A Multi-Agent Demand Forecasting and Order Generation Method and System for Front-end Warehouse Replenishment

By working collaboratively with multiple agents, the system generates collaborative demand and order quantities for forward warehouses, solving the problem of uneven inventory distribution in multi-forward warehouse replenishment systems, improving the fairness of replenishment and inventory turnover efficiency, and reducing the risk of stockouts.

CN123117977APending Publication Date: 2026-11-17JIANGSU CANGLIAN TIANXIA NETWORK TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611051953.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-15
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

When multiple forward warehouses simultaneously submit requests to the central warehouse for the same target product, the existing forward warehouse replenishment system suffers from problems such as insufficient allocable inventory in the central warehouse, difficulty in quantifying the priority of each warehouse's requests, failure to coordinate the allocation of near-expiry inventory and historical stockouts, and a disconnect between predicted quantities and actual order packaging specifications. This makes it difficult to balance replenishment fairness and inventory turnover efficiency.

Method used

Through multi-agent collaborative work, the demand forecasting agent of each front warehouse generates the single warehouse demand forecast, and the central warehouse collaborative agent summarizes and allocates the demand. Combining the proportion of near-expiry inventory, the number of historical stockouts and the minimum packaging size, the allocation weight is calculated and the collaborative order quantity is generated, and finally the order quantity is generated.

Benefits of technology

It enables coordinated allocation among multiple forward warehouses when central warehouse inventory is insufficient, improving the fairness of replenishment and inventory turnover efficiency, ensuring the matching of forecasted quantities with actual order specifications, and reducing the risk of stockouts.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123117977A_ABST
    Figure CN123117977A_ABST
Patent Text Reader

Abstract

This invention discloses a multi-agent demand forecasting and order quantity generation method and system for front-end warehouse replenishment, relating to the fields of intelligent front-end warehouse replenishment and supply chain collaborative decision-making technology. The method includes: each front-end warehouse demand forecasting agent generating a single-warehouse demand forecast based on historical sales, inventory, stockouts, product lifecycle, weather, holidays, delivery cycles, safety stock, and minimum packaging size; a central warehouse collaborative agent summarizing the demand forecasts from each single warehouse and determining whether there is an over-demand based on the available inventory; when inventory is insufficient, calculating the allocation weight based on the proportion of near-expiry inventory, historical stockout frequency, and single-warehouse demand forecast to obtain the collaborative order quantity; each front-end warehouse generating an order quantity rounded down according to the minimum packaging size, and forming a target product order plan. This invention can improve the rationality of multi-front-end warehouse replenishment allocation and the accuracy of order execution.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of intelligent replenishment of forward warehouses and supply chain collaborative decision-making technology, and in particular to a multi-agent demand forecasting and order quantity generation method and system for forward warehouse replenishment. Background Technology

[0002] With the rapid development of instant retail, fresh food e-commerce, and community pre-warehouse businesses, pre-warehouses have gradually evolved from traditional end-point storage nodes into high-frequency operating units integrating order fulfillment, inventory turnover, product preservation, and coordinated order fulfillment with central warehouses. The accuracy of pre-warehouse replenishment depends not only on the historical sales volume of a single warehouse but is also affected by factors such as weather, holidays, product lifecycle, historical stockouts, real-time inventory, safety stock, delivery cycle, and minimum packaging size. Existing replenishment systems mostly employ single-point forecasting or centralized replenishment strategies, which can alleviate stockouts and inventory backlogs to some extent. However, when multiple pre-warehouses simultaneously submit requests to the central warehouse for the same target product, problems remain, including insufficient allocable inventory in the central warehouse, difficulty in quantifying the priority of each warehouse's requests, failure to coordinate the allocation of near-expiry inventory and historical stockouts, and a disconnect between forecasted quantities and actual order packaging sizes. This makes it difficult to simultaneously ensure fairness in replenishment among pre-warehouses, control stockout risks, and maintain inventory turnover efficiency.

[0003] CN120952857B discloses a method and system for dynamic prediction of replenishment demand based on multi-node time-series data fusion. The method includes: deploying multiple lightweight store agent backends at multiple store POS terminals; updating virtual inventory across multiple chain stores based on cumulative sales volume calculations to obtain multiple sets of real-time product inventory data; predicting inventory demand based on multiple sets of time-series and real-time product inventory data to obtain multiple sets of replenishment times and predicted replenishment quantities; performing dynamic supply chain fitting to output replenishment delivery routes; and dynamically replenishing multiple chain stores in response to the replenishment delivery routes. This solution primarily addresses the problem of insufficient synchronization of sales and inventory data across multiple chain stores, making dynamic replenishment optimization difficult. However, its focus is on multi-node time-series data fusion, virtual inventory updates, and delivery route fitting. It does not perform weighted calculations for collaborative allocation among multiple front-end warehouses in scenarios where central warehouse inventory is insufficient, nor does it combine the proportion of near-expiry inventory, historical stockout frequency, and single-warehouse demand forecasts for order allocation. This makes it difficult to solve the inventory allocation problem when multiple front-end warehouses compete for the same target product.

[0004] CN114862302B discloses a method and apparatus for replenishing a front warehouse. The method includes: determining each front warehouse to be sorted based on the current sorting quantity in the sorting trigger instruction; determining each first predicted inventory quantity of each front warehouse to be sorted for the goods to be sorted at the sorting time; updating each second replenishment quantity of each front warehouse to be sorted based on each first predicted inventory quantity and each first replenishment quantity of each front warehouse to be sorted; the sum of each first replenishment quantity equals the sum of each second replenishment quantity; and replenishing each sorting area corresponding to each front warehouse to be sorted according to each second replenishment quantity. This solution improves the accuracy of replenishment in the sorting process by redetermining the replenishment quantity at sorting times closer to the target time. However, its core is to reallocate the replenishment quantity in the sorting area based on sorting trigger instructions. It does not establish a linkage mechanism between the demand forecasting agent and the central warehouse collaborative agent for the daily ordering business of the front warehouse, nor does it judge whether the aggregated demand is excessive based on the available inventory of the central warehouse, nor does it round up the order quantity based on the smallest packaging size after the collaborative order quantity is formed. Therefore, it is difficult to meet the needs of automated ordering plan generation under conditions of multiple front warehouses, the same target product, and limited inventory. Summary of the Invention

[0005] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the specification abstract and the title of the invention, to avoid obscuring the purpose of this section, the specification abstract, and the title of the invention. Such simplifications or omissions shall not be used to limit the scope of the invention.

[0006] In view of the problems of existing front-end warehouse replenishment technology, such as the disconnect between single-warehouse demand forecasting and central warehouse inventory constraints, the lack of collaborative allocation basis when multiple front-end warehouses request goods at the same time, the difficulty in balancing stockout risk and inventory turnover efficiency in the case of insufficient inventory, and the failure of collaborative allocation results to be adapted to the smallest packaging size, this invention is proposed.

[0007] Therefore, the problem to be solved by this invention is how to generate more reasonable and executable collaborative demand and order quantities when multiple front warehouses generate replenishment needs for the same target product at the same time, by combining the actual sales demand, inventory status, stockout risk and the available inventory of the central warehouse of each front warehouse.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, embodiments of the present invention provide a multi-agent demand forecasting and order quantity generation method for replenishing forward warehouses, comprising, Each front-end warehouse demand forecasting agent acquires historical sales data, inventory data, historical out-of-stock counts, lifecycle tags, weather data, holiday data, delivery cycle, safety stock, and minimum packaging size of the target product, and generates a single warehouse demand forecast. The central warehouse collaborative intelligent agent obtains the identifiers and allocable inventory of all front warehouses that have submitted single-warehouse demand forecasts for the target product, sums up the single-warehouse demand forecasts of all front warehouses to obtain the total demand for the target product, and determines whether the total demand exceeds the allocable inventory. When the total demand exceeds the available inventory, the allocation weight is calculated based on the proportion of near-expiry inventory in each forward warehouse, the number of historical stockouts, and the demand forecast for each warehouse. The available inventory is then allocated according to the allocation weight to obtain the coordinated demand. The demand forecasting agent rounds up the collaborative order quantity based on the collaborative order quantity and minimum packaging size of the current front warehouse to obtain the order quantity for each front warehouse. The central warehouse collaborative intelligent agent generates a target product requisition plan for multiple front warehouses based on the front warehouse identifier, target product identifier, single warehouse demand forecast, collaborative requisition quantity, order quantity, and collaborative allocation information.

[0009] Secondly, embodiments of the present invention provide a multi-agent demand forecasting and order quantity generation system for replenishing forward warehouses, comprising: The demand forecasting module is used by the demand forecasting agents of each front warehouse to obtain historical sales data, inventory data, historical out-of-stock times, lifecycle tags, weather data, holiday data, delivery cycle, safety stock, and minimum packaging size of the target product, and generate a single warehouse demand forecast. The demand aggregation module is used by the central warehouse collaborative intelligent agent to obtain the identifiers and allocable inventory of all front warehouses that have submitted single-warehouse demand forecasts for the target product, sum up the single-warehouse demand forecasts of all front warehouses to obtain the aggregated demand of the target product, and determine whether the aggregated demand exceeds the allocable inventory. The collaborative allocation module is used to take the single-warehouse demand forecast of each front warehouse as the collaborative order quantity of the corresponding front warehouse when the total demand does not exceed the allocable inventory. Otherwise, it calculates the allocation weight based on the near-expiry inventory ratio, historical stockout frequency and single-warehouse demand forecast of each front warehouse, and allocates the allocable inventory according to the allocation weight to obtain the collaborative order quantity. The order rounding module is used by the demand forecasting agent of each front warehouse to round up the collaborative order quantity based on the collaborative order quantity and minimum packaging size of the front warehouse, so as to obtain the order quantity of each front warehouse. The plan generation module is used by the central warehouse collaborative intelligent agent to generate a target product requisition plan for multiple front warehouses based on the front warehouse identifier, target product identifier, single warehouse demand forecast, collaborative requisition quantity, order quantity, and collaborative allocation information.

[0010] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any step of the above-described method for multi-agent demand forecasting and order quantity generation for front-end warehouse replenishment.

[0011] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the above-described method for multi-agent demand forecasting and order quantity generation for front-end warehouse replenishment.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 A flowchart for a multi-agent demand forecasting and order quantity generation method for replenishing forward warehouses; Figure 2 This is a structural diagram of a multi-agent demand forecasting and order quantity generation system for replenishing forward warehouses. Detailed Implementation

[0014] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0015] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort should fall within the scope of protection of this invention.

[0016] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0017] As mentioned in the background section, existing multi-node replenishment demand forecasting methods primarily focus on sales data synchronization, inventory demand forecasting, or delivery route fitting. Existing forward warehouse replenishment methods mainly focus on re-determining replenishment quantities at sorting times. Neither fully considers the collaborative allocation problem among multiple forward warehouses under conditions of insufficient central warehouse inventory, nor does it incorporate near-expiry inventory ratios, historical stockout occurrences, single-warehouse demand forecasts, and minimum packaging specifications into a unified requisition quantity generation process. To address these issues, this invention provides a multi-agent demand forecasting and requisition quantity generation method for forward warehouse replenishment.

[0018] Reference Figures 1-2 , Figure 1 The flowchart below illustrates a multi-agent demand forecasting and order quantity generation method for front-end warehouse replenishment according to an embodiment of the present invention. The method includes: S1: The demand forecasting agent of each front warehouse obtains historical sales data, inventory data, historical out-of-stock times, lifecycle tags, weather data, holiday data, delivery cycle, safety stock, and minimum packaging size of the target product, and generates a single warehouse demand forecast.

[0019] S1.1: The demand forecasting agent reads historical sales data, inventory data, historical out-of-stock counts, lifecycle tags, weather data, holiday data, delivery cycle, safety stock, and minimum packaging size based on the target product identifier; among which, the inventory data includes the real-time inventory and locked inventory at the current moment, and the historical sales data includes the daily sales sequence of the target product for each day within the preset historical time period.

[0020] S1.2: Sort the daily sales volume sequence according to the sales date, and remove returned sales, canceled order sales and abnormal negative sales to obtain the effective daily sales volume sequence; calculate the daily average sales volume based on the effective daily sales volume sequence, and use the daily average sales volume as the benchmark daily sales volume of the target product.

[0021] S1.3: Determine the sales adjustment factor based on the life cycle label, and multiply the base day sales by the sales adjustment factor to obtain the life cycle adjusted daily sales; where the life cycle label is the introduction stage, growth stage, maturity stage, decline stage or elimination stage, and different life cycle labels correspond to different sales adjustment factors.

[0022] In this embodiment of the application, if the lifecycle label is the introduction stage, the sales adjustment coefficient is the first preset coefficient; if the lifecycle label is the growth stage, the sales adjustment coefficient is the second preset coefficient; if the lifecycle label is the maturity stage, the sales adjustment coefficient is the third preset coefficient; if the lifecycle label is the decline stage, the sales adjustment coefficient is the fourth preset coefficient; and if the lifecycle label is the elimination stage, the sales adjustment coefficient is the fifth preset coefficient.

[0023] Preferably, the first preset coefficient and the second preset coefficient are both greater than the third preset coefficient, and the fourth preset coefficient and the fifth preset coefficient are both less than the third preset coefficient; the first preset coefficient is 1.2 to 1.5, the second preset coefficient is 1.1 to 1.3, the third preset coefficient is 1.0, the fourth preset coefficient is 0.7 to 0.9, and the fifth preset coefficient is 0.3 to 0.5.

[0024] It should be noted that the settings of the first to fifth preset coefficients are determined based on prior knowledge of the sales variation patterns at each stage of the product lifecycle. Specifically: Introduction Phase (First Preset Coefficient 1.2~1.5): When a product is first launched, the sales base is low but the growth potential is high. In order to ensure the market coverage and trial sales opportunities of new products, the front-end warehouse needs to be positively amplified when forecasting demand. The coefficient is greater than 1.0 and has the highest value among all stages because the absolute value of sales is the lowest during the introduction phase. If it is not amplified, the replenishment volume is easily underestimated, resulting in stockouts and loss of promotion opportunities. The value range of 1.2~1.5 is determined based on the safety stock threshold for new product distribution in the retail industry. If it is lower than 1.2, the amplification effect is insufficient to compensate for the distortion of the initial sales. If it is higher than 1.5, it is easy to cause excessive inventory for the first order. Growth stage (second preset coefficient 1.1~1.3): Product sales show an accelerating upward trend, but have not yet reached a stable peak. It is necessary to appropriately increase the forecast to keep up with the growth pace. The coefficient is lower than that of the introduction stage (1.1~1.3 < 1.2~1.5) because the existing historical data in the growth stage is more complete than that in the introduction stage, reducing the urgency of increasing the forecast; however, it is still greater than 1.0 to ensure that replenishment can match the growth trend. Maturity stage (third preset coefficient 1.0): Product sales fluctuate stably around the peak, and historical daily sales can accurately reflect future demand. There is no need to amplify or reduce it. Therefore, the coefficient is set to 1.0 as a benchmark. Decline phase (fourth preset coefficient 0.7~0.9): Product sales show a continuous downward trend. If replenishment is based on the daily sales of the maturity phase, it will lead to inventory backlog. Therefore, it is necessary to multiply by a coefficient less than 1.0 to reduce the inventory. A value range of 0.7~0.9 means that the predicted value is compressed by 10%~30%, which matches the gradual reduction of products in the decline phase in the retail industry and the gradual exit strategy to avoid one-time clearance.

[0025] Phase-out period (fifth preset coefficient 0.3~0.5): Product sales are extremely low and the product is about to be withdrawn from the market. Only a very small amount of inventory needs to be maintained to meet the needs of a very small number of users or for transitional sales. Therefore, the predicted value is greatly compressed to 30%~50% of the benchmark. If it is lower than 0.3, it may lead to a complete stockout and affect the user experience. If it is higher than 0.5, it does not meet the inventory control target for the upcoming phase-out.

[0026] S1.4: Determine the weather impact coefficient based on weather data and the holiday impact coefficient based on holiday data; multiply the lifecycle-corrected daily sales by the weather impact coefficient and the holiday impact coefficient respectively to obtain the comprehensive corrected daily sales.

[0027] In this embodiment of the application, the method for determining the holiday impact coefficient is as follows: if the holiday data indicates that the predicted date is a statutory holiday, then the holiday impact coefficient is the first holiday coefficient; the value range of the first holiday coefficient is 1.2 to 1.5; if the holiday data indicates that the predicted date is the day before a statutory holiday, then the holiday impact coefficient is the second holiday coefficient; the value range of the second holiday coefficient is 1.3 to 1.8, and the second holiday coefficient is greater than the first holiday coefficient; if the holiday data indicates that the predicted date is the day after a statutory holiday, then the holiday impact coefficient is the third holiday coefficient; the value range of the third holiday coefficient is 0.7 to 0.9; if the holiday data indicates that the predicted date is a regular working day or a non-holiday rest day, then the holiday impact coefficient is the fourth holiday coefficient, and the fourth holiday coefficient is equal to 1.

[0028] It should be noted that different types of statutory holidays correspond to different coefficients for the first, second, and third holidays. These coefficients are fixed values ​​that are pre-calibrated based on the historical sales fluctuations of the target product category during each statutory holiday.

[0029] In this embodiment of the application, the weather data includes the forecast weather type and the forecast temperature; the holiday data includes whether the current date is a holiday and the type of holiday; the forecast weather type includes sunny, cloudy, overcast, light rain, moderate rain, heavy rain, light snow, moderate snow, and heavy snow; different forecast weather types correspond to different weather impact coefficients.

[0030] S1.5: Determine the number of days of demand coverage based on the delivery cycle, and multiply the comprehensive adjusted daily sales by the number of days of demand coverage to obtain the cycle demand forecast; wherein, the number of days of demand coverage includes at least the number of days corresponding to the delivery cycle.

[0031] S1.6: Determine the stockout compensation coefficient based on the historical stockout frequency and the preset stockout frequency threshold, and multiply the periodic demand forecast by the stockout compensation coefficient to obtain the compensated periodic demand forecast.

[0032] Preferably, if the number of historical stockouts is greater than the preset stockout threshold, the stockout compensation coefficient is greater than 1; if the number of historical stockouts is less than or equal to the preset stockout threshold, the stockout compensation coefficient is 1.

[0033] It should be noted that the preset out-of-stock frequency threshold is determined based on the historical out-of-stock rate statistical distribution of the target product category: obtain the historical out-of-stock frequency of all products under the target product category within the preset statistical period, calculate the upper quartile of the historical out-of-stock frequency, and use the upper quartile as the preset out-of-stock frequency threshold.

[0034] S1.7: Calculate the available inventory based on the real-time inventory and locked inventory. Add the safety stock to the compensated periodic demand forecast and then subtract the available inventory to obtain the demand gap for a single warehouse. The available inventory is the value of the real-time inventory minus the locked inventory.

[0035] S1.8: Compare the single warehouse demand gap with zero. If the single warehouse demand gap is less than zero, then the single warehouse demand forecast is set to zero. If the single warehouse demand gap is greater than or equal to zero, then the single warehouse demand gap is set as the single warehouse demand forecast, and the single warehouse demand forecast, target product identifier, minimum packaging specification and corresponding front warehouse identifier are sent to the central warehouse collaborative intelligent agent.

[0036] S2: The central warehouse collaborative intelligent agent obtains the identifiers and allocable inventory of all front warehouses that have submitted single-warehouse demand forecasts for the target product, sums up the single-warehouse demand forecasts of all front warehouses to obtain the total demand for the target product, and determines whether the total demand exceeds the allocable inventory.

[0037] S2.1: The central warehouse collaborative intelligent agent receives the single warehouse demand forecast quantity, target product identifier, minimum packaging specification and corresponding front warehouse identifier sent by the demand forecasting intelligent agents of each front warehouse, and collects the received data according to the target product identifier to obtain the front warehouse demand submission set of the target product.

[0038] S2.2: Perform validity checks on the set of demand submissions for the front-end warehouse, and remove data with empty target product identifiers, empty front-end warehouse identifiers, single warehouse demand forecasts less than zero, or minimum packaging size less than or equal to zero, to obtain the set of valid demand submissions.

[0039] In this embodiment, the validity verification method is as follows: the central warehouse collaborative agent traverses each submission record in the front warehouse demand submission set, and executes the first verification condition and the second verification condition in sequence; the first verification condition is: determining whether the target product identifier in the submission record is an empty string or a null value, and determining whether the front warehouse identifier is an empty string or a null value; if the target product identifier is an empty string or a null value, or the front warehouse identifier is an empty string or a null value, then the submission record is marked as an invalid record and removed from the front warehouse demand submission set; the second verification condition is: determining whether the single warehouse demand forecast in the submission record is less than zero, and determining whether the minimum packaging specification in the submission record is less than or equal to zero; if the single warehouse demand forecast is less than zero, or the minimum packaging specification is less than or equal to zero, then the submission record is marked as an invalid record and removed from the front warehouse demand submission set; the submission records that pass the first verification condition and the second verification condition are retained to obtain the valid demand submission set.

[0040] In this embodiment of the application, when multiple single-warehouse demand forecasts exist for the same front warehouse identifier in the same forecast batch, the central warehouse collaborative agent selects the latest single-warehouse demand forecast according to the submission time and uses it as the valid single-warehouse demand forecast corresponding to the front warehouse identifier, and deletes the remaining single-warehouse demand forecasts corresponding to the front warehouse identifier.

[0041] S2.3: Extract all front warehouse identifiers from the set of valid demand submissions and establish a demand correspondence between front warehouse identifiers and single warehouse demand forecasts; the demand correspondence is used to record the demand quantity of each front warehouse for the target product.

[0042] In this embodiment, the method for establishing the demand mapping relationship is as follows: the central warehouse collaborative agent creates a key-value pair mapping table with the former warehouse identifier as the primary key; the central warehouse collaborative agent traverses the set of valid demand submissions, and for each valid submission record, extracts the former warehouse identifier and the single warehouse demand forecast from the valid submission record; the central warehouse collaborative agent stores the former warehouse identifier as the key and the single warehouse demand forecast as the value in the key-value pair mapping table; if the same former warehouse identifier already exists in the key-value pair mapping table, the central warehouse collaborative agent retains the latest key-value pair according to the submission time and overwrites the existing key-value pair; after traversing all valid submission records, the key-value pair mapping table is output as the demand mapping relationship.

[0043] S2.4: Read the central warehouse inventory record based on the target product identifier, obtain the real-time inventory quantity, locked inventory quantity, and reserved inventory quantity of the target product in the central warehouse, and calculate the allocable inventory quantity based on the real-time inventory quantity, locked inventory quantity, and reserved inventory quantity.

[0044] Preferably, the allocable inventory is the value obtained by subtracting the locked inventory and reserved inventory in the central warehouse from the real-time inventory in the central warehouse; when the calculation result is less than zero, the allocable inventory is set to zero.

[0045] S2.5: Based on the demand correspondence, sum up the single-warehouse demand forecasts corresponding to all front-end warehouse identifiers to obtain the total demand for the target product.

[0046] In this embodiment of the application, the formula for calculating the total demand is as follows: ; in, This represents the total demand for the target product. Let be the single-warehouse demand forecast for the i-th forward warehouse, n be the total number of forward warehouses that have submitted single-warehouse demand forecasts, and i be the serial number of the forward warehouse that has submitted single-warehouse demand forecasts.

[0047] S2.6: Compare the total demand with the available inventory to determine the inventory satisfaction status.

[0048] In this embodiment, if the total demand is less than or equal to the allocable inventory, the inventory satisfaction status is "sufficient inventory". The central warehouse collaborative agent determines the single warehouse demand forecast for each front warehouse as the collaborative order quantity for each front warehouse and writes the sufficient inventory status into the collaborative allocation information. If the total demand is greater than the allocable inventory, the inventory satisfaction status is "insufficient inventory". The central warehouse collaborative agent writes the effective demand submission set, demand correspondence, total demand, allocable inventory, and insufficient inventory status into the collaborative allocation information and passes the collaborative allocation information to the subsequent allocation weight calculation step.

[0049] S3: When the total demand exceeds the available inventory, the allocation weight is calculated based on the proportion of near-expiry inventory in each forward warehouse, the number of historical stockouts, and the demand forecast for each warehouse. The available inventory is then allocated according to the allocation weight to obtain the coordinated order quantity.

[0050] S3.1: After determining that the total demand exceeds the available inventory, the central warehouse collaborative intelligent agent reads the single warehouse demand forecast corresponding to each front warehouse identifier from the unallocated data set, and retrieves the near-expiry inventory ratio data and historical stockout frequency data for each front warehouse respectively.

[0051] In this embodiment of the application, the calculation method for the near-expiry inventory ratio data is pre-completed by the demand forecasting agent of each front warehouse: obtain the inventory quantity of the target product in the current front warehouse with a remaining shelf life less than or equal to a preset shelf life threshold, divide the inventory quantity with a remaining shelf life less than or equal to the preset shelf life threshold by the total inventory quantity of the target product in the current front warehouse to obtain the near-expiry inventory ratio data.

[0052] It should be noted that the preset shelf life threshold is 20% to 30% of the shelf life days of the target product. Using 20% ​​to 30% of the shelf life days as the preset shelf life threshold is based on the average turnover days of goods from warehousing to sale in the front-end warehouse instant retail scenario. If it is lower than 20%, the threshold is too strict, and most of the products that are about to expire are not included in the near-expiration range, so the near-expiration ratio data cannot effectively reflect the actual risk of near-expiration backlog. If it is higher than 30%, the threshold is too lenient, and a large number of products that still have sufficient sales time are incorrectly marked as near-expiration, resulting in the distortion of the weighting calculation.

[0053] In this embodiment of the application, the historical out-of-stock data is the total number of days that the target product was out of stock in each pre-set historical time period in each pre-positioned warehouse. This data is obtained by the demand forecasting agent of each pre-positioned warehouse in S1 and sent to the central warehouse collaborative agent along with the single warehouse demand forecast.

[0054] S3.2: According to the preset weighting rules, calculate the allocation weight of each front warehouse based on the near-expiry inventory ratio data, historical stockout data, and single warehouse demand forecast. The allocation weight is used to represent the allocation ratio of each front warehouse in the allocation of allocable inventory.

[0055] In this embodiment of the application, the preset weighting rule is a weighted summation rule: multiply the near-expiry inventory ratio data of each front warehouse by a first weighting coefficient to obtain a first weighted value; multiply the historical out-of-stock count data of each front warehouse by a second weighting coefficient to obtain a second weighted value; multiply the single-warehouse demand forecast of each front warehouse by a third weighting coefficient to obtain a third weighted value; add the first weighted value, the second weighted value and the third weighted value to obtain the weighted total score of each front warehouse; divide the weighted total score of each front warehouse by the sum of the weighted total scores of all front warehouses to obtain the allocated weight of each front warehouse.

[0056] It should be noted that the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient are all preset positive numbers and their sum is equal to 1.

[0057] In this embodiment, the first weighting coefficient ranges from 0.25 to 0.35, the second weighting coefficient ranges from 0.20 to 0.30, and the third weighting coefficient ranges from 0.40 to 0.50. It should be noted that the third weighting coefficient has the highest value among the three because the predicted demand for a single warehouse directly reflects the actual demand scale of each front-end warehouse for the target commodity, and is the most crucial factor in allocation decisions. The first weighting coefficient is second highest because the higher the proportion of near-expiry inventory, the greater the risk of expired losses faced by the front-end warehouse. In the event of inventory shortages, priority should be given to ensuring the supply of such front-end warehouses to reduce loss costs. The second weighting coefficient is the lowest because the historical number of stockouts reflects the historical performance of the supply stability of the front-end warehouse, and should be used as a supplementary but not dominant reference factor in allocation decisions. The sum of the three equals 1 to ensure that the allocation weights are normalized to the range of 0 to 1.

[0058] S3.3: Allocate the allocable inventory according to the allocation weight of each front warehouse, that is, multiply the allocable inventory by the allocation weight of each front warehouse to obtain the initial allocation amount of each front warehouse.

[0059] S3.4: Perform upper limit constraint verification on the initial allocation quantity of each forward warehouse: compare the initial allocation quantity with the single warehouse demand forecast quantity of the forward warehouse. If the initial allocation quantity is greater than the single warehouse demand forecast quantity, the single warehouse demand forecast quantity is used as the final allocation quantity of the forward warehouse, and the excess inventory quantity is included in the inventory quantity to be redistributed; if the initial allocation quantity is less than or equal to the single warehouse demand forecast quantity, the initial allocation quantity is used as the final allocation quantity of the forward warehouse.

[0060] In this embodiment of the application, after the central warehouse collaborative intelligent agent completes the upper limit constraint verification of all front warehouses, it determines whether the inventory to be redistributed is greater than zero. If the inventory to be redistributed is greater than zero, the central warehouse collaborative intelligent agent redistributes the inventory to be redistributed according to the allocation weight of each front warehouse that has not reached the single warehouse demand forecast, until the inventory to be redistributed is fully allocated or all front warehouses have reached their respective single warehouse demand forecasts.

[0061] S3.5: Determine the final allocation quantity of each front warehouse after the upper limit constraint verification as the collaborative order quantity of each front warehouse, and store the collaborative order quantity with the corresponding front warehouse identifier.

[0062] S4: The demand forecasting agent rounds up the collaborative order quantity based on the collaborative order quantity and minimum packaging size of the current front warehouse to obtain the order quantity for each front warehouse.

[0063] S4.1: The demand forecasting agent reads the collaborative order quantity corresponding to its own front warehouse identifier from the central warehouse collaborative agent, and reads the minimum packaging specification using the commodity code of the target product as an index. The minimum packaging specification is the quantity corresponding to the smallest ordering unit of the target product in the procurement and distribution process.

[0064] In this embodiment of the application, before reading the collaborative order quantity, the demand forecasting agent also reads the comparison judgment result output by the central warehouse collaborative agent. If the comparison judgment result is that the inventory is sufficient, the demand forecasting agent directly reads the single warehouse demand forecast quantity as the collaborative order quantity; if the comparison judgment result is that the inventory is short, the demand forecasting agent reads the collaborative order quantity determined after the S3 allocation process and issues an inventory shortage warning.

[0065] In this embodiment, the method for inventory shortage warning is as follows: when the comparison judgment result is inventory shortage, the demand forecasting agent calculates the difference between the single warehouse demand forecast and the collaborative order quantity to obtain the inventory gap of each front warehouse; the demand forecasting agent divides the inventory gap by the single warehouse demand forecast to obtain the demand satisfaction gap ratio of each front warehouse; if the demand satisfaction gap ratio is greater than the preset gap ratio threshold, the demand forecasting agent generates inventory shortage warning information including front warehouse identifier, target product identifier, collaborative order quantity, single warehouse demand forecast, and inventory gap, and sends the inventory shortage warning information to the management terminal and the central warehouse collaborative agent.

[0066] It should be noted that the preset gap ratio threshold ranges from 10% to 20%. When the preset gap ratio threshold is below 10%, even a small inventory gap will trigger an alert, leading to an overload of alert information and making it difficult for managers to filter out the alerts that truly need attention. When the preset gap ratio threshold is above 20%, some abnormal situations with large inventory gaps that do not exceed the threshold may be missed, leading to an accumulation of stockout risks. Therefore, setting the preset gap ratio threshold to 10% to 20% can achieve a balance between alert sensitivity and management efficiency.

[0067] The inventory shortage warning information also includes the suggested emergency replenishment quantity calculated by the demand forecasting agent based on the collaborative order quantity. The suggested emergency replenishment quantity is calculated by rounding up the inventory shortage quantity according to the smallest packaging size, and the suggested emergency replenishment quantity is sent to the management terminal as part of the inventory shortage warning information.

[0068] S4.2: Divide the collaborative order quantity by the minimum packaging size to obtain the packaging quantity ratio; the packaging quantity ratio is used to characterize the packaging quantity corresponding to the minimum packaging unit after the collaborative order quantity is converted.

[0069] S4.3: Perform an up-rounding operation on the ratio of package quantities to obtain the rounded package quantities.

[0070] In this embodiment, the rounding up operation is as follows: when the ratio of the number of packages is an integer, the number of packages after rounding is equal to the ratio of the number of packages; when the ratio of the number of packages is not an integer, the number of packages after rounding is the smallest integer greater than the ratio of the number of packages.

[0071] S4.4: Multiply the rounded-down packaging quantity by the minimum packaging size to obtain the initial order quantity; where the initial order quantity is an integer multiple of the minimum packaging size that is greater than or equal to the collaborative order quantity.

[0072] In this embodiment of the application, a shelf life upper limit constraint verification is performed on the initial order quantity: the shelf life days of the target product and the upper limit of the number of days the target product can be sold in the current front warehouse are obtained, wherein the upper limit of the number of days ... the number of days the number of the number of days the number of the number of days the number of the number of the number of days the number of the number of the number of days the number of the number of the number of days the number of the number of the number of days the number of the number of the number of the number of days the number of the number of the number of days the number of the number of the number of days the number of the number of the number of days the number of the number of the number of days the number of the number of the number of days the number of the number of the number of days the number of the number of the number of days the number of the number of the number of days the number of the number of days the number of the number of days the number of the number of days the number of the number of days the number of the number of days the number of the number of days the number of the number of days the number of the number of days the number of the number of days the number of the number of days the number of the number of days the number of the number of days the number of the number of days the number of the number of days the number of the number of days the number of days the number of days the number of days the number of days the number of days the number of days the number of days the number of days the number of days the number of days the number of days the number of days the number of days the number of days the number of days the number of days the number of days the number

[0073] It should be noted that the shelf life limit constraint verification for the initial order quantity is because products that have exceeded their shelf life cannot be sold normally. If the order quantity is too large and the products cannot be sold within the shelf life, it will cause unnecessary loss costs. Therefore, by constraining the order quantity through the upper limit of the number of days that can be sold, it can effectively suppress the risk of over-ordering caused by batch rounding and subsequent near-expiry stockpiling.

[0074] In this embodiment, the demand forecasting agent also dynamically adjusts the upper limit of the number of days available for sale based on the lifecycle label: when the lifecycle label is the growth stage or maturity stage, the upper limit of the number of days available for sale is 80% of the shelf life days to fully meet the sales demand during the growth and stability stages; when the lifecycle label is the decline stage or obsolescence stage, the upper limit of the number of days available for sale is 50% of the shelf life days to avoid a large backlog of inventory of declining products.

[0075] S4.5: Determine the initial order quantity as the final order quantity for each front warehouse, and store the final order quantity in association with the front warehouse identifier and the target product identifier. At the same time, send the final order quantity to the central warehouse collaborative intelligent agent.

[0076] S5: The central warehouse collaborative intelligent agent generates a target product requisition plan for multiple front warehouses based on the front warehouse identifier, target product identifier, single warehouse demand forecast, collaborative requisition quantity, order quantity, and collaborative allocation information corresponding to each front warehouse.

[0077] S5.1: The central warehouse collaborative agent reads the total demand, allocable inventory, and inventory fulfillment status of the target product; reads the single warehouse demand forecast, collaborative order quantity, final order quantity, and allocation weight corresponding to each front warehouse identifier; sums up the final order quantities corresponding to all front warehouse identifiers to obtain the total order quantity; and determines whether the inventory fulfillment status is an inventory shortage status and whether the total order quantity is greater than the allocable inventory.

[0078] In this embodiment of the application, the determination of the inventory satisfaction status includes: when the inventory satisfaction status is an insufficient inventory status and the total order quantity is greater than the available inventory quantity, the total order quantity is subtracted from the available inventory quantity to obtain the excess allocation quantity; when the inventory satisfaction status is an adequate inventory status or the total order quantity is less than or equal to the available inventory quantity, the excess allocation quantity is determined to be zero.

[0079] The reason for the over-allocation is that after the demand forecasting agents of each front warehouse independently perform the rounding up operation according to their respective minimum packaging specifications, the final order quantity of each front warehouse is not less than its corresponding collaborative order quantity. However, the sum of the collaborative order quantities of each front warehouse is not greater than the allocable inventory. Therefore, the sum of the final order quantities of each front warehouse may exceed the allocable inventory. The over-allocation is the quantity corresponding to this excess.

[0080] S5.2: When the excess allocation is greater than zero, the central warehouse collaborative agent calculates the difference between the final order quantity of each front warehouse and its corresponding collaborative demand quantity, and sorts all front warehouse identifiers in descending order to obtain an excess allocation sorting sequence. Following the excess allocation sorting sequence, the agent subtracts the quantity corresponding to the smallest packaging specification from the final order quantity of the front warehouse identifier at the top of the sorting and updates it. This deduction operation is repeated until the cumulative deduction reaches the excess allocation or the difference between all front warehouse identifiers in the excess allocation sorting sequence is equal to zero. The final order quantity of each front warehouse after deduction is determined as the planned order quantity for each front warehouse. When the excess allocation is equal to zero, the final order quantity of each front warehouse is directly determined as the planned order quantity for each front warehouse.

[0081] It should be noted that the difference represents the redundant order quantity generated by each front warehouse due to rounding up. The larger the difference, the higher the surplus of the final order quantity of the front warehouse relative to its coordinated demand. The quantity corresponding to the smallest packaging size is deducted first from the front warehouse with the larger difference, so that the sum of the planned order quantities of all front warehouses does not exceed the available inventory, while reducing the impact on the replenishment guarantee of the front warehouse with the smaller difference.

[0082] S5.3: Determine the plan type label for the target product requisition plan based on the inventory satisfaction status, and mark the plan type label as the plan header information.

[0083] In this embodiment of the application, when the inventory satisfaction status is "sufficient inventory", the plan type label is "sufficient inventory plan"; when the inventory satisfaction status is "insufficient inventory", the plan type label is "inventory shortage allocation plan"; when the plan type label is "inventory shortage allocation plan", the plan header information also includes the inventory gap between the allocable inventory and the total demand, as well as the specific values ​​of the first weight coefficient, the second weight coefficient and the third weight coefficient used in the allocation weight calculation.

[0084] S5.4: Based on all front warehouse identifiers, assign a plan record to each front warehouse in the target product requisition plan, and fill in the front warehouse identifier, target product identifier, single warehouse demand forecast, collaborative requisition quantity, and requisition plan order quantity of each front warehouse into the corresponding fields.

[0085] It should be noted that the planning record includes the following fields: front warehouse identifier field, target product identifier field, single warehouse demand forecast field, collaborative order quantity field, order quantity field, and collaborative allocation information field; the target product identifier field includes the product code and product name of the target product, and the front warehouse identifier field includes the front warehouse number and name.

[0086] S5.5: The central warehouse collaborative agent calculates the rounded difference between the planned order quantity and the collaborative order quantity of each front warehouse, and at the same time calculates the demand difference between the collaborative order quantity and the single warehouse demand forecast quantity of each front warehouse, and fills the rounded difference into the collaborative allocation information field of the corresponding plan record.

[0087] In this embodiment of the application, when the demand difference is not zero, a label indicating the reason for the difference is marked in the collaborative allocation information field; otherwise, a label indicating that the demand is fully met is marked in the collaborative allocation information field; the rounded difference is equal to the planned order quantity minus the collaborative order quantity; the demand difference is equal to the collaborative order quantity minus the single warehouse demand forecast quantity.

[0088] The reasons for the difference include: when the demand difference is less than zero and the plan type label is an inventory sufficient plan, the reason for the difference is a rounding loss label; when the demand difference is less than zero and the plan type label is an inventory shortage allocation plan, the reason for the difference is an insufficient inventory allocation label; when the demand difference is greater than zero and the plan type label is an inventory shortage allocation plan, the reason for the difference is a minimum satisfaction rate compensation label.

[0089] S5.6: Fill in the allocation weight of each front warehouse into the collaborative allocation information field of the corresponding plan record.

[0090] In this embodiment of the application, when the plan type label is inventory shortage allocation plan, the allocation basis of the allocable inventory is marked in the collaborative allocation information field. The allocation basis includes the proportion of near-expiry inventory of each front warehouse, the number of historical stockouts, and the specific values ​​of the single warehouse demand forecast in the allocation weight calculation process.

[0091] The central warehouse collaborative agent generates a description text of the allocation basis in a preset format in the collaborative allocation information field: The proportion of near-expiry inventory in the i-th front warehouse is denoted as... And write the allocation basis description text, recording the historical out-of-stock count of the i-th front warehouse as . And write the allocation basis description text, and record the single warehouse demand forecast of the i-th front warehouse as . And write the allocation basis description text, and record the allocation weight of the i-th front warehouse as . i and write the allocation basis description text.

[0092] S5.7: Generate an initial requisition plan table based on all planning records, and perform an integrity check on the initial requisition plan table.

[0093] In this embodiment of the application, an integrity check is performed on the initial order plan table: it is checked whether the initial order plan table contains all the planned records of the front warehouses that have submitted single warehouse demand forecasts; when there is an omission, the missing front warehouse is marked as a data missing state, and a data missing label is added to the plan header information of the initial order plan table.

[0094] It should be noted that the initial order plan includes a plan header and a plan details record. The plan header includes the target product identifier, total demand, allocable inventory, inventory fulfillment status, and plan type label. The plan details record includes the plan records for each forward warehouse.

[0095] S5.8: Format the verified initial order plan according to the preset output format to generate the target product order plan.

[0096] It should be noted that the target product order plan includes the front warehouse identifier, target product identifier, single warehouse demand forecast, collaborative order quantity, order plan quantity, and collaborative allocation information for each front warehouse to perform subsequent replenishment operations based on the order plan quantity.

[0097] In this embodiment, the preset output format is a table format, which includes a header row and a data row. The header row includes a front warehouse identifier column, a target product identifier column, a single warehouse demand forecast column, a collaborative order quantity column, an order plan order quantity column, and a collaborative allocation information column. The target product order plan is sent to the demand forecasting intelligent agent of each front warehouse in the form of electronic data through the central warehouse collaborative intelligent agent, and is also sent to the management terminal in a printable format.

[0098] In this embodiment, the preset output format is a table format, which includes a header row and a data row. The header row includes a front warehouse identifier column, a target product identifier column, a single warehouse demand forecast column, a collaborative order quantity column, an order quantity column, and a collaborative allocation information column. The target product order plan is sent to the demand forecasting intelligent agent of each front warehouse in the form of electronic data through the central warehouse collaborative intelligent agent, and is also sent to the management terminal in a printable format.

[0099] Under the teachings of the above embodiments, such as Figure 2 As shown in the embodiments of the present invention, another aspect disclosed in these embodiments also proposes a multi-agent demand forecasting and order quantity generation system for replenishing forward warehouses, comprising: The demand forecasting module is used by the demand forecasting agents of each front warehouse to obtain historical sales data, inventory data, historical out-of-stock times, lifecycle tags, weather data, holiday data, delivery cycle, safety stock, and minimum packaging size of the target product, and generate a single warehouse demand forecast. The demand aggregation module is used by the central warehouse collaborative intelligent agent to obtain the identifiers and allocable inventory of all front warehouses that have submitted single-warehouse demand forecasts for the target product, sum up the single-warehouse demand forecasts of all front warehouses to obtain the aggregated demand of the target product, and determine whether the aggregated demand exceeds the allocable inventory. The collaborative allocation module is used to take the single-warehouse demand forecast of each front warehouse as the collaborative order quantity of the corresponding front warehouse when the total demand does not exceed the allocable inventory. Otherwise, it calculates the allocation weight based on the near-expiry inventory ratio, historical stockout frequency and single-warehouse demand forecast of each front warehouse, and allocates the allocable inventory according to the allocation weight to obtain the collaborative order quantity. The order rounding module is used by the demand forecasting agent of each front warehouse to round up the collaborative order quantity based on the collaborative order quantity and minimum packaging size of the front warehouse, so as to obtain the order quantity of each front warehouse. The plan generation module is used by the central warehouse collaborative intelligent agent to generate a target product requisition plan for multiple front warehouses based on the front warehouse identifier, target product identifier, single warehouse demand forecast, collaborative requisition quantity, order quantity, and collaborative allocation information.

[0100] This embodiment also provides a computer device applicable to the multi-agent demand forecasting and order quantity generation method for front-end warehouse replenishment, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the multi-agent demand forecasting and order quantity generation method for front-end warehouse replenishment as proposed in the above embodiment.

[0101] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0102] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the multi-agent demand forecasting and order quantity generation method for front-end warehouse replenishment as proposed in the above embodiments.

[0103] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0104] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A multi-agent demand forecasting and order quantity generation method for front warehouse replenishment, characterized in that, include: Each front-end warehouse demand forecasting agent acquires historical sales data, inventory data, historical out-of-stock counts, lifecycle tags, weather data, holiday data, delivery cycle, safety stock, and minimum packaging size of the target product, and generates a single warehouse demand forecast. The central warehouse collaborative intelligent agent obtains all front warehouse identifiers and allocable inventory for submitting the single warehouse demand forecast for the target product, sums up the single warehouse demand forecasts of all front warehouses to obtain the total demand for the target product, and determines whether the total demand exceeds the allocable inventory. When the total demand exceeds the available inventory, the allocation weight is calculated based on the near-expiry inventory ratio of each front warehouse, the number of historical stockouts, and the single warehouse demand forecast. The available inventory is then allocated according to the allocation weight to obtain the coordinated order quantity. The demand forecasting agent rounds up the collaborative order quantity based on the collaborative order quantity and the minimum packaging size of the front warehouse to obtain the order quantity of each front warehouse. The central warehouse collaborative intelligent agent generates a target product requisition plan for multiple front warehouses based on the front warehouse identifier, target product identifier, single warehouse demand forecast, collaborative order quantity, order quantity, and collaborative allocation information.

2. The multi-agent demand forecasting and order generation method for front stock replenishment according to claim 1, wherein, The method for generating the target commodity order plan includes: The central warehouse collaborative intelligent agent reads the total demand, allocable inventory, and inventory satisfaction status of the target product, and reads the single warehouse demand forecast, collaborative order quantity, final order quantity, and allocation weight corresponding to each front warehouse identifier; it accumulates the final order quantities corresponding to all front warehouse identifiers to obtain the total order quantity, and determines whether the inventory satisfaction status is an inventory shortage status and whether the total order quantity is greater than the allocable inventory. Based on the inventory satisfaction status, determine the plan type label for the target product requisition plan, and mark the plan type label as the plan header information; Based on all the front warehouse identifiers, a plan record is assigned to each front warehouse in the target commodity ordering plan, and the front warehouse identifier, the target commodity identifier, the single warehouse demand forecast, the collaborative ordering quantity, and the ordering plan ordering quantity of each front warehouse are filled into the corresponding fields respectively. The central warehouse collaborative agent calculates the rounded difference between the planned order quantity and the collaborative order quantity of each front warehouse, and at the same time calculates the demand difference between the collaborative order quantity and the single warehouse demand forecast quantity of each front warehouse, and fills the rounded difference into the collaborative allocation information field of the corresponding plan record. Fill the allocation weight of each front warehouse into the collaborative allocation information field of the corresponding plan record; An initial ordering plan is generated based on all the planned records, and an integrity check is performed on the initial ordering plan. The verified initial order plan is formatted according to the preset output format to generate the target product order plan. 3.The method of claim 2, wherein, The determination of the inventory satisfaction status includes: When the inventory is in a state of insufficient inventory and the total order quantity is greater than the available inventory quantity, the total order quantity is subtracted from the available inventory quantity to obtain the excess allocation quantity. When the inventory is in a sufficient inventory state or the total order quantity is less than or equal to the available inventory quantity, the excess allocation quantity is determined to be zero. 4.The method of claim 2 or 3, wherein, The method for determining the order quantity in the demand plan includes: When the over-allocation amount is greater than zero, the central warehouse collaborative agent calculates the difference between the final order amount of each front warehouse and its corresponding collaborative demand amount, and sorts all front warehouse identifiers in descending order to obtain an over-allocation sorting sequence. Following the order of the over-allocation sorting sequence, the final order amount corresponding to the top-ranked front warehouse identifier is subtracted from the quantity corresponding to the minimum packaging specification, and the result is updated. This deduction operation is repeated until the cumulative deduction amount reaches the over-allocation amount or the difference between all front warehouse identifiers in the over-allocation sorting sequence is equal to zero. The final order amount of each front warehouse after the deduction process is determined as the planned order amount for each front warehouse. When the excess allocation is zero, the final order quantity of each forward warehouse is directly determined as the planned order quantity of each forward warehouse.

5. The multi-agent demand forecasting and order generation method for front stock replenishment of claim 2, wherein, The determination of the final order quantity includes: The demand forecasting agent reads the collaborative order quantity corresponding to its own front warehouse identifier from the central warehouse collaborative agent, and reads the minimum packaging specification using the product code of the target product as an index; Divide the coordinated order quantity by the minimum packaging size to obtain the packaging quantity ratio; wherein the packaging quantity ratio is used to characterize the packaging quantity corresponding to the minimum packaging unit after the coordinated order quantity is converted; The ratio of the package quantities is rounded up to obtain the rounded package quantities. Multiply the rounded packaging quantity by the minimum packaging size to obtain the initial order quantity; The initial order quantity is determined as the final order quantity for each front warehouse, and the final order quantity is associated with and stored with the front warehouse identifier and the target product identifier. At the same time, the final order quantity is sent to the central warehouse collaborative intelligent agent.

6. The multi-agent demand forecasting and order generation method for front stock replenishment of claim 5, wherein, The acquisition of the coordinated order quantity includes: Read the single warehouse demand forecast corresponding to each front warehouse identifier from the dataset to be allocated, and retrieve the near-expiry inventory ratio data and historical out-of-stock count data for each front warehouse respectively. According to the preset weighting rules, the allocation weight of each front warehouse is calculated based on the near-expiry inventory ratio data, the historical out-of-stock times data, and the single warehouse demand forecast. The allocable inventory is allocated according to the allocation weight of each front warehouse, that is, the allocable inventory is multiplied by the allocation weight of each front warehouse to obtain the initial allocation amount of each front warehouse. Perform an upper limit constraint check on the initial allocation amount for each forward warehouse; The final allocation quantity of each front warehouse after passing the upper limit constraint verification is determined as the collaborative order quantity of each front warehouse, and the collaborative order quantity is associated with and stored with the corresponding front warehouse identifier.

7. The multi-agent demand forecasting and order generation method for front stock replenishment of claim 6, wherein, Perform an upper limit constraint check on the initial allocation amount for each forward warehouse, including: The initial allocation amount is compared with the single-warehouse demand forecast amount of the front warehouse. If the initial allocation amount is greater than the single-warehouse demand forecast amount, the single-warehouse demand forecast amount is used as the final allocation amount for the front warehouse, and the excess inventory amount is included in the inventory to be redistributed. If the initial allocation amount is less than or equal to the single-warehouse demand forecast amount, the initial allocation amount is used as the final allocation amount for the front warehouse.

8. A multi-agent demand forecasting and order quantity generation system for front warehouse replenishment, based on any one of the multi-agent demand forecasting and order quantity generation methods for front warehouse replenishment according to claims 1-7, characterized in that, include: The demand forecasting module is used by the demand forecasting agents of each front warehouse to obtain historical sales data, inventory data, historical out-of-stock times, lifecycle tags, weather data, holiday data, delivery cycle, safety stock, and minimum packaging size of the target product, and generate a single warehouse demand forecast. The demand aggregation module is used by the central warehouse collaborative intelligent agent to obtain the identifiers and allocable inventory of all front warehouses that have submitted the single warehouse demand forecast for the target product, to accumulate the single warehouse demand forecast of all front warehouses to obtain the aggregated demand of the target product, and to determine whether the aggregated demand exceeds the allocable inventory. The collaborative allocation module is used to take the single-warehouse demand forecast of each front warehouse as the collaborative order quantity of the corresponding front warehouse when the total demand quantity does not exceed the allocable inventory quantity; otherwise, it calculates the allocation weight based on the near-expiry inventory ratio, historical stockout frequency and single-warehouse demand forecast of each front warehouse, and allocates the allocable inventory quantity according to the allocation weight to obtain the collaborative order quantity. The order rounding module is used by the demand forecasting agent of each front warehouse to round up the collaborative order quantity based on the collaborative order quantity and the minimum packaging specification of the front warehouse, so as to obtain the order quantity of each front warehouse. The plan generation module is used by the central warehouse collaborative intelligent agent to generate a target product requisition plan for multiple front warehouses based on the front warehouse identifier, target product identifier, single warehouse demand forecast, collaborative requisition quantity, order quantity, and collaborative allocation information. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: When the processor executes the computer program, it implements the steps of the multi-agent demand forecasting and order quantity generation method for front-end warehouse replenishment as described in any one of claims 1 to 7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: When the computer program is executed by the processor, it implements the steps of the multi-agent demand forecasting and order quantity generation method for front-end warehouse replenishment as described in any one of claims 1 to 7.