Intelligent production scheduling method and related equipment

Abstract

The embodiment of the invention discloses an intelligent production scheduling method and related equipment, which are used for performing intelligent production scheduling under the condition of improving the production scheduling efficiency. The method comprises the steps that configuration information input in an interface is acquired, the configuration information comprises productivity configuration of a production line and constraint conditions of production scheduling, a large model is utilized to analyze order data of multiple target plan orders in a target batch order set according to preset cue words and the configuration information, and order data of multiple target plan orders in the target batch order set are obtained; obtaining initial production scheduling strategies corresponding to a plurality of target plan orders output by the large model, and analyzing order data of the plurality of target plan orders by using the small model based on the initial production scheduling strategies, the configuration information, the current residual capacity of the current production line and an optimization algorithm, and obtaining a target production scheduling strategy corresponding to the plurality of target plan orders output by the small model, thereby performing production scheduling on the plurality of target plan orders based on the target production scheduling strategy.

Description

Technical Field

[0001] The embodiments of this application relate to the field of intelligent scheduling, and more specifically, to intelligent scheduling methods, intelligent scheduling devices, intelligent scheduling equipment, computer-readable storage media, and computer program products containing instructions. Background Technology

[0002] As business complexity increases, existing scheduling systems need to be improved to adapt to changing needs.

[0003] Currently, the goal of the production scheduling system is to find the optimal or near-optimal production plan under multiple production constraints. To achieve these goals, the system employs optimization algorithms such as genetic algorithms and annealing algorithms. In practical engineering applications, this typically requires first modeling the business process and then using a rule engine to handle multiple constraints. Next, the rule engine calls predefined underlying optimization algorithms to calculate the solution. The system pre-configures a series of rules based on the expected business scenario, and each rule triggers the corresponding algorithm for calculation.

[0004] However, whenever users have new requirements or business scenarios change, it is often necessary to rewrite the code to adapt to these changes. Therefore, the scheduling efficiency of existing scheduling methods is low. Summary of the Invention

[0005] This application provides an intelligent scheduling method, an intelligent scheduling device, an intelligent scheduling equipment, a computer-readable storage medium, and a computer program product containing instructions, for performing intelligent scheduling while improving scheduling efficiency.

[0006] In a first aspect, embodiments of this application provide an intelligent scheduling method, including:

[0007] Obtain the configuration information entered in the interface, including the production line capacity configuration and production scheduling constraints;

[0008] Using a large model, based on preset prompts and the configuration information, the order data of multiple target planned orders in the target batch order set are analyzed to obtain the initial production scheduling strategy corresponding to the multiple target planned orders output by the large model;

[0009] Using the trained small model, based on the initial production scheduling strategy, the configuration information, the current remaining capacity of the current production line, and the optimization algorithm, the order data of the multiple target planned orders are analyzed to obtain the target production scheduling strategy corresponding to the multiple target planned orders output by the small model, so as to schedule the multiple target planned orders based on the target production scheduling strategy.

[0010] Secondly, embodiments of this application provide an intelligent production scheduling device, comprising:

[0011] The acquisition unit is used to acquire the configuration information input in the interface, the configuration information including the production line capacity configuration and production scheduling constraints;

[0012] The large model processing unit is used to analyze the order data of multiple target planned orders in the target batch order set using the large model according to the preset prompt words and the configuration information, and to obtain the initial production scheduling strategy corresponding to the multiple target planned orders output by the large model.

[0013] The small model processing unit is used to analyze the order data of the multiple target planned orders based on the initial production scheduling strategy, the configuration information, the current remaining capacity of the current production line, and the optimization algorithm using the trained small model, to obtain the target production scheduling strategy corresponding to the multiple target planned orders output by the small model, so as to schedule the multiple target planned orders based on the target production scheduling strategy.

[0014] Thirdly, embodiments of this application provide an intelligent scheduling device, including:

[0015] Central processing unit, memory, input / output interfaces, wired or wireless network interfaces, and power supply;

[0016] The memory is either a short-term storage memory or a persistent storage memory;

[0017] The central processing unit is configured to communicate with the memory and execute instructions in the memory to perform the aforementioned intelligent scheduling method.

[0018] Fourthly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a computer, cause the computer to perform the aforementioned intelligent scheduling method.

[0019] Fifthly, embodiments of this application provide a computer program product containing instructions that, when run on a computer, cause the computer to execute the aforementioned intelligent scheduling method.

[0020] As can be seen from the above technical solutions, the embodiments of this application have the following advantages: By acquiring the configuration information input in the interface, including the production line's capacity configuration and scheduling constraints, and then using a large model combined with preset prompts to analyze order data, an initial scheduling strategy is generated. Next, a trained small model, based on this information and the current remaining capacity of the production line, refines the strategy through optimization algorithms, outputting the final scheduling strategy. This combination leverages the broad applicability of the large model and the optimization capabilities of the small model for specific tasks, allowing users to directly add production line capacity configurations and constraints by inputting natural language text whenever there are new requirements or changes in business scenarios, without the need for complex redevelopment or programming. Therefore, scheduling efficiency is high.

[0021] Accordingly, the intelligent scheduling device, electronic device, computer-readable storage medium, and computer program product containing instructions provided in this application also have the aforementioned technical effects. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the architecture of an intelligent scheduling system disclosed in an embodiment of this application;

[0023] Figure 2-1 This is a flowchart illustrating an intelligent scheduling method disclosed in an embodiment of this application;

[0024] Figure 2-2 This is a schematic diagram of a preset prompt word disclosed in an embodiment of this application;

[0025] Figure 2-3 This is a schematic diagram of the architecture of an intelligent scheduling system disclosed in an embodiment of this application;

[0026] Figure 2-4 This is a schematic diagram illustrating the interaction process between a user and an intelligent scheduling system as disclosed in an embodiment of this application;

[0027] Figure 3 This is a schematic diagram of the structure of an intelligent production scheduling device disclosed in an embodiment of this application;

[0028] Figure 4 This is a schematic diagram of the structure of an intelligent production scheduling device disclosed in an embodiment of this application. Detailed Implementation

[0029] This application provides an intelligent scheduling method, an intelligent scheduling device, an intelligent scheduling equipment, a computer-readable storage medium, and a computer program product containing instructions, for performing intelligent scheduling while improving scheduling efficiency.

[0030] Please see Figure 1 The architecture of the intelligent scheduling system in this application embodiment includes:

[0031] The system comprises an intelligent scheduling device 101 and a client 102. During intelligent scheduling, the intelligent scheduling device 101 connects to the client 102. The intelligent scheduling device 101 receives configuration information input into the interface by the client 102. It utilizes a large model to analyze order data from multiple target planned orders within the target batch order set based on preset prompts and configuration information. This results in an initial scheduling strategy output by the large model for each target planned order. A trained small model then analyzes the order data based on the initial scheduling strategy, configuration information, current remaining capacity of the production line, and optimization algorithms. This small model outputs a target scheduling strategy for each target planned order, enabling the scheduling of multiple target planned orders based on this strategy.

[0032] based on Figure 1 Please refer to the intelligent scheduling system shown. Figure 2-1 , Figure 2-1 This is a flowchart illustrating an intelligent scheduling method disclosed in an embodiment of this application. The method includes:

[0033] 201. Obtain the configuration information entered in the interface. The configuration information includes the production line capacity configuration and production scheduling constraints.

[0034] In one alternative implementation, scheduling constraints characterize the rules that must be followed during the scheduling process, such as priority, delivery date, and material limitations. The configuration information that users can input through the interface includes settings for production line capacity (e.g., production line 1 can produce 100 units of product per day) and constraints that must be followed during scheduling (e.g., prioritizing urgent orders or orders for specific materials). This configuration information guides the scheduling system in how to arrange production tasks. For example, a user sets the daily capacity of production line 1 to 200 units in the system and specifies the scheduling constraint as "prioritize urgent orders."

[0035] 202. Using the large model, based on preset prompts and configuration information, analyze the order data of multiple target planned orders in the target batch order set to obtain the initial production scheduling strategy corresponding to the multiple target planned orders output by the large model.

[0036] In one alternative implementation, a large model refers to a model that is typically larger and more capable, able to handle more complex tasks, such as understanding and analyzing natural language input. Preset prompts are predefined sets of instructions or questions used to guide the large model on how to understand and process specific tasks or data. A specific example of the preset prompts is as follows: "## Role: You are a manufacturing expert, proficient in production scheduling and algorithms. Based on user data and professional knowledge, provide accurate scheduling suggestions to ensure reasonable task allocation. ## Parameter Rules: 1. **Daily Capacity Calculation**: Daily capacity = Number of workers x Production line hours ÷ Single piece hours. 2. **Exception Information Priority**: When exception information exists for a certain date, it overrides the default capacity data for that date. 3. **Remaining Capacity Principle**: The daily capacity of the production line on that day minus the total scheduled output (all tasks locked on the same date and production line). The remaining capacity of the production line needs to be calculated in real time. For each successfully locked scheduling record, the remaining capacity value of the corresponding "date + production line" is immediately updated. ## Task: Assign orders to the production department / workshop according to the following rules: ### Scheduling Rules: 1. **Sorting Priority**: {{priorityData}}. 2. **Core Strategy**: {{strategyData}}" ##Note: 1. When scheduling production, ensure that the daily production line load does not exceed 100%. If it does, planned orders can be split and distributed to different production lines or dates, but ensure that the sum of the quantities of the split planned orders remains unchanged. 2. Pay attention to the specified materials when scheduling production. A planned order can only be assigned to the production department and production line that can produce the materials for that planned order. If a planned order has no production department or production line that can produce it, it does not need to be scheduled. 3. When scheduling production, start from {{minDate}} and schedule up to {{maxDate}l. If it cannot be scheduled completely, put it on the last day. ##Mandatory Constraints: 1. **Format Specifications**: The algorithm used in each step must be clearly listed. 2. **Load Limits**: After scheduling, the daily load of each production line must be ≤100%. 3. **Order Splitting Rules**: When producing across days or across production lines, split orders, keeping the order number unchanged. Use the original order number when the entire quantity can be produced in a single day. It can be seen that the preset prompts are pre-configured with sorting rules (including but not limited to priority). The role of prompts is to help the model quickly focus on the key aspects of the task, ensuring that the model's output aligns with the user's intent and needs. Order data represents specific information about the order, including order number, product type, required quantity, and delivery date. The large model uses pre-set prompts and user-input configuration information to analyze order data in the target batch and generate an initial production scheduling strategy. This process involves understanding and processing order data, as well as how to arrange production based on capacity and constraints.For example, the large model analyzes the order set based on the configuration information and prompts input by the user (such as "prioritize urgent orders"), identifies urgent orders, and generates a preliminary production scheduling strategy, such as "prioritize arranging urgent orders on production line 1".

[0037] 203. Using the trained small model, based on the initial production scheduling strategy, configuration information, current remaining capacity of the current production line, and optimization algorithm, analyze the order data of multiple target planned orders to obtain the target production scheduling strategy corresponding to the multiple target planned orders output by the small model, and schedule the production of multiple target planned orders based on the target production scheduling strategy.

[0038] In one alternative implementation, the small model represents a trained model specifically designed for a particular task (such as production scheduling optimization), capable of analysis and decision-making based on a given strategy and data. The optimization algorithm represents the mathematical methods or computational techniques used to improve the production scheduling strategy, such as genetic algorithms, annealing algorithms, etc. Based on the initial strategy generated by the large model, user configuration information, the remaining capacity of the current production line, and the optimization algorithm, the small model further analyzes order data to generate a more refined target production scheduling strategy. This strategy aims to optimize the production process, ensuring that all orders are processed effectively while meeting constraints. For example, the small model receives the initial strategy from the large model and the remaining capacity data of the production line, and uses optimization algorithms (such as annealing algorithms, genetic algorithms, etc.) to refine the strategy. For instance, it might decide to prioritize urgent orders for model A on production line 1, and then schedule orders for model B in the remaining time, thus generating the final production scheduling strategy. This approach improves the flexibility and accuracy of production scheduling, enabling the system to quickly adapt to new business needs or scenario changes without complex redevelopment or programming.

[0039] This approach involves acquiring user-input configuration information, including production line capacity configuration and scheduling constraints. A large model, combined with preset prompts, analyzes order data to generate an initial scheduling strategy. Next, a trained small model, based on this information and the remaining capacity of the current production line, refines the strategy using an optimization algorithm, outputting the final scheduling strategy. This combination leverages the broad applicability of the large model and the task-specific optimization capabilities of the small model. It allows users to directly add production line capacity configurations and constraints by inputting natural language text whenever new needs arise or business scenarios change, without complex redevelopment or programming, thus achieving high scheduling efficiency. Secondly, existing technologies use fixed rule engines, typically finding only local optima and failing to obtain global optima. In contrast, this application effectively solves the problem of connecting business scenarios with optimization algorithms by combining the versatility of the large model and the optimization capabilities of the small model. It allows for natural language input of constraint rules, making the system easier to use and more flexible in configuration, thereby quickly and accurately generating effective scheduling strategies.

[0040] In one optional implementation, before using a large model to analyze the order data of multiple target planned orders in the target batch order set according to preset prompts and configuration information, and obtaining the initial production scheduling strategy corresponding to the multiple target planned orders output by the large model, the method further includes: obtaining the order data of multiple target planned orders in the target batch order set, and the current remaining capacity of the current production line; processing the multiple initial planned orders in batches according to the configuration information, the order data of multiple target planned orders and / or the current remaining capacity of the current production line, to obtain multiple batch order sets, wherein there is no capacity dependency between the planned orders in different batch order sets, each batch order set includes multiple planned orders, and the target batch order set is any one of the multiple batch order sets.

[0041] Specifically, the current remaining capacity of a production line represents the available capacity remaining for each production line after the current scheduled production tasks. This data is crucial for the scheduling system because it determines how many additional production tasks each production line can handle. For example, suppose production line A has a daily capacity of 100 units, and 50 units have already been scheduled for production; then the current remaining capacity of production line A is 50 units. The absence of capacity dependencies between planned orders in different batch sets means that when processing orders in batches (planned orders), orders within each batch can be processed independently without affecting or depending on each other. This ensures that orders within each batch can be processed in parallel, improving production efficiency. For example, if production line B is suitable for processing orders requiring material X, and production line C is suitable for processing orders requiring material Y, then these two types of orders can be processed in different batches because they will not affect each other's capacity. Further examples: orders can be divided into several batches based on factors such as material requirements, delivery dates, and specific production line capabilities. For instance, all orders requiring material X can be assigned to one batch, while orders requiring material Y can be assigned to another batch.

[0042] It is worth noting that large language models (large models) may encounter performance degradation and accuracy reduction when processing large amounts of data (TOKEN). This application avoids processing too much data at once by processing the data in batches, thereby maintaining the model's performance and accuracy.

[0043] In one optional implementation, multiple initial planned orders are processed in batches based on configuration information, order data of multiple target planned orders, and / or the current remaining capacity of the current production line to obtain multiple batch order sets. There is no capacity dependency between the planned orders in different batch order sets, including at least one of the following situations: multiple initial planned orders are processed in batches based on whether the materials of multiple initial planned orders do not share the same production line; multiple initial planned orders are processed in batches based on whether the time intervals from the start date to the completion date of multiple initial planned orders do not overlap; multiple initial planned orders are processed in batches based on whether the materials of multiple initial planned orders are different, and whether the start date and completion date are different; multiple initial planned orders are processed in batches based on whether the constraints include a batching scheme for multiple initial planned orders.

[0044] Specifically, the batching scheme for multiple initial planned orders included in the constraints refers to the ability, based on specific business rules or constraints, to clearly define which orders need to be processed individually or which orders can be processed separately from other orders. Furthermore, if the constraints specify a specific time period, orders within that time period can be processed separately from orders outside those time periods. If multiple orders require materials that can be produced on different production lines, these orders can be processed in different batches because they do not compete for the capacity of the same production line. If the start and completion dates of the orders do not overlap, these orders can be processed in different batches because they do not conflict with each other in terms of timing. If orders require different materials or have different start and completion dates, these orders can be processed in different batches because they do not affect each other in terms of material requirements and timing. If business rules or constraints specify a batching scheme for orders, these orders can be processed in different batches according to these rules. For example: Suppose a factory has three production lines (X, Y, M, N) and needs to process the following orders: Order A: Requires material X, which can be produced on production line X or Y. Order B requires material Y, which can be produced on production line M or N. Based on the above strategy, it can be analyzed that Orders A and B can be processed in batches because they can be produced on different production lines. If the time intervals of Orders A and B do not overlap, they can also be processed in batches. If there are multiple orders requiring material X with the same time interval, they can be processed together. If the time intervals of Orders A and B are different, they can be processed in batches.

[0045] This allows the system to independently schedule orders within each batch, reducing interdependencies and conflicts between orders and improving production flexibility and efficiency. Furthermore, this approach optimizes resource allocation, ensuring full utilization of production line capacity while avoiding overload or idleness.

[0046] In one optional implementation, before using the trained small model to analyze the order data of multiple target planned orders based on the initial production scheduling strategy, configuration information, the current remaining capacity of the current production line, and the optimization algorithm to obtain the target production scheduling strategies corresponding to the multiple target planned orders output by the small model, the method further includes: obtaining a target batch order set sample, which includes order data of multiple target planned order samples, initial production scheduling strategy samples, configuration information samples, current remaining capacity samples of the current production line, and optimization algorithm samples. The configuration information samples include production line capacity configuration samples and production scheduling constraint condition samples. The target batch order set sample is labeled with the target production scheduling strategy. The target batch order set sample is input into the small model. The small model analyzes the order data of multiple target planned order samples based on the initial production scheduling strategy samples, configuration information samples, current remaining capacity samples of the current production line, and optimization algorithm samples to obtain the predicted target production scheduling strategies corresponding to the multiple target planned order samples output by the small model. When the loss between the predicted target production scheduling strategy and the labeled target production scheduling strategy reaches a preset convergence condition, the trained small model is obtained.

[0047] Specifically, the small model is designed for production scheduling task optimization. It performs further analysis based on the results of the large model and executes specific algorithm steps. This small model is generated by fine-tuning (i.e., partial training) the open-source large model, adjusting only the parameters of the top layer. The specific operation involves first freezing the pre-trained parameters of the open-source model, then adding a new layer to the output layer. The number of parameters in this new layer is based on the number of parameters in the previous layer. These parameters can be initialized with random values, and their final values ​​are determined during the training process. Training is performed using a historically accumulated set of target batch orders. The goal of the training is to enable the model to identify which optimization algorithm is more effective under different constraints.

[0048] This approach improves the accuracy and adaptability of scheduling strategies by using smaller models specifically trained for scheduling tasks. By fine-tuning the large open-source model on real-world data, the smaller models can more accurately understand and execute complex scheduling logic, thus selecting the most suitable optimization algorithm under different production constraints. Furthermore, this method is cost-effective, as it avoids the high cost of training a completely new model from scratch while ensuring the model's efficiency and flexibility.

[0049] In an optional implementation, the method further includes: determining the remaining capacity of each production line after scheduling multiple target planned orders based on the target scheduling strategy, and using the remaining capacity of each production line as the current remaining capacity of the current production line corresponding to the next batch of orders.

[0050] Specifically, after executing a batch of orders according to the target production scheduling strategy, the system needs to calculate the remaining capacity of each production line. Remaining capacity refers to how many additional production tasks a production line can handle after completing the current batch of orders. This can be achieved through a real-time production line load calculation engine, which immediately calculates the current capacity load of each production line after each batch of production scheduling results are generated, thus determining the remaining capacity of each production line. The calculated remaining capacity data will be used as input data for the next batch of order scheduling. For example: Assume a manufacturing company has three production lines (Line 1, Line 2, and Line 3), each with a daily capacity of 100 units. First Batch Order Scheduling: The system schedules the first batch of orders according to the target production scheduling strategy. For example, Line 1 processes 40 units of product A, Line 2 processes 30 units of product B, and Line 3 processes 30 units of product C. Calculating Remaining Capacity: After the first batch of orders is completed, the system calculates the remaining capacity of each production line. For example, Line 1 has 60 units of remaining capacity, Line 2 has 70 units of remaining capacity, and Line 3 has 70 units of remaining capacity. Next batch scheduling: The system uses this remaining capacity data to schedule the next batch of orders. For example, production line 1 can schedule 20 units of product D, production line 2 can schedule 10 units of product E, and production line 3 can schedule 20 units of product F.

[0051] This ensures that the production line's capacity is fully utilized while avoiding overload or idleness, thereby improving production efficiency and resource utilization. It also makes production scheduling more flexible and adaptable, enabling a rapid response to changes in production demand.

[0052] In one optional implementation, a trained small model is used to analyze order data of multiple target planned orders based on an initial production scheduling strategy, configuration information, the current remaining capacity of the current production line, and an optimization algorithm. This results in target production scheduling strategies for the multiple target planned orders output by the small model. The analysis includes: using the trained small model to analyze order data of multiple target planned orders based on the initial production scheduling strategy, configuration information, a comparison of the current remaining capacity of the current production line with a preset minimum threshold of remaining capacity, and an optimization algorithm. This results in order splitting results of the multiple target planned orders output by the small model and production scheduling plans corresponding to each split planned order. The target production scheduling strategy corresponding to the target batch planned order set includes the order splitting results of the multiple target planned orders and the production scheduling plans of the multiple split planned orders.

[0053] Specifically, if the current remaining capacity of the production line is greater than the preset minimum remaining capacity threshold, then the order is split accordingly; otherwise, it is not split. This way, splitting orders when the remaining capacity exceeds the preset minimum threshold ensures that the production line will not be interrupted due to insufficient capacity when processing subsequent orders, thereby improving overall production efficiency. By splitting orders when there is sufficient remaining capacity, production tasks can be allocated more rationally, resource waste can be avoided, and the production line's capacity can be fully utilized.

[0054] In one optional implementation, the target scheduling strategy includes order splitting results of multiple target planned orders and scheduling plans of multiple split planned orders. Scheduling multiple target planned orders based on the target scheduling strategy includes: splitting multiple target planned orders based on the order splitting results to obtain multiple split planned orders, and scheduling multiple split planned orders based on the scheduling plans of the multiple split planned orders.

[0055] Specifically, an order splitting engine can be used to split multiple target planned orders based on the order splitting results, resulting in multiple split planned orders. These split planned orders can be saved in the system. For example, suppose a manufacturing company needs to process a large order. The system first generates a target production scheduling strategy, such as "prioritizing urgent orders, then processing regular orders, and splitting the large order into two smaller orders, one produced on production line 1 and the other on production line 2." Based on this strategy, the order splitting engine might split an order for 1000 units into two smaller orders of 500 units each, with the first 500 units produced on production line 1 and the second 500 units on production line 2. The split orders and their corresponding production plans are saved in the system to guide actual production and subsequent order tracking.

[0056] This improves the flexibility and efficiency of the production process. By breaking large orders down into smaller ones, production line resources can be allocated more rationally, reducing waiting and changeover times, thereby optimizing the overall production plan. Furthermore, developing specific production schedules for each broken-down order helps ensure that each production task is completed under optimal conditions and at the best time, further enhancing production efficiency.

[0057] In one optional implementation, after obtaining the target scheduling strategies corresponding to multiple target planned orders in the target batch order set output by the small model, the method further includes: displaying the target scheduling strategies, updating the target scheduling strategies based on the user's modification information for the target scheduling strategies, obtaining the updated target scheduling strategies, and scheduling the multiple target planned orders based on the updated target scheduling strategies.

[0058] Specifically, after generating the target production scheduling strategy, the system will display the strategy for users to view. Users can modify the target production scheduling strategy as needed, and the system will then update the strategy and reschedule order production based on the updated strategy.

[0059] This allows users to adjust their production scheduling strategies based on actual conditions, thereby more accurately meeting production needs and enhancing the flexibility and adaptability of the scheduling system. Furthermore, updating strategies based on user feedback ensures more accurate and efficient scheduling, reducing uncertainty and waste in production. Ultimately, this interactivity improves user satisfaction because users can directly participate in the scheduling decision-making process, making production plans more aligned with user expectations and requirements, thus enhancing the user experience.

[0060] To facilitate understanding of the embodiments of this application, the intelligent scheduling system of this application will be described in detail below. Please refer to [link / reference needed]. Figure 2-2 , Figure 2-2 This is a schematic diagram of the architecture of an intelligent scheduling system disclosed in an embodiment of this application. Figure 2-2 As can be seen, the intelligent scheduling system comprises an intelligent scheduling system view layer, an intelligent scheduling system service layer, a large model configuration service, a large model proxy, dedicated scheduling small models, an optimization algorithm toolset, and a data storage layer. The intelligent scheduling system view layer provides the user interface, allowing users to set scheduling conditions, capacity configurations, and constraints, execute scheduling, and edit and save the scheduling results. The intelligent scheduling system service layer handles backend business logic, including order and capacity data acquisition, batch processing, and real-time calculation of production line load. The large model configuration service configures and manages the large model, including model selection, prompt word setting, and task flow configuration. The large model proxy serves as the interface between the system and external large models, handling API calls, logging, and security functions. Dedicated scheduling small models are used for fine-tuning based on the large model, specifically optimizing scheduling strategies to improve scheduling accuracy and efficiency. The optimization algorithm toolset provides various optimization algorithms for generating and optimizing scheduling strategies. The data storage layer stores all relevant data, such as orders, capacity, and task flow configurations. The specific workflow is as follows: Users set the production line capacity configuration and scheduling constraints in the view layer. The service layer obtains order and capacity data based on these configuration information, performs batch processing and load calculation, analyzes the data in the large model, generates a preliminary scheduling strategy, further optimizes the strategy in the small model, and generates the final scheduling plan. Users can edit and save the scheduling plan, and the system executes production according to the plan.

[0061] For more details, please refer to Figure 2-3 , Figure 2-3 This is a schematic diagram illustrating the interaction process between a user and an intelligent scheduling system as disclosed in an embodiment of this application. Figure 2-3The entire process of configuring the system, executing production scheduling, reviewing results, making adjustments, and finally confirming and saving the production schedule is as follows: First, the user needs to set the production line capacity, scheduling constraints, select the applicable large model, input prompts, and define the task flow in the system. The system provides some preset options, but users can also customize settings as needed. After configuration, the user can select specific constraints and task flows on the intelligent scheduling page to start the scheduling process. After the scheduling calculation is completed, the system will write the preliminary scheduling results to the scheduling process table. The user can view the data in the scheduling process table and make necessary modifications based on the actual situation. The user confirms the modified results and then saves these changes; this data will be used to guide actual production activities.

[0062] For more details, please refer to Figure 2-4 , Figure 2-4 This is a schematic diagram of the internal operation flow of an intelligent scheduling system disclosed in an embodiment of this application. This flow is driven by a pre-configured task flow. Figure 2-4 The complete process, from user input of configuration information to system generation and execution of production scheduling strategies, and then to user feedback and data saving, is as follows: The user inputs configuration information through the interface, including capacity configuration and scheduling constraints. The system initiates the task flow based on this configuration information, starting the production scheduling process. The large-scale model analyzes order data and configuration information to generate an initial production scheduling strategy. The smaller model further refines and optimizes the production scheduling strategy based on the initial strategy, current capacity, and optimization algorithms. The system executes specific production scheduling tasks according to the optimized strategy. The system provides feedback on the scheduling results to the user, who can edit and confirm them. After user confirmation, the system saves the final production scheduling data for actual production.

[0063] Understandably, this application also provides various configuration functions, including constraint configuration, large model configuration, prompt word configuration, task flow configuration, and re-editing of scheduling process data. These functions significantly improve the system's flexibility, facilitate user customization and expansion, and enhance the system's versatility. Secondly, in cases where the task is relatively simple, the combination of large and small models proposed in this application can be simplified to a single large model. This large model can simultaneously handle natural language understanding and output specific algorithm steps, thereby simplifying system deployment. Furthermore, the task flow node design in this application is flexible and can be replaced according to different business scenarios. Due to the existence of the batch algorithm engine, task flow nodes can be chained and superimposed when processing large amounts of data. For cases where the scheduling scheme is very fixed, the scheduling strategy analysis node can be omitted, and a fixed scheduling algorithm can be executed directly. Finally, after replacing some components, the system of this application can not only execute scheduling tasks but also be applied to other tasks seeking optimal solutions under multiple constraints, such as class scheduling tasks. This can be achieved simply by replacing the data retrieval engine, the scheduling process editing module, and the order splitting engine.

[0064] For further details, please refer to Figure 3 One embodiment of the intelligent scheduling device in this application includes:

[0065] The acquisition unit is used to acquire the configuration information input in the interface, the configuration information including the production line capacity configuration and production scheduling constraints;

[0066] The large model processing unit is used to analyze the order data of multiple target planned orders in the target batch order set using the large model according to the preset prompt words and the configuration information, and to obtain the initial production scheduling strategy corresponding to the multiple target planned orders output by the large model.

[0067] The small model processing unit is used to analyze the order data of the multiple target planned orders based on the initial production scheduling strategy, the configuration information, the current remaining capacity of the current production line, and the optimization algorithm using the trained small model, to obtain the target production scheduling strategy corresponding to the multiple target planned orders output by the small model, so as to schedule the multiple target planned orders based on the target production scheduling strategy.

[0068] In one alternative implementation, the acquisition unit may be used for:

[0069] Obtain order data for multiple target planned orders in the target batch order set, as well as the current remaining capacity of the current production line. Based on the configuration information, the order data of the multiple target planned orders, and / or the current remaining capacity of the current production line, process the multiple initial planned orders in batches to obtain multiple batch order sets. There is no capacity dependency between the planned orders in different batch order sets. Each batch order set includes multiple planned orders, and the target batch order set is any one of the multiple batch order sets.

[0070] In one alternative implementation, the acquisition unit may be used for:

[0071] Based on whether the materials of the multiple initial planned orders do not share the same production line, the multiple initial planned orders are divided into batches to obtain multiple batch order sets; based on whether the time intervals from the start date to the completion date of the multiple initial planned orders do not overlap, the multiple initial planned orders are divided into batches to obtain multiple batch order sets; based on whether the materials of the multiple initial planned orders are different, and whether the start date and completion date are different, the multiple initial planned orders are divided into batches to obtain multiple batch order sets; based on whether the constraints include a batching scheme for the multiple initial planned orders, the multiple initial planned orders are divided into batches to obtain multiple batch order sets.

[0072] In one alternative implementation, the acquisition unit may be used for:

[0073] Obtain a target batch order set sample, which includes order data of multiple target planned order samples, initial production scheduling strategy samples, configuration information samples, current remaining capacity samples of the current production line, and optimization algorithm samples. The configuration information samples include production line capacity configuration samples and production scheduling constraint samples. The target batch order set sample is labeled with a target production scheduling strategy. Input the target batch order set sample into a small model. The small model analyzes the order data of the multiple target planned order samples based on the initial production scheduling strategy samples, the configuration information samples, the current remaining capacity samples of the current production line, and the optimization algorithm samples to obtain the predicted target production scheduling strategy corresponding to the multiple target planned order samples output by the small model. When the loss between the predicted target production scheduling strategy and the labeled target production scheduling strategy reaches a preset convergence condition, the trained small model is obtained.

[0074] In one alternative implementation, the acquisition unit may be used for:

[0075] After the production scheduling of the multiple target planned orders based on the target scheduling strategy is completed, the remaining capacity of each production line is determined, and the remaining capacity of each production line is used as the current remaining capacity of the current production line corresponding to the next batch of orders.

[0076] In one alternative implementation, the small model processing unit can be used for:

[0077] Using the trained small model, based on the initial production scheduling strategy, the configuration information, the comparison results of the current remaining capacity of the current production line with the preset minimum remaining capacity threshold, and the optimization algorithm, the order data of the multiple target planned orders are analyzed to obtain the order splitting results of the multiple target planned orders output by the small model and the production scheduling plan corresponding to each split planned order. The target production scheduling strategy corresponding to the target batch planned order set includes the order splitting results of the multiple target planned orders and the production scheduling plans of the multiple split planned orders.

[0078] In one alternative implementation, the small model processing unit can be used for:

[0079] Based on the order splitting results, the multiple target planned orders are split into multiple split planned orders. Based on the production schedule of the multiple split planned orders, the production schedule of the multiple split planned orders is scheduled. The target production scheduling strategy includes the order splitting results of the multiple target planned orders and the production schedule of the multiple split planned orders.

[0080] In one optional implementation, the intelligent scheduling device further includes a display unit and an update unit, wherein the display unit can be used for:

[0081] This demonstrates the target production scheduling strategy.

[0082] The update unit can be used for:

[0083] In response to the modification information of the target production scheduling strategy, the target production scheduling strategy is updated to obtain the updated target production scheduling strategy, so as to schedule the multiple target planned orders based on the updated target production scheduling strategy.

[0084] For further details, please refer to Figure 4 One embodiment of the electronic device in this application includes:

[0085] Central processing unit 401, memory 405, input / output interface 404, wired or wireless network interface 403, and power supply 402;

[0086] Memory 405 is either a short-term storage memory or a persistent storage memory;

[0087] The central processing unit 401 is configured to communicate with the memory 405 and execute instructions stored in the memory 405 to perform the aforementioned operations. Figure 2-1 The method in the illustrated embodiment.

[0088] Furthermore, embodiments of this application also provide a computer-readable storage medium, which includes instructions that, when executed on a computer, cause the computer to perform the aforementioned... Figure 2-1 The method in the illustrated embodiment.

[0089] Furthermore, embodiments of this application also provide a computer program product containing instructions, which, when run on a computer, causes the computer to perform the aforementioned... Figure 2-1 The method in the illustrated embodiment.

[0090] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0091] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0092] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0093] The units described as separate components may or may not be physically separate. 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 units can be selected to achieve the purpose of this embodiment according to actual needs.

[0094] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0095] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. An intelligent scheduling method, characterized by, The method comprises the following steps: obtaining configuration information input in an interface, the configuration information comprising production capacity configuration of a production line and constraint conditions of production scheduling; using a large model to analyze order data of a plurality of target planned orders in a target batch order set according to preset prompt words and the configuration information, to obtain an initial production scheduling strategy corresponding to the plurality of target planned orders output by the large model; using a small model trained to analyze the order data of the plurality of target planned orders based on the initial production scheduling strategy, the configuration information, current remaining production capacity of a current production line and an optimization algorithm, to obtain a target production scheduling strategy corresponding to the plurality of target planned orders output by the small model, so as to schedule the plurality of target planned orders based on the target production scheduling strategy.

2. The method of claim 1, wherein, Before the step of using a large model to analyze order data of a plurality of target planned orders in a target batch order set according to preset prompt words and the configuration information, to obtain an initial production scheduling strategy corresponding to the plurality of target planned orders output by the large model, the method further comprises: obtaining order data of the plurality of target planned orders in the target batch order set, and current remaining production capacity of a current production line; batch processing a plurality of initial planned orders according to the configuration information, the order data of the plurality of target planned orders and / or the current remaining production capacity of the current production line, to obtain a plurality of batch order sets, and there is no production capacity dependency relationship between planned orders of different batch order sets, each batch order set comprises a plurality of planned orders, and the target batch order set is any one of the plurality of batch order sets.

3. The method of claim 2, wherein, The step of batch processing a plurality of initial planned orders according to the configuration information, the order data of the plurality of target planned orders and / or the current remaining production capacity of the current production line, to obtain a plurality of batch order sets, and there is no production capacity dependency relationship between planned orders of different batch order sets, comprises at least one of the following cases: batching the plurality of initial planned orders based on whether materials of the plurality of initial planned orders share the same production line, to obtain the plurality of batch order sets; batching the plurality of initial planned orders based on whether time intervals from start dates to finish dates of the plurality of initial planned orders overlap, to obtain the plurality of batch order sets; batching the plurality of initial planned orders based on whether materials of the plurality of initial planned orders are different and whether start dates and finish dates are different, to obtain the plurality of batch order sets; batching the plurality of initial planned orders based on whether a batching scheme of the plurality of initial planned orders is included in the constraint conditions, to obtain the plurality of batch order sets.

4. The method of claim 1, wherein, Before the step of using a small model trained to analyze the order data of the plurality of target planned orders based on the initial production scheduling strategy, the configuration information, current remaining production capacity of a current production line and an optimization algorithm, to obtain a target production scheduling strategy corresponding to the plurality of target planned orders output by the small model, the method further comprises: obtaining a target batch order set sample, the target batch order set sample including order data of a plurality of target planned order samples, an initial scheduling strategy sample, a configuration information sample, a current remaining capacity sample of a current production line, and an optimization algorithm sample, the configuration information sample including a capacity configuration sample of the production line and a constraint condition sample of scheduling, the target batch order set sample being labeled with a target scheduling strategy; inputting the target batch order set sample into a small model, and analyzing, by the small model, the order data of the plurality of target planned order samples based on the initial scheduling strategy sample, the configuration information sample, the current remaining capacity sample of the current production line, and the optimization algorithm sample, to obtain a predicted target scheduling strategy corresponding to the plurality of target planned order samples output by the small model; when a loss between the predicted target scheduling strategy and the labeled target scheduling strategy reaches a preset convergence condition, obtaining a trained small model.

5. The method of claim 1, wherein, The method further includes: determining remaining capacities of each production line after scheduling the plurality of target planned orders based on the target scheduling strategy, and taking the remaining capacities of the each production line as current remaining capacities of current production lines corresponding to a next batch order set.

6. The method of claim 1, wherein, The analyzing, by the trained small model, the order data of the plurality of target planned orders based on the initial scheduling strategy, the configuration information, the current remaining capacity of the current production line, and the optimization algorithm to obtain the target scheduling strategy corresponding to the plurality of target planned orders output by the small model includes: The analyzing, by the trained small model, the order data of the plurality of target planned orders based on the initial scheduling strategy, the configuration information, a comparison result of the current remaining capacity of the current production line and a preset minimum threshold of the remaining capacity, and the optimization algorithm to obtain an order splitting result of the plurality of target planned orders and a scheduling plan corresponding to each split planned order output by the small model, the target scheduling strategy corresponding to the target batch planned order set including the order splitting result of the plurality of target planned orders and the scheduling plan of the plurality of split planned orders.

7. The method of claim 1, wherein, The target scheduling strategy includes the order splitting result of the plurality of target planned orders and the scheduling plan of the plurality of split planned orders, and the scheduling the plurality of target planned orders based on the target scheduling strategy includes: splitting the plurality of target planned orders based on the order splitting result to obtain a plurality of split planned orders; scheduling the plurality of split planned orders based on the scheduling plan of the plurality of split planned orders.

8. The method of claim 1, wherein, After obtaining the target scheduling strategy corresponding to the plurality of target planned orders in the target batch order set output by the small model, the method further includes: displaying the target scheduling strategy; updating the target scheduling strategy in response to modification information of the target scheduling strategy to obtain an updated target scheduling strategy, and scheduling the plurality of target planned orders based on the updated target scheduling strategy.

9. An intelligent scheduling device, characterized by, including: An acquisition unit is configured to acquire configuration information input in an interface, the configuration information including capacity configuration of a production line and constraint conditions of production scheduling; The large model processing unit is configured to analyze order data of a plurality of target plan orders in a target batch order set according to a preset prompt word and the configuration information by using a large model, and obtain an initial production scheduling strategy corresponding to the plurality of target plan orders output by the large model; The small model processing unit is configured to analyze the order data of the plurality of target plan orders based on the initial production scheduling strategy, the configuration information, current remaining capacity of a current production line and an optimization algorithm by using a trained small model, and obtain a target production scheduling strategy corresponding to the plurality of target plan orders output by the small model, so as to schedule the plurality of target plan orders based on the target production scheduling strategy.

10. An intelligent scheduling device, characterized by Comprise: A central processing unit and a memory; The memory is a transitory storage memory or a persistent storage memory; The central processing unit is configured to communicate with the memory and execute instruction operations in the memory to perform the method of any one of claims 1 to 8.

11. A computer readable storage medium, characterized in that, The computer readable storage medium comprises instructions which, when executed on a computer, cause the computer to perform the method of any one of claims 1 to 8.

12. A computer program product comprising instructions, characterized in that, When the computer program product is executed on a computer, the computer is caused to perform the method of any one of claims 1 to 8.

Images