Intelligent warehousing prediction and replenishment method and system based on large model

Abstract

The invention is suitable for the field of intelligent warehousing, and provides an intelligent warehousing prediction and replenishment method and system based on a large model, and the method comprises the following steps: analyzing and obtaining the demand information of purchasing, selling and transportation; obtaining and processing supply end information; detecting and acquiring inventory information of a warehouse in real time; inputting the current stock information, demand information and supply end information into a trained deep learning model to obtain an optimal stock quantity; and automatically generating a purchase order according to the optimal stock quantity calculated by the deep learning model. According to the method, each order placing and stock preparation can be carried out under front-end demands, the pressure of warehouse turnover can be reduced to the minimum, the optimal stock quantity can be predicted in a personalized mode according to different commodities, and the self-adaption degree is high.

Description

Technical Field

[0001] The present invention relates to the field of intelligent warehousing technology, and in particular to an intelligent warehousing forecasting and replenishment method and system based on a large model. Background Art

[0002] Smart warehousing is an indispensable part of modern supply chain management. It uses advanced technologies such as the Internet of Things (IoT), big data, artificial intelligence (AI), and robotics to achieve automation, intelligence, and precision in warehouse management.

[0003] Traditional business rule-based approaches suffer from significant drawbacks. First, they rely too heavily on manual experience, resulting in many important decision parameters (such as safety stock thresholds) often relying on empirical assumptions rather than precise, data-driven calculations. Due to the lack of systematic quantitative assessment methods, forecast accuracy is generally low, often failing to achieve the desired level and failing to effectively respond to market changes and demand fluctuations. Furthermore, the maintenance cost of manual rules rises significantly with the increase in the number of SKUs.

[0004] Secondly, traditional algorithmic models also have limitations. While many classic models (such as ARIMA and LSTM) can process structured historical sales data, they primarily focus on a single data dimension and struggle to fully capture the diversity of market fluctuations. Furthermore, these models perform poorly with the "cold start" problem, often resulting in low accuracy in demand forecasts for new products or new markets. These traditional algorithmic models are unable to effectively integrate data from diverse sources, resulting in a lack of a holistic perspective in decision-making and difficulty meeting increasingly complex business needs. Therefore, to overcome the shortcomings of traditional methods, new technical approaches are urgently needed. Summary of the Invention

[0005] In view of the deficiencies in the prior art, the purpose of the present invention is to provide an intelligent warehousing forecasting and replenishment method and system based on a large model to solve the problems existing in the above-mentioned background technology.

[0006] The present invention is implemented as follows: a large-scale model-based intelligent warehousing forecasting and replenishment method, the method comprising the following steps:

[0007] Analyze and obtain demand information for purchase, sales and transportation;

[0008] Acquire and process supply-side information, including supply quantity, supply location, and supply time;

[0009] Real-time detection to obtain warehouse inventory information;

[0010] Input current inventory information, demand information, and supply-side information into the trained deep learning model to obtain the optimal stocking quantity;

[0011] Automatically generate purchase orders based on the optimal stocking quantity calculated by the deep learning model.

[0012] As a further solution of the present invention: the step of analyzing and obtaining the demand information for purchase, sales and transportation specifically includes:

[0013] Collect original demand data from multiple sources and perform data cleaning;

[0014] Convert cleaned data into feature data that can be used for modeling to improve prediction accuracy;

[0015] Build and train a prediction model based on the feature data, and output demand for a set time period in the future;

[0016] The prediction model is evaluated through error analysis, the best performing model is selected and the prediction results are output.

[0017] As a further solution of the present invention: the step of obtaining and processing the supply-side information specifically includes:

[0018] Use the ERP system to obtain and collect relevant raw data, and then process the raw data in a unified manner and clean up abnormal data;

[0019] Use the processed raw data to model the supply capacity of various suppliers to quantify the suppliers' fulfillment capabilities and response speed;

[0020] Aggregate the data to obtain supply-side information and present it in a structured form.

[0021] As a further solution of the present invention: the step of obtaining the optimal stocking quantity specifically includes:

[0022] Integrate inventory information, demand information, and supply-side information into a state vector;

[0023] Input the state vector into the deep learning model to output the recommended product inventory quantity;

[0024] Calculate the reward value corresponding to the recommended product stocking quantity based on the inventory fill rate and cost indicators;

[0025] Initialize the parameters of the deep learning model, use simulated data and optimize the deep learning model according to the back propagation algorithm to make the output simple commodity stocking quantity approach the optimal stocking quantity.

[0026] Another object of the present invention is to provide an intelligent warehouse forecasting and replenishment system based on a large model, the system comprising:

[0027] Demand analysis module, used to analyze and obtain demand information for procurement, sales and transportation;

[0028] A supply-side information processing module, configured to obtain and process supply-side information, including supply quantity, supply location, and supply time;

[0029] Inventory status monitoring module, used to detect and obtain warehouse inventory information in real time;

[0030] The model processing module is used to input the current inventory information, demand information, and supply-side information into the trained deep learning model to obtain the optimal stocking quantity;

[0031] The automatic ordering module is used to automatically generate purchase orders based on the optimal stocking quantity calculated by the deep learning model.

[0032] As a further solution of the present invention: the demand analysis module includes:

[0033] Data access and pre-processing unit, used to collect original demand data from multiple sources and perform data cleaning;

[0034] Feature engineering unit, used to convert cleaned data into feature data that can be used for modeling to improve prediction accuracy;

[0035] A demand forecasting modeling unit, which constructs and trains a forecasting model based on the feature data and outputs demand for a set time period in the future;

[0036] The model evaluation and output unit is used to evaluate the effectiveness of the prediction model through error analysis, select the best performing model and output the prediction results.

[0037] As a further solution of the present invention: the model processing module includes:

[0038] The state acquisition unit is used to integrate inventory information, demand information, and supply-side information into a state vector;

[0039] The action selection unit is used to input the state vector into the deep learning model to output the recommended product inventory quantity;

[0040] The reward feedback unit calculates the reward value corresponding to the recommended product stocking quantity based on the inventory fulfillment rate and cost indicators;

[0041] The intelligent agent unit is used to initialize the parameters of the deep learning model and optimize the deep learning model using simulated data and the back propagation algorithm to make the output simple commodity stocking quantity approach the optimal stocking quantity.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] The present invention calculates the optimal stocking quantity and whether replenishment is needed based on the current inventory information, market demand information, and supply-side information, and also includes the specific replenishment quantity. The most important thing is to ensure that each order is placed based on the front-end demand, which can reduce the pressure of warehouse turnover to a minimum. At the same time, it can predict the optimal stocking quantity for different commodities in a personalized way, with a relatively high degree of adaptability. It can also be connected with the order system, warehouse management system, and transportation management system to realize automated optimal ordering, with a high degree of automation and low maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a flowchart of an intelligent warehousing forecasting and replenishment method based on a large model.

[0045] Figure 2 This is a flowchart for analyzing and obtaining purchasing, sales, and transportation demand information in an intelligent warehouse forecasting and replenishment method based on a large model.

[0046] Figure 3 This is a flowchart for obtaining and processing supply-side information in an intelligent warehouse forecasting and replenishment method based on a large model.

[0047] Figure 4 This is a flowchart for obtaining the optimal inventory quantity in an intelligent warehouse forecasting and replenishment method based on a large model.

[0048] Figure 5 This is a structural diagram of an intelligent warehousing forecasting and replenishment system based on a large model.

[0049] Figure 6 This is a structural diagram of the demand analysis module in an intelligent warehousing forecasting and replenishment system based on a large model.

[0050] Figure 7 This is a structural diagram of the model processing module in an intelligent warehousing forecasting and replenishment system based on a large model. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0052] The specific implementation of the present invention is described in detail below with reference to specific embodiments.

[0053] like Figure 1 As shown, an embodiment of the present invention provides an intelligent warehousing forecasting and replenishment method based on a large model, the method comprising the following steps:

[0054] S100, analyzes and obtains demand information for purchasing, sales and transportation;

[0055] S200, obtaining and processing supply-side information, wherein the supply-side information includes supply quantity, supply location, and supply time;

[0056] S300, real-time detection and acquisition of warehouse inventory information;

[0057] S400: Input the current inventory information, demand information, and supply-side information into the trained deep learning model to obtain the optimal stocking quantity;

[0058] S500 automatically generates purchase orders based on the optimal stocking quantity calculated by the deep learning model.

[0059] It should be noted that demand information from different channels will first be collected and analyzed, including but not limited to sales forecasts (based on historical sales data, market trends, etc.), procurement plans (taking into account promotional activities, seasonal factors, etc.) and logistics arrangements (taking into account the transportation time of goods from suppliers to warehouses). For example, when preparing for the upcoming Mother's Day promotion, the system will predict a surge in demand for related products and adjust the stocking strategy accordingly; next, the system needs to understand the specific situation on the supply side, such as the quantity of goods that each supplier can provide, the supply location, and the expected delivery time; at the same time, it will continue to monitor the actual inventory level in the warehouse to ensure that the latest inventory status is always available; all of the above information is fed as input parameters into a fully trained deep learning model. This model is built based on a large amount of historical transaction data and can comprehensively consider the complex relationships between multiple variables to predict the most optimized stocking plan; based on the results output by the model, the optimal stocking quantity is determined.

[0060] In the embodiments of the present invention, the present invention calculates the optimal stocking quantity and whether replenishment is needed, including the specific replenishment quantity, based on current inventory information, market demand information, and supply-side information. Most importantly, it ensures that each order is placed based on front-end demand, minimizing warehouse turnover pressure. Furthermore, it can predict the optimal stocking quantity for each product individually, with a high degree of adaptability. It can also be connected to order systems, warehouse management systems, and transportation management systems to achieve automated optimal order placement, with a high degree of automation and low maintenance costs.

[0061] like Figure 2 As shown, as a preferred embodiment of the present invention, the step of analyzing and obtaining the demand information for purchase, sales and transportation specifically includes:

[0062] S101, collect original demand data from multiple sources and perform data cleaning;

[0063] S102, converting the cleaned data into feature data that can be used for modeling to improve prediction accuracy;

[0064] S103, constructing and training a prediction model based on the feature data, and outputting demand for a set time period in the future;

[0065] S104, evaluating the effectiveness of the prediction model through error analysis, selecting the best performing model and outputting the prediction results.

[0066] In an embodiment of the present invention, first, the system needs to collect original demand data from multiple sources. These sources may include but are not limited to historical sales records in the enterprise resource planning (ERP) system, market research reports, social media trend analysis, weather forecasts (especially important for certain seasonal products), holiday arrangements and other external factors; then, the collected data is cleaned to remove duplicate, erroneous or incomplete records, fill in missing values, and unify the data format, or use statistical methods to fill in blank values ​​in the sales data; the cleaned data is converted into feature data that can be used for modeling. This step involves selecting which variables to use as input features and how to convert these variables to improve model performance. For example, the model's predictive ability can be enhanced by calculating year-on-year and month-on-month growth rates, creating virtual variables to represent different seasons or holidays, etc.; based on the above feature data, a prediction model is constructed. You can choose a variety of machine learning algorithms such as linear regression, decision trees, random forests, XGBoost, etc., or you can try deep learning models such as LSTM (long short-term memory network); use historical data to train the selected model, and adjust hyperparameters to optimize model performance; through a series of indicators (such as mean square error MSE, mean absolute error MAE, R 2 The model's effectiveness can be evaluated using methods such as residual plots and AIC / BIC curves. Based on the evaluation results, the best performing model can be selected and used for the final demand forecasting task.

[0067] like Figure 3 As shown, as a preferred embodiment of the present invention, the step of obtaining and processing the supply-side information specifically includes:

[0068] S201, using the ERP system to obtain and collect relevant raw data, and then unify the raw data and clean up abnormal data;

[0069] S202, using the processed raw data to model the supply capabilities of various suppliers, to quantify the suppliers' fulfillment capabilities and response speed;

[0070] S203, aggregate the data to obtain supply-side information and present it in a structured form.

[0071] In an embodiment of the present invention, in the process of acquiring and processing supply-side information, the original data related to the supplier is first collected through the ERP system, and the data is uniformly processed and cleaned for exceptions to ensure the accuracy and consistency of the data. Subsequently, the cleaned data is used to model the supply capacity of each supplier, quantify its performance capabilities and response speed, and thus evaluate the reliability of the supplier. Finally, these data are summarized into structured supply-side information, clearly presenting key content such as the supplier's supply quantity, location, and time, providing a reliable basis for subsequent decision-making. For example, when preparing for holiday promotions, the system analyzes the historical delivery data of multiple suppliers to screen out core suppliers that can respond quickly and provide stable supply, thereby ensuring the efficient operation of the supply chain.

[0072] like Figure 4 As shown, as a preferred embodiment of the present invention, the step of obtaining the optimal stocking quantity specifically includes:

[0073] S401, integrating inventory information, demand information, and supply-side information into a state vector;

[0074] S402: Input the state vector into the deep learning model to output the recommended product inventory quantity;

[0075] S403, calculating a reward value corresponding to the recommended product stocking quantity based on the inventory fill rate and cost indicators;

[0076] S404, initializing the parameters of the deep learning model, using simulated data and optimizing the deep learning model according to the back propagation algorithm, so as to make the output simple commodity stocking quantity approach the optimal stocking quantity.

[0077] In the embodiment of the present invention, the main processing process of the deep learning model is described in detail, and the optimization of the intelligent agent part in the deep learning model is also described. The specific contents are as follows:

[0078] Step 1: Model Building

[0079] Define the model structure: Build a deep learning model and define the key components of the model, including the agent, state space (inventory information, demand information, supply-side information), action space (recommended product inventory), reward mechanism (indicators based on inventory fill rate and cost), and optimization goal (minimize total cost while maximizing inventory fill rate).

[0080] Step 2: Data initialization and simulation environment generation

[0081] Parameter initialization: Initialize all parameters of the learning model.

[0082] Randomly generate initial conditions: Randomly generate initial inventory, supplier maximum supply, front-end demand, and recommended stocking quantity for each product. These values ​​are reasonably estimated based on historical data or experience.

[0083] Dynamic simulation rule setting:

[0084] Replenishment delay: The bootstrap method is used to sample historical delivery data to generate recommended inventory quantities and arrival times.

[0085] Step 3: Learning and Training

[0086] State vector construction: Inventory information, demand information, and supply-side information are integrated into a state vector and provided as input to the deep learning model.

[0087] Output recommended stocking quantity: After the state vector is input into the model, the model outputs the recommended stocking quantity for each product.

[0088] Calculate the reward value: Calculate the reward value corresponding to the recommended stocking quantity based on indicators such as inventory fulfillment rate and cost. The reward value function can be obtained through a weighted algorithm.

[0089] Simulation and experience pool filling: During each round of simulation (M days), data such as state, action, reward, and new state are generated every day and then stored in the experience pool.

[0090] Policy optimization: Randomly sample n batches of data from the experience pool, use the backpropagation algorithm to derive the policy gradient, and then calculate the gradient descent to update the neural network parameters. Repeat this process until the predetermined number of training rounds and the number of simulation days per round, M, are reached. In this way, the output recommended product stocking quantity will approach the optimal stocking quantity. It should be noted that the reward value is negatively correlated with the replenishment quantity, that is, the smaller the replenishment quantity, the higher the reward value.

[0091] like Figure 5 As shown, an embodiment of the present invention further provides an intelligent warehousing forecasting and replenishment system based on a large model, the system comprising:

[0092] Demand analysis module 100, used to analyze and obtain demand information for procurement, sales and transportation;

[0093] The supply-side information processing module 200 is used to obtain and process supply-side information, including supply quantity, supply location, and supply time;

[0094] Inventory status monitoring module 300, used to detect and obtain warehouse inventory information in real time;

[0095] Model processing module 400, used to input current inventory information, demand information, and supply-side information into the trained deep learning model to obtain the optimal stocking quantity;

[0096] The automatic ordering module 500 is used to automatically generate a purchase order based on the optimal stocking quantity calculated by the deep learning model.

[0097] In the embodiment of the present invention, the large model can integrate and analyze massive amounts of historical sales data, market trends, seasonal factors, promotional activities and other multi-source data, so as to more accurately predict the future inventory demand of different categories of goods, determine the optimal inventory level, realize automatic adjustment and optimization of inventory, reduce unnecessary inventory backlogs, reduce the funds occupied by inventory, and monitor inventory data in real time. It can also quickly generate replenishment suggestions based on the prediction results, helping enterprises to respond to market changes more flexibly.

[0098] like Figure 6 As shown, as a preferred embodiment of the present invention, the demand analysis module 100 includes:

[0099] The data access and pre-processing unit 101 is used to collect original demand data from multiple sources and perform data cleaning;

[0100] Feature engineering unit 102, used to convert the cleaned data into feature data that can be used for modeling to improve prediction accuracy;

[0101] The demand forecast modeling unit 103 constructs and trains a forecast model based on the feature data and outputs the demand for a set time period in the future;

[0102] The model evaluation and output unit 104 is used to evaluate the effectiveness of the prediction model through error analysis, select the model with the best performance and output the prediction result.

[0103] like Figure 7 As shown, as a preferred embodiment of the present invention, the model processing module 400 includes:

[0104] The state collection unit 401 is used to integrate inventory information, demand information, and supply-side information into a state vector;

[0105] Action selection unit 402, configured to input the state vector into the deep learning model and output the recommended product inventory quantity;

[0106] The reward feedback unit 403 calculates a reward value corresponding to the recommended product stocking quantity based on the inventory fulfillment rate and cost indicators;

[0107] The intelligent agent unit 404 is used to initialize the parameters of the deep learning model, and use simulated data and the back propagation algorithm to optimize the deep learning model for training, so as to make the output simple commodity stocking quantity approach the optimal stocking quantity.

[0108] The above is only a detailed description of the preferred embodiments of the present invention, which is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0109] It should be understood that, although the various steps in the flow chart of each embodiment of the present invention are shown in sequence according to the indication of the arrows, these steps are not necessarily performed in sequence according to the order indicated by the arrows. Unless otherwise specified herein, the execution of these steps is not strictly limited in order, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0110] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0111] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the disclosure in the specification and examples. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered merely as exemplary, and the true scope and spirit of the present disclosure are indicated by the claims.

Claims

1. An intelligent warehousing forecasting and replenishment method based on a large model, characterized in that: The method comprises the following steps: Analyze and obtain demand information for purchase, sales and transportation; Acquire and process supply-side information, including supply quantity, supply location, and supply time; Real-time detection to obtain warehouse inventory information; Input current inventory information, demand information, and supply-side information into the trained deep learning model to obtain the optimal stocking quantity; Automatically generate purchase orders based on the optimal stocking quantity calculated by the deep learning model.

2. The intelligent warehousing forecasting and replenishment method based on a large model according to claim 1 is characterized in that: The steps of analyzing and obtaining purchasing, sales, and transportation demand information specifically include: Collect original demand data from multiple sources and perform data cleaning; Convert cleaned data into feature data that can be used for modeling to improve prediction accuracy; Build and train a prediction model based on the feature data, and output demand for a set time period in the future; The prediction model is evaluated through error analysis, the best performing model is selected and the prediction results are output.

3. The intelligent warehousing forecasting and replenishment method based on a large model according to claim 1 is characterized in that: The steps of obtaining and processing the supply-side information specifically include: Use the ERP system to obtain and collect relevant raw data, and then process the raw data in a unified manner and clean up abnormal data; Use the processed raw data to model the supply capacity of various suppliers to quantify the suppliers' fulfillment capabilities and response speed; Aggregate the data to obtain supply-side information and present it in a structured form.

4. The intelligent warehousing forecasting and replenishment method based on a large model according to claim 1 is characterized in that: The step of obtaining the optimal stocking quantity specifically includes: Integrate inventory information, demand information, and supply-side information into a state vector; Input the state vector into the deep learning model to output the recommended product inventory quantity; Calculate the reward value corresponding to the recommended product stocking quantity based on the inventory fill rate and cost indicators; Initialize the parameters of the deep learning model, use simulated data and optimize the deep learning model according to the back propagation algorithm to make the output simple commodity stocking quantity approach the optimal stocking quantity.

5. An intelligent warehouse forecasting and replenishment system based on a large model, characterized by: The system comprises: Demand analysis module, used to analyze and obtain demand information for procurement, sales and transportation; A supply-side information processing module, configured to obtain and process supply-side information, including supply quantity, supply location, and supply time; Inventory status monitoring module, used to detect and obtain warehouse inventory information in real time; The model processing module is used to input the current inventory information, demand information, and supply-side information into the trained deep learning model to obtain the optimal stocking quantity; The automatic ordering module is used to automatically generate purchase orders based on the optimal stocking quantity calculated by the deep learning model.

6. The intelligent warehouse forecasting and replenishment system based on a large model according to claim 5 is characterized in that: The demand analysis module includes: Data access and pre-processing unit, used to collect original demand data from multiple sources and perform data cleaning; Feature engineering unit, used to convert cleaned data into feature data that can be used for modeling to improve prediction accuracy; A demand forecasting modeling unit, which constructs and trains a forecasting model based on the feature data and outputs demand for a set time period in the future; The model evaluation and output unit is used to evaluate the effectiveness of the prediction model through error analysis, select the best performing model and output the prediction results.

7. The intelligent warehouse forecasting and replenishment system based on a large model according to claim 5 is characterized in that: The model processing module includes: The state acquisition unit is used to integrate inventory information, demand information, and supply-side information into a state vector; The action selection unit is used to input the state vector into the deep learning model to output the recommended product inventory quantity; The reward feedback unit calculates the reward value corresponding to the recommended product stocking quantity based on the inventory fulfillment rate and cost indicators; The intelligent agent unit is used to initialize the parameters of the deep learning model and optimize the deep learning model using simulated data and the back propagation algorithm to make the output simple commodity stocking quantity approach the optimal stocking quantity.

Images