Cold chain shelf position optimization method and system based on digital twinning

Abstract

The invention relates to the technical field of intelligent three-dimensional warehouses, and discloses a cold chain shelf position optimization method and system based on digital twinning, and the method comprises the steps: extracting historical order data, and obtaining a historical order data set; performing goods purchase correlation analysis on the historical order data set to obtain a correlation feature set of purchase correlation goods pairs; calculating a spatial adjacency coefficient of the purchased associated goods pair according to the associated feature set; establishing a digital twinborn model of the cold chain three-dimensional warehouse, planning shelf positions in the digital twinborn model according to the spatial adjacency coefficient, and obtaining a plurality of shelf position adjustment schemes; in the digital twin model, evaluating the plurality of shelf position adjustment schemes based on path planning, and obtaining an optimal shelf position layout according to an evaluation result; and executing the optimal shelf position layout in a physical domain cold chain three-dimensional warehouse. According to the method and the system, the limitation that a cold chain three-dimensional warehouse depends on artificial experience to carry out shelf position layout is solved, and the order picking distance and time are remarkably reduced.

Description

Technical Field

[0001] The present invention relates to the field of intelligent vertical warehouse technology, and in particular to a cold chain shelf location optimization method and system based on digital twins. Background Art

[0002] Cold chain logistics is a crucial component of modern logistics systems, especially when transporting and storing temperature-controlled products (such as fresh food). Its efficiency directly impacts product quality and distribution costs. Currently, cold chain vertical warehouses, as a crucial component of cold chain logistics, require rational shelf location planning to improve storage space utilization, reduce material picking time, minimize cargo damage, and enhance overall operational efficiency.

[0003] However, traditional cold chain warehouse management often relies on manual experience to arrange shelf locations, lacking scientific, data-driven optimization methods. This approach can not only lead to low space utilization and operational inconvenience, but also, due to a lack of systematic optimization thinking, can affect overall work efficiency and the quality of warehouse environment management. With the continuous development of information technology and the Internet of Things, digital twin technology is gradually emerging in the industrial field, becoming an effective tool for improving warehouse management and optimizing resource allocation.

[0004] By creating virtual models of physical objects, digital twin technology can reflect and analyze warehouse operating status in real time, helping decision-makers simulate different scenarios in a virtual environment and optimize various warehouse management indicators. In cold chain vertical warehouses in particular, digital twin technology can simulate shelf position optimization, rationally plan warehouse space, and improve warehouse operation efficiency and product access efficiency. Currently, cold chain vertical warehouse optimization solutions based on digital twin technology are still in the exploratory stage, focusing primarily on improving warehouse management and operational efficiency. However, in practical applications, specific shelf position optimization strategies and implementation paths are lacking. Summary of the Invention

[0005] To this end, the purpose of the present invention is to overcome the limitations of the cold chain warehouse shelf layout in the prior art, and to provide a cold chain shelf location optimization method and system based on digital twins, which significantly reduces the order picking distance and time.

[0006] In the first aspect, in order to solve the above technical problems, the present invention provides a cold chain shelf location optimization method based on digital twins, comprising: Extract historical order data to obtain a historical order data set; the historical order data includes order time, product category, quantity, and storage method; Performing product purchase correlation analysis on the historical order data set to obtain a set of correlation features of purchase-related product pairs; Calculating the spatial adjacency coefficient of the purchase-related product pairs according to the associated feature set; Establishing a digital twin model of the cold chain vertical warehouse, planning shelf positions in the digital twin model according to the spatial adjacency coefficient, and obtaining multiple shelf position adjustment schemes; In the digital twin model, the plurality of shelf position adjustment schemes are evaluated based on path planning, and an optimal shelf position layout is obtained according to the evaluation results; The optimal shelf position layout is implemented in the physical domain cold chain warehouse.

[0007] In one embodiment of the present invention, a product purchase association analysis is performed on the historical order data set to obtain a set of associated features of purchase-related product pairs, including assigning a time-decay weight to each piece of the historical order data to obtain a weighted order data set; frequent item sets of product combinations in the weighted order data set are mined layer by layer based on a priori algorithms to obtain a set of candidate association rules; the candidate association rule set is screened by multiple indicators to obtain a preliminary screening rule set; the screening indicators include support, confidence, and lift; density clustering analysis is performed on the preliminary screening rule set to obtain a strong association rule set; and the set of associated features of the purchase-related product pairs is extracted from the strong association rule set.

[0008] In one embodiment of the present invention, a product purchase association analysis is performed on the historical order dataset to obtain a set of association features of purchase-associated product pairs, and the process also includes calculating the association strength of each product pair in the strong association rule set; marking product pairs with an association strength greater than or equal to a preset association strength as purchase-associated product pairs; and mapping the association rules of the purchase-associated product pairs into dynamic connecting lines between virtual shelves in the digital twin model; wherein the line width of the connecting line is positively correlated with the association strength.

[0009] In one embodiment of the present invention, the spatial adjacency coefficient of the purchase-related product pairs is calculated based on the association feature set, including that each association feature of the association feature set includes the support, confidence and lift of the purchase-related product pairs; and the spatial adjacency coefficient is calculated based on the support, confidence and lift.

[0010] In one embodiment of the present invention, calculating the spatial adjacency coefficient of the purchase-related product pair further includes, if the purchase-related product pair has a bidirectional association feature, merging the bidirectional spatial adjacency coefficients of the purchase-related product pair to obtain a merged spatial adjacency coefficient; and using the merged spatial adjacency coefficient as the actual spatial adjacency coefficient. The merged spatial adjacency coefficient is obtained based on the following method: ; Among them, W XY ' represents the combined spatial adjacency coefficient of the purchase-related product pair XY; X and Y represent the products of the purchase-related product pair; WXY W represents the spatial adjacency coefficient of X as the premise product and Y as the conclusion product; YX Indicates the spatial adjacency coefficient where Y is the premise product and X is the conclusion product; Represents the attenuation coefficient.

[0011] In one embodiment of the present invention, the shelf positions are planned in the digital twin model according to the spatial adjacency coefficient, including determining whether the storage methods of the purchase-related product pairs are the same: if the storage methods are the same, the shelves of the purchase-related product pairs are adjusted to storage units of the same logical partition according to the spatial adjacency coefficient; if the storage methods are different, the shelves of the purchase-related product pairs are adjusted to storage units of different logical partitions according to the spatial adjacency coefficient.

[0012] In one embodiment of the present invention, the shelves of the purchase-related product pairs are adjusted to storage units of the same logical partition according to the spatial adjacency coefficient, including dividing the purchase-related product pairs into strongly associated product pairs, medium associated product pairs and weakly associated product pairs according to the spatial adjacency coefficient; in the digital twin model, the shelves of the strongly associated product pairs are adjusted to adjacent storage units; the shelves of the medium associated product pairs are adjusted to non-adjacent storage units; and the shelves of the weakly associated product pairs are filled with vacancies; wherein the adjacent storage units are defined as directly connected nodes in the virtual shelf topology structure; and the non-adjacent storage units are defined as connected through at least one intermediate node in the virtual shelf topology structure.

[0013] In one embodiment of the present invention, in the digital twin model, the rationality of the multiple shelf position adjustment schemes is evaluated based on path planning to obtain the optimal shelf position layout, including loading multiple shelf position adjustment schemes in the digital twin model; for each of the shelf position adjustment schemes, simulating AGV in the digital twin model to execute multiple order data picking to obtain picking evaluation results of the multiple shelf position adjustment methods; the evaluation indicators of the picking evaluation results include picking distance, temperature control stability and AGV energy consumption; and extracting the shelf position adjustment scheme with the best picking evaluation result as the optimal shelf position layout.

[0014] In one embodiment of the present invention, the AGV performs picking according to a valid path, and determining the valid path includes generating an AGV picking path based on a genetic algorithm in the digital twin model to obtain a set of candidate paths; introducing a path planning cost function, and calculating the average path cost and standard deviation of the candidate paths; screening valid paths from the candidate path set according to the average path cost and standard deviation to obtain a screened path set; arranging the path costs of the screened path set in descending order, and extracting the first path as the valid path.

[0015] In the second aspect, based on the same inventive concept, in order to solve the above technical problems, the present invention provides a cold chain shelf location optimization system based on digital twins, comprising: A data extraction module extracts historical order data and obtains a historical order data set, wherein the historical order data includes order time, product category, quantity, and storage method; An association analysis module performs product purchase association analysis on the historical order data set to obtain a set of association features of purchase-related product pairs; an adjacency coefficient calculation module for calculating the spatial adjacency coefficient of the purchase-related product pairs based on the associated feature set; Digital twin modeling module, used to build a digital twin model of the cold chain warehouse; an intelligent planning module, interacting with the adjacency coefficient calculation module and the digital twin modeling module, and configured to generate a plurality of shelf position adjustment schemes in the digital twin model according to the spatial adjacency coefficient; a scheme evaluation module, configured to evaluate the plurality of shelf position adjustment schemes according to path planning in the digital twin model, and determine an optimal shelf position layout based on multi-objective optimization; The drive execution module is configured to convert the optimal shelf position layout into a storage equipment control instruction, and drive the shelves of the cold chain warehouse to be adjusted according to the optimal shelf position layout.

[0016] The above technical solution of the present invention has the following beneficial effects compared with the prior art: The digital twin-based cold chain shelf location optimization method and system described in this paper uses purchase correlation analysis and spatial adjacency coefficient calculation based on historical order data to provide an optimized solution for shelf location layout based on actual needs, significantly reducing order picking distance and time, and lowering operating costs. Furthermore, different shelf layout solutions are verified through digital twin model simulation, reducing implementation risks while ensuring the scientific and operational feasibility of the final solution.

[0017] Among them, by exploring the purchase correlation of goods and quantifying the spatial adjacency coefficient, the shelves are adjusted to store high-frequency co-occurring goods nearby, which greatly shortens the picking path, reduces the moving distance and operation time of stacking cranes or AGV equipment, and reduces the equipment idle rate and temperature fluctuation risks in the cold chain environment.

[0018] Dynamically adjust shelf layout based on historical order data characteristics to avoid the rigidity of static zoning strategies, flexibly respond to seasonal order fluctuations or changes in category demand, and improve warehouse resource utilization.

[0019] By using the digital twin model to conduct virtual simulation and path planning evaluation of multiple sets of shelf adjustment plans, the optimal layout plan can be quickly screened before implementation in the physical domain, reducing the number of actual handling times and trial-and-error costs, while also reducing the loss of cold storage caused by frequent entry and exit of goods in cold chain warehousing. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments of the present invention in conjunction with the accompanying drawings, wherein: Figure 1 This is a flow chart of a cold chain shelf location optimization method based on digital twins in a preferred embodiment of the present invention; Figure 2 Flowchart for obtaining a set of associated features in a preferred embodiment of the present invention; Figure 3 A flow chart for obtaining a spatial adjacency coefficient in a preferred embodiment of the present invention; Figure 4 This is a flow chart of planning shelf positions according to spatial adjacency coefficients in a preferred embodiment of the present invention; Figure 5 A flow chart for obtaining an optimal shelf position layout in a preferred embodiment of the present invention; Figure 6 A flow chart for determining a valid path in a preferred embodiment of the present invention; Figure 7 This is a structural block diagram of the cold chain shelf location optimization system based on digital twin in the preferred embodiment of the present invention. DETAILED DESCRIPTION

[0021] The present invention will be further described below with reference to the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it. However, the embodiments are not intended to limit the present invention.

[0022] Example 1: The purpose of the embodiment of the present invention is to solve the limitation of cold chain warehouse relying on manual experience to arrange shelf positions. Figure 1 As shown, the embodiment of the present invention discloses a cold chain shelf location optimization method based on digital twin, comprising: S10. Extract historical order data to obtain a historical order data set; the historical order data includes order time, product category, quantity, and storage method; S20: Perform product purchase correlation analysis on the historical order dataset to obtain a set of correlation features of purchase-related product pairs; S30, calculating the spatial adjacency coefficient of the purchase-related product pairs based on the associated feature set; S40: Establish a digital twin model of the cold chain vertical warehouse, plan shelf positions in the digital twin model according to the spatial adjacency coefficient, and obtain multiple shelf position adjustment plans; S50. In the digital twin model, evaluating the multiple shelf position adjustment schemes based on path planning, and obtaining an optimal shelf position layout according to the evaluation results; S60: Execute the optimal shelf position layout in the physical domain cold chain warehouse.

[0023] In specific application scenarios, historical order data is extracted from the cold chain warehouse's management system or other related platforms. Each historical order data entry includes the order time, product category, quantity, and storage method. For example, the order time can include the specific year, month, day, and time period. The product category can cover various cold chain products such as cola, ice cream, iced tea, dairy products, beef, and steak. The quantity represents the quantity of goods per order, and the storage method includes refrigeration, freezing, or room temperature storage. By extracting historical order data, we can obtain real-time operational information of the cold chain warehouse and build a background data set for product storage, providing a foundation for subsequent analysis.

[0024] By analyzing the order content in historical order datasets, association rule analysis algorithms (such as the Apriori algorithm and FP-growth algorithm) are used to identify frequently purchased pairs of items and extract the correlation features between each pair. These correlation features can include the purchase frequency of the item pair, the overlap in purchase times, and the similarity of the product categories. For example, if a customer frequently purchases cola alongside refrigerated meat over a certain period of time, it can be assumed that there is a strong correlation between these two categories of items. This information is valuable for shelf layout in cold chain warehouses. By establishing a set of correlation features between items, it is possible to effectively identify which items should be stored close together, thereby reducing storage and retrieval time.

[0025] The spatial adjacency coefficient of pairs of items in the associated feature set is calculated to evaluate their relative positions in the warehouse. The spatial adjacency coefficient is a quantifiable priority indicator for items to be stored in similar shelf locations. For example, if two items are frequently purchased together, their spatial adjacency coefficient is high, meaning they should be placed on adjacent shelves. After calculating the spatial adjacency coefficient, it is possible to assign appropriate relative positions to items in the cold chain warehouse, optimize shelf layout, and arrange items with high adjacency coefficients in close proximity, which helps improve storage and retrieval efficiency and warehouse management.

[0026] A digital twin model of the cold chain warehouse is constructed. This model is a three-dimensional virtual warehouse model created based on the actual warehouse layout, spatial constraints, storage conditions, and other information. Subsequently, the calculated spatial adjacency coefficient is used to plan shelf locations based on the warehouse's spatial characteristics (such as the number, location, and size of shelves). The digital twin model automatically plans the shelf layout based on the spatial adjacency coefficient and generates multiple shelf position adjustment plans.

[0027] Use path planning algorithms (such as the A* algorithm and the Dijkstra algorithm) to evaluate the rationality of multiple shelf position adjustment plans. The core of the path planning evaluation is to consider factors such as picking distance, temperature control stability, and AGV energy consumption, optimize the path for storing and accessing goods, and thus obtain the optimal shelf layout plan.

[0028] Based on the above optimization results, the optimal shelf position layout is applied to the physical domain cold chain warehouse. Warehouse staff adjust shelves and store goods according to the planned optimal shelf position layout, thereby realizing the execution of the optimization plan in the digital twin model in the actual warehouse.

[0029] The digital twin-based cold chain shelf location optimization method and system described in this invention uses purchase correlation analysis and spatial adjacency coefficient calculation based on historical order data to provide an optimized solution for shelf location layout based on actual needs, significantly reducing order picking distance and time, and lowering operating costs. Furthermore, different shelf layout schemes are simulated and verified using a digital twin technology model, ensuring the scientific and operational feasibility of the final solution while reducing implementation risks.

[0030] Among them, by exploring the purchase correlation of goods and quantifying the spatial adjacency coefficient, the shelves are adjusted to store high-frequency co-occurring goods nearby, which greatly shortens the picking path, reduces the moving distance and operation time of stacker cranes or AGV equipment, and reduces the equipment idle rate and temperature fluctuation risks in the cold chain environment.

[0031] Dynamically adjust shelf layout based on historical order data characteristics to avoid the rigidity of static zoning strategies, flexibly respond to seasonal order fluctuations or changes in category demand, and improve warehouse resource utilization.

[0032] By using the digital twin model to conduct virtual simulation and path planning evaluation of multiple sets of shelf adjustment plans, the optimal layout plan can be quickly screened before implementation in the physical domain, reducing the number of actual handling times and trial-and-error costs, while also reducing the loss of cold storage caused by frequent entry and exit of goods in cold chain warehousing.

[0033] Specifically, refer to Figure 2As shown, the historical order data set is analyzed for product purchase association to obtain a set of associated features of purchase-associated product pairs, including assigning a time-decay weight to each piece of the historical order data to obtain a weighted order data set; based on a priori algorithms, frequent item sets of product combinations in the weighted order data set are mined layer by layer to obtain a set of candidate association rules; the candidate association rule set is screened by multiple indicators to obtain a preliminary screening rule set; the screening indicators include support, confidence, and lift; a density clustering analysis is performed on the preliminary screening rule set to obtain a strong association rule set; and the associated feature set of the purchase-associated product pairs is extracted from the strong association rule set.

[0034] In specific application scenarios, a time decay model is used to weight historical order data to ensure that more recent order data has a greater impact on product relevance analysis. By assigning decaying weights to each order based on its time, we ensure that more recent order data has a stronger influence on the analysis results, while more distant order data gradually loses influence. For example, weights can be calculated using an exponential decay function, giving greater weight to order data closer to the current time. By assigning time decay weights, we can effectively prevent outdated data from interfering with analysis results, improving the timeliness of data analysis. This allows cold chain warehouse shelf layout optimization to more accurately reflect current market demand and actual product storage trends.

[0035] The Apriori algorithm is used to mine the weighted order dataset layer by layer to identify frequently occurring combinations of items. First, frequent itemsets are determined based on a set minimum support threshold. Frequent itemsets are combinations of items that appear in the weighted order dataset more than the set threshold. These combinations are highly correlated.

[0036] Next, based on these frequent item sets, candidate association rule sets are generated. Once these candidate association rule sets are obtained, they are screened using multiple metrics, such as support, confidence, and lift. By setting thresholds for support, confidence, and lift, a preliminary set of rules that meet the criteria is selected. Association rules represent relationships between items (usually products, behaviors, or events) that occur under specific conditions. For example, in order data analysis, association rules can reveal that after a customer purchases certain products (e.g., cola), they are also likely to purchase other products (e.g., ice cream). The basic form of an association rule is: A→B, where A and B are product sets or item sets, and A→B means "if you purchase product A, you are very likely to purchase product B." An association rule consists of two parts: the antecedent and the consequent. The antecedent is the left side of the rule, representing a condition or event that has already occurred, for example, the customer purchased product A; the consequent is the right side of the rule, representing an event that is likely to occur if the antecedent occurs, for example, the customer is very likely to purchase product B. The goal of association rules is to discover valuable relationships hidden in large data sets, help identify potential product combinations or behavior patterns, and thus achieve rational allocation of resources.

[0037] Association rules have three main metrics: support, confidence, and lift: Support: This indicates how often the item combination included in the association rule appears in total orders. A rule with high support indicates that the combination is common in actual orders. Support (A→B) = Frequency of orders containing both A and B / Total number of orders.

[0038] Confidence: This represents the probability of the consequent occurring when the antecedent in an association rule occurs. For example, confidence represents the proportion of orders for item A that also purchase item B. Confidence (A→B) = Number of orders supporting both A and B / Number of orders supporting A.

[0039] Lift: This indicates the strength of the association between the antecedent and consequent in an association rule, measuring their independence. A lift greater than 1 indicates a positive correlation, a lift equal to 1 indicates no correlation, and a lift less than 1 indicates a negative correlation. Lift (A→B) = Confidence (A→B) / Support (B).

[0040] Multi-index screening can extract high-quality association rules from numerous candidate rules, ensuring that subsequent optimization decisions are more accurate and effective. By comprehensively considering support, confidence, and lift, it can ensure that the association rules finally screened out have strong business relevance and practical application value.

[0041] After completing the preliminary association rule screening, density clustering analysis (such as the DBSCAN algorithm or the K-means algorithm) is used to cluster the preliminary screening rules. For example, based on the DBSCAN algorithm, density clustering analysis is performed to cluster core point rules with a confidence level greater than or equal to 40% as strong association rules; cluster edge point rules with a confidence level greater than or equal to 30% and less than 40% as weak association rules; and noise point rules with a lift less than 1 as negative association rules.

[0042] Density cluster analysis can cluster rules with similar purchasing patterns and association characteristics based on the similarities between association rules. In this way, a set of strongly associated rules can be mined from a large number of preliminary rule screenings. These rules reflect the high correlation between product purchase behaviors. Density cluster analysis can effectively identify strong association patterns hidden in complex data, avoiding the redundant and low-quality rules that may exist in traditional rule screening methods. The strong association rules obtained through cluster analysis are more in line with actual needs and can better guide the optimization of cold chain warehouse shelf layout.

[0043] Finally, specific pairs of purchase-related items are extracted from the strong association rule set to form a set of associated features. By extracting specific item-pair association features from the strong association rule set, sufficient data support is provided for precise optimization of shelf layout. This set of associated features provides warehouse managers with a clear reference, enabling more scientific shelf location planning, avoiding unnecessary frequent stocking and retrieval, and improving overall operational efficiency.

[0044] Furthermore, a product purchase correlation analysis is performed on the historical order data set to obtain a set of correlation features of purchase-related product pairs, which also includes calculating the correlation strength of each product pair in the strong association rule set; marking product pairs with correlation strength greater than or equal to a preset correlation strength as purchase-related product pairs; in the digital twin model, mapping the association rules of the purchase-related product pairs into dynamic connecting lines between virtual shelves; wherein the line width of the connecting line is positively correlated with the correlation strength.

[0045] In specific application scenarios, by calculating the strength of each strong association rule in a set of strong association rules to measure their reliability and effectiveness, we can effectively screen out product pairs with strong associations. Lift can be used as the association strength metric: Lift (A→B) = Confidence (A→B) / Suppor (B), where A and B are product pairs; Confidence (A→B) represents the probability of purchasing product B after purchasing product A; and Suppor (B) represents the frequency of product B appearing in all orders.

[0046] After calculating the association strength for each item pair, pairs with an association strength greater than or equal to a preset association strength are marked as purchase-linked. The preset association strength is set based on business needs and warehouse operations to ensure that only item pairs with a strong purchase association proceed to the subsequent shelf layout optimization phase. This avoids the introduction of weak association rules and ensures the accuracy and practicality of the optimization results.

[0047] In the digital twin model, the association rules for purchasing related pairs of items are mapped as dynamic lines between virtual shelves. Each pair of association rules is virtually linked to the shelf location in the warehouse via a line. These lines represent the relationship between the items and are dynamically reflected in the digital twin warehouse model. By mapping association rules as lines between virtual shelves, the relationships and dependencies between items can be intuitively displayed in the digital twin model, making it easier for warehouse managers to observe product layout and optimization results.

[0048] The width of each connecting line is proportional to the strength of the association. That is, the greater the strength of the association, the wider the connecting line, indicating a stronger correlation between the items and the closer the shelf locations should be. In this way, the digital twin model can intuitively demonstrate which pairs of items are most closely related and which pairs should be placed on adjacent shelves. This visual display helps warehouse managers better understand the storage needs of items.

[0049] Reference Figure 3 As shown, the spatial adjacency coefficient of the purchase-related product pair is calculated based on the associated feature set, including that each associated feature of the associated feature set includes the support, confidence and lift of the purchase-related product pair; the spatial adjacency coefficient is calculated based on the support, confidence and lift.

[0050] The spatial adjacency coefficient is calculated and determined according to the following method: ; C and D represent the premise and conclusion products of the purchase of related product pairs respectively; Sc C→D S represents the spatial adjacency coefficient of goods to CD; C→D Indicates the product's support for CD; C C→D Indicates the confidence of the product on CD; L C→D Indicates the degree to which the product enhances the CD; They represent the weight coefficients of support, confidence and lift respectively.

[0051] In specific application scenarios, the spatial adjacency coefficient is used to quantify the priority index of goods stored in similar shelf locations, reflecting the strength of the association between them. The calculation of the spatial adjacency coefficient is based on three indicators: support, confidence, and lift. Assign values ​​based on business experience. For example, for promotional items, high confidence is prioritized and low support is tolerated. Take 0.3, 0.6 and 0.1 respectively; for daily goods, high support is given priority. Take 0.2, 0.2 and 0.6 respectively; for high-value goods, give priority to high correlation and weaken low-frequency rules. Take 0.5, 0.3 and 0.2 respectively.

[0052] In addition, refer to Figure 3 As shown, calculating the spatial adjacency coefficient of the purchase-related product pair also includes determining whether the purchase-related product pair has a bidirectional association feature. If so, merging the bidirectional spatial adjacency coefficients of the purchase-related product pair to obtain a merged spatial adjacency coefficient; using the merged spatial adjacency coefficient as the actual spatial adjacency coefficient; wherein the merged spatial adjacency coefficient is obtained based on the following method: ; Among them, W XY ' represents the combined spatial adjacency coefficient of the purchase-related product pair XY; X and Y represent the products of the purchase-related product pair; W XY W represents the spatial adjacency coefficient of X as the premise product and Y as the conclusion product; YX Indicates the spatial adjacency coefficient where Y is the premise product and X is the conclusion product; Represents the attenuation coefficient.

[0053] In specific application scenarios, if the purchase behavior of product X and product Y exhibits a bidirectional correlation, it indicates a strong mutual influence between the two products. For example, if a customer purchases product X, they typically also purchase product Y; and if they also purchase product Y, they are also likely to purchase product X. Such product pairs are marked as "bidirectionally correlated."

[0054] When a pair of purchase-related items has bidirectional association rules (X→Y and Y→X), the sum of the primary direction weight and the attenuated secondary direction weight is taken to avoid directly adding them together, which would result in an artificially high weight. A weight exceeding 1.0 would lead to a false judgment of "must be forced to be adjacent." By separating the primary and secondary directions and implementing a dynamic attenuation mechanism (for example, the attenuation coefficient can be dynamically adjusted based on different scenarios, with a default of 0.2 to avoid repeated accumulation), the dominance of the strong association direction is retained while partially absorbing the additional impact of the reverse rule. This balances computational efficiency and business flexibility while preserving the directional semantics of the association rules.

[0055] Further, refer to Figure 4 As shown, in the digital twin model, the shelf positions are planned according to the spatial adjacency coefficient, including determining whether the storage methods of the purchase-related product pairs are the same: if the storage methods are the same, the shelves of the purchase-related product pairs are adjusted to the storage units of the same logical partition according to the spatial adjacency coefficient; if the storage methods are different, the shelves of the purchase-related product pairs are adjusted to the storage units of different logical partitions according to the spatial adjacency coefficient.

[0056] In a specific application scenario, the warehouse's storage units are first logically partitioned. Logical partitions are divided based on the storage method of the goods. Storage units in the same logical partition have the same storage method (such as refrigerated, frozen, or ambient temperature). In the digital twin model, these shelves share the same set of storage constraints. Storage units in different logical partitions have different storage methods. In the digital twin model, these shelves have mutually exclusive storage constraints. Goods with different storage methods should be placed in separate areas to avoid improper storage caused by temperature differences. For example, frozen foods and ambient temperature foods should be placed in the freezer and ambient temperature areas, respectively. Even if they have strong correlations, shelf positions need to be adjusted based on the storage method.

[0057] Further, refer to Figure 4 As shown, the shelves of the purchase-related product pairs are adjusted to the storage units of the same logical partition according to the spatial adjacency coefficient, including dividing the purchase-related product pairs into strongly associated product pairs, medium associated product pairs and weakly associated product pairs according to the spatial adjacency coefficient; in the digital twin model, the shelves of the strongly associated product pairs are adjusted to adjacent storage units; the shelves of the medium associated product pairs are adjusted to non-adjacent storage units; and the shelves of the weakly associated product pairs are filled with vacancies; wherein the adjacent storage units are defined as directly connected nodes in the virtual shelf topology structure; and the non-adjacent storage units are defined as connected through at least one intermediate node in the virtual shelf topology structure.

[0058] In specific application scenarios, the purchase-related product pairs are divided into strongly associated product pairs, medium associated product pairs, and weakly associated product pairs based on the spatial adjacency coefficient. For example, pairs with a spatial adjacency coefficient greater than or equal to 0.8 are classified as strongly associated product pairs; pairs with a spatial adjacency coefficient less than 0.8 and greater than or equal to 0.5 are classified as medium associated product pairs; and pairs with a spatial adjacency coefficient less than 0.5 are classified as weakly associated product pairs. In the digital twin model, the shelf layout of the warehouse can be regarded as a virtual topological structure, where nodes represent storage units and edges represent connections between storage units. By setting the topological relationship of the shelf position, the shelf position can be adjusted according to the different product associations: The shelves of strongly associated product pairs are adjusted to adjacent storage units to ensure that they are close to each other in the physical warehouse and reduce the picking path; the shelves of moderately associated product pairs are adjusted to non-adjacent storage units to ensure that they are in the same logical partition for quick access and placement; the weakly associated product pairs are adjusted based on the empty shelves, that is, the empty storage units are interpolated to achieve the purpose of optimizing space utilization.

[0059] Specifically, in the digital twin model, refer to Figure 5 As shown, the rationality of the multiple shelf position adjustment schemes is evaluated based on path planning to obtain the optimal shelf position layout, including loading multiple shelf position adjustment schemes in the digital twin model; for each of the shelf position adjustment schemes, simulating the AGV in the digital twin model to execute multiple order data picking to obtain the picking evaluation results of the multiple shelf position adjustment methods; the evaluation indicators of the picking evaluation results include picking distance, temperature control stability and AGV energy consumption; and extracting the shelf position adjustment scheme with the best picking evaluation result as the optimal shelf position layout.

[0060] In specific application scenarios, multiple shelf position adjustment plans are first loaded into the digital twin model. These adjustment plans are derived from the aforementioned spatial adjacency coefficient calculation and association analysis steps. After space allocation and logical partition planning, different shelf position layout plans are generated. Each plan may include different shelf arrangements, product locations, and partition divisions. In the digital twin model, different historical order data or simulated order data are used to perform picking tasks. Each order includes multiple items. The AGV picks goods according to the optimal path plan. The AGV performs product picking according to the preset shelf layout and picking path planning, and records key indicators of each picking process, including picking distance, temperature control stability, and energy consumed by the AGV. The picking distance is the total distance the AGV moves in the process of completing the picking of each order. A shorter picking distance indicates an optimized shelf layout, which can reduce the AGV's travel time and material handling time. Consider whether the shelf layout affects the temperature control requirements of cold chain products. A reasonable shelf layout can reduce mutual interference between different temperature control zones and maintain temperature control stability. Measure the total energy consumed by AGVs while performing picking tasks. Energy consumption is affected by AGV path length, path complexity, and shelf location. Optimizing paths to reduce AGV energy consumption helps improve overall warehouse operational efficiency. For each shelf repositioning solution, calculate and comprehensively consider various evaluation indicators.

[0061] Furthermore, the AGV performs picking according to the effective path, referring to Figure 6As shown, determining the valid path includes generating an AGV picking path based on a genetic algorithm in the digital twin model to obtain a candidate path set; introducing a path planning cost function and calculating the average path cost and standard deviation of the candidate path; screening valid paths from the candidate path set according to the average path cost and standard deviation to obtain a screened path set; arranging the path costs of the screened path set in descending order, and extracting the first path as the valid path.

[0062] In a specific application scenario, a genetic algorithm (GA) is used within the digital twin model to generate AGV picking paths. GA is an optimization algorithm that simulates natural selection and genetic mechanisms, effectively exploring the solution space and finding near-optimal paths. Multiple initial paths are first generated as the initial population of the GA. Each path represents a possible picking path for the AGV in the warehouse. After running multiple generations, a set of candidate paths is generated, representing multiple possible paths from the starting location to the target location. To evaluate the quality of each path, a path planning cost function is introduced. This cost function comprehensively considers three metrics: path length, number of turns, and temperature control stability (whether the path effectively avoids unstable temperature control areas to ensure stable temperature control in the cold chain environment). The mean and standard deviation of the path costs for all candidate paths are calculated. The standard deviation measures the dispersion of the path costs; a larger standard deviation indicates greater path variation. Valid paths are screened from the candidate path set based on the average path cost and standard deviation. The screening criteria for valid paths include limits on the average path cost and standard deviation. Paths with average path costs within a certain range and low standard deviations are first selected to ensure path planning stability. Finally, a set of paths that meet the path cost requirements and have low standard deviations are selected as candidate valid paths. The selected valid paths are sorted in descending order by path cost. The path with the lowest cost is selected from the sorted valid path set, becoming the optimal path for the AGV to perform the picking task.

[0063] AGV path optimization based on genetic algorithms, combined with a multi-objective cost function, can provide the optimal picking path solution in the cold chain warehousing system, improve the efficiency of warehouse operations, reduce energy consumption, ensure temperature control stability, and enhance the adaptability and flexibility of the system.

[0064] Example 2: Based on the same inventive concept, the present invention provides a cold chain shelf location optimization system based on digital twins, referring to Figure 7 As shown, the system includes, A data extraction module extracts historical order data and obtains a historical order data set, wherein the historical order data includes order time, product category, quantity, and storage method; An association analysis module performs product purchase association analysis on the historical order data set to obtain a set of association features of purchase-related product pairs; an adjacency coefficient calculation module for calculating the spatial adjacency coefficient of the purchase-related product pairs based on the associated feature set; Digital twin modeling module, used to build a digital twin model of the cold chain warehouse; an intelligent planning module, interacting with the adjacency coefficient calculation module and the digital twin modeling module, and configured to generate a plurality of shelf position adjustment schemes in the digital twin model according to the spatial adjacency coefficient; a scheme evaluation module, configured to evaluate the plurality of shelf position adjustment schemes according to path planning in the digital twin model, and determine an optimal shelf position layout based on multi-objective optimization; The drive execution module is configured to convert the optimal shelf position layout into a storage equipment control instruction, and drive the shelves of the cold chain warehouse to be adjusted according to the optimal shelf position layout.

[0065] The cold chain shelf position optimization system based on digital twin described in the embodiment of the present invention is used to execute the cold chain shelf position optimization method of embodiment 1, and has the same technical effect, which will not be repeated here.

[0066] In summary, the digital twin-based cold chain shelf location optimization method and system described in this invention utilizes purchase correlation analysis and spatial adjacency coefficient calculation based on historical order data to provide a demand-based optimization solution for shelf location layout, significantly reducing order picking distance and time, and lowering operational costs. Furthermore, by using digital twin model simulation to verify different shelf layout solutions, the scientific nature and feasibility of the final solution are ensured while mitigating implementation risks.

[0067] Among them, by exploring the purchase correlation of goods and quantifying the spatial adjacency coefficient, the shelves are adjusted to store high-frequency co-occurring goods nearby, which greatly shortens the picking path, reduces the moving distance and operation time of stacking cranes or AGV equipment, and reduces the equipment idle rate and temperature fluctuation risks in the cold chain environment.

[0068] Dynamically adjust shelf layout based on historical order data characteristics to avoid the rigidity of static zoning strategies, flexibly respond to seasonal order fluctuations or changes in category demand, and improve warehouse resource utilization.

[0069] By using the digital twin model to conduct virtual simulation and path planning evaluation of multiple sets of shelf adjustment plans, the optimal layout plan can be quickly screened before implementation in the physical domain, reducing the number of actual handling times and trial-and-error costs, while also reducing the loss of cold storage caused by frequent entry and exit of goods in cold chain warehousing.

[0070] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0071] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0072] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0073] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0074] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. A cold chain shelf location optimization method based on digital twins, characterized by: include, Extract historical order data to obtain a historical order data set; the historical order data includes order time, product category, quantity, and storage method; Performing product purchase correlation analysis on the historical order data set to obtain a set of correlation features of purchase-related product pairs; Calculating the spatial adjacency coefficient of the purchase-related product pairs according to the associated feature set; Establishing a digital twin model of the cold chain vertical warehouse, planning shelf positions in the digital twin model according to the spatial adjacency coefficient, and obtaining multiple shelf position adjustment schemes; In the digital twin model, the plurality of shelf position adjustment schemes are evaluated based on path planning, and an optimal shelf position layout is obtained according to the evaluation results; The optimal shelf position layout is implemented in the physical domain cold chain warehouse.

2. The cold chain shelf location optimization method based on digital twin according to claim 1 is characterized in that: Perform product purchase correlation analysis on the historical order data set to obtain a set of correlation features of purchase-related product pairs, including: Assigning a time-decay weight to each piece of the historical order data to obtain a weighted order data set; Mining frequent item sets of product combinations in the weighted order data set layer by layer based on a priori algorithms to obtain a set of candidate association rules; Performing multi-index screening on the candidate association rule set to obtain a preliminary screening rule set; the screening indicators include support, confidence and lift; Performing density cluster analysis on the preliminary screening rule set to obtain a strong association rule set; A set of associated features of the purchase-associated item pairs is extracted from the set of strong association rules.

3. The cold chain shelf location optimization method based on digital twin according to claim 2 is characterized in that: Performing product purchase correlation analysis on the historical order data set to obtain a set of correlation features of purchase-related product pairs, further comprising: Calculating the association strength of each product pair in the strong association rule set; Marking product pairs with a correlation strength greater than or equal to a preset correlation strength as purchase-related product pairs; In the digital twin model, the association rules of the purchase-related product pairs are mapped into dynamic connection lines between virtual shelves; wherein the line width of the connection line is positively correlated with the association strength.

4. The cold chain shelf location optimization method based on digital twin according to claim 1 is characterized in that: Calculating the spatial adjacency coefficient of the purchase-related product pairs according to the associated feature set includes: Each correlation feature in the correlation feature set includes support, confidence, and lift for purchasing a related product pair; The spatial adjacency coefficient is obtained by calculating the support, confidence and lift.

5. The cold chain shelf location optimization method based on digital twin according to claim 4 is characterized in that: Calculating the spatial adjacency coefficient of the purchase-related product pairs also includes: If the purchase-related product pair has a bidirectional association feature, merging the bidirectional spatial adjacency coefficients of the purchase-related product pair to obtain a merged spatial adjacency coefficient; Using the combined spatial adjacency coefficient as the actual spatial adjacency coefficient; The merged space adjacency coefficient is obtained based on the following method: ； Among them, W XY ' represents the combined spatial adjacency coefficient of the purchase-related product pair XY; X and Y represent the products of the purchase-related product pair; W XY W represents the spatial adjacency coefficient of X as the premise product and Y as the conclusion product; YX Indicates the spatial adjacency coefficient where Y is the premise product and X is the conclusion product; Represents the attenuation coefficient.

6. The cold chain shelf location optimization method based on digital twin according to claim 1 or 4 is characterized in that: Planning shelf locations in the digital twin model according to the spatial adjacency coefficient includes: Determine whether the storage methods of the purchase-related product pairs are the same: If the storage modes are the same, adjusting the shelves of the purchase-related product pairs to storage units of the same logical partition according to the spatial adjacency coefficient; If the storage methods are different, the shelves of the purchase-related product pairs are adjusted to storage units of different logical partitions according to the spatial adjacency coefficient.

7. The cold chain shelf location optimization method based on digital twin according to claim 6 is characterized by: Adjusting the shelves of the purchase-related product pairs to storage units of the same logical partition according to the spatial adjacency coefficient includes: Classifying the purchase-related product pairs into strongly-related product pairs, moderately-related product pairs, and weakly-related product pairs according to the spatial adjacency coefficient; In the digital twin model, the shelves of the strongly associated product pairs are adjusted to adjacent storage units; the shelves of the moderately associated product pairs are adjusted to non-adjacent storage units; and the shelves of the weakly associated product pairs are filled with empty spaces. The adjacent storage units are defined as directly connected nodes in the virtual shelf topology structure; and the non-adjacent storage units are defined as nodes connected via at least one intermediate node in the virtual shelf topology structure.

8. The cold chain shelf location optimization method based on digital twin according to claim 1 is characterized in that: In the digital twin model, the rationality of the multiple shelf position adjustment schemes is evaluated based on path planning to obtain the optimal shelf position layout, including: Loading multiple shelf position adjustment schemes into the digital twin model; For each shelf position adjustment scheme, simulating an AGV in the digital twin model to execute multiple order data picking, and obtaining picking evaluation results for the multiple shelf position adjustment methods; the evaluation indicators of the picking evaluation results include picking distance, temperature control stability, and AGV energy consumption; The shelf position adjustment solution with the best picking evaluation result is extracted as the optimal shelf position layout.

9. The cold chain shelf location optimization method based on digital twin according to claim 8 is characterized in that: The AGV performs picking according to an effective path, and determining the effective path includes: Generate an AGV picking path based on a genetic algorithm in the digital twin model to obtain a set of candidate paths; Introducing a path planning cost function and calculating the average path cost and standard deviation of the candidate paths; Filtering valid paths from the candidate path set according to the average path cost and the standard deviation to obtain a filtered path set; Arrange the path costs of the screening path set in descending order, and extract the first path as the valid path.

10. The cold chain shelf location optimization system based on digital twin is characterized by: include, A data extraction module extracts historical order data and obtains a historical order data set, wherein the historical order data includes order time, product category, quantity, and storage method; An association analysis module performs product purchase association analysis on the historical order data set to obtain a set of association features of purchase-related product pairs; an adjacency coefficient calculation module for calculating the spatial adjacency coefficient of the purchase-related product pairs based on the associated feature set; Digital twin modeling module, used to build a digital twin model of the cold chain warehouse; an intelligent planning module, interacting with the adjacency coefficient calculation module and the digital twin modeling module, and configured to generate a plurality of shelf position adjustment schemes in the digital twin model according to the spatial adjacency coefficient; a scheme evaluation module, configured to evaluate the plurality of shelf position adjustment schemes according to path planning in the digital twin model, and determine an optimal shelf position layout based on multi-objective optimization; The drive execution module is configured to convert the optimal shelf position layout into a storage equipment control instruction, and drive the shelves of the cold chain warehouse to be adjusted according to the optimal shelf position layout.

Images