A ginseng supply chain optimization management method and system

Abstract

The application discloses a ginseng supply chain optimization management method and system, relates to the technical field of supply chain optimization management, and brings significant benefits to the ginseng supply chain through the construction of a data-driven closed-loop optimization system, realizes accurate quantification and dynamic optimization of the value of ginseng, and directly improves economic benefits: based on a quality value dynamic mapping model, high-value sources and high-end markets can be accurately matched to realize high quality and high price; through double-layer collaborative optimization decision-making, resources are globally deployed under the quality constraint, warehouse loss and ineffective logistics are greatly reduced, and the overall net income of the supply chain is improved; at the same time, the whole-chain data fusion and state prediction realize the change from 'experience management' to 'accurate prediction', and the ability to respond to market fluctuations is enhanced; the final execution feedback mechanism ensures that the model continuously optimizes itself, forms a virtuous cycle of getting smarter and smarter, and comprehensively improves the resilience, transparency and core competitiveness of the supply chain.

Description

Technical Field

[0001] This application relates to the field of supply chain optimization management technology, specifically to a ginseng supply chain optimization management method and system. Background Technology

[0002] The traditional ginseng supply chain suffers from severe data silos, difficulty in quantifying quality changes, and reliance on experience for decision-making. From planting and processing to distribution, data is fragmented across each stage, hindering traceability and collaboration. The core value of ginseng (such as saponin content) dynamically declines over time, but the relationship between this quality change and market value lacks a precise model, leading to inefficient pricing and resource misallocation. Currently, the Internet of Things (IoT), big data, and artificial intelligence (AI) technologies offer solutions to these pain points. However, existing technologies are largely limited to monitoring single stages, lacking a comprehensive technical framework that covers the entire supply chain, integrates multi-source data, and enables global optimization decisions based on quality dynamics. Therefore, a systematic solution integrating data collection, fusion, value modeling, state prediction, and collaborative optimization is needed to address the technical bottlenecks in the refined management of the ginseng industry.

[0003] Existing technology, such as the logistics supply chain management method and management system disclosed in Chinese invention application patent application CN119027017B, relates to the field of supply chain management technology. This invention includes: Step 1, acquiring logistics supply transfer parameters; Step 2, evaluating logistics transfer warehouses; Step 3, evaluating processing methods; and Step 4, displaying and processing logistics transfer warehouses. This invention ensures the overall supply and demand balance of the logistics supply chain, improves the operational efficiency of the logistics supply chain, and determines whether to perform compression, expansion, or cooling processing on the logistics supply chain based on the supply and demand balance index of the logistics supply chain and the operational performance level of each logistics transfer warehouse. This simplifies the supply chain processing process, improves the accuracy of subsequent processing of logistics supply chain transfer warehouses, reduces logistics costs, reduces cargo damage and expiration, avoids unnecessary financial waste, and simultaneously ensures inventory allocation and cargo turnover, thus ensuring the value of logistics supply chain management.

[0004] Regarding the above solutions, the inventors of this application have discovered that the above technologies have at least the following technical problems: 1. Currently, there is a lack of a two-layer structure to decouple the objective function of the upper-layer optimization model, which directly aims to maximize the upper-layer objective value (profit), and there is no embedding of the quality value dynamic mapping function as the core value calculation engine; it is impossible to enable the profit of each logistics decision variable to be dynamically evaluated based on its accurate predicted quality and transportation time; there is a lack of an objective function of the lower-layer optimization model that focuses on minimizing resource consumption (such as transportation energy consumption) and the natural loss of human value in the circulation process (lower-layer objective value); currently, there is a lack of a separate but related structure, which makes it impossible for the model to simultaneously characterize the impact of decisions on two dimensions.

[0005] 2. Currently, the system lacks a mechanism to continuously feed back actual execution deviations to the prediction and modeling stages. As a result, the system cannot continuously correct model parameters, cannot dynamically adapt to uncertainties such as market changes and process fluctuations, and cannot achieve iterative evolution of decision-making capabilities. Furthermore, the system lacks a mechanism to directly convert optimization results into actionable instructions and issue them, failing to bridge the "last mile" from decision-making to execution. At the same time, the lack of a closed-loop feedback mechanism makes it impossible to ensure that optimization strategies can be continuously updated with changes in the environment, cannot avoid the decay of the effectiveness of static optimization schemes, and cannot maintain the supply chain operation in an optimal or near-optimal state in the long term. Summary of the Invention

[0006] To address the aforementioned technical shortcomings, the purpose of this application is to provide a method and system for optimizing the ginseng supply chain management.

[0007] To solve the above-mentioned technical problems, this application adopts the following technical solution: In the first aspect, this application provides a ginseng supply chain optimization management method, which includes the following steps: S1, full-chain multi-source data collection and fusion: collect multi-source data of ginseng to generate a full-chain multi-source dataset with spatiotemporal and batch labels.

[0008] S2. Dynamic mapping modeling of ginseng quality value: Based on the full-chain multi-source dataset, a dynamic mapping function for ginseng quality value is constructed to generate the unit economic value of ginseng.

[0009] S3. Multi-step prediction of supply chain status: Based on the full-chain multi-source dataset and the quality value dynamic mapping function, the market demand and the quality of each node ginseng in several decision cycles are predicted, and the predicted supply and demand matrix and the predicted quality matrix are output.

[0010] S4. Collaborative Optimization Decision Generation: Based on the aforementioned supply and demand matrix and predicted quality matrix as input, a two-level optimization model is established to obtain the optimal logistics and allocation decision scheme.

[0011] S5. Decision Execution and Closed-Loop Feedback: Execute the optimal decision scheme and update the quality value dynamic mapping function and prediction model.

[0012] Preferably, the full-chain multi-source dataset includes environmental data, processing parameters, inventory status, logistics trajectory, and key quality indicator data.

[0013] Preferably, the collection of multi-source data on ginseng to generate a full-chain multi-source dataset with spatiotemporal and batch labels includes: real-time collection of status information of ginseng in planting, processing, warehousing and logistics based on IoT sensors and detection equipment deployed at various nodes of the supply chain to obtain raw multi-source heterogeneous data.

[0014] Based on a unified data communication protocol, the original multi-source heterogeneous data is aggregated to obtain an aggregated data stream with spatiotemporal tags.

[0015] Based on preset data quality rules and unified batch identifiers, the aggregated data stream is cleaned and aligned to obtain a standardized spatiotemporal dataset.

[0016] Based on the unified batch identifier and timestamp, the standardized spatiotemporal dataset is fused and structured to generate a full-chain multi-source dataset with spatiotemporal and batch labels.

[0017] Preferably, the step of constructing a dynamic mapping function for ginseng quality value based on the full-chain multi-source dataset to generate the unit economic value of ginseng includes: extracting the periodic quality test value sequence of each batch of ginseng and its market transaction price data at the corresponding test time point based on the full-chain multi-source dataset to form a training sample set.

[0018] Construct candidate forms of dynamic mapping functions for quality value that include quality value basis functions, time decay functions, and a set of undetermined parameters.

[0019] Based on the training sample set, the set of undetermined parameters in the candidate form of the quality value dynamic mapping function is fitted and solved using a nonlinear regression method, thereby obtaining the quality value dynamic mapping function with determined parameters, and thus generating the unit economic value of ginseng.

[0020] Preferably, the construction of a candidate form of the dynamic mapping function for quality value, which includes a quality value basis function, a time decay function, and a set of undetermined parameters, includes: In the formula This represents the predicted economic value per unit of ginseng. This is represented as a dynamic mapping function of quality value. This indicates a quality indicator for human-participating bond chemistry. This represents the elapsed time since the harvesting reference point. It is a quality value basis function. Represented as a time decay function, Represented as a linear scaling factor. This represents the offset of the base value.

[0021] Preferably, the step of predicting market demand and ginseng quality at each node for several decision-making cycles and outputting a predicted supply-demand matrix and a predicted quality matrix includes: extracting and summarizing the current inventory status and in-stock ginseng batch information at each node of the supply chain based on the full-chain multi-source dataset, so as to construct a supply chain state vector of the overall supply chain status at the current moment.

[0022] Based on historical order data, market intelligence, and the aforementioned quality value dynamic mapping function, the expected product value range is derived, and the end-market demand quantity for several decision-making cycles is generated.

[0023] Based on the supply chain state vector and the predicted market demand, the inventory level of each node is analyzed.

[0024] Based on the initial quality detection values ​​and current time points of each batch of ginseng recorded in the full-chain multi-source dataset, and combined with the time decay function in the quality value dynamic mapping function, the quality decay of each batch of ginseng at each decision time point in the future is calculated to generate a predicted quality matrix.

[0025] Based on the quantity of demand in the end market, the inventory levels at each node, and the predicted quality matrix, the data are integrated and structured to output a predicted supply and demand matrix for optimized decision-making.

[0026] Preferably, generating the predicted quality matrix includes: calculating the formula... The result is the The batch in The node and the first Predictive quality matrix for each decision cycle Represented as the first Initial quality test values ​​for each batch. Representing the current time Represented as from the current moment to the nth The time interval at the start of each decision cycle Represented as the first The harvesting reference time for each, Represented as the first The batch to the future The total storage time from the start of harvesting in each cycle.

[0027] Preferably, the step of establishing a two-layer optimization model based on the supply and demand matrix and the predicted quality matrix as input, and obtaining the optimal logistics and allocation decision scheme, includes: establishing the objective function of the upper-layer optimization model and the objective function of the lower-layer optimization model based on the predicted supply and demand matrix and the predicted quality matrix as input.

[0028] Define decision variables, which represent the logistics allocation from one node to another in each decision cycle, and set a set of constraints for the two-layer optimization model, including flow balance, capacity limitation, and quality-specific constraints.

[0029] Based on the genetic algorithm framework, a population consisting of multiple potential logistics allocation schemes is initialized, and a specific fitness function is designed. For each individual scheme in the population, the fitness value is calculated based on the fitness function. Genetic operations are then performed on the population based on the fitness value to generate a new offspring population, and this iterative process is repeated.

[0030] When the iteration process meets the preset termination condition, the individual with the highest fitness is selected from the final population, and it is decoded into a specific logistics allocation scheme to output the optimal decision scheme.

[0031] Preferably, executing the optimal decision scheme and updating the quality value dynamic mapping function and prediction model includes: based on the optimal decision scheme, analyzing the cargo allocation information of each decision cycle and each logistics path to generate a set of instructions to be executed for each physical execution node in the supply chain.

[0032] During the execution of the instructions, the actual status of each link in the supply chain is continuously collected to obtain a new round of multi-source datasets for the entire supply chain.

[0033] Based on the new round of full-chain multi-source dataset, the actual execution results are extracted and compared with the planned values ​​in the predicted supply and demand matrix, the predicted quality matrix, and the optimal decision scheme to calculate the deviation data.

[0034] Based on historical data containing the aforementioned deviation data, the parameters of the quality value dynamic mapping function are refitted and updated; simultaneously, the prediction model is retrained and updated.

[0035] In a second aspect, this application provides a system for optimizing the management of the ginseng supply chain, comprising: preferably, a full-chain multi-source data acquisition and fusion module for acquiring multi-source data of ginseng to generate a full-chain multi-source dataset with spatiotemporal and batch labels.

[0036] The ginseng quality value dynamic mapping modeling module constructs a ginseng quality value dynamic mapping function based on the full-chain multi-source dataset to generate the unit economic value of ginseng.

[0037] The multi-step supply chain status prediction module, based on the full-chain multi-source dataset and the quality value dynamic mapping function, predicts the market demand and the quality of each node's inputs for several decision cycles, and outputs a predicted supply and demand matrix and a predicted quality matrix.

[0038] The collaborative optimization decision generation module, based on the aforementioned supply and demand matrix and predicted quality matrix as input, establishes a two-layer optimization model and obtains the optimal logistics and allocation decision scheme.

[0039] The decision execution and closed-loop feedback module is used to execute the optimal decision scheme and update the quality value dynamic mapping function and prediction model.

[0040] The beneficial effects of this application are as follows: 1. The ginseng supply chain optimization management method and system provided by this application, through the construction of a data-driven closed-loop optimization system, brings significant benefits to the ginseng supply chain, realizes the accurate quantification and dynamic optimization of ginseng value, and directly improves economic benefits: Based on the quality value dynamic mapping model, it can accurately match high-value sources with high-end markets to achieve high quality and high price; through two-layer collaborative optimization decision-making, resources are allocated globally under the constraint of ensuring quality, greatly reducing warehousing losses and ineffective logistics, thereby improving the overall net profit of the supply chain; at the same time, the integration of full-chain data and status prediction realizes the transformation from "experience management" to "accurate prediction", enhancing the ability to cope with market fluctuations; the final execution feedback mechanism ensures that the model continuously self-optimizes, forming a virtuous cycle of becoming smarter the more it is used, and comprehensively improving the resilience, transparency and core competitiveness of the supply chain.

[0041] 2. This application employs a two-layer structure to decouple the upper-layer optimization model's objective function, which directly aims to maximize the upper-layer objective value (profit), while embedding a dynamic quality value mapping function as the core value calculation engine. This allows the profit of each logistics decision variable to be dynamically evaluated based on its accurate predicted quality and transportation time. The lower-layer optimization model's objective function focuses on minimizing resource consumption (such as transportation energy consumption) and the natural loss of human resources' value during the circulation process (lower-layer objective value). This separate yet interconnected structure enables the model to simultaneously characterize the impact of decisions on two dimensions. The final optimal decision output is not simply the maximum profit solution, but rather a scheduling scheme that achieves the best balance between high profit and low consumption / low loss under current predictions and constraints. This demonstrates the technological effect of this application, which, while pursuing economic benefits, intrinsically drives the supply chain towards a greener and more sustainable direction.

[0042] 3. This application continuously feeds actual execution deviations back to the prediction and modeling stages, enabling the system to continuously correct model parameters and dynamically adapt to uncertainties such as market changes and process fluctuations, thus achieving iterative evolution of decision-making capabilities. Optimization results are directly converted into actionable instructions and issued, bridging the "last mile" from decision-making to execution. Simultaneously, the closed-loop feedback mechanism ensures that optimization strategies are continuously updated with environmental changes, avoiding the effectiveness decay of static optimization schemes and maintaining the supply chain operation in an optimal or near-optimal state in the long term. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is a flowchart illustrating the implementation steps of the method described in this application.

[0045] Figure 2 This is a schematic diagram of the system structure connection of this application. Detailed Implementation

[0046] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0047] Please see Figure 1 As shown, this application provides a ginseng supply chain optimization management method in the first aspect, including: S1, full-chain multi-source data collection and fusion: collecting multi-source data of ginseng to generate a full-chain multi-source dataset with spatiotemporal and batch labels.

[0048] In a specific example, the full-chain multi-source dataset includes environmental data, processing parameters, inventory status, logistics trajectory, and key quality indicator data.

[0049] It should be noted that the full-chain multi-source dataset is a structured and indexed data resource. Its data structure is usually represented by "batch number" as the primary table key, which is associated with multiple "event" records arranged in chronological order. Each event record contains a timestamp, node position, event type (such as "environmental monitoring", "processing start", "quality inspection" and "location update") and specific multi-dimensional data values.

[0050] In a specific example, the collection of multi-source data on ginseng to generate a full-chain multi-source dataset with spatiotemporal and batch labels includes: real-time collection of ginseng status information in planting, processing, warehousing and logistics based on IoT sensors and detection equipment deployed at various nodes of the supply chain to obtain raw multi-source heterogeneous data.

[0051] Based on a unified data communication protocol, the original multi-source heterogeneous data is aggregated to obtain an aggregated data stream with spatiotemporal tags.

[0052] Based on preset data quality rules and unified batch identifiers, the aggregated data stream is cleaned and aligned to obtain a standardized spatiotemporal dataset.

[0053] Based on the unified batch identifier and timestamp, the standardized spatiotemporal dataset is fused and structured to generate a full-chain multi-source dataset with spatiotemporal and batch labels.

[0054] It should be noted that data acquisition refers to the continuous capture of physical or chemical signals reflecting the ginseng's environment, processing, storage location, and quality using sensing devices distributed throughout the supply chain, thereby converting the ginseng's physical state into initial digital information. The raw, multi-source heterogeneous data is characterized by its "multi-source" nature, originating from different types of sensors (such as thermometers, hygrometers, spectrometers, and GPS) and systems (such as Manufacturing Execution Systems, MES), and its "heterogeneous" nature, manifesting in differences in data format, sampling frequency, and semantics. Examples include continuous sensor readings, discrete process parameters, text-based status reports, and image-based quality inspection snapshots.

[0055] Furthermore, data collection is not simply a matter of data accumulation, but rather a targeted and multimodal collaborative sensing process. For example, at the planting stage, environmental data affecting ginseng growth is collected by deploying sensors for soil temperature and humidity, light intensity, and other parameters. At harvest time, rapid, non-destructive testing equipment such as near-infrared spectrometers is used to obtain initial key quality indicator data for ginseng (such as estimated saponin content), and a globally unique batch identifier is generated for each batch. In the processing workshop, processing parameters (such as temperature, duration, and pressure) for steps like washing, drying, and slicing are collected from PLCs (Programmable Logic Controllers) or dedicated sensors via networked equipment. In warehousing and logistics, inventory status (including environmental and location information) is monitored through temperature and humidity sensors and RFID (Radio Frequency Identification) readers in warehouses; and logistics trajectories and environmental changes are continuously tracked through GPS / BeiDou positioning devices and temperature and humidity recorders on transport vehicles. This cross-stage, multi-type data collection system ensures that all key factors affecting the final value of ginseng are captured.

[0056] It should be noted that aggregation refers to the real-time or near-real-time transmission and collection of raw data scattered across different geographical locations such as plantations, processing workshops, and warehouses to a central data processing unit via wired or wireless networks, following a unified communication protocol (such as MQTT and HTTP), ensuring the integrity and timeliness of the data during transmission. The aggregated data stream is a sequence of data packets arranged in chronological order of arrival. Each data packet, in addition to containing the original measurement value, is automatically appended with a timestamp of the acquisition time and the geographical coordinates or logical node identifier of the acquisition device, forming a preliminary spatiotemporal correlation of the data.

[0057] It should be noted that the cleaning operation refers to identifying and removing obvious outliers caused by sensor malfunctions and communication interference based on predefined reasonable thresholds or statistical rules, and interpolating and filling in reasonable missing values ​​caused by brief interruptions to improve the reliability and integrity of the data. The alignment operation refers to using the unique "batch identifier" of each batch of ginseng as the primary key to associate and match various types of data from different nodes and time points but belonging to the same batch, ensuring that soil data, processing parameters, storage environment, transportation trajectory, and multiple quality test results of the same batch of ginseng can accurately correspond on a timeline. A standardized spatiotemporal dataset refers to a data set that, after cleaning and alignment, is time-synchronized, format-uniformed, and where each data point can be clearly attributed to a specific batch of ginseng.

[0058] It's important to note that fusion processing refers to logically concatenating all data entries belonging to the same batch within a standardized spatiotemporal dataset, according to chronological order and flow nodes, to create a complete digital history of that batch of ginseng from planting to the current node. Structured processing involves reorganizing the fused data according to the three dimensions of "batch-time-node," creating an index, and forming a database table or similar structure that facilitates efficient querying and analysis. Each record is clearly labeled with the batch number, precise timestamp, and spatial node information. The full-chain multi-source dataset is a highly structured collection of data containing information about the entire lifecycle of the ginseng supply chain. Its "full-chain" characteristic is reflected in its coverage of all stages from planting to distribution; its "multi-source" characteristic is reflected in the integration of multi-dimensional information such as environment, process, logistics, and quality; and "spatiotemporal and batch tags" ensure that every piece of data is traceable and locatable.

[0059] Furthermore, the core of the fusion processing lies in establishing a unified data association framework. All data is assigned a unified timestamp (using UTC) and the batch number of the ginseng it belongs to upon generation. After data aggregation, it is first cleaned, for example, by removing sensor readings exceeding physical limits (such as storage temperatures of -50°C), or by using the average of preceding and following times to fill in temporary signal loss. The crucial alignment operation follows: the system uses the "batch number" as a unique key to associate all data records generated by the same batch of ginseng at different locations (plantations, processing plants, and warehouses) and at different times (harvest, processing, warehousing, and delivery). This means that for ginseng batch "B001," its growing environment on a specific day, processing parameters in a specific workshop, storage time in the warehouse, transportation route, and quality changes during each sampling inspection can be traced.

[0060] S2. Dynamic mapping modeling of ginseng quality value: Based on the full-chain multi-source dataset, a dynamic mapping function for ginseng quality value is constructed to generate the unit economic value of ginseng.

[0061] In a specific example, the step of constructing a dynamic mapping function for ginseng quality value based on the full-chain multi-source dataset to generate the unit economic value of ginseng includes: extracting the periodic quality test value sequence of each batch of ginseng and its market transaction price data at the corresponding test time point based on the full-chain multi-source dataset to form a training sample set.

[0062] Construct candidate forms of dynamic mapping functions for quality value that include quality value basis functions, time decay functions, and a set of undetermined parameters.

[0063] Based on the training sample set, the set of undetermined parameters in the candidate form of the quality value dynamic mapping function is fitted and solved using a nonlinear regression method, thereby obtaining the quality value dynamic mapping function with determined parameters, and thus generating the unit economic value of ginseng.

[0064] It should be noted that extraction refers to screening out all valid data entries from a structured, multi-source dataset covering the entire supply chain, recording key chemical quality indicators of ginseng (such as total saponin content) and the price and time of transactions for that batch of ginseng in the market. The resulting training sample set is a collection containing multiple data points, each consisting of a triple "(quality_k, duration_k, actual value_k)", where quality_k is denoted as... Let the duration _k be denoted as The actual value _k is denoted as Where "quality_k" represents the th quality_k. The measured values ​​of ginseng quality indicators (such as saponin content) at the time of this test are given. "Duration_k" represents the time elapsed from the harvest benchmark point of this batch of ginseng to the time of this test or transaction (in days). "Actual value_k" represents the actual market transaction price per unit weight (e.g., per kilogram) of this batch of ginseng at time "Duration_k". This set is the sole data basis for training the value mapping function.

[0065] Furthermore, the training data is entirely derived from a multi-source dataset across the entire supply chain; key ginsenoside content sequences and their corresponding market transaction prices are extracted from these datasets. Each valid set of "(quality, duration, actual value)" data represents a market assessment of the value of ginseng of a specific quality and "age". Through learning from a large number of such data points, the predictive model can capture the pattern that market value increases with quality, but the rate of increase gradually slows down (i.e., diminishing marginal utility), and the general trend of value decaying exponentially with time.

[0066] It should be noted that the action of constructing a candidate form of function refers to, based on the biological and market laws that the value of ginseng is mainly driven by its intrinsic quality and decays over time, designing a mathematical function framework in advance that can simultaneously describe the positive contribution relationship of quality to value and the negative decay relationship of time to value.

[0067] It should be noted that the training of the predictive model is the parameter fitting process. With the goal of minimizing the error between the predicted value and the actual transaction value, the optimal parameter set is automatically solved using a nonlinear regression algorithm (such as the Levenberg-Marquardt algorithm). After training, a definite function is obtained: the quality-value dynamic mapping function. It can be used to predict the expected value of a batch of ginseng at a certain point in the future, thereby affecting demand forecasting; the quality value dynamic mapping function is directly embedded in the upper-level optimization objective to accurately calculate the sales revenue that different logistics solutions can bring, so that optimization decisions are directly based on dynamic and accurate value assessment, rather than on rough cost or static classification.

[0068] It should be noted that the fitting and solving process refers to using optimization algorithms such as nonlinear least squares to find a set of optimal parameter values. This makes it possible for all data points in the training sample set... Function prediction value Compared with actual observed values The overall error (such as mean square error) between them is minimized, where This is represented as a parameter to be determined.

[0069] It should be noted that the dynamic mapping function for quality value with defined parameters refers to all undetermined parameters within the candidate form of the function expression. After being replaced with specific numerical values ​​obtained through solving, a completely deterministic and executable mathematical calculation formula is formed. For example, if the solution is obtained... Then the function is specified as .

[0070] In a specific example, the construction of a candidate form for the dynamic mapping function of quality value, which includes a quality value basis function, a time decay function, and a set of undetermined parameters, includes:

[0071] In the formula This represents the predicted economic value per unit of ginseng. This is represented as a dynamic mapping function of quality value. This indicates a quality indicator for human-participating bond chemistry. This represents the elapsed time since the harvesting reference point. It is a quality value basis function. Represented as a time decay function, Represented as a linear scaling factor. This represents the offset of the base value.

[0072] It should be noted that the candidate form couples the effects of quality and time by multiplying the quality value basis function and the time decay function, which can more reasonably simulate the complex relationship between the dynamic changes of ginseng value and quality and time. Furthermore, the quality value basis function is used to characterize the law of value change as quality improves, for example, by using... ,in This refers to chemical quality indicators related to ginseng bonding, such as total saponin content. This is represented as a correction parameter, affecting the curvature of the quality value curve. The quality value basis function reflects the diminishing marginal effect of quality improvement on value. The time decay function is used to characterize the law of value decay over time, for example, using... The exponential decay model simulates the natural decay process of bioactive components. Represented as a shape parameter, it determines the rate at which value decays over time. It is represented as a natural exponential function; the fundamental value offset represents the fundamental value offset that is independent of quality and time, such as including fixed value components such as packaging and brand.

[0073] S3. Multi-step prediction of supply chain status: Based on the full-chain multi-source dataset and the quality value dynamic mapping function, the market demand and the quality of each node ginseng in several decision cycles are predicted, and the predicted supply and demand matrix and the predicted quality matrix are output.

[0074] In a specific example, the process of predicting market demand and ginseng quality at each node over several decision-making cycles and outputting a predicted supply-demand matrix and a predicted quality matrix includes: extracting and summarizing the current inventory status and in-stock ginseng batch information at each node of the supply chain based on the full-chain multi-source dataset, in order to construct a supply chain state vector representing the overall state of the supply chain at the current moment.

[0075] Based on historical order data, market intelligence, and the aforementioned quality value dynamic mapping function, the expected product value range is derived, and the end-market demand quantity for several decision-making cycles is generated.

[0076] Based on the supply chain state vector and the predicted market demand, the inventory level of each node is analyzed.

[0077] Based on the initial quality detection values ​​and current time points of each batch of ginseng recorded in the full-chain multi-source dataset, and combined with the time decay function in the quality value dynamic mapping function, the quality decay of each batch of ginseng at each decision time point in the future is calculated to generate a predicted quality matrix.

[0078] Based on the quantity of demand in the end market, the inventory levels at each node, and the predicted quality matrix, the data are integrated and structured to output a predicted supply and demand matrix for optimized decision-making.

[0079] It should be noted that the extraction and aggregation operations refer to retrieving inventory data for each storage node (such as central warehouses and regional warehouses) at the current point in time from a structured, spatiotemporally and batch-labeled, multi-source dataset covering the entire supply chain. This involves querying detailed information for each batch of ginseng in storage, including its batch number, entry time, initial quality, current estimated quality (based on initial quality and storage time), and location. This forms a comprehensive digital snapshot of the supply chain. The constructed supply chain state vector is a multi-dimensional numerical vector, whose dimensions include at least the inventory quantity at each node, the current quality of each batch of ginseng, and its storage age (number of days).

[0080] It should be noted that time series analysis methods (such as ARIMA and LSTM) are employed, with historical order data as the primary input, and external market intelligence (such as promotional activities and industry reports) is integrated. A new influencing factor is introduced: the "expected product value range," calculated using a quality-value dynamic mapping function. Furthermore, for each forecast period, the system estimates the possible market price range based on the current and expected future quality distribution of ginseng on the market using the quality-value dynamic mapping function. This price range is then used as a feature input to the forecasting model to more accurately capture the impact of price (value) fluctuations on demand. The end-market demand quantity is a two-dimensional data table about future time and geographic markets, forming the core of the forecasting supply and demand matrix. Its rows represent future decision-making periods (e.g., week 1, week 2, etc.), and columns represent different end-markets or sales regions, such as matrix elements. Representatives predicted in the first The first cycle and the first The demand in each market.

[0081] It should be noted that, given the initial supply chain status and future market demand, and based on pre-defined replenishment logic and transportation lead times, computer simulations are used to calculate the expected inventory levels of each supply chain node (warehouse) at the end of each decision-making cycle. This process serves as a bridge connecting demand forecasting with production or replenishment planning. Furthermore, the future inventory levels at each node are crucial information for assessing supply chain fulfillment capabilities and identifying potential stockouts or overstocking risks. Together with the predicted demand, they constitute a complete picture of supply and demand balance analysis, providing the foundational data for flow balance constraints in the optimization model.

[0082] It should be noted that decay calculation refers to, for each specific batch of ginseng with known initial quality and harvest (or storage) time, applying the time decay component of the quality value dynamic mapping function based on its planned or predicted circulation path and time, to calculate its expected quality at a future target decision time. Specifically, "the expected quality at the target decision time is equal to the initial quality multiplied by the time decay component," where the duration of the time decay component is the total time elapsed from the initial moment to the target future moment for that batch.

[0083] In one specific instance, generating the predicted quality matrix includes:

[0084] Through calculation formula The result is the The batch in The node and the first Predictive quality matrix for each decision cycle Represented as the first Initial quality test values ​​for each batch. Representing the current time Represented as from the current moment to the nth The time interval at the start of each decision cycle Represented as the first The harvesting reference time for each, Represented as the first The batch to the future The total storage time from the start of harvesting in each cycle.

[0085] It is important to note that the predicted quality matrix ensures that subsequent optimization decisions are based on accurate predictions of future quality, rather than static or outdated quality information. The predicted quality matrix is ​​a three-dimensional or multi-dimensional data structure that describes the expected quality distribution of available ginseng at different future time points (cycles) and different supply chain nodes. Integrating the biological decay characteristics of ginseng into the key data carrier of the supply chain dynamic optimization model enables optimization decisions to consider "time value" and "quality decay".

[0086] It should be noted that the integration and structuring process refers to linking and encapsulating the demand forecast table (end-market demand quantity), the inventory forecast table (inventory level at each node), and the quality forecast matrix (cycle-node-batch) according to a unified index (time period, spatial node, or market) to form a complete and consistent supply chain status forecast dataset, i.e., the final forecast supply and demand matrix.

[0087] S4. Collaborative Optimization Decision Generation: Based on the aforementioned supply and demand matrix and predicted quality matrix as input, a two-level optimization model is established to obtain the optimal logistics and allocation decision scheme.

[0088] In a specific example, the step of establishing a two-level optimization model based on the supply and demand matrix and the predicted quality matrix as input, and obtaining the optimal logistics and allocation decision scheme, includes: establishing the objective function of the upper-level optimization model and the objective function of the lower-level optimization model based on the predicted supply and demand matrix and the predicted quality matrix as input.

[0089] Define decision variables, which represent the logistics allocation from one node to another in each decision cycle, and set a set of constraints for the two-layer optimization model, including flow balance, capacity limitation, and quality-specific constraints.

[0090] Based on the genetic algorithm framework, a population consisting of multiple potential logistics allocation schemes is initialized, and a specific fitness function is designed. For each individual scheme in the population, the fitness value is calculated based on the fitness function. Genetic operations are then performed on the population based on the fitness value to generate a new offspring population, and this iterative process is repeated.

[0091] When the iteration process meets the preset termination condition, the individual with the highest fitness is selected from the final population, and it is decoded into a specific logistics allocation scheme to output the optimal decision scheme.

[0092] It should be noted that establishing the upper-level objective function refers to formalizing the economic objectives of the supply chain as a mathematical optimization problem. The quality-value dynamic mapping function is directly integrated into the objective function to calculate the expected sales revenue generated by each planned shipment, thereby ensuring that optimization decisions are directly guided by dynamic and accurate economic value, rather than static costs or classifications. Furthermore, the mathematical expression of the established upper-level objective function is as follows: In the formula, This is expressed as the upper-level target value, i.e., the overall net benefit of the supply chain; Directly call the quality value dynamic mapping function trained in S2, based on the first Each cycle starts from the node Transported to the node Predicted quality of goods at that time and transportation time Calculate its expected unit value; Represented as a decision variable, representing the first... Each cycle starts from the node Transported to the node The volume of goods. Represented as a slave node To the node The unit transportation cost; the upper objective function maximizes net revenue by summing the expected value of all planned transport volumes and then subtracting the total transportation cost.

[0093] It should be noted that establishing the lower-level objective function means formalizing the resource efficiency and sustainability goals of the supply chain into a quantifiable optimization problem. This guides the optimization process to pursue economic benefits while also considering the resource costs of logistics activities and the natural depreciation of human resources. Furthermore, the mathematical expression of the established lower-level objective function is as follows: In the formula, Represented as the lower-level target value, it is a weighted comprehensive cost of resource consumption and quality loss; Expressed as total energy consumption, Expressed as transport distance, Expressed as energy consumption per unit distance per unit weight. This is expressed as the weighting coefficient corresponding to the total energy consumption; This is expressed as total potential quality loss. It is derived from the attenuation factor in the dynamic mapping function of quality value, and is relevant to the transportation time. The proportion of value loss within, This represents the weighting coefficient corresponding to the total potential quality loss.

[0094] It's important to note that setting decision variables and constraints defines a feasible solution space for the optimization problem. Decision variables directly reflect all optimization actions. Constraints include: the inflow and outflow of each node must be balanced (flow balance constraint); the shipment volume of each node cannot exceed its inventory or capacity limit (capacity constraint); the total volume shipped to each market must at least meet the predicted demand (demand satisfaction constraint); and the transportation time for high-value ginseng cannot exceed the maximum allowable time based on its quality (quality-specific constraint). These constraints ensure that the generated solution is business-feasible.

[0095] It should be noted that designing the fitness function is the core of achieving collaborative optimization of two-layer objectives. The fitness function constructs a mechanism that allows both upper-layer and lower-layer objectives to be considered simultaneously during the evolutionary process of a single-layer genetic algorithm. Furthermore, the specific fitness function is calculated using the following formula: In the formula, Representative proposal The fitness value is used to measure the quality of a solution. The transformation coefficient converts the lower-level objective value into a dimensionless penalty term, enabling it to numerically synergize with the upper-level objective value (the numerator). A specific fitness function allows the algorithm to automatically favor algorithms that maximize the upper-level objective value (gain) while minimizing the penalty term in the denominator during evolution (i.e.,...). Minimize the amount of the solution .

[0096] It should be noted that fitness value calculation is the sole criterion for the genetic algorithm to evaluate the "goodness" of each candidate solution in the population. The higher the fitness value of a solution, the better it performs in balancing economic benefits and resource consumption, and the greater the probability that it will be selected for reproduction in the subsequent "survival of the fittest" process.

[0097] It's important to note that genetic operations are crucial steps in simulating biological evolution. Selection, based on fitness, identifies superior individuals as "parents"; crossover simulates gene recombination, combining traits from different parents to produce "offspring"; mutation introduces small, random changes, increasing population diversity and preventing premature convergence to local optima. This process iterates continuously, driving the entire population towards a better solution. Furthermore, the solution process follows the standard genetic algorithm framework, but with the fitness function at its core. Starting with a randomly initialized population, each generation involves decoding, calculating fitness, selection, crossover, and mutation until the maximum number of iterations is reached or convergence occurs. Finally, decoding the individual with the highest fitness yields the globally or near-globally optimal solution.

[0098] It should be noted that the output optimal decision solution is a specific, quantitative logistics scheduling plan. It clarifies which warehouses (nodes) and quantities (decision variables) of ginseng should be allocated from which batches in which future decision cycles, and to which markets or processing points (another node) they should be transported to, in order to achieve Pareto optimality in terms of overall supply chain revenue and resource efficiency. This plan serves as a direct link between intelligent decision-making and physical execution.

[0099] This application employs a two-layer structure to decouple the objective function of the upper-layer optimization model, which directly aims to maximize the upper-layer objective value (profit), while embedding a dynamic quality value mapping function as the core value calculation engine. This allows the profit of each logistics decision variable to be dynamically evaluated based on its accurate predicted quality and transportation time. The objective function of the lower-layer optimization model focuses on minimizing resource consumption (such as transportation energy consumption) and the natural loss of human value during the circulation process (lower-layer objective value). This separate yet interconnected structure enables the model to simultaneously characterize the impact of decisions on two dimensions. The final optimal decision output is not simply the solution that maximizes profit, but rather a scheduling scheme that achieves the best balance between high profit and low consumption / low loss under current predictions and constraints. This demonstrates the technological effect of this application in pursuing economic benefits while intrinsically driving the supply chain towards a greener and more sustainable direction.

[0100] S5. Decision Execution and Closed-Loop Feedback: Execute the optimal decision scheme and update the quality value dynamic mapping function and prediction model.

[0101] In a specific instance, executing the optimal decision scheme and updating the quality value dynamic mapping function and prediction model includes: based on the optimal decision scheme, analyzing the cargo allocation information of each decision cycle and each logistics path to generate a set of instructions to be executed for each physical execution node in the supply chain.

[0102] During the execution of the instructions, the actual status of each link in the supply chain is continuously collected to obtain a new round of multi-source datasets for the entire supply chain.

[0103] Based on the new round of full-chain multi-source dataset, the actual execution results are extracted and compared with the planned values ​​in the predicted supply and demand matrix, the predicted quality matrix, and the optimal decision scheme to calculate the deviation data.

[0104] Based on historical data containing the aforementioned deviation data, the parameters of the quality value dynamic mapping function are refitted and updated; simultaneously, the prediction model is retrained and updated.

[0105] It should be noted that parsing refers to decomposing and translating the abstract optimal logistics allocation plan, represented by mathematical model variables, according to business logic (such as by warehouse, vehicle, and production batch), into specific operational commands that can be understood and executed by machines or humans. For example, "transporting 50 kg of ginseng in batch C001 from central warehouse A to regional market B in cycle 1" is transformed into warehouse A's "sorting, packing, and outbound" instructions and the transportation scheduling system's "dispatch and route planning" instructions. The set of instructions to be executed is a structured task list, where each instruction contains a clear operation object (such as warehouse number and vehicle ID), operation content (such as picking and shipping), operation quantity (such as 50 kg), and execution time window (such as cycle 1).

[0106] It should be noted that after generating a set of instructions to be executed for each physical execution node in the supply chain, each instruction in the set of instructions to be executed is sent to the corresponding warehouse management system, transportation management system or production execution system based on a unified instruction interface protocol. The sending action refers to accurately and timely pushing specific operation instructions to the information system responsible for controlling the corresponding physical facilities or processes through a standardized application programming interface or communication protocol. For example, the outbound instruction is sent to the task queue of the warehouse management system through API, and the transportation instruction is sent to the vehicle dispatch terminal through message queue.

[0107] It should be noted that the data sources, types, and structures continuously collected are completely consistent with those in S1. This process is a new round of perception of the supply chain status, ensuring that the system can capture the actual results generated after the plan is executed, as well as new changes in the external environment in real time.

[0108] It should be noted that the new round of full-chain multi-source datasets are digital records generated within the latest time window, reflecting the actual status after the execution of instructions. They contain key information such as actual shipment volume, actual transportation time, actual arrival time, actual warehousing environment, and actual sampling quality, and serve as the data foundation for evaluating the accuracy of predictions and plans and for discovering new knowledge.

[0109] It should be noted that the comparison and calculation of deviations is the core of the closed-loop feedback mechanism. This action systematically compares the "plan" (derived from the scheme) The difference between “predictions” (from the prediction quality matrix and the prediction supply and demand matrix) and “actual” (from the new dataset).

[0110] It should be noted that the deviation data specifically includes, but is not limited to: the deviation between actual market demand and predicted demand, the deviation between actual transportation or storage time and the planned duration, and the deviation between actual sampled quality and predicted quality. These deviation values ​​are key indicators for quantifying model prediction errors and plan execution errors, reflecting the uncertainty of the real world and the degree of disconnect between the model and reality.

[0111] It should be noted that the action of refitting and updating the parameters of the quality-value dynamic mapping function refers to using newly accumulated historical transaction and quality data, which includes deviation data, to construct an expanded training sample set. Based on this new sample set, nonlinear regression (such as S2) is performed again to resolve the parameter set in the quality-value dynamic mapping function, thereby obtaining an updated value mapping function that better reflects the current market value laws and the actual decay characteristics of ginseng.

[0112] It should be noted that the parameter update process enables the value mapping model to have adaptive learning capabilities. For example, if the market's perception of the premium for high-quality ginseng changes, or if the rate of decay of ginseng slows down due to advancements in storage technology, the model parameters will automatically adjust by incorporating the latest quality data, allowing the quality-value dynamic mapping function to more accurately reflect the latest "quality-time-value" relationship.

[0113] It should be noted that retraining and updating the predictive model refers to retraining the time series model (such as LSTM or ARIMA) used in step S3 to predict market demand and supply chain status using historical series data that include the latest actual demand and actual quality degradation data. By incorporating the latest bias information, the model's internal parameters are corrected, thereby improving its accuracy in predicting future market fluctuations and quality change trends.

[0114] It should be noted that updating the forecasting model enables the forecasting system to continuously optimize. The system can learn from historical forecasting errors and continuously adjust its forecasting logic to better adapt to changes in market demand patterns and supply chain operating rules, thereby providing more reliable input for the next round of optimization decisions.

[0115] This application continuously feeds actual execution deviations back to the prediction and modeling stages, enabling the system to constantly adjust model parameters and dynamically adapt to uncertainties such as market changes and process fluctuations, thus achieving iterative evolution of decision-making capabilities. Optimization results are directly converted into actionable instructions and issued, bridging the "last mile" from decision-making to execution. Simultaneously, the closed-loop feedback mechanism ensures that optimization strategies are continuously updated with environmental changes, avoiding the effectiveness decay of static optimization schemes and maintaining the supply chain operation in an optimal or near-optimal state in the long term.

[0116] Please see Figure 2 As shown, in a second aspect, this application provides a system for optimizing the management of the ginseng supply chain.

[0117] The system 100 of the ginseng supply chain optimization management method described in this invention can be installed in an electronic device. Depending on the functions implemented, the ginseng supply chain optimization management method 100 may include a full-chain multi-source data acquisition and fusion module 101, a ginseng quality value dynamic mapping modeling module 102, a supply chain status multi-step prediction module 103, a collaborative optimization decision generation module 104, and a decision execution and closed-loop feedback module 105. The modules described in this invention can also be called units, referring to a series of computer program segments that can be executed by the processor of an electronic device and perform a fixed function, stored in the memory of the electronic device.

[0118] In this embodiment, the functions of each module / unit are as follows: The full-chain multi-source data acquisition and fusion module is used to acquire multi-source data of ginseng to generate a full-chain multi-source dataset with spatiotemporal and batch labels.

[0119] The ginseng quality value dynamic mapping modeling module constructs a ginseng quality value dynamic mapping function based on the full-chain multi-source dataset to generate the unit economic value of ginseng.

[0120] The multi-step supply chain status prediction module, based on the full-chain multi-source dataset and the quality value dynamic mapping function, predicts the market demand and the quality of each node's inputs for several decision cycles, and outputs a predicted supply and demand matrix and a predicted quality matrix.

[0121] The collaborative optimization decision generation module, based on the aforementioned supply and demand matrix and predicted quality matrix as input, establishes a two-layer optimization model and obtains the optimal logistics and allocation decision scheme.

[0122] The decision execution and closed-loop feedback module is used to execute the optimal decision scheme and update the quality value dynamic mapping function and prediction model.

[0123] This application provides a ginseng supply chain optimization management method and system. By constructing a data-driven closed-loop optimization system, it brings significant benefits to the ginseng supply chain, achieving precise quantification and dynamic optimization of ginseng value, and directly improving economic efficiency. Based on a quality-value dynamic mapping model, it can accurately match high-value sources with high-end markets, achieving premium quality and premium prices. Through two-layer collaborative optimization decision-making, resources are allocated globally while ensuring quality constraints, significantly reducing warehousing losses and ineffective logistics, thereby improving the overall net revenue of the supply chain. At the same time, the integration of data across the entire chain and status prediction realize the transformation from "experience-based management" to "precise prediction," enhancing the ability to cope with market fluctuations. The final execution feedback mechanism ensures continuous self-optimization of the model, forming a virtuous cycle of becoming smarter with use, and comprehensively improving the resilience, transparency, and core competitiveness of the supply chain.

[0124] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0125] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0126] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0127] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0128] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0129] Finally, 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.

Claims

1. A method for optimizing the management of the ginseng supply chain, characterized in that, include: S1. Full-chain multi-source data collection and fusion: Collect multi-source data of ginseng to generate a full-chain multi-source dataset with spatiotemporal and batch labels; S2. Dynamic mapping modeling of ginseng quality value: Based on the full-chain multi-source dataset, a dynamic mapping function for ginseng quality value is constructed to generate the unit economic value of ginseng; S3. Multi-step prediction of supply chain status: Based on the full-chain multi-source dataset and the quality value dynamic mapping function, the market demand and the quality of each node's input are predicted for several decision cycles, and the predicted supply and demand matrix and the predicted quality matrix are output. S4. Collaborative Optimization Decision Generation: Based on the aforementioned supply and demand matrix and predicted quality matrix as input, a two-level optimization model is established to obtain the optimal logistics and allocation decision scheme. S5. Decision Execution and Closed-Loop Feedback: Execute the optimal decision scheme and update the quality value dynamic mapping function and prediction model.

2. The ginseng supply chain optimization management method according to claim 1, characterized in that, The full-chain multi-source dataset includes environmental data, processing parameters, inventory status, logistics trajectory, and key quality indicator data.

3. The ginseng supply chain optimization management method according to claim 1, characterized in that, The collection of multi-source data from ginseng is used to generate a full-chain multi-source dataset with spatiotemporal and batch labels, including: Based on IoT sensors and detection equipment deployed at various nodes of the supply chain, the status information of ginseng in planting, processing, storage and logistics is collected in real time to obtain raw multi-source heterogeneous data. Based on a unified data communication protocol, the original multi-source heterogeneous data is aggregated to obtain an aggregated data stream with spatiotemporal tags; Based on preset data quality rules and unified batch identifiers, the aggregated data stream is cleaned and aligned to obtain a standardized spatiotemporal dataset. Based on the unified batch identifier and timestamp, the standardized spatiotemporal dataset is fused and structured to generate a full-chain multi-source dataset with spatiotemporal and batch labels.

4. The ginseng supply chain optimization management method according to claim 1, characterized in that, The step of constructing a dynamic mapping function for ginseng quality value based on the full-chain multi-source dataset to generate the unit economic value of ginseng includes: Based on the aforementioned full-chain multi-source dataset, the periodic quality test value sequence of each batch of ginseng and its market transaction price data at the corresponding test time point are extracted to form a training sample set; Construct candidate forms of dynamic mapping functions for quality value that include quality value basis functions, time decay functions, and a set of undetermined parameters; Based on the training sample set, the set of undetermined parameters in the candidate form of the quality value dynamic mapping function is fitted and solved using a nonlinear regression method, thereby obtaining the quality value dynamic mapping function with determined parameters, and thus generating the unit economic value of ginseng.

5. The ginseng supply chain optimization management method according to claim 4, characterized in that, The construction of a candidate form for a dynamic mapping function of quality value, which includes a quality value basis function, a time decay function, and a set of undetermined parameters, includes: In the formula This represents the predicted economic value per unit of ginseng. This is represented as a dynamic mapping function of quality value. This indicates a quality indicator for human-participating bond chemistry. This represents the elapsed time since the harvesting reference point. It is a quality value basis function. Represented as a time decay function, Represented as a linear scaling factor. This represents the offset of the base value.

6. The ginseng supply chain optimization management method according to claim 1, characterized in that, The process of predicting market demand and ginseng quality at each stage of several decision-making cycles, and outputting a predicted supply-demand matrix and a predicted quality matrix, includes: Based on the full-chain multi-source dataset, the current inventory status and in-stock ginseng batch information of each node in the supply chain are extracted and summarized to construct a supply chain state vector of the overall supply chain status at the current moment. Based on historical order data, market intelligence, and the aforementioned quality value dynamic mapping function, the expected product value range is derived, and the end-market demand quantity for several decision-making cycles is generated. Based on the supply chain state vector and the predicted market demand, the inventory level of each node is analyzed and determined. Based on the initial quality detection values ​​and current time points of each batch of ginseng recorded in the full-chain multi-source dataset, and combined with the time decay function in the quality value dynamic mapping function, the quality decay of each batch of ginseng at each decision time point in the future is calculated to generate a predicted quality matrix. Based on the quantity of demand in the end market, the inventory levels at each node, and the predicted quality matrix, the data are integrated and structured to output a predicted supply and demand matrix for optimized decision-making.

7. The ginseng supply chain optimization management method according to claim 6, characterized in that, The process of generating the predicted quality matrix includes: Through calculation formula The result is the The batch in The node and the first Predictive quality matrix for each decision cycle Represented as the first Initial quality test values ​​for each batch. Representing the current time Represented as from the current moment to the nth The time interval at the start of each decision cycle Represented as the first The harvesting reference time for each, Represented as the first The batch to the future The total storage time from the start of harvesting in each cycle.

8. The ginseng supply chain optimization management method according to claim 1, characterized in that, The method, based on the supply and demand matrix and the predicted quality matrix as input, establishes a two-level optimization model and obtains the optimal logistics and allocation decision scheme, including: Using the predicted supply and demand matrix and the predicted quality matrix as input, establish the objective function of the upper-level optimization model and the objective function of the lower-level optimization model; Set decision variables, which represent the logistics allocation from one node to another in each decision cycle, and set a set of constraints for the two-level optimization model, including flow balance, capacity limitation and quality-specific constraints. Based on the genetic algorithm framework, a population consisting of multiple potential logistics allocation schemes is initialized, and a specific fitness function is designed. For each individual scheme in the population, the fitness value is calculated based on the fitness function. Genetic operations are then performed on the population based on the fitness value to generate a new offspring population, and this iterative process is repeated. When the iteration process meets the preset termination condition, the individual with the highest fitness is selected from the final population, and it is decoded into a specific logistics allocation scheme to output the optimal decision scheme.

9. The ginseng supply chain optimization management method according to claim 1, characterized in that, The execution of the optimal decision-making scheme and the updating of the quality value dynamic mapping function and prediction model include: Based on the optimal decision-making scheme, the cargo allocation information of each decision cycle and each logistics path is analyzed to generate a set of instructions to be executed for each physical execution node in the supply chain. During the execution of the instructions, the actual status of each link in the supply chain is continuously collected to obtain a new round of multi-source datasets for the entire supply chain. Based on the new round of full-chain multi-source dataset, the actual execution results are extracted and compared with the planned values ​​in the predicted supply and demand matrix, the predicted quality matrix, and the optimal decision scheme to calculate the deviation data. Based on historical data containing the aforementioned deviation data, the parameters of the quality value dynamic mapping function are refitted and updated; simultaneously, the prediction model is retrained and updated.

10. A system for implementing the ginseng supply chain optimization management method according to any one of claims 1-9, characterized in that, include: The full-chain multi-source data acquisition and fusion module is used to collect multi-source data of ginseng to generate a full-chain multi-source dataset with spatiotemporal and batch labels; The ginseng quality value dynamic mapping modeling module constructs a ginseng quality value dynamic mapping function based on the full-chain multi-source dataset to generate the unit economic value of ginseng. The supply chain status multi-step prediction module, based on the full-chain multi-source dataset and the quality value dynamic mapping function, predicts the market demand and the quality of each node's inputs in several decision cycles, and outputs the predicted supply and demand matrix and the predicted quality matrix. The collaborative optimization decision generation module, based on the aforementioned supply and demand matrix and predicted quality matrix as input, establishes a two-layer optimization model and obtains the optimal logistics and allocation decision scheme. The decision execution and closed-loop feedback module is used to execute the optimal decision scheme and update the quality value dynamic mapping function and prediction model.

Images