objective functions. Decision variables include vehicle assignment variables and charging decision variables, while the objective functions include transportation and charging time functions and grid load peak-valley difference functions. Based on these decision variables, a population initialization is performed on the transportation plan, vehicle data, charging station data, grid data, and the steel plant's physical transportation network to obtain a chromosome population. Each chromosome in the population consists of a vehicle assignment segment, a charging plan segment, and a route planning segment. The transportation and charging time values ​​and grid load peak-valley difference values ​​for each chromosome in the population are calculated based on the transportation and charging time functions and the grid load peak-valley difference function. The update count of the chromosome population is then obtained. If the update count of the chromosome population does not reach a corresponding threshold, the process is repeated based on the transportation and charging time values ​​of each chromosome in the population. The system uses the transport and charging time value and the grid load peak-valley difference value to select a portion of chromosomes in the chromosome population for genetic operations to update the chromosome population. It then returns to the previous step of calculating the transport and charging time value and the grid load peak-valley difference value for each chromosome in the chromosome population based on the transport and charging time function and the grid load peak-valley difference function. When the number of chromosome population updates reaches a corresponding threshold, the system determines the global performance index of each chromosome in the chromosome population based on the transport and charging time value and the grid load peak-valley difference value. Finally, it selects the target chromosome with the optimal global performance index from the chromosome population and outputs the vehicle assignment scheme, charging plan scheme, and path planning scheme based on the vehicle assignment segment, charging plan segment, and path planning segment of the target chromosome, respectively.

[0142] In one possible implementation, the configuration parameters also include constraints, including locomotive exclusivity constraints, charging pile exclusivity constraints, and path connectivity constraints. The transportation and charging scheduling module 503 is also used for:

[0143] Before updating the chromosome population, anomaly detection is performed on the vehicle assignment segments of some chromosomes that have undergone genetic operations based on locomotive exclusivity constraints, as well as on the charging plan segments of some chromosomes that have undergone genetic operations based on charging pile exclusivity constraints, and on the path planning segments of some chromosomes that have undergone genetic operations based on path connectivity constraints.

[0144] It should be noted that the detailed functions of each module in the embodiments of this application can be found in the corresponding disclosure of the above-mentioned embodiments of the steel plant electric locomotive transportation and charging coordinated scheduling method, and will not be repeated here.

[0145] This application also provides an electronic device in its embodiments. See also... Figure 6 , Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device in this embodiment may include, but is not limited to, fixed terminals such as mobile phones, laptops, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, etc. Figure 6 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0146] like Figure 6 As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. When the electronic device is powered on, the RAM 603 also stores various programs and data required for the operation of the electronic device. The processing unit 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0147] Typically, the following devices can be connected to I / O interface 605: input devices 606 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 607 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 608 including, for example, memory cards, hard drives, etc.; and communication devices 609. Communication device 609 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 6Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0148] This application also provides a computer program product including computer-readable instructions, which, when executed on an electronic device, cause the electronic device to implement any of the steel plant electric locomotive transportation and charging coordinated scheduling methods provided in this application.

[0149] This application also provides a computer-readable storage medium carrying one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any of the steel plant electric locomotive transportation and charging coordinated scheduling methods provided in this application.

[0150] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the device embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0151] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0152] In the above embodiments, the implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, in the form of a computer program product.

[0153] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

Claims

1. A method for collaborative scheduling of steel plant electric locomotive transportation and charging, characterized in that, The steel plant electric locomotive transportation and charging collaborative scheduling method comprises: Collecting multi-modal data of a to-be-scheduled task, the multi-modal data comprising transportation-related data and charging scheduling-related data; Performing multi-task learning on the transportation-related data to predict a transportation plan scheme of the to-be-scheduled task; Processing the transportation plan scheme and the charging scheduling-related data based on an NSGA-II optimization algorithm to output a vehicle assignment scheme, a charging plan scheme and a path planning scheme of the to-be-scheduled task.

2. The steel plant electric locomotive transportation and charging co-scheduling method according to claim 1, characterized in that, The transportation-related data comprises MES plan data, a steel plant physical transportation network and transportation task data, and the multi-task learning on the transportation-related data to predict a transportation plan scheme of the to-be-scheduled task comprises: Constructing a graph structure corresponding to the steel plant physical transportation network, the graph structure comprising a node set and an edge set, and each edge in the edge set being used to describe a track connection relationship between two connected nodes; Constructing a space-time feature vector of each node in the node set based on the MES plan data and the transportation task data, and determining a weight of each edge in the edge set based on the transportation task data, the weight being used to represent a physical distance and a transportation intensity of the edge; Inputting the space-time feature vector of each node in the node set and the weight of each edge in the edge set into a transportation plan prediction model to obtain the transportation plan scheme output by the transportation plan prediction model, the transportation plan prediction model being obtained by pre-training using a multi-task learning architecture.

3. The steel plant electric locomotive transportation and charging co-scheduling method according to claim 2, characterized in that, The transportation plan prediction model comprises a space feature extraction layer, a time feature extraction layer, a feature fusion layer and a multi-task prediction output layer, and the inputting of the space-time feature vector of each node in the node set and the weight of each edge in the edge set into the transportation plan prediction model to obtain the transportation plan scheme output by the transportation plan prediction model comprises: Extracting a corresponding space feature vector from the input space-time feature vector and the input weight through the space feature extraction layer; Extracting a corresponding time feature vector from the input space-time feature vector through the time feature extraction layer; Concatenating the space feature vector and the time feature vector to obtain a space-time feature vector through the feature fusion layer; Predicting a transportation task type, a transportation starting work station, a transportation destination work station and a transportation start time corresponding to the space-time feature vector through the multi-task prediction output layer.

4. The steel plant electric locomotive transportation and charging co-scheduling method according to claim 2, characterized in that, The steel plant electric locomotive transportation and charging collaborative scheduling method further comprises: Issuing a scheduling instruction based on the vehicle assignment scheme, the charging plan scheme and the path planning scheme, and optimizing the transportation plan prediction model according to an execution condition of the scheduling instruction.

5. The steel plant electric locomotive transportation and charging co-scheduling method according to claim 1, characterized in that, The charging scheduling-related data comprises vehicle data, charging station data, power grid data and a steel plant physical transportation network, and the processing of the transportation plan scheme and the charging scheduling-related data based on the NSGA-II optimization algorithm to output the vehicle assignment scheme, the charging plan scheme and the path planning scheme of the to-be-scheduled task comprises: obtaining configuration parameters, the configuration parameters comprising decision variables and an objective function, the decision variables comprising vehicle assignment variables and charging decision variables, the objective function comprising a transportation and charging time function and a power grid load peak-valley difference function; performing population initialization on the transportation plan scheme, the vehicle data, the charging station data, the power grid data, and the steel plant physical transportation network according to the decision variables to obtain a chromosome population, each chromosome in the chromosome population being composed of a vehicle assignment section, a charging plan section, and a path planning section; calculating a transportation and charging time value and a power grid load peak-valley difference value of each chromosome in the chromosome population according to the transportation and charging time function and the power grid load peak-valley difference function; obtaining an update number of the chromosome population; in a case where the update number of the chromosome population does not reach a corresponding threshold value, performing genetic operations on part of the chromosomes in the chromosome population based on the transportation and charging time value and the power grid load peak-valley difference value of each chromosome in the chromosome population to update the chromosome population, and returning to the step of calculating the transportation and charging time value and the power grid load peak-valley difference value of each chromosome in the chromosome population according to the transportation and charging time function and the power grid load peak-valley difference function; in a case where the update number of the chromosome population reaches the corresponding threshold value, determining a global performance index of each chromosome in the chromosome population according to the transportation and charging time value and the power grid load peak-valley difference value of each chromosome in the chromosome population; selecting a target chromosome with the optimal global performance index from the chromosome population, and outputting the vehicle assignment scheme, the charging plan scheme, and the path planning scheme according to the vehicle assignment section, the charging plan section, and the path planning section of the target chromosome, respectively.

6. The steel plant electric locomotive transportation and charging co-scheduling method according to claim 5, characterized in that, The configuration parameters further comprise constraint conditions, the constraint conditions comprising locomotive exclusivity constraint conditions, charging pile exclusivity constraint conditions, and path connectivity constraint conditions, and the processing of the transportation plan scheme and the charging scheduling associated data based on the NSGA-II optimization algorithm to output the vehicle assignment scheme, the charging plan scheme, and the path planning scheme of the task to be scheduled further comprises: before updating the chromosome population, performing anomaly detection on the vehicle assignment section of the part of the chromosomes that have been subjected to genetic operations based on the locomotive exclusivity constraint conditions, performing anomaly detection on the charging plan section of the part of the chromosomes that have been subjected to genetic operations based on the charging pile exclusivity constraint conditions, and performing anomaly detection on the path planning section of the part of the chromosomes that have been subjected to genetic operations based on the path connectivity constraint conditions.

7. A steel plant electric locomotive transportation and charging co-scheduling device, characterized in that, The steel plant electric locomotive transportation and charging collaborative scheduling device comprises: a data acquisition module configured to acquire multi-modal data of a task to be scheduled, the multi-modal data comprising transportation associated data and charging scheduling associated data; a transportation plan prediction module configured to perform multi-task learning on the transportation associated data to predict a transportation plan scheme of the task to be scheduled; The transportation and charging scheduling module is configured to process the transportation plan scheme and the charging scheduling associated data based on an NSGA-II optimization algorithm to output a vehicle assignment scheme, a charging plan scheme and a path planning scheme of the to-be-scheduled task.

8. A computer program product, characterised in that, The computer readable instructions, when executed on an electronic device, cause the electronic device to implement the steel plant electric locomotive transportation and charging collaborative scheduling method of any one of claims 1 to 6.

9. An electronic device, comprising: The memory is configured to store a computer program. The processor is configured to execute the computer program to enable the electronic device to implement the steel plant electric locomotive transportation and charging collaborative scheduling method of any one of claims 1 to 6. The storage medium carries one or more computer programs, which, when executed by an electronic device, enable the electronic device to implement the steel plant electric locomotive transportation and charging collaborative scheduling method of any one of claims 1 to 6.

10. A computer storage medium, characterized in that, ​

Images