Intelligent supply chain scheduling optimization method and system based on autonomous controllable large model

Abstract

The invention discloses an intelligent supply chain scheduling optimization method and system based on an autonomous controllable large model, and the method comprises the steps: employing a demand prediction agent to predict a future market demand condition according to product sales data; according to supplier supply product data, adopting a supplier management agent to select an optimal supplier as a target supplier; according to the predicted future market demand condition and inventory data, an inventory optimization agent is adopted to generate an optimal inventory replenishment plan; according to the inventory level after the optimal inventory replenishment plan and the logistics transportation data, adopting a logistics scheduling agent to calculate and obtain an optimal transportation path; feedback data of all the agents are collected, performance indexes of all the agents are calculated through the information feedback agents according to the feedback data, and agent parameters are adjusted according to the performance indexes of all the agents; and the strategy output of each agent is integrated into the scheduling strategy of the supply chain. The scheduling strategy can be dynamically scheduled and updated in real time.

Description

Technical Field

[0001] The present invention relates to supply chain scheduling technology, and in particular to an intelligent supply chain scheduling optimization method and system based on an autonomous and controllable large model. Background Art

[0002] With the rapid development of artificial intelligence technology, intelligent supply chain optimization and scheduling systems have received widespread attention and application in the field of supply chain management. Supply chain optimization is a complex process involving multiple links from production, inventory, logistics to final delivery. Traditional supply chain optimization strategies usually rely on fixed models and historical experience, which are difficult to cope with the current rapidly changing market environment and increasingly complex supply chain networks. At the same time, the deep learning and large model technologies that have emerged in recent years have shown great potential in fields such as data analysis and time series prediction, and are expected to further promote the development of supply chain optimization and scheduling.

[0003] Traditional supply chain optimization strategies are usually based on historical data and known logistics and inventory management principles, such as economic order quantity (EOQ), safety stock strategy, transportation optimization, etc. Although these strategies have shown a certain degree of robustness in long-term supply chain management, they are prone to failure in the face of market demand fluctuations, supply chain disruptions, and emergencies. In addition, traditional supply chain models mostly rely on expert experience and manual parameter adjustment, and it is difficult to adapt to changes in the supply chain network in real time. Especially in the context of rapid expansion of data dimensions (such as global market dynamics, market sentiment in social media, etc.), the limitations of traditional methods are becoming more and more obvious. Summary of the invention

[0004] In order to overcome the drawbacks of traditional supply chain optimization strategies, the purpose of the present invention is to provide an intelligent supply chain scheduling optimization method and system based on an autonomous and controllable large model that can be dynamically scheduled and updated in real time.

[0005] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions: An intelligent supply chain scheduling optimization method based on an autonomous and controllable large model includes the following steps: (1) Obtaining supply chain-related data at the current moment, including product sales data, supplier product supply data, inventory data, and logistics and transportation data; (2) Based on the product sales data, a demand forecasting agent is used to predict future market demand; wherein the demand forecasting agent is trained using historical product sales data as training samples and actual sales conditions at the corresponding time as labels; (3) Based on the product data supplied by the suppliers, the supplier management agent is used to select the best supplier as the target supplier; wherein the supplier management agent selects the supplier with the smallest sum of the positive weight of the product cost, the positive weight of the product delivery time, the negative weight of the product quality, and the positive weight of the supplier's supply reputation value as the best supplier; (4) Based on the predicted future market demand and inventory data, an inventory optimization agent is used to generate the optimal inventory replenishment plan; (5) using a logistics scheduling agent to calculate an optimal transportation path based on the inventory level after the optimal inventory replenishment plan and the logistics transportation data; wherein the logistics scheduling agent calculates the optimal transportation path with the lowest transportation cost as the objective function; (6) Collect feedback data from the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent, calculate the performance indicators of each agent using the information feedback agent based on the feedback data, and adjust the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent based on the performance indicators of each agent until the performance indicators meet the preset requirements; (7) The outputs of the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent whose performance indicators meet the preset requirements are combined as the scheduling strategy output of the supply chain.

[0006] An intelligent supply chain scheduling optimization system based on an autonomous and controllable large model, including: Data perception module, used to obtain supply chain-related data at the current moment, including product sales data, supplier supply product data, inventory data, and logistics transportation data; A demand forecasting agent is used to forecast future market demand based on product sales data; wherein the demand forecasting agent is trained using historical product sales data as training samples and actual sales conditions at corresponding moments as labels; A supplier management agent, used to select the best supplier as the target supplier based on the supplier supply product data; wherein the supplier management agent selects the supplier with the smallest sum of the positive weight of product cost, the positive weight of product delivery time, the negative weight of product quality and the positive weight of supplier supply reputation value as the best supplier; Inventory optimization agent, which generates optimal inventory replenishment plans based on predicted future market demand and inventory data; A logistics scheduling agent is used to calculate the optimal transportation path based on the inventory level after the optimal inventory replenishment plan and the logistics transportation data; wherein the logistics scheduling agent calculates the optimal transportation path with the lowest transportation cost as the objective function; The information feedback agent is used to collect feedback data from the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent, calculate the performance indicators of each agent based on the feedback data, and adjust the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent according to the performance indicators of each agent until the performance indicators meet the preset requirements; The scheduling strategy output module is used to merge the outputs of the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent whose performance indicators meet the preset requirements as the scheduling strategy output of the supply chain.

[0007] Compared with the prior art, the present invention has the following beneficial effects: 1. Intelligent data screening and optimization: Through an autonomous and controllable large model including several intelligent agents, the present invention can intelligently and automatically screen out the best supply chain optimization strategy from a large amount of market data, and generate an optimized scheduling plan by combining deep learning technology. Compared with the traditional supply chain management model, this system does not require human intervention, greatly improves the accuracy and efficiency of decision-making, and ensures that the strategy can adapt to the dynamic changes of the market in a timely manner; 2. Multi-data fusion and market perception: It integrates structured data (such as historical inventory data, logistics data) and unstructured data (customer feedback, etc.), and effectively captures supply chain change trends through real-time perception of market sentiment. This innovation enables the system to make more comprehensive and accurate scheduling decisions in decision-making, improving the foresight of decision-making; 3. Multi-agent collaborative working mechanism: The system uses multiple agents to work together, including demand forecasting, supplier management, inventory optimization, logistics scheduling and information feedback. Each agent has a clear division of labor and automatically handles different tasks, thus achieving efficient generation and execution of supply chain strategies, shortening the time for decision-making and implementation, and improving overall efficiency; 4. Feedback mechanism and strategy optimization: By obtaining data from information feedback agents, the actual implementation effect of the strategy can be evaluated and optimized. The feedback mechanism not only improves the adaptability of the strategy in different market environments, but also achieves self-improvement of the strategy through continuous learning and optimization, significantly enhancing the stability and responsiveness of the overall supply chain. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 It is a flow chart of the intelligent supply chain scheduling optimization method based on the autonomous and controllable large model provided by the present invention. DETAILED DESCRIPTION

[0009] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0010] The embodiment of the present invention provides an intelligent supply chain scheduling optimization method based on an autonomous and controllable large model, such as Figure 1 As shown, the following steps are included: (1) Obtain supply chain-related data at the current moment, including product sales data, supplier product supply data, inventory data, and logistics and transportation data.

[0011] Among them, product sales data includes data such as the sales volume of each product at each moment in the past; supplier supply product data includes data such as supplier delivery time, supplier quotation, supplier product quality score, supplier delivery delay rate, supplier default rate, etc.; inventory data includes inventory levels at each moment, including the inventory volume of each product, inventory turnover rate, historical sales volatility, etc.; logistics and transportation data includes vehicle capacity, transportation time, transportation cost, logistics route, transportation tools, etc., which can reflect the logistics status of the entire supply chain and provide data support for scheduling optimization.

[0012] Supply chain related data is acquired through the perception module. Due to the diversity of data sources, the perception module needs to process data of different formats and frequencies. Therefore, after data collection, a series of preprocessing operations are required to ensure the quality and consistency of the data. After the preprocessing is completed, the perception module needs to fuse different data sources to provide comprehensive and accurate information for supply chain optimization and scheduling strategies.

[0013] (2) Based on product sales data, use demand forecasting agents to predict future market demand.

[0014] Specifically, firstly, the product sales feature X is obtained according to the product sales data: , Represents product sales data before the current moment, Seasonal sales characteristics refer to the phenomenon that the sales volume of goods changes significantly with the seasons. For example, in clothing sales, spring and autumn are the seasons when clothing changes, and consumers will buy new season clothing; while summer and winter are the peak seasons for clothing retail, because consumers need to buy seasonal clothing. Indicates holiday sales characteristics, for example, consumers will buy moon cakes, fruits, gifts, etc. during the Mid-Autumn Festival; After that, the product sales characteristics Input the trained demand forecasting agent to obtain future market demand , wherein the demand forecasting agent is a combination of the LSTM network and the Transformer network: , in represents the LSTM network model parameters, Represents the Transformer network model parameters, Represents the model parameters of the demand forecasting agent.

[0015] The demand forecasting agent is trained using historical product sales data as training samples and actual sales conditions at the corresponding time as labels. The specific training method is as follows: A. Calculate the forecast error based on actual sales data and predicted future market demand: , In the formula, , Respectively , The prediction error at time is the time weighting coefficient, express Actual sales volume at the moment, Indicates the predicted Market demand at all times; B. Adjust the learning rate of the demand prediction agent model parameters according to the prediction error: , in They are , The learning rate at each moment is the basis for adjusting the learning rate, and is fine-tuned at each time step based on factors such as the current error. , , is a hyperparameter, is the time step coefficient for learning rate decay. A hyperparameter that controls the impact of the current error on the learning rate adjustment. It is usually set to a small positive number to avoid excessive fluctuations in the learning rate. A hyperparameter that controls the effect of the error change trend on the learning rate adjustment. It is usually a small positive number. Usually a small positive number is used to gradually reduce the learning rate over a longer period of time. This item is used to avoid excessive learning rate in the late stage of training, which may lead to unstable model convergence. To control the penalty The effect on the learning rate. This hyperparameter determines the degree to which the squared error value suppresses the learning rate. A higher value will put more emphasis on suppressing the growth of the learning rate, especially when the error is large. is the weight coefficient of the momentum term, which is used to smooth the change of learning rate and avoid excessive fluctuation of learning rate. represents the change in prediction error, is the momentum term at time t, and , is the weight coefficient, is the momentum term at time t-1, express Demand forecasting agent model parameters at the moment The loss function is represents the gradient; C. Update the demand forecasting agent model parameters according to the adjusted learning rate: , in, express The demand forecasting agent model parameters at the moment, is a hyperparameter, is The inverse of the BFGS (quasi-Newton method) approximation matrix at time.

[0016] (3) Based on the product data supplied by suppliers, the supplier management agent is used to select the best supplier as the target supplier.

[0017] The supplier management agent selects the supplier with the smallest sum of the positive weight of product cost, the positive weight of product delivery time, the negative weight of product quality, and the positive weight of supplier supply reputation as the optimal supplier. The specific objective function is: , In the formula, represents the best supplier, Indicates supplier The quoted cost, Indicates supplier The average delivery time, Indicates supplier The product quality score, represents the supply reputation value of supplier i, , , are the highest acceptable cost, longest delivery time and highest product quality score, respectively. , , , is the weight coefficient and satisfies the following formula: , Respectively , time The value of It is the adjustment range coefficient, which reflects the degree of change in the external market environment or internal demand. It is the adjustment increment, which can be calculated based on the supply chain strategy or actual situation (such as delivery time pressure when the project schedule is tight). is a nonlinear adjustment term, and They are the adjustment factor and feedback variable, which control the smoothness and elasticity of the weight adjustment, which helps to avoid excessive or insufficient weight adjustment due to short-term fluctuations. Reflects the adjustment sensitivity. The larger the value, the smoother the adjustment. Dynamic impact weights can be adjusted based on user feedback, market dynamics, or historical performance. represents the feedback coefficient, is with Relevant real-time feedback, which can be market dynamics (such as price fluctuations) or project progress (such as delivery progress), ensures that the weight adjustment process is not too drastic, allowing the system to respond to changes gradually; , in, is the supplier’s delivery delay rate, is the supplier’s default rate, , is the adjustment weight, , It is the penalty coefficient, which reflects the cumulative effect of historical behavior. For example, the more delays there are, the greater the impact will be in subsequent risk assessments.

[0018] In practical applications, suppose a company cooperates with multiple suppliers to obtain key raw materials. If the supplier management agent finds that a supplier has a high delivery delay rate, it will choose to increase the purchase volume from suppliers with better performance and reduce dependence on suppliers with poor performance. Through these functions, the supplier management agent effectively ensures the stability and flexibility of the supply chain and helps companies operate efficiently in a complex and changing market environment.

[0019] (4) Based on the predicted future market demand and inventory data, an inventory optimization agent is used to generate the optimal inventory replenishment plan.

[0020] Among them, the main responsibility of the inventory optimization agent is to ensure that the company minimizes inventory costs and inventory backlogs while meeting customer needs. The agent formulates and implements optimized inventory management strategies through real-time analysis of inventory data, demand forecasts and supply chain dynamic information to improve overall operational efficiency.

[0021] The inventory optimization agent generates the optimal inventory replenishment plan according to the following formula, and promptly recommends increasing replenishment or reducing inventory to reduce holding costs: , In the formula, for The replenishment quantity required at any time is used as the optimal inventory replenishment plan. is the safety stock level, To predict future market demand, express Actual sales volume at the moment, for Inventory levels at any given moment.

[0022] The inventory optimization agent is also used to monitor the inventory level of various products in real time, automatically generate inventory reports by integrating with the warehouse management system and sales data, and perform threshold detection on the current inventory level. , issue a low inventory alert when , an excess inventory alert is issued, where Represent the minimum and maximum inventory levels respectively.

[0023] The inventory optimization agent is also used to calculate the sales strategy for slow-moving products, including: A. Get the node feature vector of each product: , In the formula, For products The node feature vector of For products Current sales volume, For products inventory levels, For products The inventory turnover rate, For products Historical sales volatility; B. Calculate the similarity between any two products according to the following formula: , In the formula, For products and The similarity between Indicates the norm; C. Take each product as a base point and calculate the node feature vector of the product. , a graph convolutional neural network is used to form a correlation graph between all products and obtain the set of neighbor products associated with each product; D. Based on the node feature vector of the product Calculate the unsalable characteristic value of the product : , In the formula, , , They represent weight parameters respectively, , are the maximum sales volume and the maximum turnover rate, Representation and products The associated set of neighbor products; E. Calculate the product discount based on the product's unsalable feature value: , In the formula, Indicates products exist The discount at the moment, as the sales strategy output, , , They represent weight coefficients respectively, = represents the time decay function, is the attenuation coefficient.

[0024] The longer the time, the greater the discount. For example, when the inventory of a popular product is rapidly decreasing, the intelligent agent will analyze the market demand trend and recommend increasing the replenishment quantity, while coordinating various links in the supply chain to ensure that new goods arrive in time. For slow-moving products, the intelligent agent will identify their low turnover rate and recommend the implementation of promotional activities to increase sales, ultimately effectively optimizing the entire inventory management process, so that enterprises can respond flexibly and achieve sustained profitability in a complex market environment.

[0025] (5) Based on the inventory level after the optimal inventory replenishment plan and the logistics transportation data, the logistics scheduling agent is used to calculate the optimal transportation route.

[0026] Logistics scheduling agents play a key role in the entire supply chain management. They are mainly responsible for optimizing the cargo transportation process to ensure that logistics costs are reduced while meeting customer needs. The agent first monitors the transportation process in real time through an integrated transportation management system (TMS), including transportation methods (such as trucks, ships, and aviation), transportation routes, and cargo locations. By analyzing real-time traffic flow, weather conditions, and transportation company capabilities, the agent is able to evaluate potential transportation delays and develop alternative plans to deal with emergencies. For example, in terms of transportation task allocation, the agent will intelligently allocate transportation tasks based on order priority, customer demand, and inventory levels. For orders that need to be delivered urgently, the agent will prioritize the use of fast transportation methods such as air transportation, while for orders that can wait, it will choose a lower-cost transportation method. At the same time, the agent will also calculate the best loading plan based on the order quantity and cargo characteristics to improve the utilization rate of transportation vehicles and reduce transportation costs. It will use optimization algorithms to dynamically adjust the transportation route to ensure the shortest transportation time and the lowest transportation cost. When abnormal conditions occur during the transportation of goods, the agent can quickly notify the relevant parties and propose solutions to ensure timely response. By monitoring key performance indicators (KPIs) such as on-time delivery rate, transportation costs, and customer satisfaction, the agent is able to continuously analyze and optimize logistics strategies to respond to rapid changes in the market.

[0027] The logistics scheduling agent calculates the optimal transportation route with the lowest transportation cost as the objective function. The specific objective function is: , In the formula, After optimization time The value of is used as the decision variable output of the optimal transportation path. and are the updated cost and time estimates from the Bayesian update, Indicates whether to select the transport node To transport node No. path, if , indicating selection, Indicates not to choose, represents the prior probability, and They represent the correlation between the new observation data and the estimated transportation time and transportation cost respectively. represents the overall probability of new observations, represents new observation data, and Respectively The estimated transportation time and transportation cost of the momentary path, Indicates the number of transport nodes.

[0028] The optimal transportation path is obtained through reinforcement learning for the above objective function. The specific steps of reinforcement learning are: A. Setting status ; B. Set Action for ; C. Setting the reward function ; D. Set the update rules as: , In the formula, represents the value function, , They represent weight coefficients respectively, Indicates that The maximum value of A, A represents any action in the action space, express The state of the moment; E. Continue to execute the above update rules until the preset number of iterations is reached, and the final output { } as the optimal transportation route.

[0029] It is understandable that the objective function may also be solved in other ways to obtain the most efficient transportation path.

[0030] The logistics scheduling agent performs transport vehicle loading optimization according to the following formula: , In the formula, represents the optimization objective function, represents the transport vehicle capacity, Indicates the order quantity, Indicates The weight of the goods in the order. Indicates The loading quantity of an order, the value range is [0, ],in It is The demand for an order, Indicates business indicators for orders, and , , is the weight coefficient, Indicates the delivery deadline of the order (the number of days remaining from the current time, the shorter the time, the higher the priority), Indicates the past order volume of the order customer, which is used to reflect the customer's business volume. Indicates the order demand (the higher the demand, the higher the priority).

[0031] (6) Collect feedback data from the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent, calculate the performance indicators of each agent using the information feedback agent based on the feedback data, and adjust the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent based on the performance indicators of each agent until the performance indicators meet the preset requirements.

[0032] The information feedback agent is a crucial part of the supply chain optimization and scheduling system. It is responsible for collecting, monitoring and feedback data from various modules of the system in real time to ensure the efficient operation and continuous improvement of the supply chain. First, it obtains various data from demand forecasting, supplier management, inventory optimization and logistics scheduling agents, such as order processing time, inventory level, supplier delivery timeliness, transportation delays, and accuracy of demand forecasts. By analyzing this data, the information feedback agent can monitor key performance indicators (KPIs) in the supply chain in real time, such as inventory turnover, on-time order delivery rate, and logistics costs. Once an abnormal situation is found, such as the supplier's failure to deliver on time or the logistics route is delayed, the information feedback agent will immediately generate an alarm and notify the corresponding agent to take corrective measures, such as adjusting supplier cooperation or changing the transportation route. In addition, the information feedback agent is not limited to real-time monitoring, but also generates feedback reports and data visualization based on historical data to help companies identify potential problems and bottlenecks in the supply chain. These reports provide managers with detailed operational insights to facilitate strategic adjustments. For example, the agent may find through data analysis that the delivery performance of a supplier has declined during a specific period of time. The system can recommend that the company consider replacing suppliers or taking other preventive measures. The ultimate goal of the information feedback agent is to ensure that the entire supply chain system forms a closed-loop feedback mechanism, continuously optimize the process, improve the system's response speed and flexibility, enable the supply chain to cope with complex and changing market environments, reduce costs and improve overall efficiency.

[0033] (7) The outputs of the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent whose performance indicators meet the preset requirements are combined as the scheduling strategy output of the supply chain.

[0034] After the entire transaction is executed, the execution results of the entire supply chain are comprehensively evaluated and analyzed, and a detailed optimization report is generated. First, based on the execution data of each task, such as the accuracy of demand forecasting, the effectiveness of supplier management, the rationality of inventory optimization, and the efficiency of logistics scheduling, each link of the supply chain is evaluated and compared with the preset key performance indicators (KPIs) to identify existing problems and optimization space. Then, combined with historical data to analyze trends and patterns, summarize the performance of strategies under different conditions, and provide a basis for future improvements. Finally, a supply chain optimization report is generated, which lists problems, data analysis, and improvement suggestions in detail to help companies optimize resource allocation, improve efficiency, reduce costs, and provide decision-making support for responding to future market changes.

[0035] The embodiment of the present invention also provides an intelligent supply chain scheduling optimization system based on an autonomous and controllable large model, including: Data perception module, used to obtain supply chain-related data at the current moment, including product sales data, supplier supply product data, inventory data, and logistics transportation data; A demand forecasting agent is used to forecast future market demand based on product sales data; wherein the demand forecasting agent is trained using historical product sales data as training samples and actual sales conditions at corresponding moments as labels; A supplier management agent, used to select the best supplier as the target supplier based on the supplier supply product data; wherein the supplier management agent selects the supplier with the smallest sum of the positive weight of product cost, the positive weight of product delivery time, the negative weight of product quality and the positive weight of supplier supply reputation value as the best supplier; Inventory optimization agent, which generates optimal inventory replenishment plans based on predicted future market demand and inventory data; A logistics scheduling agent is used to calculate the optimal transportation path based on the inventory level after the optimal inventory replenishment plan and the logistics transportation data; wherein the logistics scheduling agent calculates the optimal transportation path with the lowest transportation cost as the objective function; The information feedback agent is used to collect feedback data from the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent, calculate the performance indicators of each agent based on the feedback data, and adjust the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent according to the performance indicators of each agent until the performance indicators meet the preset requirements; The scheduling strategy output module is used to merge the outputs of the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent whose performance indicators meet the preset requirements as the scheduling strategy output of the supply chain.

[0036] The system provided by the embodiment of the present invention can be used to execute the above method, and has the corresponding functions and beneficial effects of the execution method. The specific introduction of each intelligent agent is described in the method, and will not be repeated here.

[0037] It is worth noting that in the embodiment of the above system, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0038] The embodiments described above are only illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, i.e., they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, or of course, it can be implemented only by hardware, as long as the function or effect can be achieved.

[0039] The method and system provided by the embodiments of the present invention can be applied to supply chain scheduling optimization in various industries, such as the power equipment industry or the medical industry.

[0040] It should be understood that the above embodiments and descriptions only describe the principles, main features and advantages of the present invention. Without departing from the spirit and scope of the present invention, the present invention may be subject to various changes and improvements, and these changes and improvements all fall within the scope of protection of the present invention.

Claims

1. An intelligent supply chain scheduling optimization method based on an autonomous and controllable large model, characterized in that: The steps include: (1) Obtaining supply chain-related data at the current moment, including product sales data, supplier product supply data, inventory data, and logistics and transportation data; (2) Based on the product sales data, a demand forecasting agent is used to predict future market demand; wherein the demand forecasting agent is trained using historical product sales data as training samples and actual sales conditions at the corresponding time as labels; (3) Based on the product data supplied by the suppliers, the supplier management agent is used to select the best supplier as the target supplier; wherein the supplier management agent selects the supplier with the smallest sum of the positive weight of the product cost, the positive weight of the product delivery time, the negative weight of the product quality, and the positive weight of the supplier's supply reputation value as the best supplier; (4) Based on the predicted future market demand and inventory data, an inventory optimization agent is used to generate the optimal inventory replenishment plan; (5) using a logistics scheduling agent to calculate an optimal transportation path based on the inventory level after the optimal inventory replenishment plan and the logistics transportation data; wherein the logistics scheduling agent calculates the optimal transportation path with the lowest transportation cost as the objective function; (6) Collect feedback data from the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent, calculate the performance indicators of each agent using the information feedback agent based on the feedback data, and adjust the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent based on the performance indicators of each agent until the performance indicators meet the preset requirements; (7) The outputs of the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent whose performance indicators meet the preset requirements are combined as the scheduling strategy output of the supply chain.

2. The intelligent supply chain scheduling optimization method based on an autonomous and controllable large model according to claim 1 is characterized in that: Step (2) specifically includes: (2.1) Obtain product sales feature X based on product sales data: ， Represents product sales data before the current moment, Indicates seasonal sales characteristics, Indicates holiday sales characteristics; (2.2) Product sales features Input the trained demand forecasting agent to obtain future market demand , wherein the demand forecasting agent is a combination of the LSTM network and the Transformer network: , in represents the LSTM network model parameters, Represents the Transformer network model parameters, Represents the model parameters of the demand forecasting agent.

3. The intelligent supply chain scheduling optimization method based on an autonomous and controllable large model according to claim 2 is characterized in that: The demand forecasting agent is trained in the following way: Based on actual sales data and predicted future market demand, calculate the forecast error: ， In the formula, , Respectively , The prediction error at time is the time weighting coefficient, express Actual sales volume at the moment, Indicates the predicted Market demand at all times; Adjust the learning rate of the demand forecasting agent model parameters based on the prediction error: ， in They are , The learning rate at each moment, , , is a hyperparameter, is the time step coefficient of learning rate decay, is the weight coefficient of the momentum term, represents the change in prediction error, for The momentum term at time , and , is the weight coefficient, for The momentum term at time, express Demand forecasting agent model parameters at the moment The loss function is represents the gradient; Update the demand forecasting agent model parameters according to the adjusted learning rate: ， in, express The demand forecasting agent model parameters at the moment, is a hyperparameter, is The inverse of the BFGS approximation matrix of time.

4. The intelligent supply chain scheduling optimization method based on an autonomous and controllable large model according to claim 1 is characterized in that: The objective function of the supplier management agent is: ， In the formula, represents the best supplier, Indicates supplier The quoted cost, Indicates supplier The average delivery time, Indicates supplier The product quality score, represents the supply reputation value of supplier i, , , are the highest acceptable cost, longest delivery time and highest product quality score, respectively. , , , is the weight coefficient and satisfies the following formula: ， Respectively , time The value of is the adjustment amplitude coefficient, is the adjustment increment, and are the adjustment factor and feedback variable, represents the feedback coefficient, is with The amount of relevant real-time feedback; ， in, is the supplier’s delivery delay rate, is the supplier’s default rate, , is the adjustment weight, , is the penalty coefficient.

5. The intelligent supply chain scheduling optimization method based on an autonomous and controllable large model according to claim 1 is characterized in that: The inventory optimization agent generates the optimal inventory replenishment plan according to the following formula: ， In the formula, for The replenishment quantity required at any time is used as the optimal inventory replenishment plan. is the safety stock level, To predict future market demand, express Actual sales volume at the moment, for Inventory levels at any given moment.

6. The intelligent supply chain scheduling optimization method based on an autonomous and controllable large model according to claim 5 is characterized in that: The inventory optimization agent is also used to: Perform threshold detection on current inventory level. , issue a low inventory alert when , an excess inventory alert is issued, where Respectively represent the minimum and maximum inventory levels; Calculate a sales strategy for slow-moving products, including: A. Get the node feature vector of each product: ， In the formula, For products The node feature vector of For products Current sales volume, For products inventory levels, For products Inventory turnover rate, For products Historical sales volatility; B. Calculate the similarity between any two products according to the following formula: ， In the formula, For products and The similarity between Indicates the norm; C. Take each product as a base point and calculate the node feature vector of the product. , a graph convolutional neural network is used to form a correlation graph between all products and obtain the set of neighbor products associated with each product; D. Based on the node feature vector of the product Calculate the unsalable characteristic value of the product : ， In the formula, , , They represent weight parameters respectively, , are the maximum sales volume and the maximum turnover rate, Representation and products The associated set of neighbor products; E. Calculate the product discount based on the product's unsalable feature value: ， In the formula, Indicates products exist The discount at the moment, as the sales strategy output, , , They represent weight coefficients respectively, = represents the time decay function, is the attenuation coefficient.

7. The intelligent supply chain scheduling optimization method based on an autonomous and controllable large model according to claim 1 is characterized in that: The objective function of the logistics scheduling agent is: ， In the formula, After optimization time The value of is used as the decision variable output of the optimal transportation path. and are the updated cost and time estimates from the Bayesian update, Indicates whether to select the transport node To transport node No. path, if , indicating selection, Indicates not to choose, represents the prior probability, and They represent the correlation between the new observation data and the estimated transportation time and transportation cost respectively. represents the overall probability of new observations, represents new observation data, and Respectively The estimated transportation time and transportation cost of the momentary path, Indicates the number of transport nodes.

8. The intelligent supply chain scheduling optimization method based on an autonomous and controllable large model according to claim 7 is characterized in that: The objective function of the logistics scheduling agent obtains the optimal transportation path through reinforcement learning. The steps of reinforcement learning are as follows: A. Setting status ; B. Set Action for ; C. Setting the reward function ; D. Set the update rules as: ， In the formula, represents the value function, , They represent weight coefficients respectively, Indicates that The maximum value of A, A represents any action in the action space, express The state of the moment; E. Continue to execute the above update rules until the preset number of iterations is reached, and the final output { } as the optimal transportation route.

9. The intelligent supply chain scheduling optimization method based on an autonomous and controllable large model according to claim 7 is characterized in that: The logistics scheduling agent is also used to optimize the loading of transport vehicles according to the following formula: ， In the formula, represents the optimization objective function, represents the transport vehicle capacity, Indicates the order quantity, Indicates The weight of the goods in the order, Indicates The loading quantity of an order, the value range is [0, ],in It is The demand for an order, Indicates business indicators for orders, and , , is the weight coefficient, Indicates the delivery deadline for the order. Indicates the past order volume of this order customer. Indicates the order demand.

10. An intelligent supply chain scheduling optimization system based on an autonomous and controllable large model, characterized in that: include: Data perception module, used to obtain supply chain-related data at the current moment, including product sales data, supplier supply product data, inventory data, and logistics transportation data; A demand forecasting agent is used to forecast future market demand based on product sales data; wherein the demand forecasting agent is trained using historical product sales data as training samples and actual sales conditions at corresponding moments as labels; A supplier management agent, used to select the best supplier as the target supplier based on the supplier supply product data; wherein the supplier management agent selects the supplier with the smallest sum of the positive weight of product cost, the positive weight of product delivery time, the negative weight of product quality and the positive weight of supplier supply reputation value as the best supplier; Inventory optimization agent, which generates optimal inventory replenishment plans based on predicted future market demand and inventory data; A logistics scheduling agent is used to calculate the optimal transportation path based on the inventory level after the optimal inventory replenishment plan and the logistics transportation data; wherein the logistics scheduling agent calculates the optimal transportation path with the lowest transportation cost as the objective function; The information feedback agent is used to collect feedback data from the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent, calculate the performance indicators of each agent based on the feedback data, and adjust the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent according to the performance indicators of each agent until the performance indicators meet the preset requirements; The scheduling strategy output module is used to merge the outputs of the demand forecasting agent, supplier management agent, inventory optimization agent, and logistics scheduling agent whose performance indicators meet the preset requirements as the scheduling strategy output of the supply chain.

Images