Supply chain logistics intelligent scheduling method based on dynamic planning

Abstract

The invention discloses a supply chain logistics intelligent scheduling method based on dynamic planning, and relates to the technical field of logistics scheduling, and the method comprises the following steps: firstly, taking a preset data synchronization period as a basic data updating period, and executing a data synchronization task between warehouses; in the data synchronization process, warehouse-in and warehouse-out operation data of each warehouse are obtained in real time, and the actual operation conditions of the warehouses are reflected. According to the method, by extracting the order generation frequency, the completion frequency, the multi-category coverage rate and other characteristics and combining intelligent evaluation of a deep learning model, the warehouse in-out state is accurately quantified, the warehouse is divided into a high-frequency type and a stable type, and the synchronization period is intelligently regulated and controlled. The high-frequency warehouse shortens the period, ensures real-time updating of data, and avoids inventory shortage or excess; the warehouse is stabilized to maintain a preset period, and resource waste is reduced. The strategy improves the response capability of the supply chain to market fluctuation, optimizes the replenishment and transportation efficiency, reduces the synchronization cost, and achieves lean management.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics scheduling, and in particular to an intelligent scheduling method for supply chain logistics based on dynamic programming. Background Art

[0002] Intelligent scheduling of supply chain logistics based on dynamic programming refers to optimizing logistics scheduling problems in the supply chain by using the dynamic programming (DP) algorithm. Specifically, dynamic programming is an optimization method that decomposes complex problems into multiple sub-problems and solves them recursively. In supply chain logistics, intelligent scheduling means dynamically adjusting logistics transportation plans based on real-time demand, inventory, transportation routes, transportation tools, and other resource constraints to minimize costs, improve efficiency, and meet service levels. Through the optimization strategy of dynamic programming, problems such as path selection, cargo distribution, and transportation planning can be effectively solved, especially when facing uncertainties (such as demand fluctuations, traffic conditions, etc.), flexible and efficient solutions can be provided.

[0003] The prior art has the following deficiencies:

[0004] In the existing intelligent logistics scheduling system, in order to ensure the efficiency of the overall supply chain operation, multiple warehouses are usually updated periodically to ensure that key data such as inventory information, transportation plans and demand forecasts can be updated and shared in a timely manner. However, when a warehouse frequently performs in-and-out operations, if the data update cycle is not sufficient to support real-time demand, it will lead to lagging inventory information, which will affect the accuracy of decision-making. Specifically, the system may still believe that a warehouse has sufficient inventory, but in fact the inventory is close to exhaustion, resulting in the inability to execute the replenishment plan in time, causing inventory shortages. This shortage will not only affect the continuity of supply, but may also cause delayed delivery of orders, thereby affecting customer satisfaction and corporate reputation. On the other hand, outdated inventory data may cause the system to continue to execute the replenishment plan when demand decreases, resulting in excess inventory. This excess not only takes up valuable storage space, but also increases storage costs, and may result in additional handling and disposal costs, which in turn affects the cost-effectiveness of the overall supply chain. Therefore, data synchronization lag seriously affects the accuracy of inventory management, increases the risk of shortages and excess inventory problems, and ultimately has a negative impact on the efficiency and cost control of the supply chain.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a supply chain logistics intelligent scheduling method based on dynamic programming. By extracting features such as order generation frequency, completion frequency, and multi-category coverage, combined with deep learning model intelligent evaluation, this method accurately quantifies the warehouse in and out status, divides it into two categories: high frequency and stable, and intelligently adjusts the synchronization cycle. High-frequency warehouses shorten the cycle to ensure real-time data updates and avoid inventory shortages or surpluses; stable warehouses maintain preset cycles to reduce resource waste. This strategy improves the supply chain's responsiveness to market fluctuations, optimizes replenishment and transportation efficiency, while reducing synchronization costs and achieving lean management to solve the problems in the above-mentioned background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a supply chain logistics intelligent scheduling method based on dynamic programming, comprising the following steps:

[0008] First, use the preset data synchronization cycle as the basic data update cycle to perform data synchronization tasks between warehouses;

[0009] During the data synchronization process, the inbound and outbound operation data of each warehouse is obtained in real time to reflect the actual operation status of the warehouse;

[0010] After preprocessing the acquired data, features reflecting the frequent warehouse in-and-out operations are extracted from it, and the extracted features are analyzed under the detection window, and the warehouse in-and-out status is quantified through the analyzed features;

[0011] The analyzed features are input into the pre-learned deep learning model, and the deep learning model is used to conduct intelligent evaluation of the warehouse in and out conditions;

[0012] Based on the evaluation results of the deep learning model, the warehouse's inbound and outbound status is divided into two categories: high-frequency inbound and outbound and stable inbound and outbound;

[0013] For stable inbound and outbound warehousing, continue to synchronize data between warehouses at the preset data synchronization cycle;

[0014] For high-frequency warehousing and out-of-warehouse transactions, the data synchronization cycle between warehouses is intelligently adjusted based on the evaluation results of the deep learning model to meet adaptive data synchronization requirements and ensure that data between warehouses remains efficiently synchronized.

[0015] Preferably, the acquired data is preprocessed to extract features reflecting the frequent warehouse in-and-out operations, the extracted features including the frequency of order generation and completion in the warehouse and the change in the proportion of the number of commodity categories involved in the in-and-out operations to the total number of commodity categories in the warehouse. Within the detection window, the frequency of order generation and completion in the warehouse and the change in the proportion of the number of commodity categories involved in the in-and-out operations to the total number of commodity categories in the warehouse are analyzed to generate an order-driven operation quantification value and a multi-category operation coverage quantification value, respectively. The order-driven operation quantification value quantifies the ratio of the warehouse in-and-out operation frequency directly caused by order generation and completion to the total operation frequency, and the multi-category operation coverage quantification value quantifies the change in the proportion of the number of commodity categories involved in the in-and-out operations to the total number of commodity categories in the warehouse.

[0016] Preferably, the specific steps of analyzing the generation and completion frequency of orders in the warehouse under the detection window to generate the order-driven operation quantitative value are as follows:

[0017] Extract the relevant data of warehouse order generation and calculate the order generation factor to quantify the driving force of order generation on warehouse operations. The calculation expression is as follows:

[0018]

[0019] , where OG is the order generation factor, t0 is the start time of the detection window, t1 is the end time of the detection window, R(t) is the order generation weight coefficient, which indicates the importance weight of different orders to warehouse operations at time t, N(t) is the number of orders generated, is the order generation rate, which means the cumulative number of orders generated within the detection window at time t;

[0020] Analyze the frequency of order completion and the impact of operations, and calculate the order completion factor. The calculation expression is as follows:

[0021]

[0022] , where OC is the order completion factor, W(i) is the completion weight coefficient of the i-th order, F(i) is the frequency of the i-th order completion, V(i) is the number of orders involved in the i-th order, L(i) is the amount of warehouse logistics resources consumed when the i-th order is completed, and n is the total number of completed orders;

[0023] The order-driven operation quantitative value is generated by the order generation factor OG and the order completion factor OC. The generation formula is as follows:

[0024]

[0025] , where ODI is the quantitative value of order-driven operation, α is the weight coefficient of order generation factor OG, β is the weight coefficient of order completion factor OC, ΔT is the time from order generation to order completion, T max is the maximum allowed time for order processing, and π is a mathematical constant.

[0026] Preferably, the specific steps of analyzing the change in the ratio of the number of commodity categories involved in the inbound and outbound operations to the total number of commodity categories in the warehouse under the detection window to generate the quantitative value of multi-category operation coverage are as follows:

[0027] The operation frequency of each commodity category reflects its activity in the inbound and outbound operations. To describe this feature, the proportion of commodity categories involved in the operation is defined to measure the frequency of operations of the commodity category in the current time period. The formula is as follows:

[0028]

[0029] , where φ j (t) is the total number of inbound and outbound operations of the jth product category within time t, Φ j is the theoretical maximum number of operations for the jth product category, P j (t) is the operation participation ratio of the jth product category in time t;

[0030] The frequency of commodity category operation participation not only affects the circulation of a single commodity, but is also related to the distribution of the overall operation. Therefore, the dynamic distribution characteristic factor of commodity category participation in operation is calculated to quantify the breadth and dynamics of commodity operation participation. The calculation expression is as follows:

[0031]

[0032] , where D is the dynamic distribution characteristic factor, ∈ is the smoothing factor, m is the total number of commodity categories, and Γ(t) is the normalization factor of the operation intensity of all categories within time t. The calculation expression is as follows:

[0033] In the warehousing and outbound operations, the coverage fluctuation characteristics are extracted by calculating the deviation between the proportion of the current commodity category participating in the operation and its benchmark value. The calculation expression is as follows:

[0034]

[0035] , where ΔC is the operating coverage fluctuation characteristic, is the participation ratio of the benchmark operation of the jth commodity category, is the time weight decay factor, e is the natural base, λ is the decay rate control parameter, and t0 is the start time of the detection window;

[0036] The dynamic distribution characteristic factor D and the operation coverage fluctuation characteristic ΔC are combined to generate the quantitative value of multi-category operation coverage. The generation formula is as follows: Where MCOF is the quantitative value of multi-category operation coverage, ω is the weight parameter of the comprehensive dynamic distribution characteristic factor D, is the operational coverage fluctuation characteristic and ΔC is the weight parameter.

[0037] Preferably, after analyzing the extracted features, the order-driven operation quantification values ​​and multi-category operation coverage quantification values ​​generated after the analysis are input into a pre-learned deep learning model, and the in-and-out frequency index is generated by the deep learning model, and the in-and-out frequency index is used to perform an intelligent evaluation of the warehouse in-and-out situation.

[0038] Preferably, the inbound frequency index generated by the pre-trained deep learning model during the intelligent evaluation of the warehouse inbound and outbound conditions is compared and analyzed with the pre-set reference threshold of the inbound frequency index to divide the warehouse inbound and outbound status. The division steps are as follows:

[0039] If the entry frequency index is greater than the preset entry frequency index reference threshold, the current warehouse entry and exit is classified as high-frequency entry and exit;

[0040] If the inventory frequency index is less than or equal to the preset inventory frequency index reference threshold, the current warehouse inventory is classified as stable inventory.

[0041] Preferably, for high-frequency inbound and outbound warehousing, the data synchronization cycle between warehouses is intelligently adjusted according to the evaluation results of the deep learning model to achieve adaptive data synchronization requirements. The specific steps to ensure efficient synchronization of data between warehouses are as follows:

[0042] After the warehouse is determined to be in a high-frequency in-and-out state, in order to accurately quantify the degree of deviation between the in-warehousing frequency index IFI and the in-warehousing frequency index reference threshold, the data synchronization cycle is further optimized to calculate the synchronization adjustment coefficient. The adjustment coefficient calculation formula is as follows:

[0043]

[0044] , where k is the sensitivity factor, IFI ref is the reference threshold of the storage frequency index, δ is the stability adjustment factor, which controls the smoothness of the index adjustment part. is an exponential function, e is the natural base, q is the adjustment intensity factor, and H is the synchronization adjustment coefficient, which is used to adjust the proportional factor of the data synchronization period;

[0045] After obtaining the synchronization adjustment coefficient H, the intelligently adjusted data synchronization period is calculated to adapt to the synchronization requirements of the current high-frequency in-and-out storage status. The calculation expression is as follows:

[0046]

[0047] , where T adjusted is the adjusted data synchronization period, t preset is the preset data synchronization period, μ is the complexity adjustment factor, σ IFI is the standard deviation of the inventory frequency index, IFI avg is the historical average of the inventory frequency index, and ∈ is the smoothing factor.

[0048] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0049] The present invention extracts key features such as order generation frequency, completion frequency, and multi-category coverage, and combines it with the intelligent evaluation of the deep learning model. This method can accurately quantify the warehouse's in-and-out status, and divide it into high-frequency in-and-out and stable in-and-out categories, thereby realizing intelligent regulation of the data synchronization cycle. For high-frequency in-and-out warehouses, shorten the synchronization cycle to ensure real-time update of data and avoid inventory shortages or surpluses due to lagging information; while for stable in-and-out warehouses, maintain the preset synchronization cycle to reduce resource waste caused by frequent synchronization. This differentiated strategy not only improves the supply chain's ability to respond to market changes and order fluctuations, and ensures the efficient execution of replenishment, allocation, and transportation links, but also significantly reduces the overall cost of data synchronization by optimizing system load and reducing redundant synchronization, ultimately helping companies achieve efficient and lean management in complex supply chain networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0051] Figure 1 The present invention is a method flow chart of a supply chain logistics intelligent scheduling method based on dynamic programming. DETAILED DESCRIPTION

[0052] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0053] The present invention provides Figure 1The method for intelligent scheduling of supply chain logistics based on dynamic programming includes the following steps:

[0054] First, use the preset data synchronization cycle as the basic data update cycle to perform data synchronization tasks between warehouses;

[0055] This periodic synchronization ensures that key data such as inventory information, transportation plans, and demand forecasts between warehouses can be updated and shared at fixed time intervals. The preset cycle is usually set based on business needs, data change frequency, and system processing capabilities to ensure data timeliness while avoiding system load and resource waste caused by too frequent synchronization.

[0056] The preset data synchronization cycle is used as an initial setting to provide a stable synchronization framework. In actual applications, the business volume and operation frequency of different warehouses may vary, so the preset cycle needs to have a certain degree of flexibility so that it can be adjusted according to actual conditions in subsequent steps. Through periodic synchronization, it can ensure that the data of most warehouses is updated within a fixed time, thus providing a reliable data foundation for the overall operation of the supply chain.

[0057] During the data synchronization process, the inbound and outbound operation data of each warehouse is obtained in real time to reflect the actual operation status of the warehouse;

[0058] Warehouse operation data includes dynamic change information such as inventory increase and decrease, order generation and completion, and goods transfer. This step ensures that the system can capture the real-time dynamics of warehouse operations and provide the latest basic data for subsequent data analysis and processing.

[0059] Real-time acquisition of inbound and outbound data usually relies on advanced technologies such as Internet of Things (IoT) devices, RFID tags, barcode scanners, and automated warehouse management systems (WMS). These technologies can efficiently and accurately record every inbound and outbound operation of goods and transmit the data to a central database or data processing center. Through real-time data collection, the system can reflect the actual operating status of the warehouse, laying a solid foundation for subsequent preprocessing and feature extraction.

[0060] After preprocessing the acquired data, features reflecting the frequent warehouse in-and-out operations are extracted from it, and the extracted features are analyzed under the detection window, and the warehouse in-and-out status is quantified through the analyzed features;

[0061] The raw inbound and outbound data obtained often contains noise, missing values, and inconsistencies. Therefore, these data must be preprocessed to ensure their quality and usability. The preprocessing steps include data cleaning, missing value filling, data standardization, and format conversion.

[0062] Data preprocessing is a key step in the process of data analysis and machine learning. Through data cleaning, errors and abnormal data can be removed; through missing value filling, gaps in the data can be filled to avoid deviations in subsequent analysis; data standardization ensures that data from different sources have a consistent scale, which is convenient for subsequent feature extraction and model training. In addition, preprocessing may also include data integration, which unifies the format of data from different warehouses and systems to ensure data consistency and integrity. This process improves the quality of the data and ensures that the analysis and model training in the subsequent steps can be based on reliable data.

[0063] After preprocessing the acquired data, features reflecting the frequent warehouse in-and-out operations are extracted from them. The extracted features include the frequency of order generation and completion in the warehouse and the change in the proportion of the number of commodity categories involved in the in-and-out operations to the total number of commodity categories in the warehouse. Under the detection window, the frequency of order generation and completion in the warehouse and the change in the proportion of the number of commodity categories involved in the in-and-out operations to the total number of commodity categories in the warehouse are analyzed, and the order-driven operation quantification value and the multi-category operation coverage quantification value are generated respectively. The order-driven operation quantification value quantifies the ratio of the warehouse in-and-out operation frequency directly caused by order generation and completion to the total operation frequency, measures the driving force of order generation and completion on the overall operational activities of the warehouse, and reflects the impact of changes in order demand on warehouse operations. The multi-category operation coverage quantification value quantifies the change in the proportion of the number of commodity categories involved in in-and-out operations to the total number of commodity categories in the warehouse, reflecting the complexity and diversity of warehouse operations, especially when multiple categories of goods are flowing at the same time.

[0064] The significant increase in the frequency of order generation and completion in the warehouse and the large fluctuation range do indicate that the current warehouse is in a high-frequency in-and-out state. The increase in the frequency of order generation means that the warehouse needs to respond to new demands more frequently and perform picking, distribution and outbound operations; while the significant increase in the frequency of order completion reflects the acceleration of the flow of goods, which is usually closely related to the warehouse's efficient outbound capacity and rapid turnover requirements. At the same time, the large fluctuation range indicates that the changes in order demand are highly uncertain, and the warehouse needs to frequently adjust the operating process in a short period of time to cope with the rapid fluctuations in order generation and completion. The combined changes of the two indicate that the warehouse is experiencing frequent in-and-out cycles under high-load operation. This state requires not only the warehouse to respond quickly to each order, but also to synchronize inventory information at a higher frequency to avoid data lag affecting decision-making and efficiency. Therefore, the significant increase and fluctuation in the frequency of order generation and completion is one of the important characteristics for evaluating whether the warehouse has entered a high-frequency in-and-out state, which directly reflects the increase in operating pressure and dynamic management needs.

[0065] The specific steps for analyzing the generation and completion frequency of orders in the warehouse under the detection window to generate the quantitative value of order-driven operations are as follows:

[0066] Extract the relevant data of warehouse order generation and calculate the order generation factor to quantify the driving force of order generation on warehouse operations. The calculation expression is as follows:

[0067]

[0068] , where OG is the order generation factor, which quantifies the driving force of order generation on warehouse operations, t0 is the start time of the detection window, t1 is the end time of the detection window, R(t) is the order generation weight coefficient, which indicates the importance weight of different orders to warehouse operations at time t, and N(t) is the number of orders generated. is the order generation rate, which means the cumulative number of orders generated within the detection window at time t;

[0069] This step calculates the cumulative value of the driving force of order generation within the detection window by integration to reflect the impact of order generation on warehouse operations. It avoids the limitation of simply counting the number of orders, focuses on the changes in generation rate and weight, and lays the foundation for subsequent analysis.

[0070] Analyze the frequency of order completion and the impact of operations, and calculate the order completion factor. The calculation expression is as follows:

[0071]

[0072] , where OC is the order completion factor, W(i) is the completion weight coefficient of the i-th order, F(i) is the frequency of the i-th order completion, V(i) is the number of orders involved in the i-th order, L(i) is the amount of warehouse logistics resources consumed when the i-th order is completed, and n is the total number of completed orders;

[0073] This step comprehensively considers the frequency, volume, and resource utilization efficiency of order fulfillment, quantifies the dynamic impact of order fulfillment on warehouse operations, and provides a comprehensive basis for evaluating the warehouse's current operational load.

[0074] The order-driven operation quantitative value is generated by the order generation factor OG and the order completion factor OC. The generation formula is as follows:

[0075]

[0076] , where ODI is the quantitative value of order-driven operation, α is the weight coefficient of order generation factor OG, β is the weight coefficient of order completion factor OC, ΔT is the time from order generation to order completion, T max is the maximum allowed time for order processing, used to standardize the effect of time differences, and π is a mathematical constant.

[0077] π stands for the mathematical constant "pi", which is approximately equal to 3.14159. It is often used in calculations involving periodic or oscillatory properties. In the example, the role of π is to convert the time difference ΔT o The effect of is combined with the sine function to give it a periodic variation. The periodicity of the sine function is determined by π, so that the effect of the time difference is within the standardized range (0 to T max ) is manifested in the form of fluctuations. When π and ΔT / T max When combined, the dynamic changes of the time interval between order generation and completion on warehouse operation pressure can be simulated, thus reflecting the sensitivity of frequent time fluctuations to operation status. This processing method emphasizes the nonlinear characteristics of time dynamic changes, making the formula more suitable for describing complex warehouse inbound and outbound operation modes.

[0078] This step combines the order generation, completion and its dynamic fluctuation characteristics to comprehensively quantify the driving force of orders on warehouse operations. In particular, the sine function part introduces a dynamic time factor to capture the sensitivity of short-term order fluctuations to operational pressure.

[0079] The larger the performance value of the order-driven operation quantification value generated by analyzing the generation and completion frequency of orders in the warehouse under the detection window, the higher the performance value, indicating that the current warehouse is in a high-frequency in-and-out state, and vice versa, indicating that the warehouse is in a normal in-and-out state. The order-driven operation quantification value quantifies the driving force of orders on warehouse in-and-out operations by analyzing the generation and completion frequency of orders within the monitoring window. When the performance value is large, it means that within the detection window, the generation and completion frequency of orders has increased significantly, triggering a large number of in-and-out operations, such as picking, distribution, replenishment, etc., and the warehouse operation frequency and operation intensity have increased, reflecting that it is in a high-frequency in-and-out state. When the performance value is small, it means that the generation and completion frequency of orders is relatively stable and low, and the in-and-out operations of the warehouse are few or evenly distributed. At this time, the warehouse is in a normal in-and-out state.

[0080] The proportion of the number of commodity categories involved in the in-and-out operation to the total number of commodity categories in the warehouse has increased significantly and has shown dynamic fluctuations, which does indicate that the current warehouse is in a high-frequency in-and-out state, because this phenomenon reflects the significant increase in the breadth and frequency of warehouse operations. When the proportion of commodity categories involved has increased significantly, it means that the warehouse's in-and-out operations have expanded to more categories of goods, which is usually closely related to the increase in order diversity and the increase in demand complexity; and dynamic fluctuations indicate that these operations change frequently within the detection window, further revealing that the warehouse is responding quickly to order requirements or inventory adjustment tasks. The combination of high proportions and dynamic fluctuations shows that the warehouse not only has a wide range of operations, but also has a tight processing rhythm and a high frequency of goods turnover. This feature is highly correlated with the high-frequency in-and-out state of the warehouse, because it reflects that the warehouse needs to frequently perform operations such as sorting, replenishment, and distribution when responding to diversified order requirements. In short, this change is an important manifestation of the high-frequency in-and-out state, and provides a key basis for evaluating the intensity of warehouse operations and optimizing management strategies.

[0081] The specific steps for analyzing the changes in the proportion of the number of commodity categories involved in the inbound and outbound operations to the total number of commodity categories in the warehouse under the detection window to generate the quantitative value of multi-category operation coverage are as follows:

[0082] The operation frequency of each commodity category reflects its activity in the inbound and outbound operations. To describe this feature, the proportion of commodity categories involved in the operation is defined to measure the frequency of operations of the commodity category in the current time period. The formula is as follows:

[0083]

[0084] , where φ j (t) is the total number of inbound and outbound operations of the jth commodity category within time t, reflecting the operational activity of the category, Φ j is the theoretical maximum number of operations for the jth product category, determined by historical data, indicating the upper limit of operations that the category may reach, P j (t) is the operation participation ratio of the jth product category in time t, indicating the relative activity of the category in the operation;

[0085] This step quantifies the relative change in operation frequency by comparing the current number of operations for each product category with its historical theoretical maximum. j The (t) value indicates that the commodity category has frequent in-and-out warehouse operations within the monitoring window, which provides a basis for subsequent analysis of the extensiveness and volatility of commodity category operations.

[0086] The frequency of commodity category operation participation not only affects the circulation of a single commodity, but is also related to the distribution of the overall operation. Therefore, the dynamic distribution characteristic factor of commodity category participation in operation is calculated to quantify the breadth and dynamics of commodity operation participation. The calculation expression is as follows:

[0087]

[0088] , where D is the dynamic distribution characteristic factor, ∈ is the smoothing factor, which is a very small constant (usually 10 -6 ), used to prevent the calculation of ln(P j (t)) when P j (t) = 0 leads to the problem of undefined logarithm, m is the total number of commodity categories, Γ(t) is the normalization factor of the operation intensity of all categories within time t, and the calculation expression is as follows:

[0089] Through the above steps, the dynamic distribution characteristics of category operations can be captured. If the dynamic distribution characteristic factor D value is high, it means that the current warehouse's inbound and outbound operations cover more categories and have dynamic volatility, which is directly related to the high-frequency inbound and outbound status. The logarithmic term of this step amplifies the impact of low-frequency categories, making the distribution characteristics more comprehensive.

[0090] In the warehouse in and out operations, not only the operation distribution of commodity categories is important, but also its fluctuation characteristics can reveal the frequency of warehouse operations. The coverage fluctuation characteristics are extracted by calculating the deviation between the proportion of the current commodity category involved in the operation and its benchmark value. The calculation expression is as follows:

[0091]

[0092] , where ΔC is the fluctuation characteristic of operation coverage, which quantifies the degree of dynamic change of category operation. is the participation ratio of the benchmark operation of the jth commodity category, is the time weight decay factor, e is the natural base, λ is the decay rate control parameter, and t0 is the start time of the detection window;

[0093] The above steps highlight the dynamics of recent operation data through the time decay factor, and quantify the deviation amplitude of category operation proportion. High ΔC value indicates that warehouse inbound and outbound operations have large coverage volatility, which directly reflects the high-frequency circulation state.

[0094] The dynamic distribution characteristic factor D and the operation coverage fluctuation characteristic ΔC are combined to generate the quantitative value of multi-category operation coverage. The generation formula is as follows: Where MCOF is the quantitative value of multi-category operation coverage, ω is the weight parameter of the comprehensive dynamic distribution characteristic factor D, which is used to control the contribution of the comprehensive dynamic distribution characteristic factor D to the final quantitative value of multi-category operation coverage. It is the operational coverage fluctuation characteristic ΔC and is the weight parameter, which is used to adjust the influence of ΔC on the final multi-category operational coverage quantitative value.

[0095] The multi-category operation coverage quantitative value generates a unified indicator by weighted integration of the distribution characteristics and fluctuation characteristics of category operations, and comprehensively quantifies the warehouse in and out operation status within the monitoring window. A high MCOF value indicates that the warehouse operation is extensive and frequent, which is suitable for identifying high-frequency in and out status and guiding the intelligent adjustment and management optimization of the data synchronization cycle.

[0096] The multi-category operation coverage quantitative value generated by analyzing the proportion of the number of commodity categories involved in the in-and-out operation to the total number of commodity categories in the warehouse under the detection window is larger, indicating that the current warehouse is in a high-frequency in-and-out state, and vice versa, indicating that the warehouse is in a normal in-and-out state. The multi-category operation coverage quantitative value reflects the proportion of the number of commodity categories involved in the in-and-out operation to the total number of commodity categories in the warehouse within the monitoring window. When this quantitative value is large, it means that the warehouse handles a wide range of commodity categories per unit time, and the in-and-out activities cover more categories of goods, indicating that the operation frequency is high and the task load is heavy. The warehouse is dealing with diversified and intensive logistics needs, which is consistent with the characteristics of the high-frequency in-and-out state. When the performance value is low, it means that there are fewer commodity categories involved in the in-and-out operation, and the circulation activities of the warehouse are more concentrated or stable, indicating a regular operation state.

[0097] The analyzed features are input into the pre-learned deep learning model, and the deep learning model is used to conduct intelligent evaluation of the warehouse in and out conditions;

[0098] After analyzing the extracted features, the order-driven operation quantification values ​​and multi-category operation coverage quantification values ​​generated after the analysis are input into the pre-learned deep learning model. The deep learning model is used to generate the in-and-out frequency index, and the in-and-out frequency index is used to conduct an intelligent evaluation of the warehouse in-and-out situation.

[0099] A pre-learned deep learning model refers to a deep learning algorithm that can intelligently predict or evaluate new input data after the model is fully trained with a training data set. In this scenario, the deep learning model is based on the warehouse's historical in-and-out operation data, combined with the order-driven operation quantification value and multi-category operation coverage quantification value after feature extraction, and repeatedly trains the model to enable it to learn and identify the complex nonlinear relationship between in-and-out operation frequency and related features. The pre-learned model can be directly used to evaluate new data, generate a more accurate in-and-out frequency index, and thus provide intelligent analysis of the warehouse operation status.

[0100] The pre-training of deep learning models mainly relies on large-scale, structured data sets, and mines hidden patterns and key associations in the data by building multi-layer neural networks (such as convolutional neural networks or long short-term memory networks). For example, the quantitative value of order-driven operations may be highly correlated with the frequency of inbound and outbound operations, while the quantitative value of multi-category operation coverage can provide supplementary information on the complexity of warehouse operations. Through training, the model can weight the influence of these features and generate a unified inbound and outbound frequency index. The pre-learned model already has a high degree of generalization and prediction capabilities, so it can quickly and accurately complete the evaluation when faced with real-time data.

[0101] The biggest advantage of the pre-learned deep learning model is its intelligence and efficiency, which can significantly improve the evaluation accuracy of warehouse in and out inventory conditions. Through the model training process, the system learns how to predict the changing trend of in and out inventory frequency based on the key quantitative features of the input, thereby reducing manual intervention and improving the evaluation speed. This pre-training process usually involves learning thousands or even millions of data, including different order fluctuation scenarios, warehouse operation complexity, and special situations in historical operations. Through the nonlinear computing power of the deep learning model, the system can capture patterns that are difficult to discover with traditional methods, such as potential incentives for high-frequency in and out inventory or special fluctuations in multi-category operation coverage.

[0102] In addition, the pre-learned deep learning model can dynamically adapt to the data features of real-time input. When new quantitative features are input into the model, the model can quickly calculate and generate the in-and-out frequency index, providing an intelligent assessment of whether the warehouse is in a high-frequency in-and-out state. This approach not only significantly improves the evaluation efficiency, but also ensures the stability and accuracy of the results. Compared with traditional rule-making or simple statistical analysis, deep learning models have higher robustness and adaptability, and can cope with complex and changing warehouse operation scenarios. This intelligent evaluation capability provides strong support for optimizing the data synchronization cycle and improving the efficiency of supply chain management.

[0103] The deep learning model is not limited here. Any deep learning model that can perform comprehensive analysis on the order-driven operation quantification value ODI and the multi-category operation coverage quantification value MCOF to generate the inbound and outbound frequency index IFI can be used. The technical solution of the present invention has been realized. The present invention provides a specific implementation method.

[0104] The formula for generating the inventory frequency index IFI is as follows:

[0105]

[0106] , where A1 and A2 are the preset proportional coefficients of the order-driven operation quantification value ODI and the multi-category operation coverage quantification value MCOF, respectively, and A1 and A2 are both greater than 0.

[0107] It can be seen from the inventory frequency index that the larger the order-driven operation quantization value performance value generated after analyzing the generation and completion frequency of orders in the warehouse under the detection window, the larger the multi-category operation coverage quantization value performance value generated after analyzing the proportion of the number of commodity categories involved in the inventory entry and exit operations to the total number of commodity categories in the warehouse under the detection window. This indicates that the larger the inventory frequency index performance value generated when the pre-trained deep learning model is used to perform an intelligent evaluation of the warehouse's inventory entry and exit conditions, the greater the probability that the current warehouse is in high-frequency inventory entry and exit. Conversely, the smaller the probability that the current warehouse is in high-frequency inventory entry and exit.

[0108] The preset proportional coefficients A1 and A2 here refer to the weight parameters introduced in the formula to balance the influence of the order-driven operation quantification value ODI and the multi-category operation coverage quantification value MCOF. The role of these coefficients is to adaptively adjust the importance of the two quantitative values ​​according to actual business needs and warehouse operation characteristics. For example, in some scenarios, the frequency of order-driven operation may be the main factor in evaluating the frequency of warehousing. At this time, the weight of the order-driven operation quantification value ODI in the formula can be increased by increasing A1; if the multi-category coverage has a greater impact on warehouse efficiency, A2 can be adjusted appropriately. The setting of the preset proportional coefficients is usually based on historical data and experience to ensure that the warehousing frequency index IFI generated by the formula can accurately reflect the actual warehouse status in different scenarios.

[0109] Based on the evaluation results of the deep learning model, the warehouse's inbound and outbound status is divided into two categories: high-frequency inbound and outbound and stable inbound and outbound;

[0110] The inbound and outbound status of the warehouse is divided into the following steps:

[0111] If the entry frequency index is greater than the preset entry frequency index reference threshold, the current warehouse entry and exit is classified as high-frequency entry and exit;

[0112] If the inventory frequency index is less than or equal to the preset inventory frequency index reference threshold, the current warehouse inventory is classified as stable inventory.

[0113] High-frequency in-and-out refers to the frequency of warehouse in-and-out operations being significantly higher than the daily average, usually accompanied by high volatility in order generation and completion and frequent turnover of multiple categories of goods. Stable in-and-out refers to the frequency of warehouse in-and-out operations being relatively stable, close to the average level of its daily operations, with a relatively regular rhythm of order generation and completion, and a relatively balanced variety and quantity of goods involved.

[0114] For stable inbound and outbound warehousing, continue to synchronize data between warehouses at the preset data synchronization cycle;

[0115] For stable inbound and outbound storage, data synchronization between warehouses is continued at the preset data synchronization cycle. The purpose is to maintain data consistency through a fixed synchronization cycle when the operation frequency is low and the inventory changes are relatively stable, while avoiding the waste of system resources caused by unnecessary frequent synchronization. This method can not only meet the data update needs under stable inbound and outbound storage conditions, but also reduce the burden of data processing and transmission, optimize system performance and operating costs, and thus achieve efficient use of resources while maintaining supply chain efficiency.

[0116] For high-frequency inbound and outbound warehousing, the data synchronization cycle between warehouses is intelligently adjusted based on the evaluation results of the deep learning model to meet adaptive data synchronization requirements and ensure efficient synchronization of data between warehouses.

[0117] For high-frequency inbound and outbound warehousing, the data synchronization cycle between warehouses is intelligently adjusted according to the evaluation results of the deep learning model to achieve adaptive data synchronization requirements. The specific steps to ensure efficient synchronization of data between warehouses are as follows:

[0118] After the warehouse is determined to be in a high-frequency in-and-out state, in order to accurately quantify the degree of deviation between the in-warehousing frequency index IFI and the in-warehousing frequency index reference threshold, the data synchronization cycle is further optimized to calculate the synchronization adjustment coefficient. The adjustment coefficient calculation formula is as follows:

[0119]

[0120] , where k is the sensitivity factor, which is used to control the relationship between the entry frequency index IFI and the entry frequency index reference threshold IFI ref The weight of the deviation degree on the synchronization adjustment, IFI refis the reference threshold of the inventory frequency index, δ is the stability adjustment factor, which controls the smoothness of the index adjustment part to avoid the excessive impact of the inventory frequency index IFI deviation on the synchronization adjustment coefficient H. is an exponential function, e is the natural base, q is the adjustment intensity factor, which controls the intensity of synchronous adjustment in the case of high deviation. The slope of the data is affected, H is the synchronization adjustment coefficient, which is used to adjust the proportional factor of the data synchronization period;

[0121] By calculating the adjustment coefficient H, it is possible to flexibly respond to high-frequency in-and-out warehouse conditions of different intensities, providing an accurate quantitative basis for periodic adjustment.

[0122] After obtaining the synchronization adjustment coefficient H, the intelligently adjusted data synchronization period is calculated to adapt to the synchronization requirements of the current high-frequency in-and-out storage status. The calculation expression is as follows:

[0123]

[0124] , where T adjusted is the adjusted data synchronization period, T preset is the preset data synchronization period, which is the fixed synchronization interval under normal conditions. μ is the complexity adjustment factor, which is used to quantify the weight of the impact of warehouse operation complexity on the synchronization period. σ IFI is the standard deviation of the inventory frequency index, IFI avg is the historical average of the inventory frequency index, ∈ is the smoothing factor used to avoid IFI in the denominator avg Small positive values ​​near zero where the formula becomes unstable.

[0125] The final calculated T adjusted It is an intelligently adjusted data synchronization cycle that can dynamically adapt to the high-frequency state of the warehouse while taking into account volatility and complexity to ensure efficient and stable data synchronization.

[0126] For high-frequency in-and-out warehouse status, according to the evaluation results of the deep learning model, the role of intelligently adjusting the data synchronization cycle is to meet the real-time demand for data updates under high-frequency operation status, and ensure that data sharing between warehouses remains efficient and consistent. Under high-frequency in-and-out warehouse status, warehouse in-and-out operations are frequent and complex, and inventory data and order information change rapidly. If the preset fixed synchronization cycle is still used, data lag may occur, thereby affecting the accuracy of scheduling decisions. By intelligently adjusting the data synchronization cycle, the system can dynamically calculate the optimized synchronization cycle based on the current in-and-out frequency index of the warehouse, match the synchronization frequency with the warehouse operation intensity, and shorten the data synchronization interval under high-frequency in-and-out warehouse status. This not only can timely update the key data of the warehouse (such as order generation and completion frequency, multi-category operation coverage, etc.), but also can reduce the risks caused by data lag, such as replenishment delays or resource mismatches. In addition, intelligent adjustment avoids the waste of resources caused by over-frequency synchronization, improves the flexibility and efficiency of data synchronization, and thus ensures the overall operating efficiency and stability of the supply chain.

[0127] The present invention extracts key features such as order generation frequency, completion frequency, and multi-category coverage, and combines it with the intelligent evaluation of the deep learning model. This method can accurately quantify the warehouse's in-and-out status, and divide it into high-frequency in-and-out and stable in-and-out categories, thereby realizing intelligent regulation of the data synchronization cycle. For high-frequency in-and-out warehouses, shorten the synchronization cycle to ensure real-time update of data and avoid inventory shortages or surpluses due to lagging information; while for stable in-and-out warehouses, maintain the preset synchronization cycle to reduce resource waste caused by frequent synchronization. This differentiated strategy not only improves the supply chain's ability to respond to market changes and order fluctuations, and ensures the efficient execution of replenishment, allocation, and transportation links, but also significantly reduces the overall cost of data synchronization by optimizing system load and reducing redundant synchronization, ultimately helping companies achieve efficient and lean management in complex supply chain networks.

[0128] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0129] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0130] It should be noted that, in this article, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0131] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0132] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0133] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0134] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0135] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0136] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0137] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A supply chain logistics intelligent scheduling method based on dynamic programming, characterized in that: The following steps are involved: First, use the preset data synchronization cycle as the basic data update cycle to perform data synchronization tasks between warehouses; During the data synchronization process, the inbound and outbound operation data of each warehouse is obtained in real time to reflect the actual operation status of the warehouse; After preprocessing the acquired data, features reflecting the frequent warehouse in-and-out operations are extracted from it, and the extracted features are analyzed under the detection window, and the warehouse in-and-out status is quantified through the analyzed features; The analyzed features are input into the pre-learned deep learning model, and the deep learning model is used to conduct intelligent evaluation of the warehouse in and out conditions; Based on the evaluation results of the deep learning model, the warehouse's inbound and outbound status is divided into two categories: high-frequency inbound and outbound and stable inbound and outbound; For stable inbound and outbound warehousing, continue to synchronize data between warehouses at the preset data synchronization cycle; For high-frequency warehousing and out-of-warehouse transactions, the data synchronization cycle between warehouses is intelligently adjusted based on the evaluation results of the deep learning model to meet adaptive data synchronization requirements and ensure that data between warehouses remains efficiently synchronized.

2. The method for intelligent scheduling of supply chain logistics based on dynamic programming according to claim 1 is characterized in that: After preprocessing the acquired data, features reflecting the frequent warehouse in-and-out operations are extracted from them. The extracted features include the frequency of order generation and completion in the warehouse and the change in the proportion of the number of commodity categories involved in the in-and-out operations to the total number of commodity categories in the warehouse. Under the detection window, the frequency of order generation and completion in the warehouse and the change in the proportion of the number of commodity categories involved in the in-and-out operations to the total number of commodity categories in the warehouse are analyzed, and order-driven operation quantification values ​​and multi-category operation coverage quantification values ​​are generated respectively. The order-driven operation quantification value is used to quantify the ratio of the warehouse in-and-out operation frequency directly caused by order generation and completion to the total operation frequency. The multi-category operation coverage quantification value is used to quantify the change in the proportion of the number of commodity categories involved in the in-and-out operations to the total number of commodity categories in the warehouse.

3. The supply chain logistics intelligent scheduling method based on dynamic programming according to claim 2 is characterized in that: The specific steps for analyzing the generation and completion frequency of orders in the warehouse under the detection window to generate the quantitative value of order-driven operations are as follows: Extract the relevant data of warehouse order generation and calculate the order generation factor to quantify the driving force of order generation on warehouse operations. The calculation expression is as follows: Where OG is the order generation factor, t0 is the start time of the detection window, t1 is the end time of the detection window, R(t) is the order generation weight coefficient, which indicates the importance weight of different orders to warehouse operations at time t, N(t) is the number of orders generated, is the order generation rate, which means the cumulative number of orders generated within the detection window at time t; Analyze the frequency of order completion and the impact of operations, and calculate the order completion factor. The calculation expression is as follows: Where OC is the order completion factor, W(i) is the completion weight coefficient of the i-th order, F(i) is the frequency of the i-th order completion, V(i) is the number of orders involved in the i-th order, L(i) is the amount of warehouse logistics resources consumed when the i-th order is completed, and n is the total number of completed orders; The order-driven operation quantitative value is generated by the order generation factor OG and the order completion factor OC. The generation formula is as follows: Where ODI is the quantitative value of order-driven operation, α is the weight coefficient of order generation factor OG, β is the weight coefficient of order completion factor OC, ΔT is the time from order generation to order completion, T max is the maximum allowed time for order processing, and π is a mathematical constant.

4. The supply chain logistics intelligent scheduling method based on dynamic programming according to claim 2 is characterized in that: The specific steps for analyzing the changes in the proportion of the number of commodity categories involved in the inbound and outbound operations to the total number of commodity categories in the warehouse under the detection window to generate the quantitative value of multi-category operation coverage are as follows: The operation frequency of each commodity category reflects its activity in the inbound and outbound operations. To describe this feature, the proportion of commodity categories involved in the operation is defined to measure the frequency of operations of the commodity category in the current time period. The formula is as follows: In the formula, φ j (t) is the total number of inbound and outbound operations of the jth product category within time t, Φ j is the theoretical maximum number of operations for the jth product category, P j (t) is the operation participation ratio of the jth product category in time t; The frequency of commodity category operation participation not only affects the circulation of a single commodity, but is also related to the distribution of the overall operation. Therefore, the dynamic distribution characteristic factor of commodity category participation in operation is calculated to quantify the breadth and dynamics of commodity operation participation. The calculation expression is as follows: In the formula, D is the dynamic distribution characteristic factor, ∈ is the smoothing factor, m is the total number of commodity categories, Γ(t) is the normalization factor of the operation intensity of all categories within time t, and the calculation expression is as follows: In the warehousing and outbound operations, the coverage fluctuation characteristics are extracted by calculating the deviation between the proportion of the current commodity category participating in the operation and its benchmark value. The calculation expression is as follows: Where ΔC is the operating coverage fluctuation characteristic, is the participation ratio of the benchmark operation of the jth commodity category, is the time weight decay factor, e is the natural base, λ is the decay rate control parameter, and t0 is the start time of the detection window; The dynamic distribution characteristic factor D and the operation coverage fluctuation characteristic ΔC are combined to generate the quantitative value of multi-category operation coverage. The generation formula is as follows: Where MCOF is the quantitative value of multi-category operation coverage, ω is the weight parameter of the comprehensive dynamic distribution characteristic factor D, is the operational coverage fluctuation characteristic and ΔC is the weight parameter.

5. The supply chain logistics intelligent scheduling method based on dynamic programming according to claim 2 is characterized in that: After analyzing the extracted features, the order-driven operation quantification values ​​and multi-category operation coverage quantification values ​​generated after the analysis are input into the pre-learned deep learning model. The deep learning model is used to generate the in-and-out frequency index, and the in-and-out frequency index is used to conduct an intelligent evaluation of the warehouse in-and-out situation.

6. The method for intelligent scheduling of supply chain logistics based on dynamic programming according to claim 5 is characterized in that: The inbound and outbound status of the warehouse is divided into the following steps: If the entry frequency index is greater than the preset entry frequency index reference threshold, the current warehouse entry and exit is classified as high-frequency entry and exit; If the inventory frequency index is less than or equal to the preset inventory frequency index reference threshold, the current warehouse inventory is classified as stable inventory.

7. The method for intelligent scheduling of supply chain logistics based on dynamic programming according to claim 6 is characterized in that: For high-frequency inbound and outbound warehousing, the data synchronization cycle between warehouses is intelligently adjusted according to the evaluation results of the deep learning model to achieve adaptive data synchronization requirements. The specific steps to ensure efficient synchronization of data between warehouses are as follows: After the warehouse is determined to be in a high-frequency in-and-out state, in order to accurately quantify the degree of deviation between the in-warehousing frequency index IFI and the in-warehousing frequency index reference threshold, the data synchronization cycle is further optimized to calculate the synchronization adjustment coefficient. The adjustment coefficient calculation formula is as follows: Where k is the sensitivity factor, IFI ref is the reference threshold of the storage frequency index, δ is the stability adjustment factor, which controls the smoothness of the index adjustment part. is an exponential function, e is the natural base, q is the adjustment intensity factor, and H is the synchronization adjustment coefficient, which is used to adjust the proportional factor of the data synchronization period; After obtaining the synchronization adjustment coefficient H, the intelligently adjusted data synchronization period is calculated to adapt to the synchronization requirements of the current high-frequency in-and-out storage status. The calculation expression is as follows: Where, T adju[ted is the adjusted data synchronization period, t pre[et is the preset data synchronization period, μ is the complexity adjustment factor, σ IFI is the standard deviation of the inventory frequency index, IFI avg is the historical average of the inventory frequency index, and ∈ is the smoothing factor.

Images