Stereoscopic warehouse intelligent storage optimization method, engine and computer program product

Abstract

The invention discloses a stereoscopic warehouse intelligent storage optimization method, an engine and a computer program product, and belongs to the technical field of warehouse storage. The method comprises the steps of obtaining a storage information set of goods in a warehouse in a first preset historical time period; acquiring external influence information of the warehouse in a second preset future time period; processing the plurality of storage information subsets and the external influence information in the storage information set based on a time sequence prediction module in the trained prediction model to obtain cargo demand prediction information in a second preset future time period; and based on a dynamic optimization module in the trained prediction model, carrying out optimization processing on the cargo demand prediction information to obtain optimized cargo storage position information. According to the technical scheme, the future cargo demand peak or valley can be foreseen in advance based on the historical cargo in-out data, reasonable cargo layout adjustment is made in advance, and the problem that a warehouse is unsmooth in operation due to sudden demand change is avoided.

Description

Technical Field

[0001] The present invention relates to the field of warehouse storage technology, and in particular to a method, engine, device, medium and computer program product for intelligent storage optimization of a stereoscopic warehouse. Background Art

[0002] With the rapid development of e-commerce and supply chain, warehousing systems play a vital role in modern logistics. Warehouses are not only the core nodes for cargo storage, but also the key link in improving supply chain efficiency. However, in the face of increasingly complex cargo flows and changing market demands, traditional warehousing management methods can no longer meet the requirements of efficient operation of modern logistics. Especially in scenarios such as seasonal fluctuations, promotional activities and sudden demands, warehouses need to be able to quickly and intelligently adjust cargo layout and storage strategies to improve space utilization, reduce operation time and thus reduce operating costs.

[0003] Traditional warehousing systems usually use storage strategies with fixed rules, such as "first in, first out" or "most frequently used goods near the entrance and exit". These rules cannot flexibly respond to dynamic changes in the market, and it is difficult to adapt to seasonal fluctuations or dramatic changes in the frequency of goods in and out caused by promotional activities. Many existing technologies can only be optimized based on the current storage status and historical data, and cannot predict future changes in goods demand in advance. Therefore, they often cannot cope with sudden peaks or inefficiently use warehouse space. Some warehouse management systems use heuristic algorithms (such as greedy algorithms or genetic algorithms) to optimize the layout of goods, but these algorithms can often only solve local problems, lack a global perspective, and cannot dynamically respond to the constantly changing flow of goods. Summary of the invention

[0004] In order to solve the problems existing in the prior art, the present invention provides a method for optimizing intelligent storage of a stereoscopic warehouse, which comprises:

[0005] A storage information set of goods in a warehouse within a first preset historical time period is obtained, wherein the storage information set of goods includes: a plurality of storage information subsets divided sequentially according to preset time intervals in the first preset historical time period, each of the storage information subsets includes a plurality of storage information elements, and each of the storage information elements includes: a timestamp, the category of the goods, the quantity of the goods, and the in-and-out storage location of the goods; external influence information of the warehouse within a second preset future time period is obtained, wherein the external influence information includes: holiday information, weather information, and promotional activity information; a plurality of storage information subsets and the external influence information in the storage information set are processed based on a time series prediction module in a trained prediction model to obtain cargo demand prediction information within the second preset future time period; the cargo demand prediction information is optimized based on a dynamic optimization module in a trained prediction model to obtain optimized cargo storage location information, wherein the cargo storage location information includes storage location allocation information and lane layout information, and the dynamic optimization module is designed based on a heuristic algorithm.

[0006] In the method as described above, optionally, the time series prediction module includes an LSTM layer, a Transformer layer, a multimodal processing layer and an output layer connected in sequence;

[0007] It also includes: an input layer, which is connected to the LSTM layer to output the storage information set of the goods to the LSTM layer, and the input layer is also connected to the multimodal processing layer to output the external influence information to the multimodal processing layer.

[0008] In the method described above, optionally, the result obtained by the optimization processing of the dynamic optimization module is fed back to the timing prediction module to be used as a configuration file of the timing prediction module.

[0009] In the above method, optionally, before obtaining the storage information set of goods in the warehouse within the first preset historical time period, the method further includes:

[0010] Acquire historical data, the historical data including a plurality of internal factor samples and a plurality of external factor samples that are sequentially divided according to the preset time interval in a third preset historical time period of the warehouse, the internal factor samples including: the category of the goods, the quantity of the goods, and the in-and-out warehouse locations of the goods at each timestamp within the preset time interval, and the external factor samples including: holiday information, weather information, and promotional activity information within the preset time interval; normalize the historical data to obtain normalized data; use the normalized data to train an initialized prediction model to obtain a trained prediction model.

[0011] In the method described above, optionally, during the training process, the constraints determined by the dynamic optimization module include: total quantity limit of goods in the lane, length of warehouse entrance and exit paths; and the loss function is to minimize the time consumption for entering and exiting the warehouse.

[0012] In the method described above, optionally, the heuristic algorithm is one of the following algorithms:

[0013] Greedy algorithm, genetic algorithm, ant colony algorithm, simulated annealing algorithm and particle swarm optimization algorithm.

[0014] In another aspect, the present invention provides a stereoscopic warehouse intelligent storage optimization engine, which includes:

[0015] The first acquisition module is used to obtain a storage information set of goods in the warehouse within a first preset historical time period, and the storage information set of goods includes: multiple storage information subsets divided in sequence according to preset time intervals in the first preset historical time period, each of the storage information subsets includes multiple storage information elements, and each of the storage information elements includes: timestamp, category of the goods, quantity of the goods, and in-and-out locations of the goods; the second acquisition module is used to obtain external influence information of the warehouse within a second preset future time period, and the external influence information includes: holiday information, weather information, and promotional activity information; the demand prediction module is used to process the multiple storage information subsets and the external influence information in the storage information set based on the time series prediction module in the trained prediction model to obtain the goods demand prediction information within the second preset future time period; the location information acquisition module is used to optimize the goods demand prediction information based on the dynamic optimization module in the trained prediction model to obtain the optimized goods storage location information, and the goods storage location information includes storage location allocation information and lane layout information, and the dynamic optimization module is designed based on a heuristic algorithm.

[0016] In yet another aspect of the present invention, an electronic device is provided, comprising: a processor and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the above-mentioned intelligent storage optimization method for a stereoscopic warehouse.

[0017] On the other hand, the present invention provides a computer-readable storage medium, which stores at least one instruction, at least one program, code set or instruction set. The at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the above-mentioned intelligent storage optimization method for a high-bay warehouse.

[0018] In yet another aspect, the present invention provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the various steps in the aforementioned intelligent storage optimization method for a stereoscopic warehouse and various possible implementations.

[0019] The technical solution provided by the embodiment of the present invention has the following beneficial effects:

[0020] By obtaining a storage information set of goods in a warehouse within a first preset historical time period, the storage information set of goods includes: a plurality of storage information subsets divided in sequence according to preset time intervals in the first preset historical time period, each storage information subset includes a plurality of storage information elements, each storage information element includes: a timestamp, a category of goods, a quantity of goods, and an in-and-out location of goods; obtaining external influence information of the warehouse within a second preset future time period, the external influence information includes: holiday information, weather information, and promotional activity information; processing the plurality of storage information subsets and external influence information in the storage information set based on a time series prediction module in a trained prediction model to obtain cargo demand prediction information within a second preset future time period; optimizing the cargo demand prediction information based on a dynamic optimization module in a trained prediction model to obtain optimized cargo storage location information, the cargo storage location information includes storage location allocation information and lane layout information, the dynamic optimization module is designed based on a heuristic algorithm, so that it can foresee future cargo demand peaks or troughs in advance based on historical cargo in-and-out data, and make reasonable cargo layout adjustments in advance to avoid the warehouse from falling into a dilemma of poor operation due to sudden changes in demand. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of a method for optimizing intelligent storage in a stereoscopic warehouse provided by an embodiment of the present invention;

[0022] Figure 2 It is a structural diagram of a stereoscopic warehouse intelligent storage optimization engine provided by another embodiment of the present invention. DETAILED DESCRIPTION

[0023] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0024] See also Figure 1 The embodiment of the present invention provides an intelligent storage optimization method for a stereoscopic warehouse, which solves the defect that the traditional warehousing system makes instant storage optimization decisions based on the current cargo storage data, that is, it cannot predict the changes of warehouse cargo in the future and cannot flexibly respond to the dynamic changes of the market and the changes in demand caused by seasonal fluctuations. The method specifically includes the following steps:

[0025] Step 101: Obtain a set of storage information of goods in a warehouse within a first preset historical time period.

[0026] Among them, the storage information set of goods includes: a plurality of storage information subsets divided in sequence according to preset time intervals in the first preset historical time period, each storage information subset includes a plurality of storage information elements, and each storage information element includes four features, namely: timestamp, category of goods, quantity of goods, and in-and-out storage location of goods. The data acquisition method can be a manually input list of goods information or a record of goods information by an automated system, which is not limited in the present invention. The data contained in the storage information set is the historical entry and exit data of goods, which may also be called historical goods inflow and outflow data. The warehouse is a three-dimensional warehouse with configurations such as lanes and storage locations.

[0027] The first preset historical time period may be one day before the current time. In this case, the storage information set of the goods in the first preset historical time period is the storage information of the goods within one day before the current time, or it may be three days or other time lengths determined according to actual conditions. If the behavior corresponding to the timestamp is the outbound behavior, the quantity of goods is the outbound quantity of goods, and the inbound and outbound positions of goods are the outbound positions; if the behavior corresponding to the timestamp is the inbound behavior of goods, the quantity of goods is the inbound quantity of goods, and the inbound and outbound positions of goods are the inbound positions. The inbound and outbound positions of goods are the storage location allocation information and lane layout information of the goods. In other embodiments, the category of goods may also be replaced by the specification of the goods.

[0028] Step 102, obtaining external impact information of the warehouse within a second preset future time period.

[0029] The external influence information includes three features, namely: holiday information, weather information, and promotional activity information. Holiday information can be used to indicate the time of the holiday in the second preset future time period, weather information can be used to indicate the weather in the second preset future time period, and promotional activity information is used to indicate the promotional activity in the second preset future time period. The second preset future time period can be 7 days from the current time, or 30 days from the current time, which is not specifically limited in this embodiment.

[0030] Step 103: Based on the time series prediction module in the trained prediction model, multiple storage information subsets and external influence information in the storage information set are processed to obtain cargo demand prediction information within a second preset future time period.

[0031] The prediction model includes: a time series prediction module and a dynamic optimization module. The time series prediction module is a deep learning model, which includes an input layer, an LSTM (Long Short-Term Memory) layer, a Transformer layer, a multimodal processing layer and an output layer connected in sequence. The input layer is connected to the LSTM layer to output a storage information set of goods to the LSTM layer, and the input layer is also connected to the multimodal processing layer to output external influence information to the multimodal processing layer. The dynamic optimization module is a module designed based on a heuristic algorithm. The heuristic algorithm is any one of the following algorithms: greedy algorithm, genetic algorithm, ant colony algorithm, simulated annealing algorithm and particle swarm optimization algorithm. The result obtained after optimization processing by the dynamic optimization module is fed back to the time series prediction module to be used as a configuration file of the time series prediction module. The result obtained is reflected in the physical model of the stereoscopic warehouse.

[0032] The prediction model is trained using historical cargo in-and-out data. After the training is completed, the output of the time series prediction module can reflect the relationship between the historical cargo in-and-out data and the actual cargo demand. The historical cargo in-and-out data is the data within the third preset historical time period, and the third preset historical time can be the cargo in-and-out data within one year. During training, the data size of the storage information set can be N x 10x 4, where N represents the number of samples, such as N = 32, which means 32 samples, 10 time steps, and 4 features. Nx 10 means that each of the N samples has 10 data records, that is, there are N0 data records in total. The time step is the preset time interval. If a batch of historical data is divided into 10 groups according to the preset time interval, then there are 10 corresponding time steps. Therefore, the data contained in the storage information set can be called time series data. The data size of the external influence information can be N x 3, where N represents the number of samples, such as N = 32, which means 32 samples and 3 external factor features.

[0033] Before training, the historical cargo import and export data is first normalized to eliminate the differences between different cargo categories and orders of magnitude, and convert them into features that can be used by the model.

[0034] The input layer is used to receive historical cargo in and out data and external influencing factor data. The LSTM layer is used to process the time series data from the input layer and capture short-term dependencies. The LSTM layer outputs a hidden state of a sequence, which contains the dependencies between time steps and can obtain the mutual dependencies of data in a short period of time. It is suitable for processing complex demand changes such as seasonal fluctuations and promotional activities. The Transformer layer is a deep learning model based on the self-attention mechanism. It performs well in processing long-distance dependencies. Through the Transformer layer, the dependencies between data with longer time intervals in historical data can be obtained. The Transformer multi-head self-attention mechanism is applied to the output of the LSTM layer to further capture the global dependencies in long time series. By calculating the correlation between different time steps, the influence between long-term time steps can be discovered. The input of the multimodal processing layer is external factor data. The features of the external factor data can be combined with the time series data by feature fusion or embedding vectors to improve the model's ability to handle complex factors. Specifically, the features output by the Transformer layer are combined with the features of external factors through the concat operation. The output layer includes a fully connected layer, which is used to output the forecast of the demand for goods in the future and receive the output data of the aforementioned LSTM layer, Transformer layer, and multimodal processing layer. When the data size of the input layer is: 32x 10x 4, 32x 3, the data size of the LSTM layer is 32x 10x 64, the data size of the Transformer layer is 32x 10x 64, the data size of the multimodal processing layer is 32x 67, and the data size of the output layer is 32x 1.

[0035] The task of the dynamic optimization module is to intelligently optimize the dynamics of goods entering and leaving the warehouse based on the goods demand forecast information provided by the time series forecast module. By dynamically adjusting the storage location and lane allocation of goods, it ensures the optimal storage and access efficiency of goods in different time periods, reduces warehouse operating costs, and improves storage utilization.

[0036] In the process of dynamic optimization of goods, traditional heuristic algorithms (such as greedy algorithms, genetic algorithms, simulated annealing algorithms, etc.) perform well in solving arrangement and path optimization problems, but have limitations in dealing with complex time series and multivariate data. Deep learning can better understand and predict complex demand patterns through its powerful nonlinear modeling capabilities. Combining heuristic algorithms with deep learning can improve prediction accuracy while maintaining flexible optimization capabilities.

[0037] The dynamic optimization module can update the prediction results regularly or in real time using the time series prediction module to form future demand forecasts. Based on the predicted demand, the heuristic algorithm is used to generate the initial allocation plan for the storage location and the lane, while considering various constraints, such as the total amount of goods in the lane, the length of the entrance and exit path, etc. The physical model of the three-dimensional warehouse is constructed, and the optimized storage location allocation and lane arrangement are used as data, that is, in the form of a configuration file of the time series prediction module. Feedback can be used to change the number of layers of the structure of the deep model, such as the number of layers in LSTM. Combined with actual cargo flow data, the prediction accuracy of the model is continuously optimized and the weights of each constraint are adjusted. Taking the minimization of the time consumption of entering and leaving the warehouse as the loss function, an end-to-end prediction model is trained to achieve that at any time, a type of goods is input and its corresponding global optimal placement position in the warehouse is given. In addition, through real-time feedback and iterative updates of the model in actual use, the prediction results of the deep learning model are continuously optimized, and the storage location allocation of the heuristic algorithm can also be dynamically adjusted with changes in demand to improve the adaptability of the system.

[0038] After the prediction model training is completed, the time series prediction module is used to process multiple storage information subsets in the storage information set of the warehouse's goods in the first preset historical time period and the external influence information of the warehouse in the second preset future time period, so as to obtain the goods demand forecast information in the second preset future time period.

[0039] Step 104: Based on the dynamic optimization module in the trained prediction model, the cargo demand prediction information is optimized to obtain optimized cargo storage location information.

[0040] The dynamic optimization module in the trained prediction model optimizes the cargo demand prediction information to obtain optimized cargo storage location information, which includes storage location allocation information and aisle layout information.

[0041] Through the cargo demand forecast results provided by the time series forecast module, the dynamics of cargo in and out of the warehouse are intelligently optimized. By dynamically adjusting the storage location and lane allocation of cargo, the optimal storage and access efficiency of cargo is ensured in different time periods, reducing warehouse operation costs and improving storage utilization.

[0042] The method in this application can improve the space utilization and operation efficiency of the warehouse by 30%, and reduce the time and labor costs by 20%-40%. This method also has the ability to optimize globally, avoid local inefficiency, respond to seasonal or sudden demands in real time, and improve warehousing flexibility and response speed by about 50%, especially during peak periods of increased order volume, which can avoid system bottlenecks. At the same time, intelligent scheduling reduces the risk of cargo loss and expiration, and effectively reduces inventory waste by 10%-15%. Over time, the system will continue to learn and optimize, and the overall warehousing operating costs will drop by more than 20%, further promoting the intelligence and long-term benefits of warehouse management.

[0043] By introducing deep learning time series models such as LSTM, this method can accurately predict future cargo flow trends based on historical cargo inlet and outlet data, so that it can not only optimize based on the current status, but also foresee future cargo demand peaks or troughs in advance, make reasonable cargo layout adjustments in advance, and avoid the warehouse from falling into the predicament of poor operation due to sudden changes in demand.

[0044] This method combines deep learning with optimization algorithms (i.e., heuristic algorithms), which can not only dynamically optimize the storage location of goods, but also continuously adjust the warehouse layout. By learning the complex flow and storage relationships of goods through the neural network model, the system can make more intelligent global optimization of the overall warehouse layout rather than local optimization, greatly improving warehouse utilization and operational efficiency.

[0045] This method can support dynamic adjustment of cargo storage strategies in different periods and scenarios. For example, during seasonal demand peaks, promotional periods, or cargo return cycles, the system can respond quickly and provide real-time optimization solutions. In contrast, traditional systems often rely on manual intervention or periodic manual adjustments and cannot respond quickly to complex market demand changes.

[0046] Storage information elements can also include the frequency of goods entering and exiting. At this time, by predicting and dynamically adjusting the frequency of goods entering and exiting, it can ensure that high-frequency goods entering and exiting are placed in easy-to-access locations, while low-frequency goods are reasonably distributed in farther places. Through such intelligent scheduling, the system can not only reduce the time and labor costs required for goods handling, but also maximize the utilization rate within the limited storage space, significantly reducing warehouse operating costs.

[0047] Storage information elements can also include: cargo attributes, storage conditions, aisle width, warehouse equipment performance, etc., and then the multi-dimensional data is comprehensively optimized. This multi-dimensional optimization capability enables the system to better respond to the diverse needs in complex warehousing environments and further improve overall operational efficiency.

[0048] Compared with existing technologies, this method can automatically analyze the historical data and future demand changes of goods through machine learning, reducing the need for manual intervention and greatly improving the intelligent level of warehouse management. This intelligent storage optimization engine for stereoscopic warehouses can continuously learn and evolve. Over time, the optimization plan will become more and more accurate, making warehouse management more efficient.

[0049] See also Figure 2 An embodiment of the present invention provides a stereoscopic warehouse intelligent storage optimization engine, which includes: a first acquisition module 201, a second acquisition module 202, a second acquisition module 203, and a location information acquisition module 204.

[0050] Among them, the first acquisition module 201 is used to obtain a storage information set of goods in the warehouse within a first preset historical time period, and the storage information set of goods includes: a plurality of storage information subsets divided in sequence according to preset time intervals in the first preset historical time period, each of the storage information subsets includes a plurality of storage information elements, and each of the storage information elements includes: a timestamp, the category of the goods, the quantity of the goods, and the entry and exit locations of the goods.

[0051] The second acquisition module 202 is used to acquire external impact information of the warehouse within a second preset future time period, and the external impact information includes: holiday information, weather information, and promotion activity information.

[0052] The demand forecasting module 203 is used to process the multiple storage information subsets in the storage information set and the external influence information based on the time series forecasting module in the trained forecasting model to obtain the cargo demand forecasting information in the second preset future time period.

[0053] The location information obtaining module 204 is used to optimize the cargo demand forecast information based on the dynamic optimization module in the trained forecast model to obtain optimized cargo storage location information, wherein the cargo storage location information includes storage location allocation information and lane layout information, and the dynamic optimization module is designed based on a heuristic algorithm.

[0054] Optionally, the timing prediction module in the demand prediction module 203 includes an input layer, an LSTM layer, a Transformer layer, a multimodal processing layer and an output layer connected in sequence; the input layer is connected to the LSTM layer to output a storage information set of goods to the LSTM layer, and the input layer is also connected to the multimodal processing layer to output external influence information to the multimodal processing layer.

[0055] Optionally, the result obtained through the optimization processing by the dynamic optimization module is fed back to the timing prediction module to be used as a configuration file of the timing prediction module.

[0056] Optionally, the engine is also used to:

[0057] Acquire historical data, the historical data including multiple internal factor samples and multiple external factor samples that are divided sequentially according to preset time intervals in the warehouse within a third preset historical time period, the internal factor samples including: the category of goods, the quantity of goods, and the in-and-out warehouse locations of goods at each timestamp within the preset time interval, and the external factor samples including: holiday information, weather information, and promotional activity information within the preset time interval; normalize the historical data to obtain normalized data; use the normalized data to train the initialized prediction model to obtain a trained prediction model.

[0058] Optionally, during the training process, the constraints determined by the dynamic optimization module include: the total amount of goods in the aisle, the length of the warehouse entrance and exit path; the loss function is to minimize the time consumed in entering and leaving the warehouse.

[0059] Optionally, the heuristic algorithm is any one of the following algorithms:

[0060] Greedy algorithm, genetic algorithm, ant colony algorithm, simulated annealing algorithm and particle swarm optimization algorithm.

[0061] It should be noted that: the intelligent storage optimization engine for stereoscopic warehouses provided in the above embodiment only uses the division of the above functional modules as an example when performing intelligent storage optimization on stereoscopic warehouses. In actual applications, the above functional distribution can be completed by different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the intelligent storage optimization engine for stereoscopic warehouses provided in the above embodiment and the embodiment of the intelligent storage optimization method for stereoscopic warehouses belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0062] An embodiment of the present invention provides an electronic device, which includes: a memory and a processor. The processor is connected to the memory and is configured to execute the above-mentioned intelligent storage optimization method for a stereoscopic warehouse based on instructions stored in the memory. The number of processors can be one or more, and the processor can be single-core or multi-core. The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as a read-only memory (ROM) or flash RAM, and the memory includes at least one storage chip. The memory can be an example of the following computer-readable medium.

[0063] The embodiment of the present invention provides a computer-readable storage medium, on which at least one instruction, at least one program, code set or instruction set is stored, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the above-mentioned intelligent storage optimization method for a stereoscopic warehouse. Computer-readable storage media include: permanent and non-permanent, removable and non-removable media can implement information storage by any method or technology. Information can be a computer-readable instruction, a data structure, a module of a program or other data. Examples of computer storage media include, but are not limited to: phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk-read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassette, disk storage or other magnetic storage device or any other non-transmission medium, which can be used to store information that can be accessed by a computing device.

[0064] The embodiment of the present invention further provides a computer program product comprising instructions, and when the computer program product is run on a computer, the aforementioned intelligent storage optimization method for a stereoscopic warehouse is executed by the computer.

[0065] It is known from common technical knowledge that the present invention can be implemented by other embodiments that do not deviate from its spirit or essential features. Therefore, the above disclosed embodiments are only illustrative in all respects and are not exclusive. All changes within the scope of the present invention or within the scope equivalent to the present invention are included in the present invention.

Claims

1. A method for optimizing intelligent storage in a stereoscopic warehouse, characterized in that: The method comprises: Acquire a storage information set of goods in a warehouse within a first preset historical time period, the storage information set of goods comprising: a plurality of storage information subsets divided sequentially by the first preset historical time period according to preset time intervals, each of the storage information subsets comprising a plurality of storage information elements, each of the storage information elements comprising: a timestamp, a category of the goods, a quantity of the goods, and an entry / exit location of the goods; Acquire external influencing information of the warehouse in a second preset future time period, wherein the external influencing information includes: holiday information, weather information, and promotional activity information; Processing the plurality of storage information subsets in the storage information set and the external influence information based on the time series prediction module in the trained prediction model to obtain cargo demand prediction information within the second preset future time period; Based on the dynamic optimization module in the trained prediction model, the cargo demand prediction information is optimized to obtain optimized cargo storage location information, wherein the cargo storage location information includes storage location allocation information and lane layout information, and the dynamic optimization module is designed based on a heuristic algorithm.

2. The method according to claim 1, characterized in that The time series prediction module includes an input layer, an LSTM layer, a Transformer layer, a multimodal processing layer and an output layer connected in sequence; The input layer is connected to the LSTM layer to output the storage information set of the goods to the LSTM layer, and the input layer is also connected to the multimodal processing layer to output the external influence information to the multimodal processing layer.

3. The method according to claim 2, characterized in that The result obtained through the optimization processing by the dynamic optimization module is fed back to the timing prediction module to be used as a configuration file of the timing prediction module.

4. The method according to claim 3, characterized in that Before obtaining the storage information set of goods in the warehouse within the first preset historical time period, the method further includes: Acquire historical data, the historical data including a plurality of internal factor samples and a plurality of external factor samples that are sequentially divided according to the preset time interval in the third preset historical time period of the warehouse, the internal factor samples including: the category of the goods, the quantity of the goods, and the in-and-out locations of the goods at each time stamp in the preset time interval, and the external factor samples including: holiday information, weather information, and promotional activity information in the preset time interval; Normalizing the historical data to obtain normalized data; The normalized data is used to train the initialized prediction model to obtain a trained prediction model.

5. The method according to claim 4, characterized in that During the training process, the constraints determined by the dynamic optimization module include: the total amount of goods in the lane limit, the length of the warehouse entrance and exit path; The loss function is to minimize the time consumed in and out of storage.

6. The method according to claim 1, characterized in that The heuristic algorithm is one of the following algorithms: Greedy algorithm, genetic algorithm, ant colony algorithm, simulated annealing algorithm and particle swarm optimization algorithm.

7. A high-bay warehouse intelligent storage optimization engine, characterized in that: The engine comprises: A first acquisition module is used to acquire a storage information set of goods in a warehouse within a first preset historical time period, wherein the storage information set of goods includes: a plurality of storage information subsets divided sequentially according to preset time intervals in the first preset historical time period, each of the storage information subsets includes a plurality of storage information elements, and each of the storage information elements includes: a timestamp, a category of the goods, a quantity of the goods, and an entry and exit location of the goods; A second acquisition module is used to acquire external impact information of the warehouse in a second preset future time period, wherein the external impact information includes: holiday information, weather information, and promotional activity information; A demand forecasting module, configured to process the plurality of storage information subsets in the storage information set and the external influence information based on the time series forecasting module in the trained forecasting model, so as to obtain cargo demand forecasting information within the second preset future time period; The location information acquisition module is used to optimize the cargo demand forecast information based on the dynamic optimization module in the trained forecast model to obtain optimized cargo storage location information, wherein the cargo storage location information includes storage location allocation information and lane layout information, and the dynamic optimization module is designed based on a heuristic algorithm.

8. An electronic device, characterized in that: The electronic device comprises: a processor and a memory for storing executable instructions of the processor; Wherein, the processor is configured to execute the intelligent storage optimization method for a stereoscopic warehouse as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor to implement the intelligent storage optimization method for a high-bay warehouse as described in any one of claims 1 to 6.

10. A computer program product comprising instructions, characterized in that When the computer program product is run on a computer, the intelligent storage optimization method for a stereoscopic warehouse according to any one of claims 1 to 6 is executed by the computer.

Images