KR-102964011-B1 - An intelligent logistics distribution management system and its operation method that combines data-based demand forecasting and dynamic packaging optimization

KR102964011B1KR 102964011 B1KR102964011 B1KR 102964011B1KR-102964011-B1

Abstract

The present invention relates to an intelligent logistics distribution management system and a method of operation that combines data-based demand forecasting and dynamic packaging optimization. More specifically, it relates to an intelligent logistics distribution management system and a method of operation that analyzes real-time data between an offline wholesale distribution network and an online sales channel using machine learning technology to predict optimal inventory levels and prices, realizes dynamic pricing and intelligent supply chain optimization in heterogeneous distribution environments, dynamically determines an optimal packaging solution for each order that satisfies both cost efficiency and freshness preservation by linking predicted demand, real-time freshness data, and delivery environments, and establishes a sustainable logistics system by systematically managing the use of eco-friendly packaging materials. According to the present invention, by integrating and analyzing online and offline data to improve the accuracy of inventory-demand forecasting, opportunity costs resulting from excess inventory and stockouts are minimized. In particular, IoT and blockchain-based freshness management systems improve the efficiency of cold chain quality management and significantly reduce waste losses from products nearing their expiration date by linking with predicted demand. In addition, the intelligent packaging optimization system accurately predicts the required quantity of packaging materials based on predicted product demand and dynamically recommends the optimal combination of packaging materials according to the delivery environment, thereby creating additional effects such as reducing packaging costs and preventing unnecessary overpackaging.

Inventors

최현수

Assignees

주식회사 릴코컴퍼니

Dates

Publication Date: 20260513
Application Date: 20250916

Claims (2)

In an intelligent logistics distribution management system (1000) that combines data-based demand forecasting and dynamic packaging optimization, The on/offline data integration collection unit (100) is for collecting and standardizing data generated from offline wholesale distribution networks and online sales channels in real time, and An inventory demand forecasting engine unit (200) for predicting future demand and calculating the optimal inventory level by combining LSTM (Long Short-Term Memory) and GRU (Gated Recurrent Unit) networks into an ensemble, and A dynamic price optimization unit (300) for determining the optimal price in real time according to market conditions using a reinforcement learning algorithm, and A freshness management system unit (400) for ensuring freshness by monitoring the entire cold chain process with IoT sensors and recording history on a blockchain, and It is a multimodal personalized recommendation engine unit (500) for recommending customized products to customers using a graph neural network, and It is a closed mall operation system unit (600) for securing loyal customers by operating an independent membership shopping mall without platform fees, and It includes an intelligent packaging optimization system unit (700) that dynamically determines the optimal packaging solution for each order by linking the prediction data of the above-mentioned inventory demand prediction engine unit (200) with the real-time data of the freshness management system unit (400), and The data collected from the above-mentioned online/offline data integration collection unit (100) is transmitted to the inventory demand forecasting engine unit (200), and The prediction result of the above inventory demand forecasting engine unit (200) is simultaneously transmitted to the dynamic price optimization unit (300) and the freshness management system unit (400), and The freshness score of the above freshness management system unit (400) is transmitted to the multimodal personalized recommendation engine unit (500), and The recommendation result of the above multimodal personalized recommendation engine unit (500) is provided to the customer through the closed mall operation system unit (600), and An intelligent logistics distribution management system linked with data-based demand forecasting and dynamic packaging optimization, characterized by having a circular structure in which customer feedback information is fed back to the online/offline data integration collection unit (100).
In a method of operating an intelligent logistics distribution management system (1000) that combines data-based demand forecasting and dynamic packaging optimization, The online and offline data integration collection unit (100) collects various types of data, such as sales records, inventory status, and customer information, in real time through the point-of-sale (POS) system of the offline wholesale distribution network and the API of the online sales channel, and converts the data into a standardized data structure and stores it in a database in a data collection standardization step (S100). The inventory demand forecasting engine unit (200) predicts future demand and calculates the optimal inventory level through an LSTM-GRU ensemble model via a data preprocessing module (210), an LSTM network module (220), a GRU network module (230), an attention mechanism module (240), an ensemble integration layer module (250), and an uncertainty estimation module (260), and is an optimal inventory level calculation step (S200). The dynamic price optimization unit (300) determines the optimal price in real time using a reinforcement learning algorithm through an environment modeling module (310), a dual-layer network module (320), a priority-based experience reproduction module (330), a noise injection search strategy module (340), a multi-task learning module (350), and an adaptive discount rate mechanism module (360), and is an optimal price determination step (S300). The freshness management system unit (400) monitors the freshness of products through IoT sensors and dynamically adjusts delivery priority through the IoT sensor network module (410), distributed edge computing layer module (420), blockchain-based history management module (430), freshness score calculation algorithm module (440), dynamic delivery priority determination module (450), and freshness information visualization interface module (460), and is a delivery priority dynamic adjustment step (S400) in which delivery priority is dynamically adjusted. The multimodal personalization recommendation engine unit (500) recommends customized products for each customer using a graph neural network through the heterogeneous data integration layer module (510), bipartite graph structure module (520), graph attention network module (530), metapath-based aggregation module (540), cold start problem solving module (550), recommendation optimization module (560), and recommendation result explanation module (570). The customized product recommendation stage (S500) and the intelligent packaging optimization system unit (700) performs an intelligent packaging optimization stage in which the optimal packaging solution is dynamically determined through the packaging material data management module (710), the demand-linked packaging material prediction module (720), the freshness-linked dynamic packaging recommendation module (730), the packaging cost efficiency optimization module (740), and the eco-friendly packaging management module (750). The closed mall operation system unit (600) provides optimized prices and recommended products to customers through the front-end UI module (610), member management module (620), VIP membership module (630), regular delivery service module (640), and personalized recommendation linkage module (650), and performs a closed mall operation stage (S600) that reflects real-time feedback into the system. A method of operating an intelligent logistics distribution management system linked with data-based demand forecasting and dynamic packaging optimization, characterized in that all customer interactions are recorded as new real-time feedback data and transmitted back to the data collection standardization stage (S100), thereby enabling the entire system to continuously learn and complete a virtuous cycle structure.

Description

An intelligent logistics distribution management system and its operation method that combines data-based demand forecasting and dynamic packaging optimization The present invention relates to an intelligent logistics distribution management system and a method of operation that combines data-based demand forecasting and dynamic packaging optimization. More specifically, it relates to an intelligent logistics distribution management system and a method of operation that analyzes real-time data between an offline wholesale distribution network and an online sales channel using machine learning technology to predict optimal inventory levels and prices, realizes dynamic pricing and intelligent supply chain optimization in heterogeneous distribution environments, dynamically determines an optimal packaging solution for each order that satisfies both cost efficiency and freshness preservation by linking predicted demand, real-time freshness data, and delivery environments, and establishes a sustainable logistics system by systematically managing the use of eco-friendly packaging materials. As the e-commerce market grows rapidly, the online food distribution market is also expanding. In particular, while online sales of refrigerated foods such as dairy products are increasing, existing online distribution systems are facing the following technical limitations. First, existing systems operate in a data silo environment where data between online and offline sales channels is separated, making integrated inventory-demand forecasting structurally impossible. In other words, bulk B2B sales data from offline wholesale distribution networks and real-time B2C sales data from online retail channels are managed in different formats and cycles, failing to provide a comprehensive picture of overall market demand. As a result, forecasting models rely on biased or incomplete data, frequently leading to problems such as increased disposal costs due to chronic excess inventory or lost sales opportunities due to inventory shortages. Second, current systems are missing revenue generation opportunities by relying on static pricing policies. Most systems use the 'cost-plus method,' which adds a certain margin to the cost, or rule-based algorithms that simply follow competitors' prices. This is a rigid approach that fails to consider dynamic variables such as real-time demand changes, inventory levels, product freshness, and competitive intensity, making it impossible to make strategic pricing decisions, such as clearing inventory of products nearing their expiration date or maximizing margins during periods of surging demand. Third, existing distribution systems lack an intelligent approach to freshness management, which is a core value of refrigerated foods. Most cold chain management simply releases inventory based on the 'First-In, First-Out (FIFO)' principle and fails to systematically implement 'First-In, First-Out (FEFO),' which considers each product's actual temperature exposure history or remaining shelf life. In addition, issues such as temperature deviations during the delivery process are often discovered after the fact, making preemptive measures impossible, and it is difficult to determine liability when quality problems arise due to the difficulty of tracing the history. Fourth, customer personalized recommendation systems rely only on single-modal data such as purchase history, which limits the accuracy and diversity of recommendations. In other words, traditional collaborative filtering methods are vulnerable to the 'Cold Start' problem, where it is difficult to recommend new users or new products, and fail to identify customers' hidden tastes or the subtle characteristics of products. For example, because it does not understand the sentiment of customer review text or the visual characteristics of product images, it is impossible to provide truly personalized recommendations to customers who prefer cheese with a soft texture. Fifth, the packaging process, the final stage of logistics, is operated inefficiently and unscientifically. Existing systems use standardized packaging materials uniformly without considering variables such as the type and quantity of ordered products, delivery distance, and external temperature. This results in resource waste and increased logistics costs due to unnecessary overpackaging, or conversely, damage to the freshness of the product during delivery due to inadequate cooling packaging. There is a lack of dynamic, data-driven packaging solutions for maintaining optimal freshness. FIG. 1 is an overall configuration conceptual diagram of an intelligent logistics distribution management system that combines data-based demand forecasting and dynamic packaging optimization according to one embodiment of the present invention. FIG. 2 is a conceptual diagram of the configuration of an online/offline data integration collection unit (100) of an intelligent logistics distribution management sys