AI-based cloud warehousing intelligent inventory forecasting and dynamic allocation system

Abstract

This invention relates to the field of cloud warehouse intelligent management technology, specifically to an AI-based cloud warehouse intelligent inventory prediction and dynamic allocation system. This system achieves second-level synchronization across all warehouse terminals via IoT sensors, and extrapolates inventory demand at the warehouse terminals based on a full-link synchronization model. This, combined with an inventory demand prediction model built using an LSTM neural network, forms a two-dimensional demand verification, accurately capturing inventory gaps and demand. When allocation needs exist, the system uses a rule engine to transform multiple manually defined dynamic allocation rules into executable instructions. Multi-objective optimization verification is performed in a full-link simulation environment, deeply analyzing allocation paths and weather constraints at the warehouse terminals. By comprehensively considering virtual paths and actual transportation, the system selects the allocation strategy with the best overall effect, breaking through the experience-based decision-making bottleneck of traditional cloud warehousing and significantly enhancing the agility, risk resistance, and sustainable development level of cloud warehouse management.

Description

Technical Field

[0001] This invention relates to the field of cloud warehouse intelligent management technology, and more specifically, to an AI-based cloud warehouse intelligent inventory forecasting and dynamic allocation system. Background Technology

[0002] As a core node in the modern supply chain, cloud warehousing's inventory management and allocation efficiency directly impacts enterprise costs, customer experience, and sustainable development capabilities. Traditional cloud warehousing management models rely primarily on manual experience and outdated system records, making it difficult to cope with high-frequency fluctuations in market demand and complex, ever-changing supply chain environments.

[0003] With the development of IoT, AI, and digital twin technologies, some companies have attempted to collect warehouse data through sensors, but this only achieves physical device networking and fails to build a data-driven intelligent decision-making closed loop. Existing solutions generally suffer from two major technical bottlenecks: first, the accuracy of multi-source data (inventory, weather, traffic) fusion is insufficient, making it impossible to accurately map warehouse status; second, the allocation route optimization does not deeply integrate environmental factors, resulting in poor strategy robustness.

[0004] Therefore, how to build an intelligent system covering the entire process of "data collection, intelligent prediction, dynamic allocation, and simulation verification" to achieve accurate capture of inventory demand, environmental adaptation of allocation strategies, and balanced optimization of cost, timeliness, and carbon emissions across the entire chain has become the core technical challenge for the intelligent upgrade of cloud warehousing. Summary of the Invention

[0005] In view of the shortcomings of existing technologies, the purpose of this invention is to provide an AI-based cloud warehousing intelligent inventory forecasting and dynamic allocation system.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The AI-based cloud warehousing intelligent inventory forecasting and dynamic allocation system includes a full-link integrated system building unit for building a cloud warehousing full-link integrated system.

[0008] The warehouse inventory forecasting unit periodically determines the actual inventory demand for each warehouse in the cloud warehouse, and further determines the forecasted inventory demand for each warehouse.

[0009] The inventory transfer initiation unit controls the cloud warehousing full-link integration system to perform a simulation for a duration of H. After the simulation ends, it obtains the actual inventory demand projection of each warehouse in the cloud warehousing full-link integration system, and then determines the dynamic transfer demand value of the cloud warehouse. Based on the comparison result between the dynamic transfer demand value and the demand threshold, it determines whether to initiate the inventory transfer step.

[0010] After determining to initiate the inventory transfer step, the inventory transfer execution unit manually formulates multiple dynamic transfer rules, converts each dynamic transfer rule into an executable command and imports it into the cloud warehousing end-to-end integrated system, further determines the transfer feedback value of each dynamic transfer rule, selects the transfer execution rule from all dynamic transfer rules, and performs the inventory transfer of the cloud warehouse according to the transfer execution rule.

[0011] Furthermore, the method for determining the inventory demand forecast for the warehouse end is as follows: identify a warehouse end, obtain the inventory time series set of the warehouse end, import the inventory time series set into the inventory demand forecast model corresponding to the warehouse end, and the inventory demand forecast model derives the inventory demand forecast for the warehouse end.

[0012] To obtain the inventory time series set at the warehouse end, the steps are as follows: Obtain the actual inventory demand of a warehouse end within the previous t time period, and integrate the actual inventory demand within the t time period into an inventory time series set.

[0013] Furthermore, the method for determining the dynamic allocation demand value of cloud warehousing is as follows: obtain the projected inventory fulfillment quantity for each warehousing end, sum and average the projected inventory fulfillment quantities for each warehousing end to calculate the average projected inventory fulfillment quantity, compare each warehousing end pairwise, calculate the absolute difference between the projected inventory fulfillment quantities of the two compared warehousing ends to calculate the projected fulfillment gap, sum and average all projected fulfillment gaps to calculate the average projected fulfillment gap, and calculate the ratio between the average projected fulfillment gap and the average projected inventory fulfillment quantity to calculate the dynamic allocation demand value of cloud warehousing.

[0014] Furthermore, the method for obtaining the predicted inventory fulfillment quantity at the warehousing end is as follows: Select a warehousing end, calculate the difference between the cloud warehousing inventory demand forecast quantity and the actual inventory demand projection quantity at that warehousing end, and calculate the predicted inventory fulfillment quantity at that warehousing end.

[0015] Furthermore, the method for determining the allocation feedback value of the dynamic allocation rule is as follows: After a dynamic allocation rule is converted into an executable command and imported into the cloud warehousing full-link fusion system, the cloud warehousing full-link fusion system performs a simulation of duration H. During the simulation, for each allocation path generated, the meteorological impact value of the path corresponding to the allocation path is determined, and the starting and ending warehouse ends of each allocation path are determined simultaneously. After all allocation paths are generated, all allocation paths are sorted in the order of their generation to generate a path sequence. The allocation restriction value of each warehouse end is further determined, and the allocation restriction value of each warehouse end is summed and averaged to calculate the average allocation restriction value. After the simulation ends, an allocation basic dataset is generated, and the allocation basic dataset is imported into the allocation basic quantification model. The allocation basic quantification model derives the allocation basic quantification value, and the dynamic allocation demand value of the cloud warehouse is obtained simultaneously. Based on the average allocation restriction value, the allocation basic quantification value, and the dynamic allocation demand value, the allocation feedback value of the dynamic allocation rule is calculated.

[0016] Furthermore, the method for determining the allocation restriction value at the storage end is as follows: Select a storage end, mark all allocation paths in the path sequence that contain this storage end as associated paths, sort all associated paths in the order they were generated, match every two adjacent associated paths after sorting as a storage associated path group, obtain the associated restriction value of each storage associated path group, sum the associated restriction values ​​of all storage associated path groups and take the average value to calculate the allocation restriction value of the storage end.

[0017] Furthermore, the method for obtaining the associated limitation value of the warehouse associated path group is as follows: the path meteorological impact values ​​of two associated paths in the warehouse associated path group are summed to calculate the meteorological impact sum value; the absolute difference between the path meteorological impact values ​​of the two associated paths is calculated to calculate the associated limitation gap value; and the ratio between the meteorological impact sum value and the associated limitation gap value is calculated to obtain the associated limitation value of the warehouse associated path group.

[0018] Furthermore, the method for determining the path meteorological impact value of the allocation route is as follows: obtain the predicted meteorological data on the allocation route, import the predicted meteorological data into the meteorological evaluation model, and derive the path meteorological impact value of the allocation route from the meteorological evaluation model.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] The system of this invention, through full-domain coverage by IoT sensors, can achieve second-level synchronization of all warehouse terminals in cloud warehousing. It extrapolates the inventory demand of the warehouse terminals based on the end-to-end synchronization model, forming a two-dimensional demand verification with the inventory demand prediction model built by LSTM neural network. This accurately captures inventory gaps and inventory demands. When there is a transfer requirement, the system uses a rule engine to transform multiple sets of manually defined dynamic transfer rules into executable instructions. Multi-objective optimization verification (balancing cost, timeliness, and carbon emissions) is performed in the end-to-end simulation environment. The system deeply analyzes the transfer path and the weather constraints of the warehouse terminals, and selects the transfer strategy with the best overall effect by combining virtual paths and actual transportation. This breaks through the experience-based decision-making bottleneck of traditional cloud warehousing, and significantly enhances the agility, risk resistance, and sustainable development level of cloud warehouse management. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the principle of an AI-based cloud warehousing intelligent inventory forecasting and dynamic allocation system.

[0022] Figure 2 This is a flowchart illustrating the operation of an AI-based cloud warehousing intelligent inventory forecasting and dynamic allocation system.

[0023] Figure 3 This is a flowchart of the construction process for a cloud warehousing end-to-end integrated system. Detailed Implementation

[0024] Reference Figures 1 to 3 The AI-based cloud warehousing intelligent inventory forecasting and dynamic allocation system includes a full-link integrated system construction unit, a warehouse inventory forecasting unit, an inventory allocation initiation unit, and an inventory allocation execution unit.

[0025] The end-to-end integrated system building unit constructs a cloud warehousing end-to-end integrated system. First, it performs physical device networking, business system integration, and external data integration. Physical device networking includes the deployment of sensors at each warehouse and transportation end (sensors need to be deployed at each warehouse end corresponding to the cloud warehousing system). An edge data gateway is built using Node-RED, supporting protocols such as MQTT and OPCUA. Sensors required for deployment at the warehouse end include RFID readers (at shelf entrances and exits to collect inventory inbound and outbound data), smart meters (deployed in different areas to collect lighting and equipment energy consumption data, accuracy ±1%), and temperature and humidity sensors (e.g., one sensor per 50㎡ in a fresh produce warehouse, accuracy ±0.5℃). Sensors required for deployment at the transportation end include GPS... + Beidou dual-mode positioning (real-time vehicle location, error ≤ 5 meters), OBD interface (collecting vehicle fuel consumption and mileage, converting carbon emissions), etc.; the business systems connected include ERP / WMS (for obtaining inventory ledgers, order batches, and supplier information), TMS (for obtaining transportation orders, logistics provider quotations, and route planning data), carbon footprint platform (for connecting to third-party data services (such as CarbonTrust) and obtaining a carbon factor library for transportation modes (such as 0.015 kg CO2 / km·ton for railways)); the integrated external data includes weather API (for obtaining regional precipitation and wind speed data (affecting road transportation timeliness, delay coefficient ±10-30%)), and policy data (for accessing carbon emission trading market prices (such as 1 kg CO2 = 50 yuan, used for cost-carbon emission conversion)); constructing 3D physical models for each warehouse end, including geometric modeling (constructing a precise digital mirror of the warehouse space, such as using LiDAR scanning to obtain warehouse point cloud data, and using Blender / B IM (Revit) is used to construct the warehouse structure (shelves, aisles, loading docks) and equipment models (AGVs, stacker cranes). Unity's ProBuilder plugin is used to simplify the number of model faces (≤100,000 triangles) to ensure a browser rendering frame rate of ≥30fps. Attribute tags are added to each model, such as shelf ID, AGV number, and carbon emission factor. Dynamic attribute binding is also used (associating the 3D model with the real-time status, business data, and calculated attributes of physical entities. Real-time status data items include inventory quantity, equipment operating status (operating / faulty), location coordinates, energy consumption, temperature, and humidity. Business data items include location SKU information, order priority, equipment model, maintenance cycle, and operator ID. Calculated attribute data items include inventory turnover rate, carbon emissions, energy consumption, equipment utilization rate, and order response time). Finally, a cloud warehousing end-to-end integrated system is built (the cloud warehousing end-to-end integrated system is updated in real time to ensure that its functionality, performance, security, and business adaptability are always consistent with the real cloud warehousing situation).

[0026] The warehouse inventory forecasting unit periodically determines the actual inventory demand for each warehouse in the cloud warehouse (the time interval is determined according to the inventory management needs of the cloud warehouse; the actual inventory demand is the actual inventory demand of the warehouse based on the ongoing business activities). It further determines the inventory demand forecast for each warehouse (the inventory demand forecast is the actual inventory demand of the warehouse after H hours (which can be 1 hour or 2 hours).

[0027] The method for determining the inventory demand forecast for the warehouse end is as follows: identify a warehouse end, obtain the inventory time series set of the warehouse end, import the inventory time series set into the inventory demand forecast model corresponding to the warehouse end, and derive the inventory demand forecast quantity for the warehouse end from the inventory demand forecast model.

[0028] To obtain the inventory time series set at the warehouse end, the steps are as follows: Obtain the actual inventory demand of a warehouse end within the previous t time period, and integrate the actual inventory demand within the t time period into an inventory time series set.

[0029] Each warehouse has an independent inventory demand forecasting model. All inventory demand forecasting models are built based on LSTM models. This embodiment takes warehouse A as an example and discloses the process of building the inventory demand forecasting model: Multiple inventory time series sets are collected from warehouse A, an LSTM model is built, and the built LSTM model is trained using the inventory time series sets as the basic data. During this process, each inventory time series set is assigned an inventory demand forecast quantity, which is the predicted actual inventory demand of warehouse A after time H. Then, the multiple inventory time series sets are divided into training, validation, and test sets according to a specific ratio of 70%:15%:15%. The training set accounts for 70% to ensure the model has sufficient data to learn patterns, and the validation and test sets... Each set accounts for 15%, balancing parameter tuning and generalization assessment. For time-series data, it needs to be divided according to time order (e.g., January-July 2023 for training, August-September for validation, and October-December for test) to avoid future information leakage. The validation set is used to evaluate model performance (e.g., MSE, MAE) after each training round to avoid overfitting. Hyperparameter tuning involves adjusting parameters such as the number of LSTM layers, number of neurons, learning rate, and sequence length to find the optimal combination. Overfitting prevention includes adding dropout layers (e.g., adding 20% ​​dropout after the LSTM layer) and early stopping mechanisms. Using test set data that was not used in training, the generalization performance index of the model is calculated. Finally, the inventory demand prediction model for warehouse A is completed.

[0030] The inventory transfer initiation unit controls the cloud warehousing end-to-end integrated system to perform a simulation for a duration of H. After the simulation, it obtains the actual inventory demand projection of each warehouse in the cloud warehousing end-to-end integrated system (i.e., the cloud warehousing end-to-end integrated system performs a time-dimensional simulation according to conventional transfer, replenishment, and delivery strategies, and analyzes the actual inventory demand of each warehouse in the cloud warehousing end-to-end integrated system after H hours under conventional strategies). Then, it determines the dynamic transfer demand value of the cloud warehouse. When the dynamic transfer demand value of the cloud warehouse is higher than the demand threshold (the demand threshold is the key trigger condition for initiating inventory transfer), it determines to initiate the inventory transfer step.

[0031] The method for determining the dynamic allocation demand value of cloud warehousing is as follows: Obtain the projected inventory fulfillment amount for each warehouse, sum and average the projected inventory fulfillment amounts for each warehouse to calculate the average projected inventory fulfillment amount (if the calculated average projected inventory fulfillment amount is 0, adjust the value of the average projected inventory fulfillment amount to 0.05), compare each warehouse pairwise, calculate the absolute difference between the projected inventory fulfillment amounts of the two compared warehouses to calculate the projected fulfillment gap, sum and average all projected fulfillment gaps to calculate the average projected fulfillment gap, and calculate the ratio of the average projected fulfillment gap to the average projected inventory fulfillment amount to calculate the dynamic allocation demand value of cloud warehousing.

[0032] How to obtain the predicted inventory demand fulfillment amount for the warehouse end: Select a warehouse end, calculate the difference between the cloud warehouse inventory demand forecast and the actual inventory demand forecast for that warehouse end, and calculate the predicted inventory demand fulfillment amount for that warehouse end (the larger the predicted inventory demand fulfillment amount, the stronger the coverage of the actual inventory demand under the current rules).

[0033] After determining to initiate the inventory transfer step, the inventory transfer execution unit manually formulates multiple dynamic transfer rules, transforms each dynamic transfer rule into an executable command, and imports it into the cloud warehousing end-to-end integrated system. (The methods for transforming dynamic transfer rules into executable commands include rule element decomposition and digital representation of rules (using structured languages ​​(JSON / XML) or decision tables). The rules are defined using a table, and the rule engine is used to parse them (converting natural language rules into executable logic). Instruction format standardization (generating a system-recognizable instruction set, typically including: transfer orders (containing outgoing and incoming warehouses, SKUs, quantities, transportation methods, etc.); route planning instructions (transmitting transportation routes and capacity allocation to the TMS); and equipment scheduling instructions (notifying the WMS to schedule AGVs, sorting robots, etc.). Rule elements are broken down into conditions, actions, parameters, and constraints. For example: Condition: Warehouse inventory ≤ safety line and adjacent warehouse inventory ≥ transfer amount; Action: Transfer a specified SKU from an adjacent warehouse; Parameter: Transfer amount = safety line - current inventory; Constraint: Single transfer cost ≤ 500 yuan, timeliness ≤ 12 hours). Further, the transfer feedback value for each dynamic transfer rule is determined, and the dynamic transfer rule with the largest transfer feedback value is marked as the transfer execution rule. Inventory transfers are then performed on the cloud warehouse according to the transfer execution rule.

[0034] The method for determining the allocation feedback value of dynamic allocation rules: After a dynamic allocation rule is converted into an executable command and imported into the cloud warehousing end-to-end fusion system, the cloud warehousing end-to-end fusion system performs a simulation of duration H. During the simulation, for each allocation path generated (from warehouse A to warehouse D, i.e., an allocation path is generated), the meteorological impact value corresponding to the allocation path is determined, and the starting and ending warehouses of each allocation path are determined simultaneously. After all allocation paths are generated, all allocation paths are sorted in the order of their generation (according to the sorting order) to generate a path sequence, which is then further... Determine the allocation restriction value for each warehouse, sum and average these values ​​to calculate the average allocation restriction value, labeled as EnC. After the simulation, generate an allocation base dataset (containing transportation costs, warehousing costs, and total carbon emissions of the cloud warehousing end-to-end integrated system throughout the simulation). Import the allocation base dataset into the allocation base quantification model. The allocation base quantification model derives the allocation base quantification value, labeled as Dsp. Simultaneously, obtain the dynamic allocation demand value of the cloud warehouse after the simulation, labeled as Wzr, and use the formula... Calculate the allocation feedback value of the dynamic allocation rule; ag1 represents the first correction coefficient (which can be 1.15), and ag2 represents the second correction coefficient (which can be 1.08).

[0035] The method for determining the allocation restriction value at the warehouse end is as follows: Select a warehouse end, and mark all allocation paths in the path sequence that contain this warehouse end as associated paths (regardless of whether the warehouse end is the starting warehouse end or the terminal warehouse end in the allocation path, it is marked as an associated path). Sort all associated paths in the order they were generated, and match every two adjacent associated paths after sorting into a warehouse associated path group. Obtain the associated restriction value of each warehouse associated path group, sum the associated restriction values ​​of all warehouse associated path groups and take the average value to calculate the allocation restriction value of the warehouse end.

[0036] The method for obtaining the associated constraint value of the warehouse associated path group is as follows: Sum the meteorological impact values ​​of the two associated paths in the warehouse associated path group to calculate the meteorological impact sum. Calculate the absolute difference between the meteorological impact values ​​of the two associated paths to calculate the associated constraint gap value (when the calculated associated constraint gap value is 0, adjust the value of the associated constraint gap value to 0.1). Calculate the ratio between the meteorological impact sum value and the associated constraint gap value to obtain the associated constraint value of the warehouse associated path group.

[0037] The method for determining the path meteorological impact value of the allocation path is as follows: (based on the meteorological system) obtain the predicted meteorological data for the allocation path (the predicted meteorological data is the meteorological data of the allocation path within the H-hour period predicted by the meteorological system, including precipitation, visibility, and wind speed), import the predicted meteorological data into the meteorological evaluation model, and the meteorological evaluation model derives the path meteorological impact value of the allocation path.

[0038] The meteorological evaluation model is built by: constructing a neural network model, collecting multiple forecast meteorological data, using the forecast meteorological data as the basic data, and training the constructed neural network model. In this process, each forecast meteorological data is assigned a path meteorological impact value, which is set between 1 and 20. The larger the value, the greater the impact of the forecast meteorological data on the allocation path. For example, the greater the precipitation, the greater the driving danger on the allocation path, and the greater the impact on the allocation path. Then, the multiple forecast meteorological data are divided into training set, validation set and test set according to a specific ratio, which is determined to be 70%:15%:15%. The meteorological evaluation model is completed through neural network training.

[0039] The method for building the basic quantification model of allocation is as follows: A neural network model is built, and multiple basic allocation datasets are collected. Using these datasets as the base data, the built neural network model is trained. During this process, each basic allocation dataset is assigned a basic allocation quantification value, with the value ranging from 1 to 100. The larger the value, the better the basic allocation effect of the dataset, such as lower transportation costs, lower warehousing costs, and lower total carbon emissions. Then, the multiple basic allocation quantification values ​​are divided into training, validation, and test sets according to a specific ratio of 60%:20%:20%. Weighted mean square error (WMSE) is used, and high-risk samples (index > 80) are given higher weights to optimize the model structure, making it more accurate and stable. Finally, the basic allocation quantification model is completed.

[0040] Based on the aforementioned system, through full-domain coverage by IoT sensors, second-level synchronization of all warehouse terminals in the cloud warehouse can be achieved. The system then extrapolates inventory demand at the warehouse terminals based on a full-link synchronization model, forming a two-dimensional demand verification with an inventory demand prediction model built using an LSTM neural network. This accurately captures inventory gaps and demand. When there is a need for allocation, the system uses a rule engine to transform multiple manually defined dynamic allocation rules into executable instructions. Multi-objective optimization verification (balancing cost, timeliness, and carbon emissions) is performed in a full-link simulation environment. The system deeply analyzes the allocation path and weather constraints at the warehouse terminal, comprehensively selecting the allocation strategy with the best overall effect by integrating virtual paths and actual transportation. This breaks through the experience-based decision-making bottleneck of traditional cloud warehousing, significantly enhancing the agility, risk resistance, and sustainable development level of cloud warehouse management.

[0041] The above formulas are all dimensionless calculations, and the preset parameters in the formulas should be set by those skilled in the art according to the actual situation.

[0042] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0043] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0044] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0045] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0046] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0047] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An AI-based cloud warehousing intelligent inventory forecasting and dynamic allocation system, characterized in that, Includes end-to-end integrated system building units, used to build a cloud warehouse end-to-end integrated system; The warehouse inventory forecasting unit periodically determines the actual inventory demand for each warehouse in the cloud warehouse, and further determines the forecasted inventory demand for each warehouse. The inventory transfer initiation unit controls the cloud warehousing full-link integration system to perform a simulation for a duration of H. After the simulation ends, it obtains the actual inventory demand projection of each warehouse in the cloud warehousing full-link integration system, and then determines the dynamic transfer demand value of the cloud warehouse. Based on the comparison result between the dynamic transfer demand value and the demand threshold, it determines whether to initiate the inventory transfer step. The method for determining the dynamic allocation demand value of cloud warehousing is as follows: Obtain the predicted inventory projection fulfillment amount for each warehouse, sum the predicted inventory projection fulfillment amounts for each warehouse and take the average value to calculate the average predicted inventory projection fulfillment amount, compare each warehouse pairwise, calculate the absolute difference between the predicted inventory projection fulfillment amounts of the two compared warehouses to calculate the projection fulfillment gap amount, sum all projection fulfillment gap amounts and take the average value to calculate the average projection fulfillment gap amount, and calculate the ratio between the average projection fulfillment gap amount and the average predicted inventory projection fulfillment amount to calculate the dynamic allocation demand value of cloud warehousing. How to obtain the predicted inventory fulfillment quantity for the warehouse end: Select a warehouse end, calculate the difference between the cloud warehouse inventory demand forecast quantity and the actual inventory demand projection quantity for that warehouse end, and calculate the predicted inventory fulfillment quantity for that warehouse end. After determining to initiate the inventory transfer step, the inventory transfer execution unit manually formulates multiple dynamic transfer rules, converts each dynamic transfer rule into an executable command and imports it into the cloud warehousing end-to-end integrated system, further determines the transfer feedback value of each dynamic transfer rule, selects the transfer execution rule from all dynamic transfer rules, and performs the inventory transfer of the cloud warehouse according to the transfer execution rule.

2. The AI-based cloud warehousing intelligent inventory forecasting and dynamic allocation system according to claim 1, characterized in that, How to determine the inventory demand forecast for the warehouse: Identify a warehouse, obtain the inventory time series set of the warehouse, import the inventory time series set into the inventory demand forecast model corresponding to the warehouse, and derive the inventory demand forecast for the warehouse from the inventory demand forecast model. To obtain the inventory time series set at the warehouse end, the steps are as follows: Obtain the actual inventory demand of a warehouse end within the previous t time period, and integrate the actual inventory demand within the t time period into an inventory time series set.

3. The AI-based cloud warehousing intelligent inventory forecasting and dynamic allocation system according to claim 1, characterized in that, The method for determining the allocation feedback value of dynamic allocation rules is as follows: After a dynamic allocation rule is converted into an executable command and imported into the cloud warehousing full-link fusion system, the cloud warehousing full-link fusion system performs a simulation of duration H. During the simulation, for each allocation path generated, the meteorological impact value of the path corresponding to the allocation path is determined, and the starting and ending warehouse ends of each allocation path are determined simultaneously. After all allocation paths are generated, all allocation paths are sorted in the order of their generation to generate a path sequence. The allocation restriction value of each warehouse end is further determined, and the allocation restriction value of each warehouse end is summed and averaged to calculate the average allocation restriction value. After the simulation ends, an allocation basic dataset is generated, and the allocation basic dataset is imported into the allocation basic quantification model. The allocation basic quantification model derives the allocation basic quantification value, and the dynamic allocation demand value of the cloud warehouse is obtained simultaneously. Based on the average allocation restriction value, the allocation basic quantification value, and the dynamic allocation demand value, the allocation feedback value of the dynamic allocation rule is calculated.

4. The AI-based cloud warehousing intelligent inventory forecasting and dynamic allocation system according to claim 3, characterized in that, Method for determining the allocation restriction value of the storage end: Select a storage end, mark all allocation paths in the path sequence that contain the storage end as associated paths, sort all associated paths in the order of their generation, match every two adjacent associated paths after sorting as storage associated path groups, obtain the associated restriction value of each storage associated path group, sum the associated restriction values ​​of all storage associated path groups and take the average value to calculate the allocation restriction value of the storage end.

5. The AI-based cloud warehousing intelligent inventory forecasting and dynamic allocation system according to claim 4, characterized in that, The method for obtaining the associated limit value of the warehouse associated path group is as follows: the path meteorological impact values ​​of the two associated paths in the warehouse associated path group are summed to calculate the meteorological impact sum value. The absolute difference between the path meteorological impact values ​​of the two associated paths is calculated to calculate the associated limit gap value. The ratio between the meteorological impact sum value and the associated limit gap value is calculated to obtain the associated limit value of the warehouse associated path group.

6. The AI-based cloud warehousing intelligent inventory forecasting and dynamic allocation system according to claim 5, characterized in that, The method for determining the path meteorological impact value of the allocation route is as follows: obtain the predicted meteorological data on the allocation route, import the predicted meteorological data into the meteorological evaluation model, and derive the path meteorological impact value of the allocation route from the meteorological evaluation model.

Images