Warehousing intelligent replenishment scheduling system

Abstract

The invention relates to the technical field of warehouse management, and discloses a warehouse intelligent replenishment scheduling system, which comprises a popularity analysis unit, a demand correction unit, a distribution optimization unit, an evaluation unit and a scheduling generation unit, and is characterized in that a popularity pulse sequence is generated through multi-platform real-time data feature extraction, timeliness weighting and trend anomaly decomposition; platform popularity dynamic states are accurately tracked, historical replenishment tasks and popularity pulses are fused, virtual inventory distribution is adjusted in real time, replenishment urgency is accurately evaluated, scheduling priorities are constructed based on the urgency, position information and inventory health indexes, an optimal cross-warehouse scheduling instruction set is generated with the lowest cost and the fastest time efficiency as targets, warehousing operation benefits are improved, and the system is suitable for large-scale popularization and application. The platform popularity is tracked in real time, the replenishment quota is adjusted in time, and it is guaranteed that real-time scheduling of storage adapts to new market conditions.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of warehouse management, and particularly relates to a warehouse intelligent replenishment scheduling system. BACKGROUND

[0002] In modern warehouse management, big data analysis and machine learning algorithms are used to realize accurate prediction of future commodity demand according to historical sales data, market trends, promotional activities and other influencing factors, so as to maximize the efficiency of warehouse operation, and at the same time, the multi-objective of transportation cost, warehouse cost, shortage cost and time window constraint is considered, and the optimal replenishment strategy is found through optimization algorithm. Once the optimal replenishment strategy is determined, the system can automatically generate replenishment instructions and schedule and distribute goods in each warehouse, so as to realize intelligent replenishment.

[0003] However, with the change of commodity sales mode to unified sales on multiple platforms, the existing warehouse replenishment scheduling system has exposed significant shortcomings. Specifically, although the warehouse replenishment scheduling system can predict future demand through big data and machine learning, and make replenishment plans accordingly, it is seriously insufficient in real-time tracking of platform heat in the face of dynamic changes in platform heat in the multi-platform sales scenario. This results in the system being unable to adjust the replenishment quota in a timely manner when the platform heat changes, and the real-time scheduling of the warehouse cannot adapt to the new market situation. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a warehouse intelligent replenishment scheduling system, which solves the above problems.

[0005] The above technical purpose of the present application is realized by the following technical scheme:

[0006] A warehouse intelligent replenishment scheduling system comprises:

[0007] A heat analysis unit is configured to determine a heat pulse sequence according to real-time data of multiple platforms;

[0008] A demand correction unit is configured to obtain historical replenishment tasks of a target object, fuse the historical replenishment tasks and the heat pulse sequence, and obtain a real-time demand correction coefficient;

[0009] A distribution optimization unit is configured to calculate a dynamic platform quota weight according to the real-time demand correction coefficient and the real-time data, and adjust a virtual inventory allocation target of each target object facing different platforms in real time according to the dynamic platform quota weight;

[0010] An evaluation unit is configured to calculate an inventory turnover rate of the target object, analyze the inventory turnover rate and the virtual inventory allocation target, and obtain a replenishment urgency;

[0011] The scheduling generation unit is configured to acquire position information of the target object, and obtain a dynamic optimal cross-warehouse scheduling instruction set based on the replenishment urgency and the position information.

[0012] Further, the heat pulse sequence is determined according to real-time data of multiple platforms, including:

[0013] The real-time data is subjected to feature extraction to generate a cross-platform original feature data set;

[0014] The data of different time periods is subjected to time effectiveness weighting according to the interval between the time stamp of the data in the cross-platform original feature data set and the current time to obtain a time effectiveness correction feature matrix;

[0015] The influence weight vectors of different platforms are determined based on the real-time data of multiple platforms;

[0016] The time effectiveness correction feature matrix is decomposed to obtain a trend abnormal double-track feature vector;

[0017] The influence weight vectors of different platforms and the trend abnormal double-track feature vector are fused to generate a heat pulse sequence reflecting real-time heat fluctuation.

[0018] Further, the time effectiveness correction feature matrix is decomposed to obtain a trend abnormal double-track feature vector, including:

[0019] The time effectiveness correction feature matrix is preliminarily decomposed to obtain a basic feature component matrix and a corresponding singular value vector;

[0020] Based on the basic feature component matrix and the singular value vector, the data of each time dimension is fitted to generate a trend feature vector;

[0021] The time effectiveness correction feature matrix and the trend feature vector are calculated to obtain an abnormal fluctuation feature vector;

[0022] The trend feature vector and the abnormal fluctuation feature vector are combined to obtain a trend abnormal double-track feature vector.

[0023] Further, the real-time demand correction coefficient is obtained by fusing the historical replenishment task and the heat pulse sequence, including:

[0024] The historical replenishment task of the target object is subjected to time sequence decomposition to obtain a multi-dimensional replenishment feature vector;

[0025] The heat pulse sequence is subjected to grid division according to the space-time dimension to generate a space-time heat matrix;

[0026] Adaptively adjust the time window width according to the execution cycle of the historical replenishment task, and calculate the synchronization fluctuation coefficient of the historical replenishment task and the heat pulse sequence;

[0027] Analyze the multi-dimensional replenishment feature vector and the space-time heat matrix to obtain a historical influence weight vector and a heat correction weight vector, and fuse the historical influence weight vector, the heat correction weight vector, and the synchronization fluctuation coefficient to obtain a real-time demand correction coefficient.

[0028] Further, according to the real-time demand correction coefficient and real-time data, a dynamic platform quota weight is calculated, including:

[0029] Identify the real-time data of each platform to obtain a cost sensitivity matrix;

[0030] Based on the real-time demand correction coefficient, the elastic demand threshold of each platform is calculated in combination with the historical replenishment task;

[0031] Taking the minimization of total cost and the maximization of demand satisfaction rate as the target, the cost sensitivity matrix and the elastic demand threshold are taken as the constraint condition, and the replenishment quantity of each platform is dynamically optimized to form a Pareto optimal solution set;

[0032] According to the cost sensitivity matrix, the initial platform quota vector is generated by solving the Pareto optimal solution set;

[0033] Obtain the inventory data of each target object, analyze the inventory data, and obtain an inventory health index;

[0034] According to the inventory health index, the initial platform quota vector is modified to generate a dynamic platform quota weight.

[0035] Further, according to the cost sensitivity matrix, the initial platform quota vector is generated by solving the Pareto optimal solution set, including:

[0036] Based on the cost sensitivity matrix, a response surface between the replenishment quantity of each platform and the total cost is constructed to generate a marginal cost gradient matrix;

[0037] Optimize the Pareto optimal solution set to generate a weighted Pareto front;

[0038] Calculate the comprehensive utility value of each point on the weighted Pareto front, select the point with the maximum utility as the optimal solution, and take the replenishment quantity proportion of each platform corresponding to the point as the initial platform quota vector.

[0039] Further, according to the dynamic platform quota weight, the virtual inventory allocation target of each target object facing different platforms is adjusted in real time, including:

[0040] Obtain the virtual inventory allocation target of each platform in each target object;

[0041] determining a maximum capacity coefficient in a unit time based on the inventory data;

[0042] based on the dynamic platform quota weight and the real-time data of the plurality of platforms, analyzing a demand urgency index of each platform;

[0043] calculating the demand urgency index of each platform and the maximum capacity coefficient to obtain a virtual inventory allocation adjustment coefficient;

[0044] adjusting the virtual inventory allocation target according to the virtual inventory allocation adjustment coefficient to obtain an adjusted virtual inventory allocation target.

[0045] Further, the inventory turnover rate of the target object is calculated, including:

[0046] analyzing the historical replenishment task and the space-time heat matrix of the target object to form a space-time value decay matrix;

[0047] weighting and aggregating the inventory quantity of the target object and the space-time value decay matrix to obtain a dynamic inventory equivalent;

[0048] analyzing the delay response coefficient of the historical replenishment task and the heat pulse sequence, and combining the real-time demand correction coefficient to generate a demand response efficiency coefficient;

[0049] fusing the dynamic inventory equivalent, the demand response efficiency coefficient and the sales cost to obtain a corrected inventory turnover rate.

[0050] Further, the inventory turnover rate and the platform virtual inventory allocation target are analyzed to obtain a replenishment urgency, including:

[0051] aligning the corrected inventory turnover rate and the platform virtual inventory allocation target in the space-time dimension to generate a multi-dimensional matching degree matrix;

[0052] based on the multi-dimensional matching degree matrix, the real-time demand correction coefficient and the demand urgency index of each platform, constructing a demand pressure index;

[0053] determining a replenishment elasticity threshold according to the historical replenishment task, the inventory health index and the maximum capacity coefficient;

[0054] analyzing the demand pressure index and the replenishment elasticity threshold to generate a replenishment urgency.

[0055] Further, based on the replenishment urgency and the position information, a set of dynamic optimal cross-warehouse scheduling instructions is obtained, including:

[0056] constructing a spatial distance weight matrix according to the position information of the target object;

[0057] According to the replenishment urgency, the space distance weight matrix and the inventory health index, the scheduling priority coefficient of each target object is calculated.

[0058] With the lowest scheduling cost and the fastest delivery time efficiency as the target and combining the scheduling priority coefficient, the dynamic optimal cross-warehouse scheduling instruction set is obtained.

[0059] In summary, the present application mainly has the following beneficial effects:

[0060] By extracting cross-platform original features, combining data timeliness weighting, and then decomposing the matrix to obtain trend anomaly double-track feature vectors, and fusing platform influence weight, a heat pulse sequence that can accurately reflect real-time heat fluctuation is generated. Compared with traditional systems, the present scheme can track the heat changes of each platform in real time, avoiding the problem of replenishment quota lag caused by heat mutation. For example, during the e-commerce promotion period, a platform suddenly becomes popular due to a promotion strategy, and the heat analysis unit can quickly perceive and generate the corresponding heat pulse sequence, providing accurate basis for subsequent replenishment scheduling, ensuring that warehouse replenishment can adapt to market dynamic changes in a timely manner.

[0061] By time series decomposition of historical replenishment tasks, combined with spatio-temporal dimension analysis of heat pulse sequence, the time window width is adaptively adjusted and the synchronous fluctuation coefficient is calculated, and finally the correction coefficient considering historical regularity and real-time heat is obtained. This makes the system in the multi-platform sales scene not only refer to past experience, but also combine current heat changes to accurately correct commodity demand. In complex scenarios such as different seasons and different promotion activities, it can effectively avoid the replenishment deviation caused by single dependence on historical data or inability to capture real-time heat, realize accurate matching of demand and replenishment quantity, and reduce inventory accumulation or shortage risk.

[0062] By calculating the dynamic platform quota weight according to the real-time demand correction coefficient and real-time data, adjusting the virtual inventory allocation target combined with factors such as inventory health index, and finally generating a dynamic optimal cross-warehouse scheduling instruction set according to the location information, the whole process fully considers the complex factors of multi-platform sales, and schedules with the lowest cost and fastest time efficiency. Not only can it adapt to demand fluctuations caused by changes in platform heat, but also can effectively balance transportation, warehousing and other costs, improve inventory turnover rate, improve warehouse operation efficiency, and enhance the competitiveness of enterprises in the multi-platform sales market. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 It is a schematic diagram of the warehouse intelligent replenishment scheduling system of the present application. DETAILED DESCRIPTION

[0064] Clearly, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present application.

[0065] Reference Figure 1 A warehouse intelligent replenishment scheduling system comprises:

[0066] A heat analysis unit is configured to determine a heat pulse sequence based on real-time data of multiple platforms.

[0067] A demand correction unit is configured to obtain historical replenishment tasks of a target object, fuse the historical replenishment tasks and the heat pulse sequence, and obtain a real-time demand correction coefficient, wherein the target object is a warehouse.

[0068] A distribution optimization unit is configured to calculate a dynamic platform quota weight based on the real-time demand correction coefficient and the real-time data, and adjust a virtual inventory distribution target of each target object facing different platforms in real time based on the dynamic platform quota weight.

[0069] An evaluation unit is configured to calculate an inventory turnover rate of the target object, analyze the inventory turnover rate and the virtual inventory distribution target, and obtain a replenishment urgency.

[0070] A scheduling generation unit is configured to obtain position information of the target object, and obtain a dynamic optimal cross-warehouse scheduling instruction set based on the replenishment urgency and the position information.

[0071] The real-time data includes sales, sales orders, sales times, sales locations, access traffic of platforms, user view counts, add-to-cart counts, favorite counts, search heat, discount strength, activity times, and activity ranges.

[0072] The historical replenishment tasks include replenishment time records, replenishment quantity records, replenishment source records, replenishment cost records, and replenishment delay records.

[0073] The heat analysis unit determines the heat pulse sequence based on the real-time data of the multiple platforms, thereby breaking through the real-time tracking bottleneck of the traditional system. The demand correction unit fuses the historical replenishment tasks and the heat pulse sequence. The distribution optimization unit calculates the dynamic platform quota weight and adjusts the virtual inventory distribution target in real time. The evaluation unit analyzes the inventory turnover rate and the virtual inventory distribution target to obtain the replenishment urgency. The scheduling generation unit generates the dynamic optimal cross-warehouse scheduling instruction set in combination with the position information, effectively balances the demand of the multiple platforms, maximizes the operation benefits, and reduces the loss caused by the untimely replenishment.

[0074] In one case of the embodiment, the heat pulse sequence is determined according to real-time data of multiple platforms, including:

[0075] Feature extraction is performed on the real-time data to generate a cross-platform original feature dataset, specifically including: preprocessing the real-time data of each platform, extracting time domain features using a 5-minute sliding window, decomposing the time series into frequency components using short-time Fourier transform, identifying daily, weekly, and monthly periodic fluctuation features, Z-score standardizing the feature vectors of each platform, calculating a cross-platform feature correlation matrix through cosine similarity, clustering features based on the matrix, and finally generating a cross-platform original feature dataset containing multiple dimensions (time domain traffic features, frequency domain periodic features, cross-platform correlation features, inventory correlation features, and comprehensive scheduling features), wherein the time domain traffic features include order peak value within a time window, average browsing volume, add-to-cart conversion rate fluctuation, time of collection volume, and platform traffic proportion; the frequency domain periodic features include daily periodic fluctuation intensity, periodic traffic distribution, and seasonal periodic fluctuation components; the cross-platform correlation features include correlation between different platforms, cross-platform conversion rate, interest class feature clustering, and abnormal traffic; the inventory correlation features include historical restocking response delay rate, inventory turnover rate deviation, virtual inventory consumption rate, and platform distribution time efficiency requirement; and the comprehensive scheduling features include geographical distance between target warehouse and each platform, restocking urgency, and demand surge probability.

[0076] According to the interval between the timestamp of the data in the cross-platform original feature dataset and the current time, the data in different time periods is weighted for timeliness to obtain a timeliness corrected feature matrix, specifically including: dividing the data into multiple intervals according to the interval between the timestamp and the current time, i.e. the last 1 hour, 1-6 hours, 6-24 hours, 1-3 days, 3 days or more, assigning a fixed weight coefficient to each time interval, the weight coefficient of the last 1 hour is 0.9, the weight coefficient of 1-6 hours is 0.7, the weight coefficient of 6-24 hours is 0.5, the weight coefficient of 1-3 days is 0.3, and the weight coefficient of 3 days or more is 0.1, for each data in the cross-platform original feature dataset, determining the time interval to which it belongs according to its timestamp, multiplying all feature values of the data by the weight coefficient of the corresponding interval, after completing the weighting processing of all data, standardizing all weighted values of the same feature dimension, first calculating the sum of all values of the dimension, then dividing each value by the sum to obtain the relative importance of the feature after timeliness correction, repeating the above steps to process all feature dimensions, and finally forming a timeliness corrected feature matrix, each element in the matrix represents a feature value after timeliness weighting and standardization, reflecting the influence degree of data in different time periods on current decision-making.

[0077] The influence weight vector of different platforms is determined based on real-time data from multiple platforms. Specifically, this involves: selecting key indicators that reflect the influence of a platform from the real-time data, including sales indicators, traffic indicators, conversion indicators, search popularity indicators, and promotion indicators. For each key indicator, the data of each platform is divided by the sum of the data of all platforms for that indicator to calculate the proportion of each platform under that indicator. Then, the indicator is assigned a corresponding weight according to the proportion. The proportion of each platform under each key indicator is multiplied by the weight of the corresponding indicator, and the products are added together to obtain the influence score of each platform. The influence scores of all platforms are arranged and combined in order to form the influence weight vector of different platforms.

[0078] The timeliness-corrected feature matrix is ​​decomposed to obtain the trend. Abnormal dual-track feature vector;

[0079] Influence weight vector and trend across different platforms The abnormal dual-track feature vectors are fused to generate a heat pulse sequence reflecting real-time heat fluctuations. Specifically, this includes: merging time... Time The trend component of each platform divided by the first The normalized trend term is obtained by taking the historical maximum value of the absolute value of the trend component of each platform. , time Time The abnormal components of the platform divided by the first The normalized anomaly term is obtained by finding the historical maximum value of the absolute value of the outlier components on each platform. , time Time The real-time volatility of each platform divided by the first The historical maximum volatility of each platform is used to obtain the normalized volatility term. ; platform The normalized trend term, normalized outlier term, and normalized fluctuation term are assigned weights respectively. , and Then, a weighted sum is performed to obtain the linear fusion factor. The linear fusion factor is passed through the hyperbolic tangent function. A nonlinear compression transformation is performed to obtain the platform heat compression term. ; platform The influence weight is multiplied by the platform's popularity compression factor to obtain the platform's corrected popularity contribution. ;Will The final result is obtained by summing the correction popularity contributions from each platform. Global heat pulse sequence When applied specifically, it can be achieved by the following calculation formula, for example: ;

[0080] In the formula, represents the heat pulse sequence at time , represents the total number of platforms, represents the index of , represents the influence weight of the th platform, taking the value of , represents the hyperbolic tangent function, represents the trend component of the th platform at time , represents the historical maximum value of the absolute value of the trend component of the th platform, represents the abnormal component of the th platform at time , represents the historical maximum value of the absolute value of the abnormal component of the th platform, represents the real-time volatility of the th platform at time , represents the historical maximum volatility of the th platform, represents the trend weight (default 0.6), represents the abnormal weight parameter (default 0.3), represents the volatility weight (default 0.1).

[0081] Through multi-dimensional feature extraction and timeliness weighting mechanism, the real-time tracking ability of the system to the heat of multiple platforms is significantly enhanced. The original feature data set is constructed from the time domain traffic, frequency domain period, and cross-platform association dimensions. The data dynamic characteristics are captured by combining short-time Fourier transform and sliding window technology. At the same time, cross-platform feature association analysis is realized through Z-score standardization and cosine similarity. On this basis, the data is time-weighted according to the timestamp, so that the system can real-time perceive the change of platform heat. When the traffic of a certain platform suddenly increases or the conversion rate is abnormal, the timeliness correction feature matrix can quickly reflect the change, providing data support for dynamically adjusting the replenishment quota, solving the problem of lag response of traditional system to platform heat, and ensuring that the warehouse scheduling can adapt to market fluctuations in time.

[0082] The influence weight vector is fused with the heat pulse sequence to realize intelligent optimization and distribution of the replenishment quota of multiple platforms. The influence weight of each platform is calculated based on key indicators such as sales, traffic and conversion, and the high-value platform can be accurately positioned. The trend anomaly double-track feature vector and the weighted fusion of the fluctuation term are combined to generate a pulse sequence reflecting real-time heat. When a platform has a heat peak due to a promotion activity, the system can dynamically increase the replenishment quota according to the influence weight and heat compression term, while balancing the transportation cost and shortage cost under the constraint conditions of inventory turnover rate and distribution time efficiency, to maximize the warehouse operation benefit and realize multi-objective optimization.

[0083] In one case of the embodiment, the time-sensitive correction feature matrix is decomposed to obtain a trend Anomaly double-track feature vector, comprising:

[0084] The time-sensitive correction feature matrix is preliminarily decomposed to obtain a basic feature component matrix and a corresponding singular value vector, specifically including: regarding the time-sensitive correction feature matrix as a two-dimensional matrix, the rows of the matrix represent different features, and the columns represent different time points or platforms. The matrix is decomposed into the product of three matrices through the SVD algorithm: left singular matrix, singular value matrix and transpose of right singular matrix. The column vectors of the left singular matrix constitute the basic feature component matrix, which is used to reflect the main feature direction of the data. The non-zero elements on the diagonal of the singular value matrix are the corresponding singular value vectors. The singular value vectors are arranged in descending order, and the size of the singular value vector can represent the importance degree of each feature component;

[0085] Based on the basic feature component matrix and the singular value vector, the data of each time dimension is fitted to generate a trend feature vector, specifically including: performing matrix multiplication operation based on the basic feature component matrix and the singular value matrix to reconstruct the feature space, so that the reduced matrix is restored to the structure of the original feature matrix. Then, for the column data corresponding to each time dimension in the reconstructed matrix (i.e. the feature set of each time point), the seasonal and trend decomposition algorithm is used to decompose the time series data into three parts: long-term trend component, seasonal fluctuation component and random noise component. Only the long-term trend component that can reflect the overall evolution direction of the data is retained and the seasonal and noise interference terms are removed. Then, the retained long-term trend component is smoothed by using the moving average algorithm (taking the mean value of the latest N time points). Finally, all trend features are dimensionless processed by Z-score standardization method. All standardized trend features are combined to generate a trend feature vector that can accurately reflect the long-term trend of the data.

[0086] The time-effectiveness correction feature matrix and the trend feature vector are calculated to obtain an abnormal fluctuation feature vector, specifically including: performing matrix multiplication on the trend feature vector and the basic feature component matrix to reconstruct a trend matrix reflecting long-term trend of data, subtracting the trend matrix from the original time-effectiveness correction feature matrix to obtain a residual matrix containing data fluctuation information, calculating the mean and standard deviation of the data of each time dimension in the residual matrix, taking the mean 2 times the standard deviation as an abnormal fluctuation threshold, extracting the fluctuation components in the residual matrix that exceed the threshold, and finally performing Z-score standardization processing on the extracted abnormal fluctuation components, combining the standardized abnormal fluctuation components to generate an abnormal fluctuation feature vector representing sudden abnormal changes in data;

[0087] The trend feature vector and the abnormal fluctuation feature vector are combined to obtain a trend abnormal double-track feature vector, specifically including: aligning the trend feature vector and the abnormal fluctuation feature vector by time dimension to ensure that the trend and abnormal value at each time point correspond one by one, using vector splicing algorithm to connect the trend feature and the abnormal feature at the same time point head to tail to form a double-dimensional composite vector, and performing normalization processing on the composite vector, specifically including: first calculating the mean and standard deviation of each dimension feature, then subtracting the mean and dividing by the standard deviation of each element, and finally generating a trend abnormal double-track feature vector, the trend The first half of the trend

[0088] abnormal double-track feature vector represents the long-term evolution trend of data, and the second half reflects short-term sudden fluctuations.

[0089] Through the construction of the double-track feature vector, dynamic optimization of multi-platform replenishment quota is realized. After the trend and abnormal features are spliced and normalized by time dimension, the first half represents the trend direction (such as the quarterly traffic growth trend of a platform), and the second half reflects short-term abnormalities (such as the instantaneous traffic peak of live streaming e-commerce). The system can adjust the replenishment strategy in real time by combining platform influence weight, specifically including: when the trend feature shows that a platform is in continuous hot sales, the long-term replenishment quota is increased, and when the abnormal feature detects sudden traffic, the temporary replenishment mechanism is triggered, balancing transportation cost and shortage risk, and making warehouse scheduling accurately adapt to the dynamic changes of multi-platform heat, improving operational efficiency.

[0090] In one case of the embodiment, the historical replenishment task and the heat pulse sequence are fused to obtain a real-time demand correction coefficient, including:

[0091] The historical replenishment task of the target object is time-series decomposed to obtain a multi-dimensional replenishment feature vector, specifically including: the historical replenishment task data of the warehouse is arranged in time sequence and cleaned, the missing values are processed by linear interpolation, and the abnormal values are removed, then the data is converted to a suitable time granularity according to business requirements, the time series is decomposed into long-term trend, seasonal cycle and fluctuation by using the STL algorithm, wherein the long-term trend reflects the overall change direction, the seasonal cycle reflects the fixed cycle fluctuation law, and the fluctuation contains random fluctuation and abnormal event influence, then the trend line slope, turning point and stability are calculated based on the long-term trend feature, the cycle length, fluctuation amplitude and phase shift are identified based on the seasonal cycle feature, the residual standard deviation, abnormal event frequency and influence degree are calculated based on the fluctuation feature, finally the extracted features are standardized by Z-score to eliminate the dimension influence, the importance weight of each feature is determined by principal component analysis, the features and the corresponding importance weight are weighted and combined to form a multi-dimensional replenishment feature vector including trend strength, cycle stability and fluctuation amplitude;

[0092] The heat pulse sequence is grid divided according to the time-space dimension to generate a time-space heat matrix, specifically including: when the heat pulse sequence is grid divided according to the time-space dimension to generate a time-space heat matrix, the time dimension of the heat pulse sequence is discretized at a fixed time interval (1 hour), the space dimension is divided according to the geographical region (latitude and longitude grid) to form a time-space grid unit, then each heat pulse data is mapped to the corresponding grid according to the time stamp and the sales location, the data in the same grid is aggregated by weighting, finally the time grid is taken as the row and the space grid is taken as the column to construct the time-space heat matrix composed of each unit heat value;

[0093] Adaptively adjusting the time window width according to the execution period of the historical replenishment task, and calculating the synchronization fluctuation coefficient of the historical replenishment task and the heat pulse sequence, specifically comprising: performing autocorrelation analysis on the historical replenishment time sequence, calculating the autocorrelation coefficients of different time intervals (lag order) in the historical replenishment time sequence, drawing an autocorrelation graph to identify the maximum lag order corresponding to the significant non-zero correlation coefficient, determining the initial time window width based on the maximum lag order (if the maximum lag order corresponds to 7 days, the initial window width is set to 7 days), and calculating the standard deviation of the replenishment quantity using a sliding window, when the standard deviation change rate of adjacent windows exceeds a preset threshold (20%), triggering the window width adjustment mechanism, fitting the historical window width data using an exponential smoothing method to obtain the optimal window width of the next period (if it is a sales peak season, the window width is automatically shortened to 3 days), when calculating the synchronization fluctuation coefficient, first align the heat pulse sequence in the adjusted optimal window width with the historical replenishment quantity data in time, decompose the two sets of data into high-frequency (sudden fluctuations) and low-frequency (long-term trend) components through wavelet transform, calculate the Pearson correlation coefficient of each frequency component, and then weight and sum the contribution of the frequency to the demand, finally obtain the synchronization fluctuation coefficient reflecting the consistency of the fluctuations of the two;

[0094] Wherein, by calculating the autocorrelation coefficients of different time intervals (lag order) in the historical replenishment time sequence, an autocorrelation graph is drawn to identify the maximum lag order corresponding to the significant non-zero correlation coefficient, specifically: arrange the historical replenishment data in chronological order, then calculate the correlation at different time intervals, for example, calculate the correlation of the replenishment quantity today and the replenishment quantity of different days before, such as yesterday, the day before yesterday, and the day before the day before, to obtain a series of correlation values, draw a scatter plot of these correlation values and corresponding time intervals, and draw two boundaries (confidence interval) on the graph to determine whether the correlation is significant, if the correlation value corresponding to a certain time interval exceeds the two boundaries, it means that the correlation at this time interval is significant and non-zero, that is, the replenishment quantity has obvious correlation at this time interval, then find the time interval from the graph where the correlation value first changes from exceeding the boundary to falling within the boundary as the time interval increases, this time interval is the maximum time interval, which can be used as the initial time window width;

[0095] Analyzing the multi-dimensional replenishment feature vector and the space-time heat matrix to obtain a historical influence weight vector and a heat correction weight vector, fusing the historical influence weight vector, the heat correction weight vector, and the synchronization fluctuation coefficient to obtain a real-time demand correction coefficient, specifically comprising: normalizing the historical replenishment feature vector to obtain a normalized historical replenishment feature vector , transposing the attention key vector to obtain , and fusing the normalized historical replenishment feature vector and multiplying the original similarity score dividing the original similarity score by the square root of the key dimension scaling factor to obtain a final similarity score applying a function to the final similarity score to convert the similarity into a probability distribution transposing the historical influence weight vector to obtain multiplying by the probability distribution to obtain a historical feature attention term normalizing the heat feature vector to obtain a normalized heat feature vector transposing the heat correction weight vector to obtain multiplying the normalized heat feature vector by to obtain a heat correction term multiplying the dynamic time warping function-based synchronization fluctuation coefficient by the synchronization fluctuation weight to obtain a synchronization fluctuation term dividing the adaptive time window width by the historical maximum window width of the adaptive time window width to obtain multiplying by the window width weight to obtain a time window adjustment term adding the feature attention term, the heat correction term, the synchronization fluctuation term, and the time window adjustment term to obtain a summation result, and mapping the summation result to by a sigmoid function to obtain a real-time demand correction coefficient When specifically applied, the following calculation formula can be used, for example: ;

[0096] In the formula, represents the real-time demand correction coefficient, and the value is , represents a sigmoid function, represents a historical influence weight vector, represents a transpose operation, represents a normalized exponential function, represents a normalization function, represents a historical replenishment feature vector, represents an attention key vector, represents a key dimension scaling factor, represents a heat correction weight vector, represents a heat feature vector, represents a synchronous fluctuation weight, the value range of which is 0.3 0.9, represents a synchronous fluctuation coefficient based on a dynamic time warping function, the value range of which is 0 1, represents a window width weight, the value range of which is 0.3 0.6, represents an adaptive time window width, represents a historical maximum window width;

[0097] wherein, when 0.3, it indicates that the demand for the commodity is not high, at which time the replenishment needs to be reduced, when 0.7, it indicates that the demand is relatively stable, and the original replenishment plan needs to be maintained, when 0.7, it indicates that the demand is enhanced, at which time the replenishment needs to be increased, and the closer the value is to 1, the more urgent the real-time demand correction is;

[0098] The following is an explanation of the value range of :

[0099] When the fluctuation of multi-platform sales is highly synchronous, and the deviation of the sales fluctuation amplitude of each platform from the average level of the sales fluctuation of each platform is less than 5%, at this time the value range of which is 0.3-0.4, because the multi-platform fluctuation is highly consistent, the weight of real-time correction needs to be suppressed, so as to strengthen the reliability of historical data;

[0100] When the fluctuation of multi-platform sales is moderately synchronous, and the deviation of the sales fluctuation amplitude of each platform from the average level of the sales fluctuation of each platform is less than 10%, at this time the value range of which is 0.41-0.7, because the multi-platform fluctuation is in a moderate fluctuation, at this time the weight of the real-time fluctuation coefficient needs to be increased, so as to capture the differential demand between platforms;

[0101] When the fluctuation of multi-platform sales is highly asynchronous, and the deviation of the sales fluctuation amplitude of each platform from the average level of the sales fluctuation of each platform is less than 15%, at this time the value range of which is 0.7-0.9, because the multi-platform heat difference is obvious, the real-time correction needs to be actively strengthened, so as to maximize the response to the real-time fluctuation of each platform;

[0102] The following is an explanation of the value range of :

[0103] When the turnover rate of goods sold is slow, at this time The value range is 0.3-0.35, thereby avoiding short-term fluctuations from interfering with long-term warehousing planning and reducing the window width weight to a minimum;

[0104] When the turnover rate of goods sold is moderate, at this time The value range is 0.36-0.45, which can take into account both trend stability and seasonal fluctuations, thereby maintaining the response of the baseline window width;

[0105] When the turnover rate of goods sold is relatively fast, at this time The value range is 0.46-0.6. At this point, the sales of goods fluctuate greatly in a short period of time, so it is necessary to increase the sensitivity of the window width to facilitate order tracking.

[0106] The following is about Explanation of the range of values:

[0107] When multiple platforms are experiencing stable sales, the level of interest is relatively stable. The value range is 0-0.3, because the real-time sales volume is relatively close to the historical trend at this time, so the original replenishment plan is maintained.

[0108] When sales status fluctuates across a single platform, and popularity also experiences localized fluctuations, at this time... The value range is 0.31-0.6. Because the platform is experiencing regional fluctuations at this time, in order to ensure the rationality of replenishment, the replenishment quota is increased on the original replenishment plan.

[0109] When sales status fluctuates significantly across multiple platforms, while popularity remains consistently high, at this time... The value range is 0.61-1, because the popularity is very high at this time, and the surface sales volume may increase significantly, requiring the initiation of cross-warehouse transfer and incremental replenishment.

[0110] By decomposing historical replenishment tasks over time, key features such as trend strength and periodic stability are extracted. Principal component analysis is used to determine weights, allowing for a precise understanding of historical replenishment patterns. Furthermore, the heat pulse sequences are aggregated in a grid to construct a spatiotemporal heat matrix, enabling real-time capture of sales dynamics across multiple platforms. Autocorrelation analysis and exponential smoothing are used to adaptively adjust the time window width, and wavelet transform is combined to calculate the synchronization fluctuation coefficient. This allows for flexible matching of analysis granularity based on market fluctuation characteristics. Finally, historical influence weights, heat correction weights, and synchronization fluctuation coefficients are integrated to generate a real-time demand correction coefficient. This ensures that the system can leverage historical experience to guarantee replenishment stability while rapidly responding to sudden changes in platform heat, effectively addressing the shortcomings of traditional systems in real-time heat tracking.

[0111] By understanding the synchronization of multi-platform sales fluctuations, commodity turnover efficiency and platform popularity, different threshold intervals are set for the correction coefficient. When the multi-platform sales fluctuations are highly synchronized, the real-time correction weight is reduced to rely on historical data to ensure the stability of the replenishment strategy. When the popularity difference is significant, the real-time correction is actively strengthened to quickly respond to the individual needs of each platform. For different commodity turnover efficiency, the time window sensitivity is dynamically adjusted to avoid short-term fluctuations or timely capture high-frequency demand changes. This fine-grained response mechanism enables the warehouse system to dynamically balance historical experience and real-time data when facing complex multi-platform sales scenarios, optimizes the allocation of replenishment quotas, and significantly improves the market adaptability and operational efficiency of warehouse scheduling.

[0112] In one case of the embodiment, the dynamic platform quota weight is calculated according to the real-time demand correction coefficient and real-time data, including:

[0113] The real-time data of each platform is identified to obtain a cost sensitivity matrix, specifically including: extracting discount strength, activity cost, delivery time requirement, inventory turnover rate deviation and other cost-related features from the pre-processed real-time data of each platform, using principal component analysis for dimension reduction, calculating the contribution weight of each feature to the cost, summing the feature values and corresponding weights by platform to obtain the cost sensitivity value of each platform, and arranging the cost sensitivity values of all platforms by platform as rows and features as columns to form a cost sensitivity matrix.

[0114] The elastic demand threshold of each platform is calculated based on the real-time demand correction coefficient combined with the historical replenishment task, specifically including: arranging the historical replenishment task of the warehouse for each platform in time sequence to form a replenishment amount time sequence of each platform, performing kernel density estimation on the replenishment amount time sequence of each platform respectively, taking the historical replenishment amount of each platform as a sample, calculating the probability density of different replenishment amount intervals, generating a historical replenishment probability density distribution dedicated to each platform, based on the Bayesian inference principle, taking the historical probability density distribution of each platform as the prior probability, for each platform, converting the real-time demand correction coefficient into a likelihood function, specifically: if the real-time demand correction coefficient of a platform is 0.8, the probability density of the real-time demand fluctuation interval of the platform is strengthened, so that the probability value of the likelihood function in the interval is higher than that in the historical normal interval, and the Markov chain Monte Carlo algorithm is used to iteratively sample each platform: generating initial samples from the historical prior distribution of each platform, and calculating the posterior probability of the new samples of each platform (i.e. the probability of combining historical rules and current real-time correction), after 500 iterations or more, the sample distribution of each platform converges to the posterior distribution (for example, the posterior distribution of a certain platform shows that the probability of replenishment amount 150-200 is the highest), and the 95% quantile of the converged posterior sample set of each platform is calculated: the values of all posterior samples of a platform are sorted by value, and the value at the 95% position is taken as the basic threshold value of the platform, and the time decay weight is calculated according to the historical replenishment timestamp of each platform: the historical replenishment data of a platform within 3 days is given a weight of 1.0, the data from 3 to 7 days is given a weight of 0.7, and the data 7 days ago is given a weight of 0.3, the posterior samples of each platform are reordered by time weight, and the 95% quantile is taken as the elastic demand threshold dynamically adjusted with real-time demand of the platform;

[0115] The replenishment amount of each platform is dynamically optimized to form a Pareto optimal solution set with the minimum total cost and the maximum demand satisfaction rate as the target, and the cost sensitivity matrix and the elastic demand threshold as the constraint condition, specifically including: taking the minimum total cost and the maximum demand satisfaction rate as the target, taking the sum of the product of the cost sensitivity value and the replenishment amount of each platform in the cost sensitivity matrix as the total cost, and taking the average of the ratio of the replenishment amount of each platform to the elastic demand threshold as the demand satisfaction rate, constraining the replenishment amount of each platform to be not less than the elastic demand threshold, and the product of the cost sensitivity value and the replenishment amount not to exceed the total budget, initializing the replenishment amount particle group by using the particle swarm optimization algorithm, calculating the total cost and demand satisfaction rate of each particle, updating the particle position by iteration, retaining the solution that cannot deteriorate another solution in one target optimization, and forming a Pareto optimal solution set;

[0116] According to the cost sensitivity matrix, the Pareto optimal solution set is solved to generate an initial platform quota vector;

[0117] Obtaining inventory data of each target object, analyzing the inventory data to obtain an inventory health index, specifically including: extracting current inventory, total inventory, maximum inventory, turnover rate, batch and distribution data from the inventory data of each warehouse, calculating the turnover rate deviation rate index, the inventory saturation index (current / maximum inventory), the batch health index, the distribution balance index (variance of the proportion of inventory in each warehouse), standardizing each index, and then summing up using the weighted average method to obtain the inventory health index, wherein the inventory data includes: current inventory quantity, total inventory quantity, maximum inventory quantity, inventory turnover rate, inventory batch information, and inventory distribution data, etc.

[0118] According to the inventory health index, the initial platform quota vector is corrected to generate a dynamic platform quota weight, specifically including: mapping the health index to a correction factor through linear transformation, so that the platform with better health state obtains stronger correction gain, then performing point multiplication operation on the initial platform quota vector according to the platform dimension and the corresponding correction factor to obtain a preliminary adjusted quota vector, then verifying whether the adjusted quota satisfies the constraint conditions that the product of the cost sensitivity value of each platform and the adjusted replenishment quantity is not more than the total budget and the replenishment quantity is not less than the elastic demand threshold, if not, the correction factor is fine-tuned through gradient descent method, and finally the adjusted quota vector is normalized to ensure that the sum of the weights of each platform is 1, and then the dynamic platform quota weight is generated.

[0119] The sensitivity matrix is constructed by extracting the discount strength, distribution timeliness and other cost characteristics, and the elastic demand threshold is generated by combining Bayesian inference and time decay weight, so that the system can correct the real-time demand according to the correction coefficient (dynamic calibration of replenishment benchmark), when the heat of a platform suddenly increases due to promotion, the elastic threshold will be quickly corrected combined with the high weight historical data within 3 days, and the particle swarm optimization algorithm is used to optimize the total cost and demand satisfaction rate to form a Pareto optimal solution, solving the problem of lag response of traditional system to platform heat, and ensuring that the replenishment quota can adapt to market changes in real time.

[0120] The health index is calculated by the turnover rate deviation and the inventory saturation, the quota correction factor of the platform with high health degree (such as turnover rate meeting the standard and inventory balance) is increased by 20%, and the cost sensitivity value constraint (such as the total budget not exceeding the threshold) is fine-tuned, when the inventory health degree of a platform decreases, the system automatically reduces its quota weight and triggers cross-warehouse allocation, balancing the transportation cost and the risk of out-of-stock, and making the multi-platform replenishment strategy not only respond to real-time heat fluctuation, but also guarantee the efficiency of inventory operation.

[0121] In one case of the embodiment, the initial platform quota vector is generated by solving the Pareto optimal solution set according to the cost sensitivity matrix, including:

[0122] Based on the cost sensitivity matrix, a response surface between the replenishment quantity of each platform and the total cost is constructed, and a marginal cost gradient matrix is generated, specifically including: based on the cost sensitivity matrix, the replenishment quantity and the cost sensitivity value of each platform are extracted, the nonlinear relationship between the replenishment quantity and the total cost is fitted through a multiple regression algorithm, a response surface is generated, and then the change amplitude of the total cost when the replenishment quantity of each platform increases or decreases by one unit is calculated through a finite difference method, that is, the marginal cost, the marginal cost of each platform is arranged in order, and a marginal cost gradient matrix representing the cost change rate is generated;

[0123] Optimizing the Pareto optimal solution set to generate a weighted Pareto front, specifically including: based on the Pareto optimal solution set, setting the importance proportion of total cost minimization and demand satisfaction rate maximization (cost accounts for 60%, demand accounts for 40%), adding the total cost and demand satisfaction rate of each solution after being converted according to the proportion, obtaining a comprehensive evaluation value, repeatedly calculating and adjusting the evaluation value of each solution through a particle swarm optimization algorithm, selecting the solution with the best comprehensive performance under the proportion, and arranging the evaluation values to form a weighted Pareto front reflecting different target importance preferences;

[0124] Calculating the comprehensive utility value of each point on the weighted Pareto front, and selecting the point with the maximum utility as the optimal solution, and taking the replenishment quantity proportion of each platform corresponding to the point as the initial platform quota vector, specifically including: based on the weighted Pareto front, standardizing the total cost and demand satisfaction rate of each point, weighting and summing the standardized indexes according to the cost weight 0.6 and the demand weight 0.4, obtaining the utility value of each point, comparing the utility values of all points, selecting the point corresponding to the maximum value as the optimal solution, and extracting the proportion of the replenishment quantity of each platform in the total replenishment quantity to form an initial platform quota vector.

[0125] By constructing the cost response surface and the marginal cost matrix, dynamic cost optimization of multi-platform replenishment quota is realized, the nonlinear relationship between the replenishment quantity and the total cost is fitted based on the cost sensitivity matrix, the response surface is generated to intuitively reflect the cost change law, and then the influence of unit change of the replenishment quantity of each platform on the total cost is quantified through the marginal cost gradient matrix, when the distribution cost of a platform increases sharply due to a promotion activity, the system can quickly identify the high-cost-sensitive platform according to the marginal cost gradient, and preferentially adjust the replenishment quota of the platform in the Pareto optimization, so that the total cost control and the demand satisfaction rate reach a dynamic balance, and the problem that the cost control of the traditional system lags behind the change of the platform popularity is solved.

[0126] By setting the weight ratio of cost 60% and demand 40%, the Pareto optimal solution is converted into a comprehensive evaluation value, and the optimal solution considering the minimization of cost and the maximization of demand satisfaction rate is screened out by the particle swarm algorithm. For example, when the demand satisfaction rate weight increases due to the sudden increase in the heat of a platform, the system automatically increases the demand target proportion, and recalculates the utility value and adjusts the quota, so that the proportion of replenishment of the platform is increased, and the increase of the total cost is controlled. This dynamic weight mechanism makes the warehouse scheduling not only respond to the real-time heat of the platform, but also maintain the overall benefit of multi-objective optimization.

[0127] In one case of the embodiment, the virtual inventory allocation target of each target object facing different platforms is adjusted in real time according to the dynamic platform quota weight, including:

[0128] The virtual inventory allocation target of each platform in each target object is obtained, specifically including: first extracting the dynamic quota weight of each platform, the real-time inventory data of each warehouse, and the geographical distance between the warehouse and the platform to quantify the distribution time limit (wherein the time limit weight of the near warehouse is 0.7 and the far warehouse is 0.3), then according to the quota weight of each platform, the total virtual inventory target is allocated to each platform in proportion, then for each platform, the allocation weight of each warehouse is calculated: the available capacity of the warehouse divided by the total available capacity of all warehouses multiplied by the distribution time limit weight, then the allocation weights are normalized to ensure the sum is 1, finally the total virtual inventory of the platform is split to each warehouse according to the warehouse allocation weight, and it is checked whether the allocation exceeds the available capacity of the warehouse, if so, it is reduced in proportion, and finally the virtual inventory allocation target of each platform in each warehouse is determined;

[0129] The maximum capacity coefficient in unit time is determined based on the inventory data, specifically including: after the deviation rate, the available capacity proportion, the batch turnover efficiency and the distribution balance degree are standardized, the weighted sum is calculated according to the turnover rate deviation weight 40%, the available capacity weight 30%, the batch efficiency weight 20% and the distribution balance degree weight 10%, and then multiplied by the time decay factor (the weight of the data in the last 3 days is 1.0, the weight of the data from 3 to 7 days is 0.8, and the weight of the data before 7 days is 0.5), that is, the maximum capacity coefficient in unit time is obtained;

[0130] Based on the dynamic platform quota weight and real-time data of multiple platforms, the demand urgency index of each platform is analyzed, specifically including: extracting real-time data of each platform, including sales, sales order quantity, access traffic, search popularity, discount intensity and other indicators that can reflect demand, standardizing these indicators of different dimensions, extracting key principal components from numerous indicators through principal component analysis, determining the weight of each principal component according to the variance contribution rate of each principal component, obtaining the influence weight of each indicator on demand urgency, and combining the dynamic platform quota weight to weight and sum the standardized real-time indicators (influence weight multiplied by dynamic platform quota weight multiplied by real-time indicator), forming the initial demand urgency base value, taking inventory health index and delivery time as constraint conditions, using particle swarm optimization algorithm for iterative optimization, adjusting parameters to make the urgency index more in line with actual demand, and generating an urgency index that comprehensively reflects the demand urgency of each platform.

[0131] The demand urgency index and maximum capacity coefficient of each platform are calculated to obtain a virtual inventory allocation adjustment coefficient, specifically including: standardizing and converting the demand urgency index and maximum capacity coefficient of each platform to between 0 and 1, weighting and summing the standardized demand urgency index and maximum capacity coefficient according to the weight of demand urgency 0.6 and maximum capacity coefficient 0.4 to obtain an initial adjustment coefficient, comparing the deviation data of historical adjustment coefficient and actual inventory turnover rate, and dynamically adjusting the weight proportion (5% weight adjustment each time) through iterative optimization until the matching degree of the adjustment coefficient and the inventory fluctuation reaches a preset threshold (90%), and finally generating a virtual inventory allocation adjustment coefficient.

[0132] The virtual inventory allocation target is adjusted according to the virtual inventory allocation adjustment coefficient to obtain an adjusted virtual inventory allocation target, specifically including: multiplying the initial virtual inventory allocation target value of each platform in each warehouse by the virtual inventory allocation adjustment coefficient of the corresponding platform to obtain a preliminary adjusted allocation amount, verifying whether the adjustment amount exceeds the available capacity of the warehouse (the available capacity of the warehouse is the maximum inventory minus the current inventory) for each warehouse, and if it exceeds, reducing the allocation amount by the proportion of "available capacity divided by adjusted allocation amount", and finally determining the adjusted virtual inventory allocation target of each platform in each warehouse.

[0133] By allocating the initial target, the dynamic platform quota weight, real-time inventory of the warehouse and geographical distance are comprehensively considered to divide the virtual inventory of each platform in different warehouses, and the maximum capacity coefficient including turnover rate deviation and available capacity is calculated, and through the time decay factor, the inventory allocation is fully matched with the real-time operation state of the warehouse, effectively avoiding the problem of inventory backlog or shortage caused by ignoring the dynamic capacity of the warehouse in traditional systems, and ensuring that the inventory resource allocation always matches the actual warehousing capacity.

[0134] The demand urgency is calculated by combining the principal component analysis of multiple indexes such as sales and search popularity, and the inventory health index and distribution time limit is taken as a constraint, and the particle swarm algorithm is used for iterative optimization to accurately capture the real-time demand changes of each platform. In the adjustment link, the adjustment coefficient is generated according to the weight of the demand urgency 0.6 and the maximum capacity coefficient 0.4, and the weight is iteratively corrected based on historical data, so that the adjustment strategy is highly consistent with the actual inventory fluctuation. For example, the demand urgency of a platform increases sharply due to a promotion activity, and the system will preferentially increase the virtual inventory allocation proportion of the platform, while combining the maximum capacity coefficient of the warehouse to avoid over-allocation, breaking the static allocation limitation of the traditional system. In the scene where the popularity of multiple platforms changes instantaneously, the intelligent replenishment scheduling of inventory resources is realized.

[0135] In one case of the embodiment, the inventory turnover rate of the target object is calculated, including:

[0136] The historical replenishment task and the space-time popularity matrix of the target object are analyzed to form a space-time value decay matrix, specifically including: after preprocessing the historical replenishment task, sorting by time, and extracting features from the preprocessed popularity pulse sequence, dividing the time into 1-hour granularity and the space into latitude and longitude to form a space-time grid, calculating the trend slope, turning point and stability features from the long-term trend, seasonal cycle and fluctuation components decomposed from the historical replenishment data, assigning decay weights from the time dimension, assigning a weight of 1.0 to data within 3 days, a weight of 0.7 to data from 3 to 7 days, and a weight of 0.3 to data 7 days ago, constructing a multi-dimensional feature vector based on the trend slope, turning point density and fluctuation stability, and performing tensor product operation with the weight in the segmented time to form a time decay factor; at the same time, from the spatial dimension, the geographical distance between each warehouse and platform is calculated based on the latitude and longitude coordinates of the replenishment location, and the decay coefficient is set according to the distance (near distance weight is high, far distance weight is low), and the space-time decay matrix is formed by multiplying the geographical distance and the decay coefficient; finally, the time decay factor and the space decay factor are spliced by dimension to construct a space-time decay feature containing time decay and space decay, and the space-time decay feature of the historical replenishment is matched with the space-time popularity matrix according to the grid to calculate the value decay coefficient of each grid, and the decay coefficients of all grids are arranged according to the coordinates of the time grid (row) and the space grid (column) to form a space-time value decay matrix;

[0137] The inventory of the target object is weighted and aggregated with the space-time value decay matrix to obtain a dynamic inventory equivalent, specifically including: mapping the inventory of each warehouse to the corresponding grid according to the space-time coordinates to obtain the value decay coefficient in the grid, multiplying the inventory of each warehouse with the corresponding grid decay coefficient, and weighting and aggregating according to the time grid and space grid dimensions, i.e. summing up the inventory decay values of each grid to obtain the dynamic inventory equivalent of the comprehensive space-time value decay;

[0138] The delay response coefficient of the historical replenishment task and the heat pulse sequence is analyzed, and a demand response efficiency coefficient is generated in combination with a real-time demand correction coefficient, specifically including: after pre-processing the delay record of the historical replenishment task, the heat pulse sequence is time-aligned, the time difference between the heat change and the replenishment response at each time point is calculated, the delay response coefficient is calculated by using the Pearson correlation coefficient, the delay response coefficient and the real-time demand correction coefficient are weighted and fused according to a preset weight (the delay response coefficient weight is 0.6, and the real-time demand correction coefficient correction weight is 0.4), and the demand response efficiency coefficient is generated after normalization processing.

[0139] The dynamic inventory equivalent, the demand response efficiency coefficient and the sales cost are fused to obtain a corrected inventory turnover rate, specifically including: the dynamic inventory equivalent is taken as an adjusted inventory base, the sales cost is divided by the dynamic inventory equivalent to obtain a basic turnover rate, the demand response efficiency coefficient is taken as a correction factor and multiplied by the basic turnover rate, and the product result is normalized to obtain a corrected inventory turnover rate that fuses the space-time value, the response efficiency and the cost factor.

[0140] By assigning a decay factor to historical replenishment data according to time and space, a space-time decay feature is constructed and fused with a heat matrix, so that the inventory equivalent calculation can accurately reflect the space-time differences of multi-platform heat, and the inventory base accounting is dynamically matched with the real-time heat of multi-platform.

[0141] The historical replenishment delay and the real-time demand correction coefficient are quantified by using the Pearson correlation coefficient, a response efficiency factor is generated and embedded in the turnover rate calculation, so that the turnover rate index can reflect the inventory turnover speed and market response ability at the same time. When a sudden traffic peak occurs on a platform, the system adjusts the inventory turnover rate expectation of the platform based on the demand response efficiency coefficient, and at the same time reduces the weight of the far warehouse inventory through the space-time decay matrix, and preferentially guarantees the near-warehouse replenishment. Both the traditional system turnover rate calculation lags behind the heat change, and the space-time-efficiency double-dimensional correction is realized, so as to realize the dynamic optimization allocation of warehouse resources among multi-platforms.

[0142] In one case of the embodiment, the inventory turnover rate and the platform virtual inventory allocation target are analyzed to obtain a replenishment urgency, including:

[0143] The revised inventory turnover rate is spatio-temporally aligned with the platform virtual inventory allocation target to generate a multi-dimensional matching degree matrix, specifically including: the revised inventory turnover rate is divided into 1 hour as a time unit, and is divided into a grid area in space according to latitude and longitude, and the platform virtual inventory allocation target is also divided according to the same time and space standard, so that the spatio-temporal coordinates of the two can be one-to-one corresponding, in the time dimension, due to the length and rhythm of different time data may not be consistent, the dynamic time warping algorithm is used to find the optimal matching path between the two time series, and the misalignment difference in time is eliminated; in the spatial dimension, the actual distance between different warehouses and platform spatial positions is calculated, and then the distance is converted into a similarity value, the closer the distance, the higher the similarity, and finally the time grid is taken as the row and the space grid is taken as the column, the inventory turnover rate data and the virtual inventory allocation target data in each spatio-temporal unit after the time dimension optimal matching path alignment and the spatial dimension similarity processing are placed together, the difference degree index of the two values is calculated, the difference degree index is filled into the matrix according to the corresponding spatio-temporal unit, and finally a multi-dimensional matching degree matrix is constructed;

[0144] Based on the multi-dimensional matching degree matrix, the real-time demand correction coefficient and the demand urgency index of each platform, a demand pressure index is constructed, specifically including: based on the multi-dimensional matching degree matrix, the difference degree of each spatio-temporal unit is extracted, and the weighted sum is obtained by weighting the time decay factor and the spatial distance similarity, to obtain the spatio-temporal pressure reference value, the real-time demand correction coefficient and the demand urgency index are standardized to 0 1 interval, the demand pressure index is generated by weighting and fusing the spatio-temporal pressure reference value 40%, the real-time correction coefficient 30% and the urgency index 30% by weight;

[0145] According to the historical replenishment task, the inventory health index and the maximum capacity coefficient, a replenishment elasticity threshold is determined, specifically including: the historical replenishment task data is subjected to kernel density estimation to generate a historical replenishment probability distribution, the 95% quantile is taken as a basic threshold, the inventory health index and the maximum capacity coefficient are standardized, and a correction factor is obtained by weighting the inventory health 40% and the maximum capacity 60%, the basic threshold is linearly adjusted by the correction factor, and a replenishment elasticity threshold considering the historical law and the current inventory state is generated;

[0146] The demand pressure index and the replenishment elasticity threshold are analyzed to generate a replenishment urgency, specifically including: the demand pressure index and the replenishment elasticity threshold are uniformly converted to the 0-1 interval, the difference value is obtained by subtracting the replenishment elasticity threshold from the demand pressure index, if the difference value is , the urgency is mapped by the ratio of "(difference value+1) / 2", for example, when the difference value is 1, the urgency is 1, and when the difference value is -1, the urgency is 0, the numerical difference is converted into the urgency of 0-1 by linear proportion, and the replenishment urgency is generated, wherein the larger the difference value, the higher the replenishment urgency.

[0147] Through the alignment of the time and space dimensions, the dynamic time warping algorithm solves the time sequence misalignment problem of the inventory turnover rate and the virtual inventory allocation target, and the spatial similarity calculation eliminates the evaluation deviation caused by the geographical distance, ensuring the accuracy of the data comparison. When constructing the demand pressure index, the multi-dimensional matching degree matrix, the real-time demand correction coefficient and the demand urgency index are combined to organically combine the platform popularity changes, historical demand rules and real-time demand dynamics. Compared with the traditional scheme which relies on a single index, the actual replenishment pressure can be more comprehensively and timely reflected.

[0148] The base threshold is determined by kernel density estimation of historical replenishment data, and is dynamically adjusted in combination with the inventory health index and the maximum capacity coefficient, so that the threshold can reflect both the historical replenishment rules and the current warehouse operation state. When the demand pressure index is compared with the replenishment elasticity threshold, the difference is linearly mapped to the urgency in the 0-1 interval, which accurately distinguishes the replenishment priority. For example, the inventory health of a certain warehouse decreases, causing the elasticity threshold to increase, while the demand pressure index of a certain platform increases due to a sudden surge in traffic, and the system determines through difference calculation that the replenishment urgency of the platform is close to 1 in time, and preferentially triggers cross-warehouse allocation or urgent replenishment processes, avoiding replenishment lag or excessive replenishment caused by traditional static threshold setting, and effectively improving the efficiency of warehouse resource scheduling and operation benefits.

[0149] In one case of the embodiment, based on the replenishment urgency and the location information, a set of dynamic optimal cross-warehouse scheduling instructions is obtained, including:

[0150] According to the location information of the target object, a spatial distance weight matrix is constructed, specifically including: when the spatial distance weight matrix is constructed according to the location information of the target object, the latitude and longitude coordinates of each warehouse and platform are first obtained, the actual distance between each other is calculated using geographic information, the actual distance is normalized to obtain a distance coefficient in the interval of 0-1, and then the distance is converted to a similarity weight by "1 minus the distance coefficient", the closer the distance, the higher the weight, and the weight value is filled into the corresponding position according to the warehouse as the row and the platform as the column, to generate a spatial distance weight matrix;

[0151] According to the replenishment urgency, the spatial distance weight matrix and the inventory health index, the scheduling priority coefficient of each target object is calculated, specifically including: normalizing the inventory health index, extracting the weight value in the spatial distance weight matrix corresponding to the target object, and weighting and summing the replenishment urgency, the spatial distance weight matrix and the inventory health index according to the weights of 40% for the replenishment urgency, 30% for the spatial distance weight and 30% for the inventory health index, to obtain the scheduling priority coefficient. The higher the coefficient, the higher the scheduling priority.

[0152] The dynamic optimal cross-warehouse scheduling instruction set is obtained by taking the lowest scheduling cost and the fastest delivery time efficiency as the target and combining the scheduling priority coefficient, and specifically includes: sorting the scheduling cost and delivery time efficiency of each supply warehouse to the target warehouse according to the scheduling priority coefficient of the target warehouse from high to low, for the target warehouse with the highest priority, traversing all supply warehouses, dividing the scheduling cost of each supply warehouse by its own highest scheduling cost and normalizing, dividing the delivery time efficiency by the fastest time efficiency of the supply warehouse to the target warehouse and normalizing, calculating the cost and time weighted sum of each supply warehouse according to the cost and time weight set by the business, selecting the supply warehouse with the smallest weighted sum as the scheduling source, sequentially processing all target warehouses, and generating the dynamic optimal scheduling instruction set.

[0153] By constructing a spatial distance weight matrix, the geographical distance between the warehouse and the platform is converted into a similarity weight, so that the scheduling decision gives priority to near-warehouse allocation, effectively shortening the distribution path. When calculating the scheduling priority coefficient, the replenishment urgency, spatial distance weight and inventory health index are combined to realize the organic combination of demand urgency, geographical advantage and warehouse state, and thus the scheduling demand can be comprehensively and dynamically evaluated, avoiding scheduling delay or resource mismatch caused by one-sided information.

[0154] By taking the lowest scheduling cost and the fastest delivery time efficiency as the target, combining the scheduling priority coefficient to sort the target warehouses, and preferentially processing urgent demands, when selecting the supply warehouse, the scheduling cost and delivery time efficiency are normalized, and the weighted sum is calculated according to the business weight, so as to accurately select the optimal scheduling source. For example, although a certain target warehouse has high priority, the scheduling cost of the nearest warehouse is too high, and the system will comprehensively weigh the cost and time of the next nearest warehouse to select the more optimal one. This dynamic weighing strategy can not only guarantee the rapid response of urgent demand, but also control the operating cost, avoid the cost surge caused by blindly pursuing time efficiency, or the delivery delay caused by excessive compression of cost, and effectively improve the overall replenishment scheduling efficiency of the warehouse system in the multi-platform sales scenario.

[0155] Although embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A warehouse intelligent replenishment scheduling system, characterized in that, include: The heat analysis unit is used to determine the heat pulse sequence based on real-time data from multiple platforms; The demand correction unit is used to obtain the historical replenishment tasks of the target object, and fuse the historical replenishment tasks and the heat pulse sequence to obtain the real-time demand correction coefficient. The allocation optimization unit is used to calculate the dynamic platform quota weight based on the real-time demand correction coefficient and real-time data, and adjust the virtual inventory allocation target of each target object to different platforms in real time according to the dynamic platform quota weight. The evaluation unit is used to calculate the inventory turnover rate of the target object, analyze the inventory turnover rate and the virtual inventory allocation target, and determine the replenishment urgency. The scheduling generation unit is used to obtain the location information of the target object and, based on the replenishment urgency and location information, obtain the dynamic optimal cross-warehouse scheduling instruction set.

2. The intelligent warehouse replenishment scheduling system according to claim 1, characterized in that, The heat pulse sequence was determined based on real-time data from multiple platforms, including: Extract features from real-time data to generate cross-platform raw feature datasets; Based on the interval between the timestamps generated by the data in the cross-platform original feature dataset and the current time, the data in different time periods are weighted according to their timeliness to obtain a timeliness-corrected feature matrix. The influence weight vector of different platforms is determined based on real-time data from multiple platforms. The timeliness-corrected feature matrix is ​​decomposed to obtain the trend. Abnormal dual-track feature vector; Influence weight vector and trend across different platforms The abnormal dual-track feature vectors are fused to generate a heat pulse sequence that reflects real-time heat fluctuations.

3. The intelligent warehouse replenishment scheduling system according to claim 2, characterized in that, The timeliness-corrected feature matrix is ​​decomposed to obtain the trend. The abnormal dual-track feature vector includes: The timeliness-corrected feature matrix is ​​initially decomposed to obtain the basic feature component matrix and the corresponding singular value vector. Based on the basic feature component matrix and singular value vector, the data for each time dimension is fitted to generate a trend feature vector; The abnormal fluctuation feature vector is obtained by calculating the timeliness correction feature matrix and the trend feature vector; By combining the trend feature vector and the abnormal fluctuation feature vector, the trend can be obtained. Abnormal dual-track feature vector.

4. The intelligent warehouse replenishment scheduling system according to claim 2, characterized in that, By fusing historical replenishment tasks and heatwave pulse sequences, a real-time demand correction coefficient is obtained, including: The historical replenishment tasks of the target object are decomposed in time series to obtain a multi-dimensional replenishment feature vector; The heat pulse sequence is divided into grids according to the spatiotemporal dimension to generate a spatiotemporal heat matrix; The time window width is adaptively adjusted based on the execution cycle of historical replenishment tasks, and the synchronization fluctuation coefficient between historical replenishment tasks and the heat pulse sequence is calculated. By analyzing the multidimensional replenishment feature vector and the spatiotemporal heat matrix, the historical influence weight vector and the heat correction weight vector are obtained. The historical influence weight vector, the heat correction weight vector and the synchronous fluctuation coefficient are then fused to obtain the real-time demand correction coefficient.

5. The intelligent warehouse replenishment scheduling system according to claim 4, characterized in that, Based on the real-time demand correction factor and real-time data, the dynamic platform quota weight is calculated, including: The cost sensitivity matrix is ​​obtained by identifying real-time data from each platform. Based on the real-time demand correction coefficient and historical replenishment tasks, the elastic demand threshold for each platform is calculated. With the goals of minimizing total cost and maximizing demand satisfaction, the replenishment volume of each platform is dynamically optimized using the cost sensitivity matrix and elastic demand threshold as constraints, forming a Pareto optimal solution set. The Pareto optimal solution set is solved based on the cost sensitivity matrix to generate the initial platform quota vector; Obtain inventory data for each target object, analyze the inventory data, and derive the inventory health index; The initial platform quota vector is adjusted based on the inventory health index to generate dynamic platform quota weights.

6. The intelligent warehouse replenishment scheduling system according to claim 5, characterized in that, Based on the cost sensitivity matrix, the Pareto optimal solution set is solved to generate an initial platform quota vector, including: Based on the cost sensitivity matrix, a response surface between replenishment volume and total cost for each platform is constructed to generate a marginal cost gradient matrix. Optimize the Pareto optimal solution set to generate a weighted Pareto front. Calculate the overall utility value of each point on the weighted Pareto front, select the point that maximizes utility as the optimal solution, and use the replenishment volume ratio of each platform corresponding to that point as the initial platform quota vector.

7. A warehouse intelligent replenishment scheduling system according to claim 5, characterized in that, The virtual inventory allocation targets for each target object across different platforms are adjusted in real time based on the dynamic platform quota weights, including: Obtain the virtual inventory allocation target that should be allocated to each platform in each target object; Determine the maximum capacity coefficient per unit time based on inventory data; Based on dynamic platform quota weights and real-time data from multiple platforms, the urgency index of demand for each platform is analyzed. The virtual inventory allocation adjustment coefficient is obtained by calculating the demand urgency index and maximum capacity coefficient of each platform. The virtual inventory allocation target is adjusted based on the virtual inventory allocation adjustment coefficient to obtain the adjusted virtual inventory allocation target.

8. The intelligent warehouse replenishment scheduling system according to claim 4, characterized in that, Calculate the inventory turnover rate of the target object, including: Analyze the historical replenishment tasks and spatiotemporal heat matrix of the target object to form a spatiotemporal value decay matrix; The dynamic inventory equivalent is obtained by weighted aggregation of the target object's inventory quantity and the spatiotemporal value decay matrix. Analyze the delay response coefficients of historical replenishment tasks and heat pulse sequences, and combine them with real-time demand correction coefficients to generate demand response efficiency coefficients; By integrating dynamic inventory equivalent, demand response efficiency coefficient, and sales cost, a corrected inventory turnover rate is obtained.

9. A warehouse intelligent replenishment scheduling system according to claim 8, characterized in that, Analyzing inventory turnover rate and platform virtual inventory allocation targets yields replenishment urgency, including: The revised inventory turnover rate is aligned with the platform's virtual inventory allocation target in both time and space to generate a multi-dimensional matching matrix. A demand pressure index is constructed based on a multidimensional matching degree matrix, real-time demand correction coefficient, and demand urgency indicators of each platform. The replenishment elasticity threshold is determined based on historical replenishment tasks, inventory health index, and maximum capacity coefficient. The demand pressure index and replenishment elasticity threshold are analyzed to generate a replenishment urgency level.

10. A warehouse intelligent replenishment scheduling system according to claim 9, characterized in that, Based on replenishment urgency and location information, a dynamically optimal cross-warehouse scheduling instruction set is obtained, including: Construct a spatial distance weight matrix based on the location information of the target object; Based on replenishment urgency, spatial distance weight matrix, and inventory health index, calculate the scheduling priority coefficient for each target object; With the goal of minimizing scheduling costs and maximizing delivery time, and by combining scheduling priority coefficients, a dynamic optimal cross-warehouse scheduling instruction set is derived.

Images