An intelligent scheduling method applied to a bulk material conveying system of a steel enterprise

Abstract

The application discloses an intelligent scheduling method applied to a bulk material conveying system of a steel enterprise, and relates to the technical field of industrial automatic control and intelligent scheduling. The method collects and analyzes multi-source operation data, constructs a residual service time prediction model of a target silo, and predicts the residual service time of each target silo; calculates the urgency score of a corresponding replenishment task of each target silo, and generates a candidate replenishment task list; establishes a conveyor belt network directed graph and initializes a space-time occupation table, searches for a candidate space-time path and constructs an initial scheduling scheme for each replenishment task in the candidate replenishment task list, and uses an adaptive large neighborhood search algorithm to optimize and update the scheduling scheme; in the scheduling execution process, the target silo material level state, the equipment operation state, the production plan data and the space-time occupation table are updated through real-time monitoring and a dynamic rescheduling mechanism, so that closed-loop response to changes in the bulk material conveying site of the steel enterprise is realized.

Description

Technical Field

[0001] This invention relates to the field of industrial automatic control and intelligent scheduling technology, and in particular to an intelligent scheduling method for bulk raw material conveying systems in steel enterprises. Background Technology

[0002] The bulk raw material conveying system in steel enterprises is a crucial system that is prevalent in steel production and directly impacts the stability of continuous production. During steel production, bulk raw materials such as iron ore, coke, pulverized coal, limestone, and flux typically need to be transported to target silos via pick-up points, transfer nodes, conveyor belts, and distribution equipment. Due to the continuous and high-load operation characteristics of steel production, the material level in the target silos directly affects the raw material supply to downstream production units. If the target silo fails to receive timely replenishment, problems such as material shortages, waiting for materials, or reduced load operation can easily occur, thereby affecting the steel enterprise's production rhythm and equipment operating efficiency. The risk of material shortages in target silos stems from the fact that raw material consumption in target silos is influenced by factors such as production plans, downstream production loads, historical replenishment intervals, and actual discharge rates. Simply relying on the current material level is insufficient to accurately reflect the true material supply risk of the target silo in subsequent periods.

[0003] For scheduling issues in bulk raw material conveying systems of steel enterprises, on-site methods typically include manual experience-based scheduling, fixed-rule scheduling, or inventory threshold-triggered scheduling. Manual experience-based scheduling involves dispatchers assigning replenishment tasks based on the target silo level, production plan, and equipment operating status. Fixed-rule scheduling determines the execution order of replenishment tasks based on preset priorities, safety stock thresholds, or fixed process sequences. Inventory threshold-triggered scheduling generates a replenishment task after the target silo level drops to a preset threshold and selects a conveyor path based on currently available conveyors. However, these methods have the following shortcomings: (1) They mainly rely on the current material level of the target silo to make judgments, lacking the prediction of subsequent material level changes in the target silo, which can easily lead to a delay in identifying material shortage risks; (2) They usually use fixed rules to determine the priority of replenishment tasks, which makes it difficult to comprehensively reflect the impact of the remaining available time of the target silo, the planned demand and the inventory safety margin on the urgency of the replenishment task; (3) They usually select the conveyor path temporarily for a single replenishment task, lacking unified handling of multiple replenishment tasks, conveyor belt occupation, equipment maintenance and path conflicts, which can easily cause conveyor belt resource conflicts, increased task waiting time and insufficient stability of scheduling scheme execution. Summary of the Invention

[0004] This invention addresses the problems of existing scheduling methods in the bulk raw material transportation process of steel plants, which rely on manual experience, lack forward-looking prediction, and struggle to balance the needs of multiple silos and conveyor network conflicts. It proposes an intelligent scheduling method for bulk raw material transportation systems in steel enterprises. This method first analyzes historical usage and feeding data of target silos to obtain the remaining usage time of each silo. Then, based on the remaining usage time, current inventory status, task demand, and production consumption rhythm, the urgency of the replenishment tasks to be executed is scored. Furthermore, by combining the conveyor network topology, equipment operating status, maintenance plans, and path mutual exclusion constraints, the task priority is dynamically optimized for production scheduling, generating conflict-free transportation paths and execution sequences. This achieves forward-looking prediction, globally optimized resource allocation, and stable and efficient automatic execution, reducing the workload of managers and effectively ensuring continuous production and improving production efficiency.

[0005] The technical solution of this invention is:

[0006] An intelligent scheduling method for bulk raw material conveying systems in steel enterprises, the method comprising:

[0007] Step 1: Construct a prediction model for the remaining usage time of the target silo;

[0008] Step 2: Use the target silo remaining usage time prediction model to predict the remaining usage time of each target silo in real time;

[0009] Step 3: Based on the predicted remaining usage time of each target silo, the current material level of each target silo, and the production plan data, calculate the urgency score of the corresponding replenishment task for each target silo, and arrange all replenishment tasks in order of urgency score to form a candidate replenishment task list.

[0010] Step 4: Abstract the bulk raw material conveying system of the steel enterprise into a directed graph of conveyor belt network. ,in Represents the set of nodes in a conveyor belt network. Represents the set of edges in a conveyor belt network; for each node and each conveyor belt network edge Establish a spatiotemporal occupancy table; the spatiotemporal occupancy table is used to record whether the corresponding node or the corresponding conveyor belt network edge is occupied in different time intervals;

[0011] Step 5: For each replenishment task in the candidate replenishment task list, based on... Search for candidate spatiotemporal paths using the spatiotemporal occupancy table, and then construct the current scheduling scheme;

[0012] Step 6: Use the adaptive large neighborhood search algorithm to optimize and update the current scheduling scheme, and output the current globally optimal scheduling scheme;

[0013] Step 7: Execute the replenishment task in the current global optimal scheduling scheme and update the global optimal scheduling scheme according to the dynamic rescheduling conditions. The last updated global optimal scheduling scheme is the intelligent scheduling scheme of the bulk raw material transportation system of the steel enterprise.

[0014] Optionally, according to the intelligent scheduling method, step 1 includes:

[0015] Step 1.1: Collect multi-source operational data, including target silo-related data, production plan data, downstream equipment operation data, conveyor network topology data, equipment status data, and maintenance plan data; the target silo-related data includes attribute data, historical material level data, historical feed data, and historical discharge data for each target silo; the target silo attribute data includes the historical material level of each target silo, as well as the rated capacity, low material level threshold, safety stock threshold, and rated high material level for each target silo.

[0016] Step 1.2: Using a fixed time granularity The raw multi-source runtime data collected in step 1.1 is time-aligned to generate a standard time-series dataset;

[0017] Step 1.3: Identify low material level moments based on standard time series datasets. ;

[0018] Step 1.4: Define the target silo At a historical moment The training samples are composed of predicted feature vectors and training sample labels The training samples for each target silo are constructed based on the standard time series dataset and the low material level time. The training samples of each target silo constitute the training sample set.

[0019] Step 1.5: Use the training sample set to train a regression prediction model. The trained regression prediction model will be used as the target silo remaining usage time prediction model.

[0020] Optionally, according to the intelligent scheduling method, step 1.4 includes:

[0021] Step 1.4.1: Use the remaining usage time of the target silos as training sample labels to generate training sample labels for each target silo;

[0022] If the target silo At a historical moment There will be a low material level moment afterwards. Then according to Generate target silo The remaining usage time; if at time Replenishment occurred before a low material level was reached, and the most recent replenishment started at [time missing]. Then according to Generate target silo Remaining usage time, Indicates target silo At the start of feeding The average consumption rate before; This indicates a very small positive number that prevents the denominator from being zero;

[0023] Step 1.4.2: Construct each target silo according to the following formula. At a historical moment Predicted feature vector :

[0024] ;

[0025] In the formula, Indicates target silo At any moment Material level; Indicates target silo The average consumption rate within a preset historical time window; Indicates target silo Standard deviation of consumption rate within a preset historical time window; Indicates target silo Planned consumption within the rolling scheduling window; Indicates the length of the rolling scheduling window; Indicates target silo From the time the last refueling ended to the current time The interval time; Indicates target silo The amount of the most recent replenishment; This represents the hour code corresponding to the current time t; Indicates the current time Corresponding production shift code; Indicates target silo Corresponding to the load rate of downstream equipment; Indicates target silo The corresponding raw material consumption ratio for downstream equipment; Indicates target silo At any moment Theoretical remaining usage time, Calculate according to the following formula:

[0026] .

[0027] Optionally, according to the intelligent scheduling method, step 2 includes:

[0028] Step 2.1: Obtain the current time Each target silo material level The average consumption rate calculated based on a preset historical time window. Production planning data, downstream equipment operation data, and equipment status data;

[0029] Step 2.2: Construct each target silo At the present moment Real-time feature vectors ;

[0030] Step 2.3: Transfer the real-time feature vector Input the target silo's remaining usage time prediction model to obtain the results for each target silo. Predicted remaining usage time initial value ;

[0031] Step 2.4: For each target silo Predicted remaining usage time initial value The rationality is verified using the following formula to obtain the target silos. Predicted remaining usage time ;

[0032] ;

[0033] In the formula, Indicates target silo The theoretical upper limit of the predicted remaining usage time is calculated according to the following formula:

[0034] ;

[0035] In the formula, Indicates target silo The minimum effective consumption rate.

[0036] Optionally, according to the intelligent scheduling method, step 3 includes the following steps:

[0037] Step 3.1: Calculate the target silos according to the following formula. The corresponding material replenishment task is at time Urgency score ;

[0038] ;

[0039] In the formula, Indicates target silo The remaining usage time; Indicates target silo The planned consumption amount, Represents the target silo set The number of any target silo. Indicates target silo Planned consumption within the rolling scheduling window; Indicates target silo Safety stock gap, Indicates target silo Low material level threshold, Indicates target silo Safety stock threshold; , , These represent the weights of remaining usage time, planned consumption, and safety stock gap, respectively, and satisfy the following conditions: ;

[0040] Step 3.2: Identify the target silos that satisfy the following formula as the target silos for which candidate feeding tasks need to be generated:

[0041] ;

[0042] In the formula, This represents the urgency score threshold for each replenishment task corresponding to each target silo. This represents the remaining usage time threshold for each target silo;

[0043] Step 3.3: Calculate the target silo according to the following formula. The required amount of material for the corresponding candidate material replenishment task If calculated If so, no candidate refueling task will be generated for the target silo;

[0044] ;

[0045] In the formula, Indicates target silo At any moment Material level; Indicates target silo Rated high material level; Target silos Rated capacity; Indicates the target silo within the rolling scheduling window. The maximum amount of material replenishment allowed to be completed;

[0046] Step 3.4: Sort the candidate material replenishment tasks by urgency score Sort in descending order to generate a list of candidate replenishment tasks. ,in, Indicates the first One candidate material replenishment task Indicates the first Each refueling task corresponds to a target silo at time [time]. The urgency score.

[0047] Optionally, according to the intelligent scheduling method, step 5 includes:

[0048] Step 5.1: For candidate material replenishment tasks Based on the directed graph of the conveyor belt network, the spatiotemporal A* algorithm is used to search for the set of candidate spatiotemporal paths from the material picking node to the target silo node. , Indicates material replenishment task The There are 10 candidate spatiotemporal paths; each spatiotemporal path consists of a sequence of directed edges from the material picking node to the target silo; for the material replenishment task... Any conveyor path Then the conveyor belt path The conveying capacity is ,in Represents the edge of the conveyor belt network The conveying capacity; and the replenishment task is calculated according to the following formula. On the conveyor belt path Duration on :

[0049] ;

[0050] In the formula, Represents the edge of the conveyor belt network Delivery time; Indicates material replenishment task The amount of material to be added; Represents a node The corresponding material sorting or transfer equipment switches from the previous conveying direction to the material replenishment task. The time required to switch the transmission direction;

[0051] Step 5.2: According to the candidate replenishment task list The urgency scores are sorted in descending order to determine the replenishment tasks for each candidate replenishment task. The overall start time of the plan Material replenishment task The overall end time of the plan and conveyor belt path To form an initial scheduling scheme ;

[0052] Step 5.3: [The sentence is incomplete and requires more context to be translated accurately.] The replenishment tasks that have been determined are written into the spacetime occupancy table. During the execution of the replenishment task, all conveyor network edges and nodes on the corresponding conveyor path are considered to be occupied.

[0053] Optionally, according to the intelligent scheduling method, step 6 includes:

[0054] Step 6.1: Construct the current scheduling scheme objective function :

[0055] ;

[0056] In the formula, Indicates the latest completion time of the scheduling scheme; Indicates material replenishment task The delay time; Indicates the conveyor belt path Path cost; Represents the edge of the conveyor belt network Load rate; These represent the weights of the objective function;

[0057] Material replenishment task Delay time Calculate according to the following formula:

[0058] ;

[0059] In the formula, Indicates task The predicted remaining usage time of the corresponding target silo; This indicates that a safety time has been reserved;

[0060] Conveyor Belt Path Path cost Calculate according to the following formula:

[0061] ;

[0062] In the formula, This represents the switching time penalty coefficient when the corresponding material sorting or transfer equipment at a node switches from the previous conveying direction to the current conveying path direction.

[0063] Step 6.2: Initialize the set of destructive operators for the adaptive large neighborhood search algorithm Repair operator set The weights of each destruction operator and each repair operator;

[0064] Step 6.3: Select the destroying operator and the repairing operator according to the operator weight;

[0065] Step 6.4: Generate a neighborhood scheduling scheme: First, use the selected disruption operator to generate a neighborhood scheduling scheme from the current scheduling scheme. Remove some replenishment tasks from the process, and then reinsert the removed replenishment tasks using the selected repair operator. During each insertion, the spatiotemporal A* algorithm from step 5.1 is called to re-search the conveyor belt path to obtain the neighborhood scheduling scheme. ;

[0066] Step 6.5: Use simulated annealing criteria to determine whether to accept the neighborhood scheduling scheme. If satisfied If so, the neighborhood scheduling scheme will be accepted. If not satisfied, then according to probability. Accept neighborhood scheduling scheme ,in, Indicates neighborhood scheduling scheme The probability of acceptance; Indicates the current temperature;

[0067] Step 6.6: Update the operator weights according to the following formula based on the improvement effect of the neighborhood scheduling scheme:

[0068] ;

[0069] In the formula, Indicates the operator weight learning rate; Operator Improved value of the objective function Indicates neighborhood scheduling scheme The objective function value;

[0070] Step 6.7: Update the current temperature in the simulated annealing criteria. ,in, This represents the cooling coefficient, and satisfies... ;

[0071] Step 6.8: Repeat steps 6.3 to 6.7 for one iteration until the preset iteration stopping condition is met. Then, output the current neighborhood scheduling scheme as the globally optimal scheduling scheme. Global optimal scheduling scheme This includes the target silo number, replenishment amount, planned start time, planned end time, conveyor path, and execution status for each replenishment task.

[0072] Optionally, according to the intelligent scheduling method, step 7 includes:

[0073] Step 7.1: From the globally optimal scheduling scheme Extract replenishment tasks, generate a list of replenishment tasks to be executed, and initialize the list of replenishment tasks in progress, as shown below:

[0074] ;

[0075] ;

[0076] In the formula, Indicates time List of pending material replenishment tasks; Indicates time List of material replenishment tasks in progress; Indicates material replenishment task The corresponding target silo number; Indicates material replenishment task At any moment The execution status; Indicates material replenishment task Currently in a pending execution state. Indicates material replenishment task Currently in execution status. Indicates material replenishment task It is in a completed state;

[0077] Step 7.2: List of replenishment tasks to be performed The replenishment task k in the calculation starts the judgment value. If the material replenishment task When the following conditions are met ,otherwise ;

[0078] ;

[0079] In the formula, Represents the edge of the conveyor belt network Spacetime occupancy table; Representative node Spacetime occupancy table; This indicates the number of replenishment tasks being executed at the current time t; Indicates the maximum number of concurrent material replenishment tasks;

[0080] The list of startable material replenishment tasks that meet the start conditions at time t is as follows: ;

[0081] Step 7.3: Determine if If yes, then the replenishment task will not be started and the process will proceed to step 7.5 to determine whether the dynamic rescheduling trigger condition is met; if no, then replenishment tasks will be selected from the list of startable replenishment tasks to form a set of startable replenishment tasks. And satisfy:

[0082] ;

[0083] For any replenishment task To replenish materials The corresponding conveyor network edge, transfer node, and material distribution equipment are issued start control commands, and the conveyor network edge, transfer node, and material distribution equipment are started sequentially from the target silo side to the material collection point side, and the material replenishment task is assigned. Move from the list of pending replenishment tasks to the list of ongoing replenishment tasks:

[0084] ;

[0085] In the formula, Indicates time Select the set of replenishment tasks to be initiated; This indicates the actual start time of the material replenishment task k; Indicates material replenishment task Updated to "In Progress" status; This indicates the moment before the replenishment task status was updated; Indicates the moment after the replenishment task status was updated;

[0086] Step 7.4: Monitor the ongoing replenishment task and stop the replenishment task when the stopping conditions are met, including:

[0087] List of material replenishment tasks in progress The replenishment task Calculate the stopping threshold:

[0088] ;

[0089] In the formula, Indicates material replenishment task At any moment The stop judgment value; Indicates material replenishment task The corresponding target silo number; Indicates material replenishment task From the actual start time up to the current moment The cumulative actual amount of material replenished;

[0090] when Stop the material replenishment task at that time. The corresponding conveyor network edge, transfer node, and distribution equipment are stopped sequentially from the material intake point to the target silo side, and the replenishment task is then completed. Remove from the list of ongoing replenishment tasks:

[0091] ;

[0092] In the formula, Indicates material replenishment task The actual end time; Indicates material replenishment task Updated to completed status;

[0093] Step 7.5: Determine whether the dynamic rescheduling trigger condition is met. If it is met, proceed to step 7.6; otherwise, proceed to step 7.7.

[0094] Step 7.6: Lock the list of ongoing material replenishment tasks The replenishment task in the process returns to step 2 based on the updated target silo current material level, equipment status, production plan data and time and space occupancy table, and executes steps 2 to 6 in sequence to update the global optimal scheduling scheme.

[0095] Step 7.7: Determine if If yes, proceed to step 7.2; otherwise, determine that the replenishment task in the global optimal scheduling scheme has been completed.

[0096] Optionally, according to the intelligent scheduling method, if any of the following conditions occur, the dynamic rescheduling trigger condition is determined to be met:

[0097] (1) The conveyor network edge or device corresponding to the material replenishment task to be executed is unavailable, which makes the original path unable to be executed;

[0098] (2) The production plan has changed, and the planned consumption of the target silo in the rolling scheduling window deviates from the original scheduling plan;

[0099] (3) The target silo is in an abnormal state and the silo is not covered by the currently pending or executing material replenishment task;

[0100] (4) There is a deviation between the actual execution of the material replenishment task and the plan.

[0101] The beneficial effects of adopting the above technical solution are as follows:

[0102] (1) Improves the foresight of identifying material shortage risks in target silos and reduces the lag caused by relying solely on the current material level. This invention collects historical material level data, historical feed data, historical discharge data, and real-time operation data of target silos, and uses a predictive model to calculate the remaining usage time of the target silos. This enables the scheduling process to be expanded from judging the current material level to judging the subsequent material level change trend. Thus, the material replenishment demand can be identified in advance before the target silo actually drops to a low material level, reducing the problems of material shortage, waiting for material, or reduced load operation caused by the lag in the generation of material replenishment tasks.

[0103] (2) It realizes the quantitative evaluation of the urgency of replenishment tasks and improves the rationality of the replenishment task ranking. This invention incorporates the remaining usage time of the target silo, the planned demand, and the inventory safety margin into the calculation of the urgency of the replenishment task. This makes the priority of the replenishment task no longer solely dependent on the task generation time, fixed silo level, or single material level threshold, but can comprehensively reflect the material shortage risk of the target silo, the intensity of production demand, and the inventory safety status. This allows emergency replenishment tasks to be prioritized and reduces the risk of untimely material supply to key target silos.

[0104] (3) Enhanced the coordination of conveying resources among multiple replenishment tasks, reducing the possibility of path conflicts and equipment conflicts. This invention establishes a network model including conveyor belts, transfer nodes, material distribution equipment and target silos, and combines equipment occupancy status, maintenance plans and path mutual exclusion constraints to uniformly plan the conveying path and execution sequence of replenishment tasks, so that multiple replenishment tasks can be coordinated and executed in the same scheduling window according to resource availability, thereby reducing the problems of repeated conveyor belt occupation, upstream and downstream waiting and local congestion.

[0105] (4) The scheduling scheme is improved to adapt to dynamic changes on site and the failure of static scheduling is reduced. The present invention updates the scheduling scheme by adopting a rolling time domain method and triggers dynamic rescheduling when equipment failure, path unavailability, equipment maintenance plan changes, production plan adjustment, new emergency replenishment tasks, or actual execution deviation exceeds a preset threshold. This enables the scheduling scheme to be recalculated and adjusted according to the real-time status on site, thereby reducing scheduling failure and task interruption caused by changes in equipment status or production demand.

[0106] (5) This invention establishes a data loop between task execution and subsequent scheduling optimization, thereby improving the automation management level of bulk raw material transportation systems in steel enterprises. By recording the actual start time, actual end time, actual replenishment amount, actual occupied path, and abnormal event information during the replenishment task execution process, and writing the aforementioned execution data back to the database, the execution results can serve as the data basis for subsequent prediction model updates and scheduling performance analysis. This reduces the workload of scheduling personnel repeatedly relying on experience to correct solutions and improves the reliability of subsequent replenishment prediction and scheduling decisions. Attached Figure Description

[0107] Figure 1 This is a flowchart illustrating the intelligent scheduling method applied to the bulk raw material conveying system of a steel enterprise according to this embodiment.

[0108] Figure 2 This is a schematic diagram of a bulk cargo handling scenario at a steel raw material plant.

[0109] Figure 3This embodiment illustrates the data acquisition and training process for predicting the remaining usage time of the target silo.

[0110] Figure 4 This implementation method is a flowchart illustrating the process of optimizing and updating the current scheduling scheme based on an adaptive large neighborhood search algorithm. Detailed Implementation

[0111] To facilitate understanding of this application, a more comprehensive description of this application will be provided below with reference to the accompanying drawings.

[0112] like Figure 1 As shown, this invention provides an intelligent scheduling method for bulk raw material conveying systems in steel enterprises. The method collects and analyzes multi-source data to construct a prediction model of the remaining usage time of target silos and predicts the remaining usage time of each target silo. Based on this, a candidate replenishment task list is generated by combining the current material level of the target silos, production plan data, and predicted remaining usage time, and the urgency score of each candidate replenishment task is calculated. Furthermore, a directed graph of the conveyor network is established and a spatiotemporal occupancy table is initialized. Through constrained spatiotemporal path search and adaptive large neighborhood search algorithms, the execution priority, plan start time, plan end time, and conveyor path of the candidate replenishment tasks are optimized. During the scheduling execution process, the target silo material level status, equipment operating status, production plan data, and spatiotemporal occupancy table are updated through real-time monitoring and dynamic rescheduling mechanisms, thereby achieving a closed-loop response to changes in the bulk raw material conveying site of steel enterprises.

[0113] To enable those skilled in the art to better understand the technical solutions in this specification, the following example, using a steel plant bulk raw material transportation scenario with 18 target silos and the system allowing simultaneous execution of two replenishment tasks, clearly and completely describes the intelligent scheduling method applied to a steel enterprise bulk raw material transportation system. The task scenario in this embodiment is as follows: Figure 2 As shown, the application of the method of the present invention is not limited to the number of silos, the number of devices, or the number of concurrent tasks in this embodiment. Clearly, the described embodiments are only a part of the embodiments of the present invention, and not all of them.

[0114] refer to Figure 1 The intelligent scheduling method applied to the bulk raw material transportation system of steel enterprises in this embodiment includes the following steps:

[0115] Step 1: Construct a prediction model for the remaining usage time of the target silo;

[0116] In this embodiment, reference Figure 3 Step 1 includes the following steps:

[0117] Step 1.1: Collect multi-source operational data required for modeling. Specifically, the multi-source operational data includes target silo-related data, production plan data, downstream equipment operation data, conveyor network topology data, equipment status data, and maintenance plan data; the target silo-related data includes attribute data, historical material level data, historical feed data, and historical discharge data for each target silo; the target silo attribute data includes the historical material level of each target silo at a given time, as well as the rated capacity, low material level threshold, safety stock threshold, and rated high material level for each target silo.

[0118] In this embodiment, a target silo set is defined. Where n represents the target number of silos, in this embodiment Target silo At any moment The current material level is recorded as Target silo The rated capacity is denoted as Target silo The low material level threshold is , Indicates the proportion of low material level threshold; target silo The safety stock threshold is , Indicates the safety stock threshold ratio; target silo The rated high material level is , This indicates the rated high material level ratio. In this embodiment, examples of multi-source data are shown in Table 1.

[0119] Table 1 Example of Multi-Source Operation Data

[0120] Step 1.2: According to fixed time granularity Time alignment is performed on multi-source runtime data to output a standardized time-series dataset; this dataset will be referred to as the standard time-series dataset below.

[0121] in, This indicates the time granularity, measured in minutes. In this embodiment, Differentiated processing for data sources with different sampling periods: If the sampling period of a certain type of data is greater than... If the sampling period of a certain type of data is less than 10 ... Then, mean aggregation is used to complete the downsampling and resampling process.

[0122] Step 1.3: Identify low material level moments based on standard time-series datasets;

[0123] Set the current material level of target silo i. First drop to low material level threshold The time at or below the specified level is identified as a low level event, and this time is called the low level moment, denoted as . If target silo i satisfies equation (1) at time t, then time t is identified as the low material level time of target silo i:

[0124]

[0125] Step 1.4: Define the target silo At a historical moment The training samples are composed of predicted feature vectors and training sample labels The training samples of each target silo at each time step are constructed, and the training samples of each target silo at each time step form a training sample set.

[0126] Step 1.4.1: Use the remaining usage time of the target silos as training sample labels to generate training sample labels for each target silo at each historical moment;

[0127] For the target silo At a historical moment The training samples, if at time The recent low material level event occurred afterward. Then according to Generate target silo The remaining usage time, in hours; if at time Replenishment occurred before any low material level event occurred, and the most recent replenishment started at [time missing]. Then according to Generate remaining usage time tags, where, Indicates target silo The latest time when refueling began; Indicates target silo At the moment when the recent resupply began Material level; Indicates target silo At the start of feeding The average consumption rate before, in t / h; This indicates a very small positive number that prevents the denominator from being zero. In this embodiment, ;

[0128] Step 1.4.2: Construct each target silo At a historical moment Predicted feature vector ;

[0129] Based on the preset historical time window length Construct the target silo according to the eigenvector construction method shown in equation (2). At any moment Predicted feature vector ,in, The preset historical time window length represents the length of a continuous time interval used to extract features such as historical material levels and consumption rates, in hours (h). In this embodiment, This value can be determined based on the average replenishment cycle of the target silo, the material level sampling cycle, and the model verification error; in this embodiment, it is determined to be 8h through cross-validation of historical samples, while in other embodiments it can be adjusted according to the on-site working conditions.

[0130] Predicting feature vectors It can be expressed as follows:

[0131]

[0132] In the formula, Indicates target silo At any moment Material level; Indicates target silo The average consumption rate within a preset historical time window, in t / h; Indicates target silo Standard deviation of consumption rate within a preset historical time window; Indicates target silo Planned consumption within the rolling scheduling window, in tons; This indicates the length of the rolling scheduling window, in units of h. In this embodiment... ; Indicates target silo From the time the last refueling ended to the current time The interval time, in hours; Indicates target silo The most recent replenishment quantity, in tons; This represents the hour code corresponding to the current time t; Indicates the current time Corresponding production shift code; Indicates target silo Corresponding to the load rate of downstream equipment; Indicates target silo The corresponding raw material consumption ratio for downstream equipment; Indicates target silo At any moment The theoretical remaining usage time, in hours;

[0133] Target silo At a historical moment Theoretical remaining usage time Calculate according to the following formula:

[0134]

[0135] Examples of the feature parameters involved in this step are shown in Table 2:

[0136] Table 2 Example of Predicted Feature Parameters

[0137] Step 1.5: Use the training sample set to train a regression prediction model. The trained regression prediction model will be used as the target silo remaining usage time prediction model.

[0138] Specifically, the predicted feature vectors in each training sample Input remaining usage time As output, train a regression prediction model. The training objective function is shown in the following equation:

[0139]

[0140] In the formula, This represents the model parameters after training is complete; Indicates the total number of training samples; Indicates the first The predicted feature vector of each training sample; Indicates the first The training sample labels are specified; the regression prediction model can be a random forest regression model, a gradient boosting regression model, a long short-term memory network model, or other continuous value prediction models. This implementation preferably uses a gradient boosting regression model.

[0141] Step 2: Using the target silo remaining usage time prediction model, predict the remaining usage time of each target silo in real time to obtain the predicted remaining usage time of each target silo;

[0142] In this embodiment, step 2 deploys the trained target silo remaining usage time prediction model to the scheduling server of the bulk raw material transportation system of the steel enterprise, enabling the target silo remaining usage time prediction model to receive real-time multi-source operating data and output the predicted remaining usage time of the target silo. Specifically, this includes the following steps:

[0143] Step 2.1: Obtain the current time Each target silo material level The average consumption rate calculated according to the preset historical time window Production planning data, downstream equipment operation data, and status data of all relevant equipment;

[0144] Step 2.2: Construct each target silo according to the eigenvector construction method shown in equation (2). At the present moment Real-time feature vectors ;

[0145] Step 2.3: Transfer the real-time feature vector Input the target silo's remaining usage time prediction model to obtain the results for each target silo. Predicted remaining usage time initial value As shown in the following formula:

[0146]

[0147] Step 2.4: For each target silo Predicted remaining usage time initial value The reasonableness is verified according to formula (6), and the predicted remaining usage time is obtained. The unit is hours; the predicted remaining usage time is determined according to the following formula:

[0148]

[0149] In the formula, Indicates target silo The theoretical upper limit of the predicted remaining usage time, in hours, is calculated using the following formula:

[0150]

[0151] In the formula, Indicates target silo The minimum effective consumption rate, expressed in t / h.

[0152] Table 3 shows an example of real-time prediction of the target silos in this step:

[0153] Table 3 Example of Real-Time Prediction for Target Silos

[0154] Step 3: Based on the remaining usage time of each target silo, the current material level of each target silo, and the production plan data obtained in Step 2, calculate the urgency score of the corresponding replenishment task for each target silo, arrange all replenishment tasks in order of urgency score, and generate a candidate replenishment task list.

[0155] Step 3.1: Calculate the target silos The corresponding material replenishment task is at time Urgency score The higher the urgency score, the higher the priority of the corresponding material replenishment task, and the more important the target silo. The urgency score is calculated using the following formula:

[0156]

[0157] In the formula, Indicates target silo The remaining usage time; Indicates target silo The planned consumption; Indicates target silo Safety stock gap; , , These represent the weights of remaining usage time, planned consumption, and safety stock deficit, respectively, and the weights satisfy the following conditions: ;

[0158] The remaining usage time risk is calculated using the following formula:

[0159]

[0160] Planned consumption is calculated using the following formula:

[0161]

[0162] In the above formula, Represents the target silo set The number of any target silo; Indicates target silo Planned consumption within the rolling scheduling window;

[0163] Safety stock gap is calculated using the following formula:

[0164]

[0165] In this embodiment, , , The above values ​​are preferred parameters for this implementation method and can be adjusted according to the supply requirements, transportation capacity and historical scheduling effects of steel enterprises, as long as the weights are non-negative and the sum is 1. Indicates target silo Low material level threshold; Indicates target silo Safety stock threshold;

[0166] Step 3.2: Filter the target silos for which candidate replenishment tasks need to be generated. Specifically, the target silos that satisfy the following formula are identified as the target silos for which candidate replenishment tasks need to be generated:

[0167]

[0168] In the formula, This represents the urgency score threshold for each replenishment task corresponding to each target silo. This represents the remaining usage time threshold for each target silo; in this embodiment, , This is the preferred value for this implementation method, and can be adjusted according to the target silo capacity, average conveying time, safety stock strategy, and historical shortage risk.

[0169] Step 3.3: Apply formula (13) to the target silo. The corresponding candidate material replenishment task determines the material replenishment amount. ;

[0170]

[0171] In the formula, Indicates target silo At any moment Material level; Indicates target silo Rated high material level; Target silos Rated capacity; Indicates the target silo within the rolling scheduling window. The maximum allowable replenishment amount, in tons; if calculated... If so, no candidate refueling task will be generated for the target silo;

[0172] Step 3.4: Generate a candidate replenishment task list; specifically, rank the candidate replenishment tasks according to their urgency score. Sort in descending order to generate a list of candidate replenishment tasks. ,in, Indicates time The list of candidate replenishment tasks, Indicates the first One candidate material replenishment task Indicates the first Each refueling task corresponds to a target silo at time [time]. Urgency score

[0173] The list of candidate replenishment tasks for this step is shown in Table 4:

[0174] Table 4 Example of Candidate Replenishment Task List

[0175] Step 4: Abstract the bulk raw material conveying system of the steel enterprise into a directed graph of conveyor belt network. ,in Represents the set of nodes in a conveyor belt network. Represents the set of edges in a conveyor belt network; for each node and each conveyor belt network edge Establish a spatiotemporal occupancy table and determine the edges of the conveyor network based on the spatiotemporal occupancy table. and nodes The availability status;

[0176] Step 4.1: Abstract the bulk raw material conveying system of the steel enterprise into a directed graph of conveyor belt network. .in, Represent a directed graph of a conveyor belt network; Represents the set of nodes in a conveyor belt network; This represents the set of edges in a conveyor belt network. The set of nodes... This includes material ingestion points, transfer nodes, material distribution equipment, conveyor belt intersections, and target silos; conveyor belt network edges. This represents the connection relationship of a conveyor belt in a conveyor belt network, and the edges of the conveyor belt network. The conveying capacity is denoted as The unit is t / h; conveyor belt network edge The delivery time is recorded as The unit is h; conveyor belt network edge The weight is denoted as ;

[0177] Table 5 shows an example of the conveyor network edge parameters for this step:

[0178] Table 5 Example of Conveyor Belt Network Side Parameters

[0179] The equipment status in Table 5 refers to whether the conveyor belt equipment, transfer node equipment, or material sorting equipment corresponding to the edge of the conveyor belt network is available at the current moment.

[0180] Step 4.2: For each node and each conveyor belt network edge Establish a spacetime occupancy table, nodes The spacetime occupancy table is denoted as Conveyor belt network edge The spacetime occupancy table is denoted as If the time interval If a task is already occupied by a replenishment task that has been executed, is in progress, has been scheduled, or is part of a maintenance plan for the equipment corresponding to the edge or node of the conveyor network, then it should be recorded according to the following formula:

[0181]

[0182] In the formula, Indicates the start time of occupation. Indicates the end time of occupation;

[0183] In summary, the spacetime occupancy table is a data table defined in this embodiment, used to record whether a node or conveyor network edge is occupied by a replenishment task or maintenance plan in different time intervals.

[0184] Step 4.3: Determine the resource availability status. Specifically, if the conveyor network edge... exist Inner and Existing Edge Spacetime Occupancy Table There is no intersection, and the edges of the conveyor belt network are in If the internal network is in an available state, then the edge of the conveyor network is determined to be... Available:

[0185]

[0186] In the formula, Represents the edge of the conveyor belt network At any moment The device status; Represents the edge of the conveyor belt network At any moment It is in an available state; Represents the edge of the conveyor belt network At any moment It is currently unavailable.

[0187] If node In the time interval Internal and Existing Node Spacetime Occupancy Table There is no intersection, and the edges of the conveyor belt network are in If the node is in an available state, then the determination node is determined. Available:

[0188]

[0189] In the formula, Represents a node At any moment The device status; Represents a node At any moment It is in an available state; Represents a node At any moment It is currently unavailable.

[0190] Step 5: For each material replenishment task in the candidate replenishment task list, search for candidate spatiotemporal paths and construct an initial scheduling scheme based on the directed graph of the conveyor network and the spatiotemporal occupancy table.

[0191] Step 5.1: For candidate material replenishment tasks Based on the established directed graph of the conveyor network, the spatiotemporal A* algorithm is used to search for the set of candidate spatiotemporal paths from the material picking node to the target silo node. , Indicates material replenishment task The There are 10 candidate spatiotemporal paths; each spatiotemporal path consists of a sequence of directed edges from the material picking node to the target silo; for the material replenishment task... Any conveyor path Then the conveyor belt path The conveying capacity is ,in Represents the edge of the conveyor belt network Conveying capacity; replenishment tasks On the conveyor belt path Duration on Calculate according to formula (17):

[0192]

[0193] In the formula, Represents the edge of the conveyor belt network Delivery time; Indicates material replenishment task On the conveyor belt path The duration of the event, in hours; Indicates material replenishment task The amount of material to be added, in tons; Represents a node The corresponding material sorting or transfer equipment switches from the previous conveying direction to the material replenishment task. The time required to switch the transmission direction, in hours;

[0194] Step 5.2: Construct the initial scheduling scheme ;

[0195] Specifically, according to the candidate replenishment task list The urgency scores are sorted in descending order to determine the replenishment tasks for each candidate replenishment task. The overall start time of the plan Material replenishment task The overall end time of the plan and conveyor belt path To form an initial scheduling scheme ;

[0196] Step 5.3: Initial scheduling scheme The already determined replenishment tasks are written into the spacetime occupancy table to constrain the path search of subsequent replenishment tasks. In this embodiment, in order to improve the execution stability of the scheduling scheme, all conveyor network edges and nodes on the conveyor path of the replenishment task are considered to be occupied during the execution of the replenishment task.

[0197] Specifically, for the material replenishment task Occupied conveyor network edge The time interval occupied is denoted as And write it according to the following formula:

[0198]

[0199] In the formula, Indicates material replenishment task Occupy conveyor network edge The beginning moment; Indicates material replenishment task Occupy conveyor network edge The end time;

[0200] For replenishment tasks Occupied nodes The time interval occupied is denoted as And write it according to the following formula:

[0201]

[0202] In the formula, Indicates material replenishment task Occupying nodes The beginning moment; Indicates material replenishment task Occupying nodes The end time;

[0203] Examples of candidate spatiotemporal paths for this step are shown in Table 6:

[0204] Table 6 Examples of Candidate Spatiotemporal Paths

[0205]

[0206] Step 6: Use the adaptive large neighborhood search algorithm to optimize and update the current scheduling scheme, and output the globally optimal scheduling scheme;

[0207] In this embodiment, reference Figure 4 Step 6 includes the following steps:

[0208] Step 6.1: Evaluate the initial scheduling scheme; specifically, construct the objective function of the scheduling scheme. And evaluate the scheduling scheme according to formula (20). .in, Represents any scheduling scheme; Representing the scheduling scheme The objective function value;

[0209]

[0210] In the formula, Indicates the latest completion time of the scheduling scheme; Indicates material replenishment task The delay time; Indicates the conveyor belt path The path cost is the cost of the conveyor network. The higher the path cost, the higher the cost of the conveyor path to the network resources, and the less likely it is to be selected in scheduling optimization. Represents the edge of the conveyor belt network Load rate; These represent the weights of the objective function.

[0211] Material replenishment task Delay time Calculate according to the following formula:

[0212]

[0213] In the formula, Indicates task The predicted remaining usage time of the corresponding target silo; This indicates the reserved safety time, in hours (h).

[0214] Conveyor Belt Path Path cost Calculate according to the following formula:

[0215]

[0216] In the formula, This represents the switching time penalty coefficient when the corresponding material sorting or transfer equipment switches from the previous conveying direction to the current conveying path direction.

[0217] Step 6.2: Initialize the search operators and their weights; specifically, initialize the set of destructive operators for the adaptive large neighborhood search algorithm. Repair operator set The weights of each destruction operator and each repair operator are listed in Table 7. Examples of search operators for this step are shown in Table 7.

[0218] Table 7 Examples of operators for the adaptive large neighborhood search algorithm

[0219] Step 6.3: Select the destroying operator and the repairing operator according to the operator weight; Operator The probability of selection is ,in, Represents the search operator The current weight; This indicates whether to destroy or repair the set of operators. express Any search operator in the range.

[0220] Step 6.4: Generate a neighborhood scheduling scheme; first, use the selected disruption operator to generate a scheme from the current scheduling scheme. Remove some replenishment tasks from the process, and then reinsert the removed replenishment tasks using the selected repair operator. During each insertion, the spatiotemporal A* algorithm from step 5.1 is called to re-search the conveyor belt path to obtain the neighborhood scheduling scheme. .

[0221] Step 6.5: Use simulated annealing criteria to determine whether to accept the neighborhood scheduling scheme. If satisfied If so, the neighborhood scheduling scheme will be accepted. If not satisfied, then according to probability. Accept neighborhood scheduling scheme ,in, Indicates neighborhood scheduling scheme The probability of acceptance; This indicates the current temperature.

[0222] Step 6.6: Update the operator weights based on the improvement effect of the neighborhood scheduling scheme. Operator The weights are updated according to the following formula:

[0223]

[0224] In the formula, Indicates the operator weight learning rate; Operator Improved value of the objective function Indicates the current scheduling scheme The objective function value; Indicates neighborhood scheduling scheme The objective function value. Since the objective function of the scheduling scheme is to minimize, therefore The larger the value, the better the neighborhood scheduling scheme. The better the improvement.

[0225] Step 6.7: Update the current temperature in the simulated annealing criteria. ,in, This represents the cooling coefficient, and satisfies... ;

[0226] Step 6.8: Repeat steps 6.3 to 6.7 for one iteration cycle until the maximum number of iterations is reached. Or, the number of consecutive times the current globally optimal scheduling scheme has not been updated has reached the maximum number of times it has not been improved. At that time, output the globally optimal scheduling scheme. Global optimal scheduling scheme This includes the target silo number, replenishment quantity, planned start time, planned end time, conveyor path, and execution status for each replenishment task. An example of a scheduling scheme is shown in Table 8.

[0227] Table 8. Example of Scheduling Scheme

[0228] Step 7: Execute the replenishment task in the current global optimal scheduling scheme, generate replenishment task execution data, and update the global optimal scheduling scheme according to the dynamic rescheduling conditions. The last updated global optimal scheduling scheme is the intelligent scheduling scheme for the bulk raw material transportation system of the steel enterprise.

[0229] Step 7.1: The globally optimal scheduling scheme output from Step 6 Extract replenishment tasks, generate a list of replenishment tasks to be executed, and initialize the list of replenishment tasks in progress;

[0230] Specifically, let's assume a globally optimal scheduling scheme. The time of execution is From the globally optimal scheduling scheme Extract replenishment tasks and generate a list of replenishment tasks to be executed:

[0231]

[0232]

[0233] In the formula, Indicates time List of pending material replenishment tasks; Indicates time List of material replenishment tasks in progress; Indicates material replenishment task The corresponding target silo number; Indicates material replenishment task At any moment The execution status; Indicates material replenishment task Currently in a pending execution state. Indicates material replenishment task Currently in execution status. Indicates material replenishment task It is in a completed state;

[0234] Step 7.2: Determine the replenishment tasks in the list of tasks to be executed. Check if the startup conditions are met, and generate a list of startup replenishment tasks;

[0235] Specifically, the list of tasks to be performed for replenishing materials. The replenishment task k in the calculation starts the judgment value. If the material replenishment task When the following conditions are met ,otherwise ;

[0236]

[0237] In the formula, This indicates the number of replenishment tasks being executed at the current time t; This indicates the maximum number of concurrent material replenishment tasks. In this implementation, ;

[0238] The list of startable material replenishment tasks that meet the start conditions at time t is as follows:

[0239] Step 7.3: Start the replenishment task according to the list of startable replenishment tasks, or wait for the next judgment time when the list of startable replenishment tasks is empty;

[0240] Specifically, if Select replenishment tasks from the list of initiable replenishment tasks to form an initiation replenishment task set. And satisfy:

[0241]

[0242] For any replenishment task To replenish materials The corresponding conveyor network edge, transfer node, and material distribution equipment are issued start control commands, and the conveyor network edge, transfer node, and material distribution equipment are started sequentially from the target silo side to the material collection point side, and the material replenishment task is assigned. Move from the list of pending replenishment tasks to the list of ongoing replenishment tasks:

[0243]

[0244] In the formula, Indicates time Select the set of replenishment tasks to be initiated; This indicates the actual start time of the material replenishment task k; Indicates material replenishment task Updated to "In Progress" status; This indicates the moment before the replenishment task status was updated; This indicates the moment after the replenishment task status was updated.

[0245] like If the material replenishment task is not initiated, proceed to step 7.5 to determine whether the dynamic rescheduling trigger condition is met.

[0246] Step 7.4: Monitor the material replenishment task in progress and stop the material replenishment task when the stopping conditions are met;

[0247] List of material replenishment tasks in progress The replenishment task Calculate the stopping threshold:

[0248]

[0249] In the formula, Indicates material replenishment task At any moment The stop judgment value; Indicates material replenishment task The corresponding target silo number; Indicates material replenishment task From the actual start time up to the current moment The cumulative actual amount of material replenished.

[0250] when Stop the material replenishment task at that time. The corresponding conveyor network edge, transfer node, and distribution equipment are stopped sequentially from the material intake point to the target silo side, and the replenishment task is then completed. Remove from the list of ongoing replenishment tasks:

[0251]

[0252] In the formula, Indicates material replenishment task The actual end time; Indicates material replenishment task Updated to completed status.

[0253] Step 7.5: Determine whether the dynamic rescheduling trigger condition is met based on the real-time running data. If it is met, proceed to step 7.6; otherwise, proceed to step 7.7.

[0254] Dynamic rescheduling is triggered when the real-time operating status deviates significantly from the original scheduling scheme or when key constraints change.

[0255] Specifically, the following situations apply: First, the conveyor network edge or equipment corresponding to the replenishment task to be executed becomes unavailable, causing the original path to be unexecutable; second, the production plan changes, and the planned consumption of the target silo within the rolling scheduling window deviates significantly from the original scheduling scheme; third, the target silo experiences an abnormal state, such as the remaining usage time of the target silo being lower than the safety threshold, and the silo not being covered by the currently pending or executing replenishment task; fourth, the actual execution of the executing replenishment task deviates significantly from the plan, such as the actual start time deviating significantly from the planned start time. When any of the above situations occur, the dynamic rescheduling trigger condition is determined to be met; otherwise, the dynamic rescheduling trigger condition is determined not to be met.

[0256] Step 7.6: Continue executing the replenishment task or update the globally optimal scheduling scheme based on the dynamic rescheduling trigger condition; when dynamic rescheduling is triggered, lock the list of ongoing replenishment tasks. The replenishment task in the process returns to step 2 based on the updated target silo's current material level, equipment status, production plan data, and time and space occupancy table, and then executes steps 2 to 6 in sequence to generate an updated global optimal scheduling scheme.

[0257] Step 7.7: When the dynamic rescheduling trigger condition is not met, and When the condition is met, proceed to the next judgment point and continue executing steps 7.2 to 7.6. When the time is right, the replenishment task in the globally optimal scheduling scheme is determined to be completed.

[0258] It should be understood that, inspired by the technical concept of this invention, those skilled in the art can make various improvements or modifications based on the above content without departing from the scope of this invention, and these modifications still fall within the protection scope of this invention.

Claims

1. An intelligent scheduling method for bulk raw material conveying systems in steel enterprises, characterized in that, The method includes: Step 1: Construct a prediction model for the remaining usage time of the target silo; Step 2: Use the target silo remaining usage time prediction model to predict the remaining usage time of each target silo in real time; Step 3: Based on the predicted remaining usage time of each target silo, the current material level of each target silo, and the production plan data, calculate the urgency score of the corresponding replenishment task for each target silo, and arrange all replenishment tasks in order of urgency score to form a candidate replenishment task list. Step 4: Abstract the bulk raw material conveying system of the steel enterprise into a directed graph of conveyor belt network. ,in Represents the set of nodes in a conveyor belt network. Represents the set of edges in a conveyor belt network; for each node and each conveyor belt network edge Establish a spatiotemporal occupancy table; the spatiotemporal occupancy table is used to record whether the corresponding node or the corresponding conveyor belt network edge is occupied in different time intervals; Step 5: For each replenishment task in the candidate replenishment task list, based on... Search for candidate spatiotemporal paths using the spatiotemporal occupancy table, and then construct the current scheduling scheme; Step 6: Use the adaptive large neighborhood search algorithm to optimize and update the current scheduling scheme, and output the current globally optimal scheduling scheme; Step 7: Execute the replenishment task in the current global optimal scheduling scheme and update the global optimal scheduling scheme according to the dynamic rescheduling conditions. The last updated global optimal scheduling scheme is the intelligent scheduling scheme of the bulk raw material transportation system of the steel enterprise.

2. The intelligent scheduling method according to claim 1, characterized in that, Step 1 includes: Step 1.1: Collect multi-source operational data, including target silo-related data, production plan data, downstream equipment operation data, conveyor network topology data, equipment status data, and maintenance plan data; the target silo-related data includes attribute data, historical material level data, historical feed data, and historical discharge data for each target silo; the target silo attribute data includes the historical material level of each target silo, as well as the rated capacity, low material level threshold, safety stock threshold, and rated high material level for each target silo. Step 1.2: Using a fixed time granularity The raw multi-source runtime data collected in step 1.1 is time-aligned to generate a standard time-series dataset; Step 1.3: Identify low material level moments based on standard time series datasets. ; Step 1.4: Define the target silo At a historical moment The training samples are composed of predicted feature vectors and training sample labels The training samples for each target silo are constructed based on the standard time series dataset and the low material level time. The training samples of each target silo constitute the training sample set. Step 1.5: Use the training sample set to train a regression prediction model. The trained regression prediction model will be used as the target silo remaining usage time prediction model.

3. The intelligent scheduling method according to claim 2, characterized in that, Step 1.4 includes: Step 1.4.1: Use the remaining usage time of the target silos as training sample labels to generate training sample labels for each target silo; If the target silo At a historical moment There will be a low material level moment afterwards. Then according to Generate target silo The remaining usage time; if at time Replenishment occurred before a low material level was reached, and the most recent replenishment started at [time missing]. Then according to Generate target silo Remaining usage time, Indicates target silo At the start of feeding The average consumption rate before; This indicates a very small positive number that prevents the denominator from being zero; Step 1.4.2: Construct each target silo according to the following formula. At a historical moment Predicted feature vector : ； In the formula, Indicates target silo At any moment Material level; Indicates target silo The average consumption rate within a preset historical time window; Indicates target silo Standard deviation of consumption rate within a preset historical time window; Indicates target silo Planned consumption within the rolling scheduling window; Indicates the length of the rolling scheduling window; Indicates target silo From the time the last refueling ended to the current time The interval time; Indicates target silo The amount of the most recent replenishment; This represents the hour code corresponding to the current time t; Indicates the current time Corresponding production shift code; Indicates target silo Corresponding to the load rate of downstream equipment; Indicates target silo The corresponding raw material consumption ratio for downstream equipment; Indicates target silo At any moment Theoretical remaining usage time, Calculate according to the following formula: 。 4. The intelligent scheduling method according to claim 2, characterized in that, Step 2 includes: Step 2.1: Obtain the current time Each target silo material level The average consumption rate calculated based on a preset historical time window. Production planning data, downstream equipment operation data, and equipment status data; Step 2.2: Construct each target silo At the present moment Real-time feature vectors ; Step 2.3: Transfer the real-time feature vector Input the target silo's remaining usage time prediction model to obtain the results for each target silo. Predicted remaining usage time initial value ; Step 2.4: For each target silo Predicted remaining usage time initial value The rationality is verified using the following formula to obtain the target silos. Predicted remaining usage time ; ； In the formula, Indicates target silo The theoretical upper limit of the predicted remaining usage time is calculated according to the following formula: ； In the formula, Indicates target silo The minimum effective consumption rate.

5. The intelligent scheduling method according to claim 4, characterized in that, Step 3 includes the following steps: Step 3.1: Calculate the target silos according to the following formula. The corresponding material replenishment task is at time Urgency score ; ； In the formula, Indicates target silo The remaining usage time; Indicates target silo The planned consumption amount, Represents the target silo set The number of any target silo. Indicates target silo Planned consumption within the rolling scheduling window; Indicates target silo Safety stock gap, Indicates target silo Low material level threshold, Indicates target silo Safety stock threshold; , , These represent the weights of remaining usage time, planned consumption, and safety stock gap, respectively, and satisfy the following conditions: ; Step 3.2: Identify the target silos that satisfy the following formula as the target silos for which candidate feeding tasks need to be generated: ； In the formula, This represents the urgency score threshold for each replenishment task corresponding to each target silo. This represents the remaining usage time threshold for each target silo; Step 3.3: Calculate the target silo according to the following formula. The required amount of material for the corresponding candidate material replenishment task If calculated If so, no candidate refueling task will be generated for the target silo; ； In the formula, Indicates target silo At any moment Material level; Indicates target silo Rated high material level; Target silos Rated capacity; Indicates the target silo within the rolling scheduling window. The maximum amount of material replenishment allowed to be completed; Step 3.4: Sort the candidate material replenishment tasks by urgency score Sort in descending order to generate a list of candidate replenishment tasks. ,in, Indicates the first One candidate material replenishment task Indicates the first Each refueling task corresponds to a target silo at time [time]. The urgency score.

6. The intelligent scheduling method according to claim 5, characterized in that, Step 5 includes: Step 5.1: For candidate material replenishment tasks Based on the directed graph of the conveyor belt network, the spatiotemporal A* algorithm is used to search for the set of candidate spatiotemporal paths from the material picking node to the target silo node. , Indicates material replenishment task The There are 10 candidate spatiotemporal paths; each spatiotemporal path consists of a sequence of directed edges from the material picking node to the target silo; for the material replenishment task... Any conveyor path Then the conveyor belt path The conveying capacity is ,in Represents the edge of the conveyor belt network The conveying capacity; and the replenishment task is calculated according to the following formula. On the conveyor belt path Duration on : ； In the formula, Represents the edge of the conveyor belt network Delivery time; Indicates material replenishment task The amount of material to be added; Represents a node The corresponding material sorting or transfer equipment switches from the previous conveying direction to the material replenishment task. The time required to switch the transmission direction; Step 5.2: According to the candidate replenishment task list The urgency scores are sorted in descending order to determine the replenishment tasks for each candidate replenishment task. The overall start time of the plan Material replenishment task The overall end time of the plan and conveyor belt path To form an initial scheduling scheme ; Step 5.3: [The sentence is incomplete and requires more context to be translated accurately.] The replenishment tasks that have been determined are written into the spacetime occupancy table. During the execution of the replenishment task, all conveyor network edges and nodes on the corresponding conveyor path are considered to be occupied.

7. The intelligent scheduling method according to claim 6, characterized in that, Step 6 includes: Step 6.1: Construct the current scheduling scheme objective function : ； In the formula, Indicates the latest completion time of the scheduling scheme; Indicates material replenishment task The delay time; Indicates the conveyor belt path Path cost; Represents the edge of the conveyor belt network Load rate; These represent the weights of the objective function; Material replenishment task Delay time Calculate according to the following formula: ； In the formula, Indicates task The predicted remaining usage time of the corresponding target silo; This indicates that a safety time has been reserved; Conveyor Belt Path Path cost Calculate according to the following formula: ； In the formula, This represents the switching time penalty coefficient when the corresponding material sorting or transfer equipment at a node switches from the previous conveying direction to the current conveying path direction. Step 6.2: Initialize the set of destructive operators for the adaptive large neighborhood search algorithm Repair operator set The weights of each destruction operator and each repair operator; Step 6.3: Select the destroying operator and the repairing operator according to the operator weight; Step 6.4: Generate a neighborhood scheduling scheme: First, use the selected disruption operator to generate a neighborhood scheduling scheme from the current scheduling scheme. Remove some replenishment tasks from the process, and then reinsert the removed replenishment tasks using the selected repair operator. During each insertion, the spatiotemporal A* algorithm from step 5.1 is called to re-search the conveyor belt path to obtain the neighborhood scheduling scheme. ; Step 6.5: Use simulated annealing criteria to determine whether to accept the neighborhood scheduling scheme. If satisfied If so, the neighborhood scheduling scheme will be accepted. If not satisfied, then according to probability. Accept neighborhood scheduling scheme ,in, Indicates neighborhood scheduling scheme The probability of acceptance; Indicates the current temperature; Step 6.6: Update the operator weights according to the following formula based on the improvement effect of the neighborhood scheduling scheme: ； In the formula, Indicates the operator weight learning rate; Operator Improved value of the objective function Indicates neighborhood scheduling scheme The objective function value; Step 6.7: Update the current temperature in the simulated annealing criteria. ,in, This represents the cooling coefficient, and satisfies... ; Step 6.8: Repeat steps 6.3 to 6.7 for one iteration until the preset iteration stopping condition is met. Then, output the current neighborhood scheduling scheme as the globally optimal scheduling scheme. Global optimal scheduling scheme This includes the target silo number, replenishment amount, planned start time, planned end time, conveyor path, and execution status for each replenishment task.

8. The intelligent scheduling method according to claim 6, characterized in that, Step 7 includes: Step 7.1: From the globally optimal scheduling scheme Extract replenishment tasks, generate a list of replenishment tasks to be executed, and initialize the list of replenishment tasks in progress, as shown below: ； ； In the formula, Indicates time List of pending material replenishment tasks; Indicates time List of material replenishment tasks in progress; Indicates material replenishment task The corresponding target silo number; Indicates material replenishment task At any moment The execution status; Indicates material replenishment task Currently in a pending execution state. Indicates material replenishment task Currently in execution status. Indicates material replenishment task It is in a completed state; Step 7.2: List of replenishment tasks to be performed The replenishment task k in the calculation starts the judgment value. If the material replenishment task When the following conditions are met ,otherwise ; ； In the formula, Represents the edge of the conveyor belt network Spacetime occupancy table; Representative node Spacetime occupancy table; This indicates the number of replenishment tasks being executed at the current time t; Indicates the maximum number of concurrent material replenishment tasks; The list of startable material replenishment tasks that meet the start conditions at time t is as follows ; Step 7.3: Determine if If yes, then the replenishment task will not be started and the process will proceed to step 7.5 to determine whether the dynamic rescheduling trigger condition is met; if no, then replenishment tasks will be selected from the list of startable replenishment tasks to form a set of startable replenishment tasks. And satisfy: ； For any replenishment task To replenish materials The corresponding conveyor network edge, transfer node, and material distribution equipment are issued start control commands, and the conveyor network edge, transfer node, and material distribution equipment are started sequentially from the target silo side to the material collection point side, and the material replenishment task is assigned. Move from the list of pending replenishment tasks to the list of ongoing replenishment tasks: ； In the formula, Indicates time Select the set of replenishment tasks to be initiated; This indicates the actual start time of the material replenishment task k; Indicates material replenishment task Updated to "In Progress" status; This indicates the moment before the replenishment task status was updated; Indicates the moment after the replenishment task status was updated; Step 7.4: Monitor the ongoing replenishment task and stop the replenishment task when the stopping conditions are met, including: List of material replenishment tasks in progress The replenishment task Calculate the stopping threshold: ； In the formula, Indicates material replenishment task At any moment The stop judgment value; Indicates material replenishment task The corresponding target silo number; Indicates material replenishment task From the actual start time up to the current moment The cumulative actual amount of material replenished; when Stop the material replenishment task at that time. The corresponding conveyor network edge, transfer node, and distribution equipment are stopped sequentially from the material intake point to the target silo side, and the replenishment task is then completed. Remove from the list of ongoing replenishment tasks: ； In the formula, Indicates material replenishment task The actual end time; Indicates material replenishment task Updated to completed status; Step 7.5: Determine whether the dynamic rescheduling trigger condition is met. If it is met, proceed to step 7.6; otherwise, proceed to step 7.

7. Step 7.6: Lock the list of ongoing material replenishment tasks The replenishment task in the process returns to step 2 based on the updated target silo current material level, equipment status, production plan data and time and space occupancy table, and executes steps 2 to 6 in sequence to update the global optimal scheduling scheme. Step 7.7: Determine if If yes, proceed to step 7.2; otherwise, determine that the replenishment task in the global optimal scheduling scheme has been completed.

9. The intelligent scheduling method according to claim 7, characterized in that, The dynamic rescheduling trigger condition is determined to be met if any of the following conditions are met: (1) The conveyor network edge or device corresponding to the material replenishment task to be executed is unavailable, which makes the original path unable to be executed; (2) The production plan has changed, and the planned consumption of the target silo in the rolling scheduling window deviates from the original scheduling plan; (3) The target silo is in an abnormal state and the silo is not covered by the currently pending or executing material replenishment task; (4) There is a deviation between the actual execution of the material replenishment task and the plan.

Images