Logistics digital management system based on AI large model

Abstract

The invention relates to the technical field of logistics digital management, in particular to a logistics digital management system based on an AI large model, and the system comprises a demand prediction module which collects market trend data, seasonal change information and promotion activity records, combines the market trend data, seasonal change information and promotion activity records into a market data set, calculates the current commodity demand probability, and generates a market demand matrix; and adjusting Bayesian network parameters according to the market demand matrix to obtain a demand prediction value of each commodity. According to the method, the demand prediction result is dynamically generated after the market trend data, the seasonal change information and the promotion activity records are integrated, and the fluctuation characteristics of the market environment can be captured, so that the commodity demand change is predicted, and the problem of information loss in traditional single-point prediction is avoided. In the aspect of inventory management, real-time updating of the inventory state can be achieved by means of ideal inventory level calculation and execution of a dynamic inventory adjustment plan, and the risks of resource waste and inventory structure imbalance are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of digital logistics management, and in particular to a digital logistics management system based on an AI big model. Background Art

[0002] Digital logistics management belongs to the technical field of logistics management and information technology, focusing on improving the operational efficiency and management capabilities of logistics systems through information, digitalization and intelligent means. Its core lies in combining various links in the logistics process (such as demand forecasting, inventory management, warehouse configuration, transportation scheduling, etc.) with advanced technical means, including the Internet of Things, big data analysis, artificial intelligence, blockchain, cloud computing and other technologies, to achieve real-time data collection, analysis, prediction and optimization of the logistics process.

[0003] However, the existing technologies are more based on static inventory management models with fixed parameters, which makes it difficult to quickly adjust the inventory status, and inventory backlogs and shortages often coexist. In the storage configuration link, the existing technologies lack dynamic analysis of real-time cargo flow information, and the planning of cargo storage locations usually adopts a fixed allocation method, which makes it difficult to optimize the turnover efficiency of cargo, thereby affecting the use of storage space. Therefore, improvements are needed. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and propose a logistics digital management system based on an AI big model.

[0005] In order to achieve the above objectives, the present invention adopts the following technical solutions: The logistics digital management system based on the AI ​​big model includes:

[0006] The demand forecasting module collects market trend data, seasonal change information, and promotion activity records, merges them into a market data set, calculates the current commodity demand probability, and generates a market demand matrix; according to the market demand matrix, adjusts the Bayesian network parameters, obtains the demand forecast value of each commodity, and generates a demand forecast result;

[0007] A dynamic inventory management module calculates the ideal inventory level of each commodity based on the demand forecast result and generates an inventory adjustment plan; executes the inventory adjustment plan, updates the inventory status, and generates an inventory optimization result;

[0008] The intelligent storage configuration module collects real-time data of goods entering and leaving the warehouse, analyzes the goods frequency data, and generates goods flow analysis results; based on the goods flow analysis results, an optimization algorithm is used to recalculate the goods storage location, perform dynamic configuration of warehouse space, and generate space optimization configuration results;

[0009] The performance monitoring and feedback module monitors the warehouse operation efficiency, records the picking time and error rate based on the inventory optimization results and the space optimization configuration results, compares them with standard indicators, and generates optimized performance results.

[0010] Preferably, the steps of obtaining the market demand matrix are:

[0011] Collect market trend data, seasonal change information and promotion activity records, extract time series characteristics, regional distribution characteristics and fluctuation trends in the data, organize them into time characteristic matrix, spatial distribution matrix and fluctuation matrix respectively, perform matrix integration, and generate market basic characteristic matrix;

[0012] According to the market basic characteristic matrix, the current commodity demand probability is calculated using the following formula:

[0013]

[0014] Among them, P d represents the demand probability of goods, T s represents the time fluctuation mean of the corresponding commodity in the time feature matrix, S r represents the market share of commodities in the spatial distribution matrix, W m Indicates the fluctuation range of the corresponding commodity in the volatility matrix, F t -F b Represents the difference between the high-frequency and low-frequency market characteristic difference matrices, G f represents the market growth factor matrix, L m Represents the life cycle factor matrix of the product;

[0015] According to the commodity demand probability, combined with the category distribution structure of each commodity, a matrix is ​​formed by combining the commodity demand classification to generate a market demand matrix.

[0016] Preferably, the steps of obtaining the demand forecast result are:

[0017] According to the market demand matrix, matching the market demand matrix with the initial condition nodes of the Bayesian network to generate initial Bayesian network parameter settings;

[0018] Based on the initial Bayesian network parameter setting, using the joint distribution of the market demand matrix and the historical demand data, gradually adjusting the conditional probability table of each Bayesian network node, updating the parameter configuration of the Bayesian network, and generating adjusted Bayesian network parameters;

[0019] According to the adjusted Bayesian network parameters, the demand forecast value of each commodity is calculated one by one, and the demand forecast values ​​of all commodities are combined to form a demand forecast matrix to generate a demand forecast result.

[0020] Preferably, the steps for obtaining the inventory adjustment plan are:

[0021] Based on the demand forecast results, the forecast demand value of each commodity is extracted, and the basic inventory level is obtained by combining the historical sales data and the existing inventory of the commodity;

[0022] Based on the basic inventory level, the ideal inventory level of each commodity is calculated using the following formula:

[0023]

[0024] Among them, I s is the ideal inventory level, D p is the predicted demand value of the product, H s and H l is the short-term inventory and long-term inventory of the commodity, R c is the supply chain responsiveness factor of the commodity, C m and C c The maximum storage capacity and current storage capacity of the product;

[0025] Based on the ideal inventory level of each product, the ideal inventory level is compared with the existing inventory, and an inventory adjustment plan is generated based on the product classification priority.

[0026] Preferably, the steps of obtaining the inventory optimization result are:

[0027] Based on the inventory adjustment plan, the inventory increase and decrease requirements of each commodity are analyzed item by item, the current inventory status data is called, the inventory adjustment plan is compared with the current inventory status, and the inventory adjustment details of the commodity are generated;

[0028] Based on the inventory adjustment details of the goods, increase and decrease operations are performed, the new demand is transmitted to procurement, and transfer or relocation instructions are generated for the goods with excess inventory. At the same time, the inventory database is updated to generate executed inventory adjustment records;

[0029] Based on the executed inventory adjustment records, the inventory status of each commodity in the inventory database is updated to obtain an inventory optimization result.

[0030] Preferably, the steps for obtaining the cargo flow analysis results are:

[0031] Collect real-time data on goods entering and leaving the warehouse, extract the timestamp, number of times goods enter and leave the warehouse, and the source and destination of each batch of goods, and generate cargo operation record results;

[0032] Based on the cargo operation record results, the cargo frequency data is calculated using the following formula:

[0033]

[0034] Among them, F q is the cargo frequency data, N i is the number of goods entering the warehouse, T o is the total time for goods to be shipped out of the warehouse, |T o -T i | is the absolute difference between the time of entry and exit, M s is the storage interval of the goods, L d is the storage distance factor of the goods, W v is the transport volume of the goods, G r is the average transportation distance of goods, T s is the total storage time of the goods;

[0035] Based on the cargo frequency data, the cargo frequency data is combined with the inbound and outbound source information of each cargo, the dynamic circulation trend of the cargo is analyzed, and the cargo flow analysis results are generated.

[0036] Preferably, the steps of obtaining the space optimization configuration result are:

[0037] Extracting the circulation characteristics of goods according to the goods flow analysis results, including the goods entry frequency, exit frequency, goods type and storage priority, and generating goods dynamic characteristics results by integrating the circulation characteristics data;

[0038] Based on the cargo dynamic characteristics results, the cargo storage priority score is calculated using the following formula:

[0039]

[0040] Among them, P s Score the cargo storage priority, F i is the frequency of goods entering the warehouse, F o is the frequency of goods leaving the warehouse, D s is the average storage and retrieval distance of goods, T c is the storage and retrieval time of goods, Q d is the demand for goods, S t is the storage time of the goods, L s is the shelf space occupied by the goods, H v is the stacking height of the goods, T p G is the handling priority of the goods. c is the total category quantity of the goods;

[0041] Based on the cargo storage priority score, the cargo storage priority score is matched with the existing shelf space in the warehouse, and the storage location of the cargo is adjusted according to the score to generate a space optimization configuration result.

[0042] Preferably, the steps of obtaining the optimized performance results are:

[0043] Based on the inventory optimization results and the space optimization configuration results, the warehouse operation data is extracted, the storage location, inventory change status and picking path of the goods are analyzed, and the dynamic changes of each goods are recorded one by one to obtain the warehouse operation status;

[0044] According to the warehouse operation status, the picking time and picking error rate of the goods are calculated item by item, the picking points and picking sequence associated in the picking path are extracted, the path deviation is analyzed by comparing each record with the standard indicators, and the optimized performance results are generated.

[0045] Compared with the prior art, the advantages and positive effects of the present invention are:

[0046] In the present invention, by integrating market trend data, seasonal change information and promotion activity records to dynamically generate demand forecast results, it is possible to capture the fluctuation characteristics of the market environment, thereby predicting changes in commodity demand and avoiding the problem of missing information in traditional single-point forecasting. In terms of inventory management, with the help of ideal inventory level calculation and the execution of dynamic inventory adjustment plans, real-time updates of inventory status can be achieved, reducing waste of resources and the risk of imbalanced inventory structure. By collecting cargo frequency data in real time, analyzing cargo flow characteristics and optimizing cargo storage locations, warehouse space configuration can be dynamically adjusted, improving the overall utilization efficiency of storage resources. At the same time, cargo dynamic data is combined with picking path analysis to identify inefficient links and sources of errors in the picking process, thereby reducing error rates and picking time through path optimization and feedback mechanisms, and achieving an improvement in the efficiency of logistics operations throughout the entire process. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0049] See also Figure 1 The present invention provides a technical solution: a logistics digital management system based on an AI big model includes:

[0050] The demand forecasting module collects market trend data, seasonal change information, and promotion activity records, merges them into a market data set, calculates the current commodity demand probability, and generates a market demand matrix; according to the market demand matrix, adjusts the Bayesian network parameters to obtain the demand forecast value for each commodity and generates the demand forecast result;

[0051] The dynamic inventory management module calculates the ideal inventory level of each commodity based on the demand forecast results and generates an inventory adjustment plan; executes the inventory adjustment plan, updates the inventory status, and generates inventory optimization results;

[0052] The intelligent storage configuration module collects real-time data on goods entering and leaving the warehouse, analyzes the frequency data of goods, and generates the results of goods flow analysis. Based on the results of goods flow analysis, it uses optimization algorithms to recalculate the storage location of goods, performs dynamic configuration of warehouse space, and generates space optimization configuration results.

[0053] The performance monitoring and feedback module monitors warehouse operation efficiency, records picking time and error rate based on inventory optimization results and space optimization configuration results, compares them with standard indicators, and generates optimized performance results.

[0054] The steps to obtain the market demand matrix are:

[0055] Collect market trend data, seasonal change information and promotion activity records, extract time series characteristics, regional distribution characteristics and fluctuation trends in the data, organize them into time characteristic matrix, spatial distribution matrix and fluctuation matrix respectively, perform matrix integration, and generate market basic characteristic matrix;

[0056] According to the market basic characteristic matrix, the current commodity demand probability is calculated using the following formula:

[0057]

[0058] Among them, P d represents the demand probability of goods, T s represents the time fluctuation mean of the corresponding commodity in the time feature matrix, S r represents the market share of commodities in the spatial distribution matrix, W m Indicates the fluctuation range of the corresponding commodity in the volatility matrix, F t -F b Represents the difference between the high-frequency and low-frequency market characteristic difference matrices, G f represents the market growth factor matrix, L m Represents the life cycle factor matrix of the product;

[0059] According to the commodity demand probability, combined with the category distribution structure of each commodity, the commodity demand classification is combined to form a matrix to generate a market demand matrix.

[0060] Specifically, based on the market trend data, seasonal change information and promotion activity records obtained previously, by reading the data items involving time series characteristics, regional distribution characteristics and fluctuation trends, and making one-to-one correspondence with other related database data that have been obtained, they are uniformly identified in a data management environment and loaded into a visualization interface, and the units and dimensions of each data item are checked item by item to confirm that they are within a reasonable range. For example, the calendar range for date and time data is compared from 1 AD to 9999, the numerical data is compared from 0 to 999999, and the text fields are compared with the established string parsing rules, and a character encoding consistency check is performed. Data that does not meet the range or format is recorded in an exception list, and then combined with the fluctuation trend Disassemble again to extract whether the regional distribution characteristics contain duplication or missing situations, and adapt them uniformly in the fluctuation trend curve by interpolation. Use the artificial neural network model to analyze the adjacent data distribution of each time segment in the time feature matrix. By updating the weights of data from different time segments for multiple iterations during the training phase, the model can learn more seasonal pattern differences. Finally, after identifying no obvious anomalies, all processed data are assigned to corresponding data labels for storage, so that the market basic feature matrix that can be further called and calculated is obtained. After obtaining this matrix, it may be necessary to merge new auxiliary fields again in the subsequent calculation process. Therefore, continue to reserve the merged field position and fill in the field interpretation. Finally, a unified data set is constructed to generate a market basic feature matrix.

[0061] The usefulness of the formula lies in that it combines factors such as time fluctuation, market share, fluctuation amplitude, and high-low frequency difference. By introducing market growth factors and product life cycle factors in the same framework, it can achieve a quantitative assessment of the current product demand probability;

[0062] T s The acquisition steps are as follows: discrete segmentation operation is performed on the time fluctuation sequence of the corresponding product in the time feature matrix, and then the maximum and minimum values ​​are recorded in each time period. The mean of the time fluctuation is calculated based on the statistical method after segmentation averaging. The parameter value is obtained by segmentation calculation of all timestamp data in the past three months. Here, the value is 2.30;

[0063] S r The steps to obtain are to record the ratio of commodity sales volume to total sales volume in each region, and calculate the market share ratio based on weekly data collection, and take the average value of multiple cycles as 0.25;

[0064] W m The steps to obtain are to perform a square sum of deviation analysis on the daily sales volume of the corresponding products in the volatility matrix, take the square sum and then calculate the actual volatility with the average daily sales volume, and then perform overall weighting to obtain 0.17;

[0065] |F t -F b The steps to obtain | are to calculate the absolute value of the previous peak value minus the later valley value in the high-frequency and low-frequency market characteristic difference matrix according to the daily or weekly frequency, and then compensate it with the correction factor of the current cycle. The calculation result here is 0.12;

[0066] G f The steps to obtain it are to calculate the year-on-year growth rate after logarithmic processing of the total sales volume in the most recent quarter, and average the growth values ​​collected multiple times to obtain a market growth factor of 1.05;

[0067] L m The steps to obtain are to determine its life cycle score by observing the complete sales cycle of the product and comparing its duration, and then comprehensively analyze the multiple repurchase situations in different time periods to obtain 0.90 as the product's life cycle factor;

[0068] Calculation process:

[0069] The first step is to calculate the molecular part (T s ·S r +W 2 m ):

[0070] 2.30×0.25+0.17 2 =0.575+0.0289=0.6039

[0071] The second step is to multiply the above result by |F t -F b |:

[0072] 0.6039×0.12=0.072468

[0073] The third step is to calculate the denominator (G f ·L m ):

[0074] 1.05×0.90=0.945

[0075] Step 4: Divide the result of step 2 by the result of step 3:

[0076]

[0077] Step 5: Take the square root of the result:

[0078]

[0079] The result shows that the current demand probability of the product is about 0.277. A high value indicates that the product will experience a certain degree of demand increase in the subsequent sales cycle, while a value below 0.1 indicates that the demand is relatively insufficient;

[0080] According to the commodity demand probability obtained above, the category distribution structure of each commodity can be integrated. During the execution process, the demand probability of each commodity is first numbered and identified, and associated with the commodity type table to which it belongs, and the combination of the associated results is mapped into several classification groups. Then, when performing demand probability stratification for each classification group, the actual sales situation of the commodity and the average demand value of the corresponding category are checked for each stratification, and each record is compared with the established transaction quantity standard range of 0 to 100,000 to check for abnormal situations where the transaction quantity exceeds 100,000 in a short period of time, and the data with inconsistent dimensions are specially marked. By comparing the actual quantity and demand value of the corresponding commodity in each stratification, the situation with obvious differences is sampled again, and the number of occurrences and distribution ranges are recorded. These stratification results are then comprehensively mapped with the above-mentioned demand probabilities to generate a more intuitive demand classification combination matrix. Finally, all classification groups are summarized and updated synchronously based on the stratification and category structure to obtain the final market demand matrix.

[0081] The steps to obtain demand forecast results are:

[0082] According to the market demand matrix, the market demand matrix is ​​matched with the initial condition nodes of the Bayesian network to generate initial Bayesian network parameter settings;

[0083] Based on the initial Bayesian network parameter settings, the joint distribution of the market demand matrix and the historical demand data is used to gradually adjust the conditional probability table of each Bayesian network node, update the parameter configuration of the Bayesian network, and generate the adjusted Bayesian network parameters;

[0084] According to the adjusted Bayesian network parameters, the demand forecast value of each commodity is calculated one by one, and the demand forecast values ​​of all commodities are combined to form a demand forecast matrix to generate the demand forecast result.

[0085] Specifically, according to the market demand matrix obtained above, referring to the product identification and demand value sequence contained therein, firstly, the demand value distribution of all products is numbered and compared with the interval, and the demand value of each product is compared with the range from 0 to 500,000, and recorded in the number table. Then, the data items with abnormalities or missing conditions in the number table are checked one by one. If any record with a demand value higher than 500,000 or lower than 0 is found, it is marked separately and the frequency of occurrence is counted by hierarchical counting. The frequency result is manually verified, and the part that cannot pass the verification is re-compared or eliminated. Then, the demand value information after the comparison is associated with the initial condition node list of the Bayesian network obtained above. Matching, read the node name and condition description corresponding to the product one by one, clarify the specific fields of input data required by each node in subsequent operations by parsing the probability range and dependency relationship set in the node, and then uniformly map the product demand values ​​according to the dependency relationship between the nodes, and combine the conditional weights of the nodes and the set interval checks, such as checking that the demand probability falls between 0.01 and 0.99, and then compare the node status to verify whether the mapping process exceeds the specified range. Once it exceeds 0.99 or is lower than 0.01, backtrack to check the corresponding demand value record, and confirm again whether it matches the actual distribution in the market demand matrix. Finally, the above matching process is completed and the initial Bayesian network parameter settings are generated.

[0086] Based on the initial Bayesian network parameter settings obtained above, the market demand matrix and historical demand data are called and stored in the corresponding data table. In the data table, various fields such as product sales time, sales volume, related channels and customer types are recorded. All fields are interactively compared with the previously defined acceptable range. For example, the sales volume is compared with 0 to 999999, the customer type is compared with a customer category list established in the previous step, and the entries that do not exist in the customer category are excluded. The time field is monitored against the 24-hour period and date sequence. If the time field is found to be confused, it is corrected again by comparing it with the external data source, and then After confirming that there is no obvious inconsistency in the data table, multiple iterative operations are performed against the joint distribution of historical demand data. In each iteration, several demand records are randomly selected to perform local updates on the node conditional probability table. The dependency relationship of each node is adjusted according to the probability of the extracted sales time or customer type and other fields. In order to control the number of iterations, an entropy threshold is set during the iteration process and the change in the conditional probability of the current iteration is used as the basis. If the change is less than the entropy threshold, the iteration is terminated, otherwise the iterative update is continued. After completion, the dependency statistics of each node are performed again to finally obtain the adjusted Bayesian network parameters.

[0087] According to the adjusted Bayesian network parameters obtained above, the product list is traversed in turn and the updated conditional probability table of each product in the network node is extracted. By selecting the statistical data of the corresponding sales period and customer type from the historical demand data, the data is compared with the network node dependency. If a condition item is found to be higher than the previously set probability range of 0.95 to 1.0 during the comparison process, the product is specially marked at the output result and its actual sales fluctuation is checked. If the corresponding extreme value performance is not found in the actual sales fluctuation record, the confidence of the product is marked as a general level in the node, otherwise it is marked as a high value level. After completing the above verification and matching the conditional probability, the demand forecast value of each product is smoothed once, the forecast values ​​of all products are summarized and a demand forecast matrix is ​​formed to generate the demand forecast result.

[0088] The steps to obtain the inventory adjustment plan are:

[0089] Based on the demand forecast results, the forecast demand value of each commodity is extracted, and the basic inventory level is obtained by combining the historical sales data and existing inventory of the commodity;

[0090] Based on the basic inventory level, calculate the ideal inventory level for each product. The calculation formula is:

[0091]

[0092] Among them, I s is the ideal inventory level, D p is the predicted demand value of the product, H s and H l is the short-term inventory and long-term inventory of the commodity, R c is the supply chain responsiveness factor of the commodity, C m and C c The maximum storage capacity and current storage capacity of the product;

[0093] Based on the ideal inventory level of each product, the ideal inventory level is compared with the existing inventory, and an inventory adjustment plan is generated based on the product classification priority.

[0094] Specifically, based on the demand forecast results obtained previously, by reading the demand values ​​of each commodity contained therein and correlating and identifying them with the historical sales data of the commodity, the fields such as the sales date, order quantity, and corresponding return record recorded in the historical sales data are retrieved together and corresponded with the existing inventory. After the product number is identified, the information of each field is concentrated in a product information table, and all demand values ​​are compared with the historical sales quantity according to the preset range of 0 to 999999. If any value is found to be lower than 0 or higher than 999999, it is marked in the abnormal item record and submitted for manual verification. By retrieving the actual number in the sequence, the abnormal item is checked. The actual sales fluctuations and inventory turnover in the corresponding time period are used to check for possible identification errors or unit conversion deviations. After confirming the data that meets the reasonable range, stratified statistics are performed to check the time periods when the sales volume of the goods fluctuates more frequently in the short term. At the same time, the existing inventory is compared with the historical peak value obtained previously. If the inventory is obviously low, the subsequent turnover cycle is continued to be queried to determine whether the time interval between the last entry into the warehouse is too long. Finally, a comprehensive record table is formed by combining all confirmed demand values ​​and inventory information, and then the average sales range and current saleable quantity of each product in the near future are statistically calculated, so as to obtain the basic inventory level at the time of output.

[0095] The formula is useful in that it takes into account the forecasted demand value of the commodity, the sum of short-term and long-term inventory, the supply chain responsiveness factor, and the difference between the maximum and current storage capacity, and can fully reflect the inventory dynamics at different stages through a single expression;

[0096] D p The acquisition steps are as follows: retrieve the product ID and predicted quantity data one by one according to the demand forecast results obtained previously, compare the predicted values ​​in the past week with the actual sales records, and after excluding the cases where the sales records are unclear, perform cumulative average calculation on the valid predicted values ​​to obtain the actual value, which is determined to be 1200 after monitoring and calculation here;

[0097] H s The steps to obtain are: for short-term inventory, summarize the actual inventory data of the goods in the past week, and record the number of shelves and the remaining inventory during daily inventory, and calculate the daily average to get 100;

[0098] H l The steps to obtain are: for long-term inventory, calculate the weekly inventory average within three months, use the weekly inventory results to collect inventory data one by one and calculate the average, then perform a weight calculation for the concentrated purchase period during holidays, separate the weight value from the regular period and finally integrate it to get 200;

[0099] R cThe acquisition steps are as follows: first, obtain the original data from the three dimensions of supply chain transportation time, replenishment frequency and the number of resources that can be deployed, and calculate the monthly supply time distribution. If the supply time is stable at about 2 days and the replenishment cycle is about 10 days, then score the responsiveness value of this type of goods. After comprehensive scoring, 0.80 is obtained as the supply chain responsiveness factor;

[0100] C m The acquisition steps are to select the maximum capacity suitable for the product from the warehouse layout and shelf weight limit data, combined with the possible stacking height of the product in different cargo locations, and determine it to be 3000 after actual verification;

[0101] C c The acquisition steps are to directly read the remaining capacity of the warehouse for the commodity on that day, obtain the current occupancy by counting the incoming and outgoing warehouse flows, and then deduct the actual remaining space from the maximum capacity for conversion. The final value is 2800.

[0102] Calculation process:

[0103] The first step is to calculate the numerator:

[0104] D p ·(H s +H l )=1200×(100+200)=1200×300=360000

[0105] The second step is to calculate the denominator:

[0106] R c =0.80

[0107] The third step is to divide the numerator by the denominator:

[0108]

[0109] The fourth step is to square the above result:

[0110]

[0111] Step 5: Calculate the capacity difference:

[0112] |C m -C c |=|3000-2800|=200

[0113] Step 6. Add the results:

[0114] I s =670.82+200=870.82

[0115] The result shows that when the ideal inventory level is approximately equal to 870.82, the goods can maintain a relatively stable supply under the current supply chain capacity and inventory status. If the value is greater than 1000, it means that the goods require a higher storage capacity for reserve, while a value below 300 indicates that the difference between the demand and inventory of the goods in the recent stage is small.

[0116] Based on the ideal inventory level of each commodity, when comparing the ideal inventory level with the existing inventory, it is necessary to first read the commodity and the corresponding calculation results from the ideal inventory level table generated earlier, and then match these results with the real-time inventory values ​​recorded on site. If the comparison finds that the existing inventory of some commodities is significantly less than the lower limit of the ideal inventory level range, the difference will be marked in a shortage registration item to trigger subsequent procurement or transfer operations. If the comparison finds that the inventory exceeds the upper limit of the ideal inventory level, it will be registered in the temporary allocation item, and a digital comparison will be made between the upper limit threshold and the actual number of commodities in stock to confirm The excess quantity is identified. If the excess quantity is significantly higher than 1,000, it is listed as excess inventory that needs to be handled urgently and recorded in a distribution list. Then, referring to the commodity classification priority list, each priority commodity is matched to a specific shelf and replenishment resource or allocation destination. For commodities that have not appeared in the priority list, their sales activity in the past 90 days is searched in the aforementioned historical sales database, and the inventory is sorted in descending order when the activity is low. Then a horizontal comparison is made for the relatively high activity situation, so as to complete the update and classification of commodities of different priorities, and finally form an inventory adjustment plan.

[0117] The steps to obtain inventory optimization results are:

[0118] Based on the inventory adjustment plan, analyze the inventory increase and decrease requirements of each product item by item, call the current inventory status data, compare the inventory adjustment plan with the current inventory status, and generate the inventory adjustment details of the product;

[0119] Based on the inventory adjustment details of the goods, increase and decrease operations are performed, the new demand is passed to procurement, and transfer or relocation instructions are generated for the goods with excess inventory. At the same time, the inventory database is updated and the executed inventory adjustment records are generated;

[0120] Based on the executed inventory adjustment records, the inventory status of each commodity in the inventory database is updated to obtain the inventory optimization result.

[0121] Specifically, based on the inventory adjustment plan obtained above, the inventory increase and decrease demand of each commodity is first marked in a commodity record table, and these demand information are checked one by one with the current inventory status data according to the commodity number and the corresponding increment or decrement value. When comparing one by one, it is first necessary to extract the real-time inventory quantity, the most recent entry time and the outbound frequency of each commodity from the inventory management system. When comparing, if it is found that the incremental demand is greater than 1000, it is registered in the incremental verification area and a short-cycle inspection mark is assigned. If it is found that the reduction demand exceeds 500, it is registered in the reduction verification area and a removal priority mark is assigned. These numerical thresholds are obtained through monthly inventory fluctuation statistics and modified in combination with inventory turnover indicators of different industries. Then all comparison results are collected and compared with the sales activity table of the goods to confirm one by one whether there are obvious sales fluctuations of the goods. When the sales volume is compared with the range of 0 to 999999, if the value deviates from 999999 or is lower than 0, it will be eliminated, and the remaining data will be merged with the increase and decrease demand to form a detailed and operational list, and the storage location, outbound time and available shelf space of each product and other detailed information will be noted in the list. Finally, the above information is summarized to generate the inventory adjustment details of the goods.

[0122] Based on the commodity inventory adjustment details obtained above, first retrieve all commodities marked with incremental demand and read their incremental values ​​one by one. When comparing these incremental values ​​with the order quantity, check the difference between the incremental and order quantities. If the difference is greater than 200, mark a quick purchase tag when the new demand is transmitted. If the difference is between 50 and 200, mark a regular purchase tag. If the difference is less than 50, mark a small batch purchase tag. The commodities with excess inventory are recorded in a surplus statistics table and compared with the weekly average sales according to the excess range to determine whether to perform transfer or warehouse transfer operations on the commodity. When the weekly average sales are lower than 10 and the inventory exceeds the upper limit of the ideal inventory level of 500, it will be included in the high priority transfer range. If the weekly average sales are higher, enter the warehouse transfer operation process for redistribution to the nearest area. At the same time, the increment or decrement execution action is recorded as an inventory operation event in the existing inventory database. After recording, the operation events are concentrated and the executed inventory adjustment records are formed.

[0123] Based on the executed inventory adjustment records obtained earlier, the product numbers and time information involved in each operation event are checked in turn. When checking, the increment or decrement values ​​in the records are compared with the latest inventory status in the inventory database. If it is found that the inventory value of a certain product in the database has not been updated or an abnormal deviation has occurred, a supplementary record is made in the abnormal log and the database information of the remaining products is checked simultaneously. Subsequently, the increase or decrease range of the verified product items is read one by one and their actual inventory quantity is updated. The new inventory value of each item is compared with the range of 0 to 999999. If a value exceeds 999999 or is lower than 0, it is uniformly put into the manual review list. The items outside the review list are then checked for storage location and shelf remaining quantity to ensure that there are no duplications or omissions when compared with the last inventory information. These finally updated data are unified and integrated into an inventory status update table and handed over to the subsequent execution department for secondary retrieval. After completion, the inventory optimization result is obtained.

[0124] The steps to obtain the results of cargo flow analysis are as follows:

[0125] Collect real-time data on goods entering and leaving the warehouse, extract the timestamp, number of times goods enter and leave the warehouse, and the source and destination of each batch of goods, and generate cargo operation record results;

[0126] Based on the cargo operation record results, the cargo frequency data is calculated using the following formula:

[0127]

[0128] Among them, F q is the cargo frequency data, N i is the number of goods entering the warehouse, T o is the total time for goods to be shipped out of the warehouse, |T o -T i | is the absolute difference between the time of entry and exit, M s is the storage interval of the goods, L d is the storage distance factor of the goods, W v is the transport volume of the goods, G r is the average transportation distance of goods, T s is the total storage time of the goods;

[0129] Based on the cargo frequency data, the cargo frequency data is combined with the inbound and outbound source information of each cargo, the dynamic circulation trend of the cargo is analyzed, and the cargo flow analysis results are generated.

[0130] Specifically, based on the real-time data of goods entering and leaving the warehouse obtained above, the timestamp, entry and exit times of each batch of goods are first classified and marked in a data table, and the source and destination information of the goods are matched with each other through coding. If it is found in the comparison process that the source and destination of the goods are repeated many times and the timestamp distribution is too concentrated, these batches are marked in a centralized registration table to check for duplicate records or data anomalies. In order to confirm the validity of the data, the timestamp of each record needs to be compared with the current calendar range of 0 to 9999, and the entry or exit times need to be compared with 0 to 999999. Any records that exceed the range are regarded as invalid data, and these records are moved to the abnormal project storage area, and then the original registration documents of the batch of goods are queried one by one. If there is still missing or wrong information in the registration document, it will be reviewed in the manual review stage, and then all normal data will be double-sorted by daily aggregation and weekly aggregation. The cargo numbers that appear in or out of the warehouse multiple times in a week will be screened out, and matched with the source and destination one by one. The timeliness of the cargo batch will be checked by comparing the comparison results with the seasonal sales records of the past week. At the same time, the time intervals of entry and exit will be counted. If the difference between the entry and exit time in a single batch is less than 2 hours, it will be marked as short-cycle circulation. If the difference is greater than 72 hours, it will be marked as long-cycle storage and summarized into a time difference distribution table. Subsequently, the above analysis results will be uniformly encoded and attached to the cargo number to form an identifiable data entry. These data entries will be grouped and stored in the final output to obtain the cargo operation record results.

[0131] The usefulness of the formula is that it takes into account factors such as the number of times goods enter the warehouse, the total time of leaving the warehouse, the difference between the time of entering and leaving the warehouse, the transportation volume, the average transportation distance and the total storage time, and provides a more comprehensive measurement of the flow frequency of goods;

[0132] N i The acquisition steps are as follows: through the previously obtained cargo operation record results, the warehousing actions of each batch are grouped and counted by cargo number, and then the number of warehousing records that have occurred in the group is accumulated, and finally the number of warehousing records corresponding to each cargo number is obtained. After summarizing them one by one, the value in this example is 12;

[0133] T o The steps to obtain are: first record the cumulative time consumed when the outbound operation occurs, and when the goods are outbound, perform the difference calculation between the outbound start time and the outbound end time and accumulate them to form the total outbound time. In this example, the value is 48;

[0134] |T o -T i The steps to obtain | are to read the total time T generated by the storage operation.i , and compare it with the total outbound time T o Take the absolute value of the difference. If a negative value appears, take a positive value. Combined with T in the example i =40, in this example, the term is |48-40|=8;

[0135] M s The steps of obtaining are to perform a statistical algorithm on the time interval between the goods entering the warehouse and the next outbound or next entry into the warehouse. The statistical cycle refers to the latest 30 days to obtain the average storage interval of the goods in the warehouse, which is determined to be 24 in this example;

[0136] L d The steps of obtaining are as follows: according to the geographical location records, the actual walking or mechanically transported distance of each cargo when it is transferred in the warehouse is counted, and the average is calculated by segment calibration records to obtain the storage distance factor 1.2;

[0137] W v The steps of obtaining are as follows: according to the volume measurement records obtained previously, the box dimensions are measured when the goods enter the warehouse and the transport volume is obtained by multiplying them, and then the average value of multiple measurements within a month is combined to finally calculate 2.5;

[0138] G r The steps of obtaining are to accumulate the average transportation distance of the goods on each transportation route in a period of time, and finally get 100;

[0139] T s The steps of obtaining are to obtain the complete storage time of the goods from entering the warehouse to the final shipment, and make a weighted average based on multiple batches of data. In this example, the result is 72;

[0140] Calculation process:

[0141] The first step is to calculate the molecular part

[0142] (12+48 2 )×8=(12+2304)×8=2316×8=18528

[0143] The second step is to calculate the denominator M s ·L d :

[0144] 24×1.2=28.8

[0145] The third step is to divide the numerator and denominator and take the square root:

[0146]

[0147] Step 4: Calculate

[0148]

[0149] Step 5: Divide the result of step 4 by T s :

[0150]

[0151] Step 6. Add the result of step 3 to the result of step 5:

[0152] F q =25.36+0.2196≈25.58

[0153] The result shows that the current frequency of goods flow is 25.58. When the value is greater than 30, it means that the goods flow is particularly frequent, and when it is less than 10, it means that the goods flow is relatively sparse.

[0154] Based on the goods frequency data obtained above, the frequency values ​​of the goods are matched with the corresponding records of the incoming source area and the outgoing destination area. In the matching process, first check whether there is a corresponding relationship between the goods number and the regional code, and compare each record with the range of 0 to 999999 in the regional code table. If the regional code is found to be not within this range, it is marked in a list of abnormal records, and after verification, the frequency value and the source and destination are combined for statistics. If the statistical results show that the frequency of some goods exceeds 30, they are marked in a high circulation list, and if the frequency is less than 10, they are marked in a low circulation list. For goods between 10 and 30, the frequency of the goods is not within this range. The entries are marked with regular circulation labels, and then the timestamp distribution in the actual operation records is compared with the number of times and intervals of the goods entering and exiting, and the minimum circulation interval in the past 10 days is compared one by one to see if it meets the set range of 2 hours to 720 hours. By comparing with the pre-established transportation time or storage time threshold, if it is found that it exceeds the preset time threshold, it is marked as delayed or detained. If frequent entry and exit are found and the duration is less than 2 hours, attention is paid to the high concurrency area. By re-associating all the sorted entry and exit information with the corresponding regional distribution entries, the circulation trend of each cargo is finally compared horizontally and summarized to obtain the cargo flow analysis results.

[0155] The steps to obtain the space optimization configuration results are as follows:

[0156] According to the results of cargo flow analysis, the circulation characteristics of the cargo are extracted, including the frequency of cargo entry and exit, cargo type and storage priority. By integrating the circulation characteristic data, the dynamic characteristics of the cargo are generated.

[0157] Based on the dynamic characteristics of the cargo, the cargo storage priority score is calculated using the following formula:

[0158]

[0159] Among them, P s Score the cargo storage priority, F i is the frequency of goods entering the warehouse, F o is the frequency of goods leaving the warehouse, D s is the average storage and retrieval distance of goods, T c is the storage and retrieval time of goods, Q d is the demand for goods, S t is the storage time of the goods, L s is the shelf space occupied by the goods, H v is the stacking height of the goods, T p G is the handling priority of the goods. c is the total category quantity of the goods;

[0160] Based on the cargo storage priority score, the cargo storage priority score is matched with the existing shelf space in the warehouse, and the storage location of the cargo is adjusted according to the score to generate a space optimization configuration result.

[0161] Specifically, according to the cargo flow analysis results obtained previously, the frequency of entry and exit of each cargo is first read one by one in a cargo record table, and then these values ​​are compared with the cargo type for coding, and abnormal or duplicate records are screened by comparing the valid range of 0 to 999999 in the coding table. If there is a value lower than 0 or higher than 999999, it is registered in an abnormal list and subsequent processing is suspended. Then, the corresponding storage priority data is matched in the confirmed records, and the storage priority data is interactively compared with the cargo type in a segmented manner. If the priority value exceeds 80, it is included in the high priority group, if it is lower than 20, it is included in the low priority group, and if it is between 20 and 80, it is positioned as the middle priority group, and then this high, middle and low priority group is The statistics are carried out in parallel with the frequency of goods entering and leaving the warehouse. The statistics need to be combined with the distribution of the timestamps of entry and exit in the past 30 days, and the frequency of entry and exit of each batch of goods in each time period is associated with its corresponding group. By comparing the frequency distribution of these groups in the past 30 days, the activity of the goods in different seasons or activity cycles is confirmed. If the activity of a certain product in the high-priority group is too low, it is recorded in the key checklist to find out the cause of the numerical imbalance. If a certain product in the low-priority group has frequent entry and exit, it is also recorded in the list. Then all the goods on the list are checked one by one, and the daily entry and exit times are compared with the interval between 0 and 10,000 and it is marked whether there are extreme values. Finally, all processed data are integrated to output the dynamic characteristics of the goods.

[0162] The usefulness of the formula is that it takes into account key factors such as the frequency of goods entering and leaving the warehouse, the average storage and retrieval distance, the storage and retrieval time, the demand, the storage time, the shelf space occupied, the stacking height, the processing priority and the total number of categories, and gives a numerical measure of the storage priority of the goods;

[0163] F i The acquisition step is to filter out the number of entry records of each product in the most recent period through the previously obtained dynamic characteristics of the goods, and add up these numbers to get the entry frequency. After testing and summarizing, the parameter value is determined to be 25;

[0164] F o The acquisition steps are as follows: by recording the number of outbound actions, various outbound situations corresponding to the same cargo number are counted and summarized to obtain the outbound frequency, and the current value is determined to be 40;

[0165] D s The acquisition steps are: using distance measuring tools to record the length of the transportation route between multiple stacking locations or cargo locations in the warehouse environment, and superimposing and calculating the average walking distance during each access, and finally summarizing the average access distance 120;

[0166] T c The acquisition steps are as follows: add up the time consumed by each access and average it, extract the main time period and remove the extreme value operation to obtain the access time 95;

[0167] Q d The steps to obtain are to count the demand for goods every week, accumulate the number of orders for the goods in a week, and then average the records in a month for multiple weeks. The current example value is 1800;

[0168] S t The steps of obtaining are: according to the previously obtained results of the dynamic characteristics of the goods, the continuous storage time of the goods in the storage log is counted and the cycle average is calculated, and the final value is 72;

[0169] L s The steps of obtaining are to obtain the unit area occupied by a single piece of goods on the shelf through on-site measurement, and then convert the total area according to the number of stacking layers and summarize it. The current result is 12;

[0170] H v The steps of obtaining are: according to the height limit of the stacking area and the packaging size data of the goods, the height of multiple placement situations is measured and averaged, and the final value is determined to be 2.5;

[0171] T pThe steps for obtaining the priority score are as follows: referring to the processing priority table defined in the cargo category management in the early stage, a higher priority score is marked for high-turnover cargo, and a relatively lower priority score is marked for low-turnover cargo, which is determined to be 3 here;

[0172] G c The acquisition steps are to extract the type codes of all types of goods in the warehouse and calculate the total number. The current detected value is 50;

[0173] Calculation process:

[0174] The first step is to calculate the molecular part (F i +F o )×|D s -T c |×Q d :

[0175] (25+40)×|120-95|×1800=65×25×1800=65×45000=2925000

[0176] The second step is to calculate the denominator (S t +L s ):

[0177] 72+12=84

[0178] Step 3, divide the numerator by the denominator:

[0179]

[0180] The fourth step is to square the result:

[0181]

[0182] Step 5: Calculate

[0183]

[0184] Step 6: Calculate ln(G c +1):

[0185] ln(50+1)=ln(51)≈3.93

[0186] Step 7. Divide the result of step 5 by the result of step 6:

[0187]

[0188] Step 8. Add the result of step 4 to the result of step 7:

[0189] P s =186.59+0.697≈187.287

[0190] The result shows that the current storage priority score of the goods is about 187.287. If the score is greater than 200, it means that the goods have a very high processing and access demand. If the score is less than 50, it means that the goods can be arranged for storage in a farther or low-priority location in the near future.

[0191] Based on the cargo storage priority scores obtained above, we first retrieve the storage capacity, layer height limit and unit area available for stacking corresponding to each shelf number in a warehouse shelf space registration table, and check whether there are overflow or negative values ​​in the range of 0 to 999999. For the confirmed normal shelf data, we perform a bidirectional mapping with the cargo priority scores. Specifically, we first sort the data from high to low according to the scores, temporarily put the cargo with scores higher than 300 into the high-score concentration area, and the cargo with scores lower than 50 into the low-score concentration area. The cargo with scores between 50 and 300 is marked as the regular score concentration area, and then Combine the capacity of each shelf in the shelf space registration table with the layer height limit to match. For example, shelves with a capacity of more than 1,000 units and a layer height of more than 2 meters are grouped as stackable shelves, shelves with a capacity of less than 500 units or a layer height of less than 1 meter are grouped as compact shelves, and shelves with values ​​between the two are grouped as ordinary shelves. Then read each cargo number in the high-scoring concentration area in turn and compare it with the stackable shelf group. If the cargo stacking height data is greater than 2, it is preferentially allocated to the stackable shelf group. If the cargo stacking height data is between 0.5 and 2 and the demand is greater than 1,000, it is also allocated to the stackable shelf group. If the demand is If the quantity is less than 1000 or the occupied area is greater than 500, it will be moved to the ordinary shelf group for matching. Then, the goods in the regular scoring concentration area will be compared with the demand according to the stacking height. If it is found that the shelf occupancy of any goods exceeds 80% of the total shelf capacity, an additional record will be made in the overflow registration table, and the shelf will be skipped in the next round of matching to avoid congestion. At the same time, if there is still space on the shelf, other goods will continue to be matched and allocated to the same shelf until the available capacity of the shelf is less than the set minimum surplus standard of 100. Here, 100 comes from the statistics of the average annual inbound and outbound cargo data of the warehouse and combined with the actual safety of containers and shelves. The balance is confirmed, and finally the high-scoring and regular-scoring goods are allocated. The goods in the low-scoring concentrated area are processed. These goods usually have low stacking requirements and are less frequently put into storage. Therefore, they are mostly assigned to compact shelf groups or ordinary shelf groups. For goods with a demand greater than 500, check again whether the occupied area exceeds the remaining space on the shelf. If it exceeds, check the shelf information again to find out whether there is a more suitable location. Otherwise, join the waiting allocation queue and allocate it after the next round or the next batch of space is released. Finally, all matched goods are bundled with the shelf numbers and recorded, and a location index list is listed to output the space optimization configuration result.

[0192] The steps to obtain the optimized performance results are:

[0193] Based on the inventory optimization results and space optimization configuration results, the warehouse operation data is extracted to analyze the storage location, inventory change status and picking path of the goods, and the dynamic changes of each cargo are recorded one by one to obtain the warehouse operation status;

[0194] According to the warehouse operation status, the picking time and picking error rate of goods are calculated item by item, and the picking points and picking sequence associated with the picking path are extracted. By comparing each record with the standard indicators, the path deviation is analyzed to generate the optimized performance results.

[0195] Specifically, based on the inventory optimization results and space optimization configuration results obtained previously, the real-time location of all goods and the corresponding inventory change status are first read from a warehouse operation record table, and the time difference between the time of entry and exit of the goods is compared one by one to check whether the value falls within the range of 0 to 999999. If the value is less than 0 or greater than 999999, it is marked as an abnormal item. Then, it is checked whether the movement frequency of the goods in a day is more than 5 times or less than 1 time. If it is more than 5 times, it is recorded in a high-frequency movement list, and if it is less than 1 time, it is recorded in a low-frequency movement list. The data of the high-frequency and low-frequency lists are then matched with the shelf numbers and channel numbers in the picking path table to confirm the direction of the goods on the warehouse plan, and the number of times the goods pass through each channel is summarized, and the number of times the goods pass through is compared with the range of 0 to 999999. The channel is compared with the pre-established channel usage frequency range of 500 times. If the value is greater than 500, the channel is checked in detail to check whether there are abnormal duplicate records. The round-trip path length and moving time of the corresponding goods in the records are analyzed one by one, and the moving distance and time are accumulated in each record to calculate the average moving speed. For records with an average moving speed lower than 0.1 or higher than 10, which is a predetermined standard, they are stored in a speed abnormality list. Subsequently, the specific goods numbers in the abnormality list are compared with the warehouse operation record table to confirm whether there are missing or duplicate annotations in the previous data. After manually checking the missing or duplicate annotations one by one, it is determined that all records conform to the basic logic and this part of the data is fully integrated. On this basis, the warehouse operation status including each dynamic change of the goods is formed.

[0196] According to the warehouse operation status obtained above, the picking information of goods at different time points is retrieved one by one and compared according to the picking order and the picking point location. The distance between the starting point and the end point of each picking task is recorded in segments, and the picking time is accumulated. If the picking time is shorter than 30 seconds or longer than 7200 seconds, it is recorded in the abnormal time file, and the corresponding operation time is retrieved from the goods number and the picker operation number for comparison. If the comparison result shows that there are more than 5 repeated picking actions in a single operation, it is recorded in the repeated operation list. If the picking position is found to be different from the marked position during the picking process If the difference exceeds 20 meters, it will be recorded in the positioning gap list. All the data involved in these lists will be checked uniformly and compared with the standard indicators. If the difference with the standard indicators is obvious, it will be recorded in the path deviation summary column. The picking error rate of each cargo will be statistically analyzed and calculated as the ratio of the number of errors to the total number of picking times. If the error rate exceeds 5%, it will be marked as a high error rate interval. If it is lower than 1%, it will be marked as a low error rate interval. The error rate between 1% and 5% will be marked as a regular error rate interval. Finally, the different error rate intervals will be integrated with the distance difference reflected in the path deviation summary column to generate the optimized performance results.

Claims

1. A digital logistics management system based on AI big model, characterized by: The system comprises: The demand forecasting module collects market trend data, seasonal change information, and promotion activity records, merges them into a market data set, calculates the current commodity demand probability, and generates a market demand matrix; according to the market demand matrix, adjusts the Bayesian network parameters, obtains the demand forecast value of each commodity, and generates a demand forecast result; A dynamic inventory management module calculates the ideal inventory level of each commodity based on the demand forecast result and generates an inventory adjustment plan; executes the inventory adjustment plan, updates the inventory status, and generates an inventory optimization result; The intelligent storage configuration module collects real-time data of goods entering and leaving the warehouse, analyzes the goods frequency data, and generates goods flow analysis results; based on the goods flow analysis results, an optimization algorithm is used to recalculate the goods storage location, perform dynamic configuration of warehouse space, and generate space optimization configuration results; The performance monitoring and feedback module monitors the warehouse operation efficiency, records the picking time and error rate based on the inventory optimization results and the space optimization configuration results, compares them with standard indicators, and generates optimized performance results.

2. The logistics digital management system based on AI big model according to claim 1 is characterized in that: The steps for obtaining the market demand matrix are: Collect market trend data, seasonal change information and promotion activity records, extract time series characteristics, regional distribution characteristics and fluctuation trends in the data, organize them into time characteristic matrix, spatial distribution matrix and fluctuation matrix respectively, perform matrix integration, and generate market basic characteristic matrix; According to the market basic characteristic matrix, the current commodity demand probability is calculated using the following formula: Among them, P d represents the demand probability of goods, T s represents the time fluctuation mean of the corresponding commodity in the time feature matrix, S r represents the market share of commodities in the spatial distribution matrix, W m Indicates the fluctuation range of the corresponding commodity in the volatility matrix, F t -F b Represents the difference between the high-frequency and low-frequency market characteristic difference matrices, G f represents the market growth factor matrix, L m Represents the life cycle factor matrix of the product; According to the commodity demand probability, combined with the category distribution structure of each commodity, a matrix is ​​formed according to the commodity demand classification and combination to generate a market demand matrix.

3. The logistics digital management system based on AI big model according to claim 1 is characterized in that: The steps for obtaining the demand forecast result are: According to the market demand matrix, matching the market demand matrix with the initial condition nodes of the Bayesian network to generate initial Bayesian network parameter settings; Based on the initial Bayesian network parameter setting, using the joint distribution of the market demand matrix and the historical demand data, gradually adjusting the conditional probability table of each Bayesian network node, updating the parameter configuration of the Bayesian network, and generating adjusted Bayesian network parameters; According to the adjusted Bayesian network parameters, the demand forecast value of each commodity is calculated one by one, and the demand forecast values ​​of all commodities are combined to form a demand forecast matrix to generate a demand forecast result.

4. The logistics digital management system based on AI big model according to claim 1 is characterized in that: The steps for obtaining the inventory adjustment plan are: Based on the demand forecast results, the forecast demand value of each commodity is extracted, and the basic inventory level is obtained by combining the historical sales data and the existing inventory of the commodity; Based on the basic inventory level, the ideal inventory level of each commodity is calculated using the following formula: Among them, I s is the ideal inventory level, D p is the predicted demand value of the product, H s and H l is the short-term inventory and long-term inventory of the commodity, R c is the supply chain responsiveness factor of the commodity, C m and C c The maximum storage capacity and current storage capacity of the product; Based on the ideal inventory level of each product, the ideal inventory level is compared with the existing inventory, and an inventory adjustment plan is generated based on the product classification priority.

5. The logistics digital management system based on AI big model according to claim 1 is characterized in that: The steps for obtaining the inventory optimization result are: Based on the inventory adjustment plan, the inventory increase and decrease requirements of each commodity are analyzed item by item, the current inventory status data is called, the inventory adjustment plan is compared with the current inventory status, and the inventory adjustment details of the commodity are generated; Based on the inventory adjustment details of the goods, increase and decrease operations are performed, the new demand is transmitted to procurement, and transfer or relocation instructions are generated for the goods with excess inventory. At the same time, the inventory database is updated to generate executed inventory adjustment records; Based on the executed inventory adjustment records, the inventory status of each commodity in the inventory database is updated to obtain an inventory optimization result.

6. The logistics digital management system based on AI big model according to claim 1 is characterized in that: The steps for obtaining the cargo flow analysis results are as follows: Collect real-time data on goods entering and leaving the warehouse, extract the timestamp, number of times goods enter and leave the warehouse, and the source and destination of each batch of goods, and generate cargo operation record results; Based on the cargo operation record results, the cargo frequency data is calculated using the following formula: Among them, F q is the cargo frequency data, N i is the number of goods entering the warehouse, T o is the total time for goods to be shipped out of the warehouse, |T o -T i | is the absolute difference between the time of entry and exit, M s is the storage interval of the goods, L d is the storage distance factor of the goods, W v is the transport volume of the goods, G r is the average transportation distance of goods, T s is the total storage time of the goods; Based on the cargo frequency data, the cargo frequency data is combined with the inbound and outbound source information of each cargo, the dynamic circulation trend of the cargo is analyzed, and the cargo flow analysis results are generated.

7. The logistics digital management system based on AI big model according to claim 1 is characterized in that: The steps for obtaining the space optimization configuration result are: Extracting the circulation characteristics of goods according to the goods flow analysis results, including the goods entry frequency, exit frequency, goods type and storage priority, and generating goods dynamic characteristics results by integrating the circulation characteristics data; Based on the cargo dynamic characteristics results, the cargo storage priority score is calculated using the following formula: Among them, P s Score the cargo storage priority, F i is the frequency of goods entering the warehouse, F o is the frequency of goods leaving the warehouse, D s is the average storage and retrieval distance of goods, T c is the storage and retrieval time of goods, Q d is the demand for goods, S t is the storage time of the goods, L s is the shelf space occupied by the goods, H v is the stacking height of the goods, T p G is the handling priority of the goods. c is the total category quantity of the goods; Based on the cargo storage priority score, the cargo storage priority score is matched with the existing shelf space in the warehouse, and the storage location of the cargo is adjusted according to the score to generate a space optimization configuration result.

8. The logistics digital management system based on AI big model according to claim 1 is characterized in that: The steps for obtaining the optimized performance results are: Based on the inventory optimization results and the space optimization configuration results, the warehouse operation data is extracted, the storage location, inventory change status and picking path of the goods are analyzed, and the dynamic changes of each goods are recorded one by one to obtain the warehouse operation status; According to the warehouse operation status, the picking time and picking error rate of the goods are calculated item by item, the picking points and picking sequence associated in the picking path are extracted, the path deviation is analyzed by comparing each record with the standard indicators, and the optimized performance results are generated.

Images