A big data intelligent warehouse operation management method and system

Abstract

The application discloses a kind of big data intelligent warehouse operation management method and system, it is related to data processing and intelligent warehouse management technical field, including, acquisition multi-source heterogeneous data and divide into grid unit, while fusion time, space and environmental characteristics, generate space-time data matrix, input space-time data matrix into space-time graph convolution network model, through multilayer space-time convolution analysis the relevance between grid unit, obtain the safety stock prediction value of each area, the present application can accurately obtain the safety stock prediction value of each area by integrating multi-source heterogeneous data and using space-time graph convolution network model to analyze the relevance between these data, improve the accuracy of inventory management decision, use generative adversarial network model to deeply analyze the distribution characteristics of abnormal candidate area, and accurately locate abnormal area and its deviation level by calculating deviation degree, greatly reduce the false positive rate, reduce the dependence on artificial review.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing and intelligent warehouse management, and particularly relates to a big data intelligent warehouse operation management method and system. BACKGROUND

[0002] With the rapid development of algorithm technology, the continuous progress of big data analysis and artificial intelligence technology, the traditional warehouse management method gradually changes to the direction of intelligence and automation. Modern intelligent warehouse not only needs to process massive data, but also needs to quickly and accurately respond to market changes. Early warehouse management mainly relies on manual recording, which is low in efficiency. With the passage of time, the application of bar code scanning technology greatly improves the speed and accuracy of data acquisition.

[0003] However, the prior art still has some deficiencies. On the one hand, the fusion means of multi-source heterogeneous data has limited ability, and cannot effectively fuse multi-dimensional information such as geographical position, environmental monitoring and historical inventory, resulting in inaccurate safety inventory prediction value. On the other hand, the traditional warehouse management mechanism usually lacks self-learning and optimization ability, and cannot automatically adjust parameters to improve performance according to the deviation data in actual operation. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a big data intelligent warehouse operation management method to solve the problems of insufficient fusion of multi-source heterogeneous data and inaccurate anomaly detection.

[0006] To solve the above technical problems, the present application provides the following technical scheme:

[0007] In a first aspect, the present application provides a big data intelligent warehouse operation management method, which comprises collecting multi-source heterogeneous data and dividing it into grid units, simultaneously fusing time, space and environmental characteristics, and generating a space-time data matrix;

[0008] The space-time data matrix is input into a space-time graph convolution network model, the correlation between grid units is analyzed through multi-layer space-time convolution, and the safety inventory prediction value of each region is obtained;

[0009] The current inventory of each grid unit is obtained by using an IoT sensor network, and compared with the safety inventory prediction value to obtain an abnormal candidate region;

[0010] A generative adversarial network model is constructed to analyze the distribution characteristics of the abnormal candidate region in the generative adversarial network model space, and to calculate the deviation degree. When the deviation value exceeds the statistical threshold, the abnormal region coordinates and deviation level are output;

[0011] Different response processes are started according to the deviation level, and deviation data in the deviation level are input to the feedback spatiotemporal graph convolution network model, parameters are dynamically adjusted, and an optimized spatiotemporal graph convolution network model is obtained.

[0012] Based on the optimized spatiotemporal graph convolution network model, the optimal inventory quantity of each region is calculated, and a replenishment instruction is generated, and an operation management report is generated based on the replenishment instruction.

[0013] As a preferred scheme of the operation management method of the big data intelligent warehouse, the multi-source heterogeneous data includes geographic location data, environmental monitoring data and historical inventory data.

[0014] As a preferred scheme of the operation management method of the big data intelligent warehouse, the step of obtaining the safety inventory prediction value of each region is as follows,

[0015] The spatiotemporal data matrix is input into the spatiotemporal graph convolution network model, the spatial correlation features between adjacent grids are extracted through the first layer spatiotemporal convolution, and the cooperative change law of the geographically adjacent regions is analyzed.

[0016] The dynamic trend features of each grid unit in the time sequence are captured through the second layer time convolution, and the long-term and short-term fluctuation modes are identified.

[0017] The weights of the spatial correlation features and the dynamic trend features are dynamically adjusted by using the attention mechanism, and the future safety inventory prediction value of each grid unit is output.

[0018] As a preferred scheme of the operation management method of the big data intelligent warehouse, the step of obtaining the abnormal candidate region is as follows,

[0019] The warehouse data is collected in real time by using the IoT sensor network, the original inventory log is generated, the interference is removed by using the multi-reading filtering and signal compensation method, and the inventory data is obtained.

[0020] The safety inventory prediction value obtained by the inventory data and the spatiotemporal graph convolution network model is dynamically matched through real-time stream window calculation and dynamic time warping algorithm, and the grid unit exceeding the safety inventory threshold is marked as a preliminary screening abnormality by using a stream processing engine.

[0021] The preliminary screening abnormality is combined with a rule engine, false positives are excluded through a context-aware rule chain, and the preliminary screening abnormality is merged into an abnormal candidate region by using a density clustering algorithm combined with a dynamic neighborhood radius, and the coordinates and the type of the abnormality are output after cross-validation of the multi-source sensors.

[0022] As a preferred scheme of the operation management method of the big data intelligent warehouse, the step of outputting the coordinates of the abnormal region and the deviation level includes the following steps,

[0023] Based on the generator and the discriminator, a conditional generative adversarial network model is constructed through contrastive learning supervision and double-path adversarial training.

[0024] Based on the generative adversarial network model, the spatio-temporal features of the abnormal candidate region are mapped and matched with the features of the normal distribution through a spatio-temporal graph attention mechanism, and the spatial deviation degree is calculated based on a ternary network.

[0025] Based on the spatial deviation degree, a deviation value is generated, when the deviation value exceeds a statistical threshold, the grid coordinates are located through inverse mapping, and the abnormal region coordinates and deviation level are output according to the over-limit ratio.

[0026] As a preferred scheme of the operation management method of the big data intelligent warehouse, the deviation data is fed back to the spatio-temporal graph convolutional network model, the parameters are dynamically adjusted, and the optimized spatio-temporal graph convolutional network model is obtained, which comprises the following steps,

[0027] After aligning the deviation data based on the Flink time window, the periodic and trend features are extracted through the Prophet residual decomposition, and the gated attention mechanism is fed back to the spatio-temporal graph convolutional network model.

[0028] Based on the deviation data, the spatio-temporal graph convolutional network model updates the parameter direction through EWC constraint, dynamically adjusts the spatio-temporal attention weight by combining small batch incremental training, separates the EWC importance mask and momentum term in gradient clipping, and updates the convolution kernel parameters using the AdamW optimizer to generate the optimized spatio-temporal graph convolutional network model.

[0029] As a preferred scheme of the operation management method of the big data intelligent warehouse, the optimized spatio-temporal graph convolutional network model is used to calculate the optimal inventory of each region, and a replenishment instruction is generated, and an operation management report is generated based on the replenishment instruction, which comprises the following steps,

[0030] Based on the optimized spatio-temporal graph convolutional network model, the data of inventory, order and environment are fused through the time window sliding alignment mechanism, and the safety inventory prediction value is combined to input the dynamic programming algorithm to calculate the optimal inventory value under the spatio-temporal constraint;

[0031] The replenishment path is generated by integrating the A algorithm and the ant colony algorithm, the AGV is dynamically scheduled to execute the task through the priority queue, and the replenishment time limit requirement and the goods category are injected into the instruction template to generate the replenishment instruction.

[0032] Based on the replenishment instruction, the historical replenishment records and the inventory fluctuation curve are aggregated through the time series database, and the Grafana tool is used to automatically generate a multi-dimensional analysis board to output the operation management report.

[0033] In a second aspect, the present application provides an operation management system of a big data intelligent warehouse, comprising a collection module, a prediction module, an abnormality positioning module, a deviation calculation module, a model optimization module and a report generation module.

[0034] The collection module is configured to collect multi-source heterogeneous data and divide the multi-source heterogeneous data into grid units, and generate a time-space data matrix by fusing time, space and environmental characteristics.

[0035] The prediction module is configured to input the time-space data matrix into a time-space graph convolution network model, analyze the correlation between the grid units through multi-layer time-space convolution, and obtain a safety stock prediction value of each region.

[0036] The abnormality positioning module is configured to obtain a current inventory of each grid unit by using an IoT sensor network, compare the current inventory with the safety stock prediction value, and obtain an abnormal candidate region.

[0037] The deviation calculation module is configured to construct a generative adversarial network model, analyze the distribution characteristics of the abnormal candidate region in the generative adversarial network model space, and calculate a deviation degree, and output an abnormal region coordinate and a deviation level when the deviation value exceeds a statistical threshold.

[0038] The model optimization module is configured to start different response processes according to the deviation level, feed the deviation data in the deviation level back to the time-space graph convolution network model, dynamically adjust parameters, and obtain an optimized time-space graph convolution network model.

[0039] The report generation module is configured to calculate an optimal inventory of each region based on the optimized time-space graph convolution network model, generate a replenishment instruction, and generate an operation management report based on the replenishment instruction.

[0040] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, any step of the operation management method of the big data intelligent warehouse according to the first aspect of the present application is implemented.

[0041] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, any step of the operation management method of the big data intelligent warehouse according to the first aspect of the present application is implemented.

[0042] The application has the beneficial effects that: by integrating multi-source heterogeneous data and using a spatio-temporal graph convolution network model to analyze the correlation between the data, the application can more accurately obtain the safety stock prediction value of each region. This not only improves the accuracy of inventory management decisions, but also enables the warehouse to meet customer demand while reducing the occurrence of excessive inventory or out-of-stock phenomena. By using a generative adversarial network model to deeply analyze the distribution characteristics of abnormal candidate regions and calculating the deviation degree to accurately locate the abnormal regions and their deviation levels, the false positive rate is greatly reduced, the dependence on manual review is reduced, manpower resources are saved, and the problem response speed is accelerated. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0044] Fig. 1 Flowchart of the operation and management method of the big data intelligent warehouse in embodiment 1;

[0045] Fig. 2 Schematic diagram of the operation and management method of the big data intelligent warehouse in embodiment 1;

[0046] Fig. 3 Flowchart of the operation and management overview of the big data intelligent warehouse in embodiment 1;

[0047] Fig. 4 Flowchart of processing abnormal candidate regions in embodiment 1. DETAILED DESCRIPTION

[0048] In order to make the above-mentioned purposes, features and advantages of the application more obvious and easy to understand, the specific embodiments of the application will be described in detail below with reference to the drawings in the specification.

[0049] In the following description, many specific details are set forth in order to provide a thorough understanding of the application, but the application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the application, therefore the application is not limited to the specific embodiments disclosed below.

[0050] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0051] Embodiment 1, Reference Figs. 1-4 As a first embodiment of the present application, the embodiment provides a big data intelligent warehouse operation management method, comprising the following steps:

[0052] S1, collect multi-source heterogeneous data and divide into grid cells, and simultaneously fuse time, space and environmental characteristics to generate a space-time data matrix.

[0053] Specifically, the following steps are included,

[0054] In the process of realizing from collecting multi-source heterogeneous data to generating a space-time data matrix, a comprehensive data collection architecture is first constructed. The core of this architecture is to integrate multiple sensors to ensure that key information such as geographic location data, environmental monitoring data and historical inventory data can be obtained. For example, by deploying GPS devices to collect location information, using temperature and humidity sensors, air quality detectors and other environmental monitoring devices to record environmental conditions in real time, and accessing ERP to extract historical inventory records, providing raw materials for subsequent data processing and analysis;

[0055] After the multi-source heterogeneous data collection is completed, preprocessing is performed, including data cleaning, format conversion and time synchronization steps. Since multi-source heterogeneous data may have noise, missing values or inconsistencies, appropriate data cleaning techniques must be used, such as using interpolation method to fill in missing values, applying statistical methods to identify and correct outliers, etc. At the same time, considering that data from different sources may have different formats, it needs to be converted into a standard format suitable for further analysis. In addition, in order to ensure consistency in the time dimension, the time stamps of all multi-source heterogeneous data need to be calibrated. This step is particularly important for subsequent fusion of time and space characteristics;

[0056] After completing the preprocessing of multi-source heterogeneous data, the next step is to divide the grid cells and generate a space-time data matrix based on this. First, the size and distribution of the grid cells are determined according to the actual layout, and each grid represents a specific area within the warehouse. Then, the preprocessed geographic location, environmental monitoring and historical inventory data are mapped to the corresponding grid cells. In this process, not only the physical space division needs to be considered, but also the time series, i.e. the state changes of each grid cell in different time periods. In order to more accurately capture these dynamic changes, the concept of time window can be introduced, and by setting different time intervals, the state evolution of the grid cells can be observed.

[0057] The grid cells and time, space and environmental features are fused to generate a space-time data matrix. First, the location information, environmental conditions and historical inventory are converted into numerical features, which not only reflect the current state of each grid cell, but also contain their historical records over time. For each grid cell, the geographical location, environmental conditions and inventory level are collected and recorded at a set time interval, for example, every hour or every day. Then, for each time point, a vector is constructed to represent the state of all grid cells at that time point. This vector contains the geographical location, environmental parameters and inventory of each grid cell. Subsequently, the vectors are arranged in chronological order to form a multi-dimensional array, i.e. a space-time data matrix, which can clearly show the state changes of each grid cell at different time periods and capture the dynamic evolution process of the environmental conditions and inventory levels in the entire warehouse.

[0058] S2, inputting the space-time data matrix into a space-time graph convolution network model to analyze the correlation between grid cells through multi-layer space-time convolution and obtaining a safety inventory prediction value of each region.

[0059] Specifically, the method comprises the following steps:

[0060] The space-time graph convolution network model comprises three layers. The first layer is a spatial correlation feature extraction layer, which captures the spatial correlation between adjacent grid cells through graph convolution operation. The second layer is a time series dynamic trend feature capturing layer, which identifies potential trends and periodic patterns. The third layer is an attention mechanism layer, which dynamically adjusts the weights of the spatial correlation features and the time series dynamic trend features. In order to improve the performance of the space-time graph convolution network model, a multi-scale time convolution strategy and a residual connection are used to prevent gradient vanishing problem. In the training process, an attention mechanism is adopted to ensure that the space-time graph convolution network model can adapt to various situations. Finally, the loss function is minimized through the back propagation algorithm combined with the Adam optimizer to continuously adjust the parameters of the space-time graph convolution network model to improve the prediction accuracy.

[0061] After the construction and training of the spatio-temporal graph convolution network model, the first stage extracts spatial correlation features. The spatio-temporal data matrix is input into the spatio-temporal graph convolution network model, and the spatial correlation between adjacent grid cells is captured through spatio-temporal convolution operation. The key to this stage is how to effectively use the graph convolution network to process the relationship between grid cells. In the specific operation, each grid cell is regarded as a node in the graph, and the physical proximity relationship between them is represented as an edge. These edges are usually defined by an adjacency matrix, which encodes the connection relationship between all grid cells. When performing graph convolution, the spatio-temporal graph convolution network model not only considers the information of a single grid cell, but also integrates the data of its neighbor nodes to identify the collaborative change law in the geographical adjacent area. For example, if the storage capacity in a certain grid cell suddenly increases, it may affect the demand pattern of its adjacent area. In this way, the spatio-temporal graph convolution network model can learn and predict the mutual influence between grid cells, laying the foundation for subsequent time series analysis.

[0062] The second stage captures the dynamic trend features on the time series. In this process, time convolution is used to analyze the trend of each grid cell over time. A one-dimensional convolution filter slides along the time axis to extract dynamic trend features in different time periods from the spatio-temporal data matrix. This method allows the spatio-temporal graph convolution network model to automatically learn which time intervals are most critical for predicting future inventory levels, and then identify potential trends and periodic patterns. For example, the periodic fluctuations in the demand for certain inventory items can be effectively captured through time convolution. To improve the generalization ability and accuracy of the spatio-temporal graph convolution network model, a multi-scale time convolution strategy can be used, which uses convolution kernels of different lengths to capture short-term fluctuations and long-term trends. In addition, techniques such as residual connection can be combined to avoid the gradient vanishing problem in deep network training.

[0063] The third stage uses attention mechanism to dynamically adjust the weights of spatial correlation features and time series dynamic trend features to obtain comprehensive feature vectors. The core idea of attention mechanism is to let the spatio-temporal graph convolution network model assign different weights to different features according to specific circumstances, so that the final output can better reflect the actual situation. For example, when obtaining the safety stock prediction value, if the weather conditions are exceptionally harsh in a certain period, the importance of environmental monitoring data may significantly increase. Conversely, in the case of stable demand, historical inventory data may be more critical. By introducing the attention mechanism, the spatio-temporal graph convolution network model can flexibly adapt to various situations and highlight the information that is most valuable for the current task. This step greatly improves the prediction accuracy.

[0064] The comprehensive feature vector is standardized or normalized, and a PCA algorithm is used for feature compression. The compressed comprehensive feature vector is input into the full connection layer. The full connection layer maps the high-dimensional comprehensive feature vector to a low-dimensional space, thereby realizing the final safety stock prediction. Each full connection layer is composed of multiple neurons, and each neuron is connected to all neurons of the previous layer. In this process, the spatio-temporal graph convolutional network model performs a weighted sum linear transformation on the input comprehensive feature vector, and then passes it through a ReLU activation function to enable the spatio-temporal graph convolutional network model to learn more complex patterns. To ensure the accuracy of the safety stock prediction value, the parameters need to be optimized. The backpropagation algorithm combined with the Adam optimizer is usually used to minimize the loss function. The loss function measures the difference between the safety stock prediction value and the actual inventory value. The optimization process aims to find a set of parameters that minimize this difference to obtain the final safety stock prediction value. The mathematical formula is as follows:

[0065]

[0066] wherein, represents the safety stock prediction value, V represents the weight matrix of the output layer, W represents the weight matrix of the full connection layer, P represents the principal component analysis transformation matrix, k represents the number of retained principal components, P k represents the transformation matrix composed of the first k principal components, T represents the transpose operation, x represents the comprehensive feature vector, μ represents the mean of the comprehensive feature vector, σ represents the standard deviation of the comprehensive feature vector, b represents the bias term value of the full connection layer, and c represents the bias term value of the output layer.

[0067] Through the above operations, the spatio-temporal graph convolutional network model can efficiently and accurately complete the safety stock prediction task, providing strong support for intelligent warehouse management.

[0068] S3, use IoT sensor network to obtain the current inventory of each grid unit, and compare it with the safety stock prediction value to obtain the abnormal candidate area.

[0069] Specifically, the following steps are included,

[0070] The warehouse data is collected by the IoT sensor network and integrated to generate the original inventory log. Due to possible signal interference or noise in the actual environment, the original warehouse data needs to be preprocessed to improve data quality. The specific operation includes using multiple reading filtering technology, i.e. reading the data at the same position multiple times and taking the average value to reduce random errors, and using signal compensation method to correct the signal attenuation problem caused by physical obstacles to ensure that the final obtained inventory data is as accurate as possible.

[0071] The preprocessed inventory data is matched and analyzed with the safety inventory prediction value obtained by the spatio-temporal graph convolution network model, a real-time streaming window calculation method is adopted, i.e. a fixed time window is set, for example, every 10 minutes or every hour, the inventory data of all grid units in the time window is collected, and the safety inventory prediction value obtained by the spatio-temporal graph convolution network model is compared, in order to further improve the matching accuracy, a dynamic time warping (DTW) algorithm is applied to process the nonlinear deformation problem in the time series data, the DTW algorithm can find the best alignment path between time series of different lengths, so that the distance between the sequences is minimized, so that whether the current inventory level exceeds the safety inventory prediction value is more accurately judged, once the inventory of a certain grid unit exceeds the preset safety inventory prediction value, the stream processing engine will mark the grid unit as a preliminary screening anomaly;

[0072] In order to reduce the false positive rate, the preliminary screening anomaly needs to be combined with the rule engine for secondary screening, the rule engine evaluates the authenticity and severity of each preliminary screening anomaly through the context-aware rule chain, which considers various factors such as historical inventory data, external environmental changes, special circumstances, etc. to determine which anomalies need to be focused on, in addition, the density clustering algorithm combined with the dynamic neighborhood radius method can be used to merge preliminary screening anomalies that are close in geographic location and have similar abnormal characteristics into abnormal candidate regions. This method not only effectively reduces the influence of isolated points, but also better discovers potential large-scale abnormal phenomena.

[0073] The identified abnormal candidate regions are cross-verified by multiple sensors to ensure the accuracy of the detection results, which means that multiple sensor data such as camera monitoring data and temperature and humidity sensor data need to be combined for comprehensive analysis, for example, if a certain area is marked as an abnormal candidate region, but the camera monitoring shows that there is no obvious logistics activity anomaly in the area, it may indicate that the anomaly is caused by data transmission error or other non-critical factors, rather than a real abnormal area. After cross verification, the exact coordinates of each abnormal candidate region and its corresponding abnormal type, such as excessive inventory, shortage, and substandard environmental conditions, are output, providing a basis for subsequent targeted measures;

[0074] Through the above operations, potential problems in warehouse operation can be effectively identified, and the overall management level and efficiency can be improved, not only the accuracy of anomaly detection is improved, but also the possibility of false positives is reduced, providing strong support for intelligent warehouses.

[0075] S4, a generative adversarial network model is constructed, the distribution characteristics of the abnormal candidate region in the generative adversarial network model space are analyzed, and the deviation degree is calculated, when the deviation value exceeds the statistical threshold, the abnormal region coordinates and deviation level are output.

[0076] Specifically, the steps include the following:

[0077] The Conditional Generative Adversarial Network (CGAN) model consists of a generator and a discriminator. The generator's task is to generate new samples that resemble the real data distribution based on input multi-source heterogeneous data. The discriminator is responsible for distinguishing the generated samples from the actual samples. To improve the accuracy and robustness of the CGAN model, contrastive learning supervision and dual-path adversarial training methods are used. Specifically, contrastive learning supervision utilizes the similarity between positive and negative sample pairs for training, enhancing the CGAN's understanding of normal patterns. Normal patterns are defined based on statistical regularities in historical inventory data. Dual-path adversarial training refers to... The optimization of both the generator and the discriminator is considered simultaneously, so that they can promote each other and improve performance together. For example, in contrastive learning supervision, the conditional generative adversarial network model learns which feature combinations are closer to the normal pattern, thereby helping the generator to generate more realistic samples. In dual-path adversarial training, the loss functions of the generator and the discriminator are optimized at the same time to ensure that the samples generated by the generator are as difficult as possible to be distinguished by the discriminator, while the discriminator can more accurately distinguish between the actual samples and the newly generated samples. In this way, the conditional generative adversarial network model can learn the data features in the normal pattern and provide a foundation for subsequent anomaly detection.

[0078] Based on a pre-trained conditional generative adversarial network (GAN) model, a spatiotemporal graph attention mechanism is used to map and match the spatiotemporal features of anomalous candidate regions with the data features in the normal pattern. The spatiotemporal graph attention mechanism is a technique specifically designed for processing datasets with complex spatiotemporal relationships. It considers not only spatial proximity relationships but also dynamic changes in the temporal dimension. Specifically, spatiotemporal features of anomalous candidate regions are extracted from the original multi-source heterogeneous data, including location information, environmental parameters, and historical inventory records. These spatiotemporal features are then input into the spatiotemporal graph attention mechanism and compared with the data features in the normal pattern. A triplet network method is used here to calculate the spatial deviation between the anomalous candidate region and the normal data distribution. The triplet network learns the relative distance between the actual sample and the newly generated sample by constructing three inputs (anchor point, positive sample, and negative sample), thus better capturing anomalous patterns. For example, if the inventory level of a certain grid cell is significantly higher or lower than the average level of its surrounding grid cells, that cell may be considered an anomalous candidate region. This mechanism can effectively identify regions that are significantly different from the normal pattern, laying the foundation for further analysis.

[0079] Based on the mapping result, the spatial deviation degree of each abnormal candidate region is calculated, which is a quantitative indicator reflecting the difference between the abnormal region and the data feature in the normal mode. For each abnormal candidate region, a specific deviation value is calculated based on the difference between its spatio-temporal features and the normal distribution. This value can be defined in various ways, such as Euclidean distance or cosine similarity. For example, the average Euclidean distance between the abnormal candidate region and the actual sample set can be calculated as the deviation value. The larger the deviation value, the more obvious the difference between the region and the normal mode. In order to determine which regions should be considered as real abnormalities, a statistical threshold needs to be set. This threshold can be determined according to the normal fluctuation range in historical data, usually using statistical methods such as standard deviation or percentile. For example, a deviation value exceeding twice the standard deviation of historical data is considered abnormal. When the deviation value of a region exceeds this statistical threshold, it is considered that the region has an abnormality. This process not only relies on accurate mathematical calculations, but also needs to consider various variables in the actual situation to ensure the accuracy and reliability of the final results.

[0080] After determining the abnormal regions that exceed the statistical threshold, the specific grid coordinates of these regions are located through inverse mapping, and their deviation levels are output according to the over-limit ratio. Through inverse mapping technology, the calculated deviation value can be remapped back to the original multi-source heterogeneous data space, so as to accurately locate the specific grid coordinates of the abnormal region. Inverse mapping relies on the spatio-temporal graph attention mechanism constructed earlier to ensure the accuracy and reliability of the mapping. For example, the deviation value can be converted back to the coordinate position in the original multi-source heterogeneous data space through backpropagation algorithm, so as to determine the specific position of the abnormal region.

[0081] Finally, the deviation level of each abnormal region is evaluated. The deviation level is usually divided according to the size of the deviation value, such as mild, moderate and severe. In addition, other factors such as influence range and duration can be combined for comprehensive evaluation. For example, if the deviation value of a certain abnormal region is large and the duration is long, it may be classified as a severe abnormality. The final output not only includes the accurate coordinates of the abnormal region, but also includes its corresponding deviation level, which facilitates subsequent targeted measures. The whole process aims to provide detailed information support to ensure that potential problems in warehouse operation can be discovered and handled in a timely manner.

[0082] Through the above operations, the efficiency and safety of inventory management are ensured, and the strong application potential of modern algorithm analysis technology in intelligent warehouse management is also reflected.

[0083] S5, according to the deviation level, start different response processes, and feed the deviation data to the feedback spatio-temporal graph convolution network model, dynamically adjust the parameters, and obtain the optimized spatio-temporal graph convolution network model.

[0084] Specifically includes the following operations,

[0085] Based on the deviation value and its corresponding deviation level, the corresponding response process is started. For slight abnormalities, the monitoring device is used to continuously track and record deviation data, and an alarm mechanism is triggered by setting a threshold to ensure that any subtle changes are recorded in a timely manner. For moderate abnormalities, data analysis tools such as Python's Pandas are introduced for in-depth analysis to identify potential patterns and causes, and simulation techniques are used to evaluate the effectiveness of different preventive measures to select the optimal solution. For severe abnormalities, real-time decision support mechanisms such as rule-based rapid response mechanisms are immediately activated, combined with historical deviation data for rapid diagnosis and automatic execution of pre-set correction measures such as adjusting inventory allocation or optimizing logistics routes to ensure rapid and effective problem solving and avoid greater losses.

[0086] In the process of starting the response process, the deviation data is input into the Apache Flink stream processing framework, and the time window function is used to align the deviation data. The specific operation includes using Flink's sliding window function to group deviation data according to fixed time intervals, for example, an hour can be set as a time window to ensure that all deviation data is compared and analyzed within the same time period, effectively synchronizing data from sensors and eliminating errors caused by time differences in data collection. Subsequently, the deviation data aligned by time window is input into the Prophet model for residual decomposition to extract periodic and trend features. Prophet is a tool specifically designed for time series prediction, which can effectively capture seasonal and long-term trends in deviation data. Through residual decomposition, the periodic component and trend component can be separated from the original deviation data, which not only helps to identify potential abnormal patterns, but also helps to understand normal fluctuations in deviation data, thereby improving the accuracy of overall analysis.

[0087] In order to feed the deviation data to the spatio-temporal graph convolution network model, a gated attention mechanism is adopted. The gated attention mechanism allows the spatio-temporal graph convolution network model to dynamically determine the importance of different features according to the current context, and weights the periodic and trend features in the deviation data through the gating mechanism, highlighting the information that is most critical to the current task. This not only improves the flexibility of the spatio-temporal graph convolution network model, but also enhances its ability to capture complex patterns. In specific operations, the gated attention mechanism weights the inventory level features, anomaly detection features, and periodic and trend features at each time point, so that the spatio-temporal graph convolution network model can pay more attention to areas with significant changes or anomalies, while ignoring relatively stable areas. Then the features weighted by the gated attention mechanism are fed back to the spatio-temporal graph convolution network model. These weighted features contain rich spatio-temporal information, which helps the spatio-temporal graph convolution network model better understand and predict the inventory changes in the warehouse, and effectively improves the adaptability of the spatio-temporal graph convolution network model, enabling it to maintain high prediction accuracy in a changing environment.

[0088] Based on the feedback of the deviation data, the parameters of the spatio-temporal graph convolution network model are dynamically adjusted. First, the Elastic Weight Consolidation (EWC) technique is used to constrain the parameter update direction. EWC is a technique to prevent catastrophic forgetting, which protects the parameters trained by historical multi-source heterogeneous data by constraining the parameter update direction. EWC can help the spatio-temporal graph convolution network model maintain its understanding of existing parameters while adapting to new changes when inputting new multi-source heterogeneous data. Then, the mini-batch incremental training method is used to gradually and dynamically adjust the spatio-temporal attention weights, so that they can more accurately capture the spatio-temporal correlation between grid cells. The mini-batch incremental training method can continuously improve the performance of the spatio-temporal graph convolution network model without significantly increasing computational resources. Each training only uses a small amount of new multi-source heterogeneous data, so that the spatio-temporal graph convolution network model can quickly adapt to environmental changes. In addition, during the gradient clipping process, the EWC importance mask and the momentum term are separated to ensure that they do not interfere with each other.

[0089] Finally, the AdamW optimizer is used to adjust the convolution kernel parameters in the spatio-temporal graph convolution network model. AdamW is a variant of the Adam optimizer that effectively prevents overfitting by improving weight decay, generating an optimized spatio-temporal graph convolution network model.

[0090] Through the above operations, the accuracy and efficiency of anomaly detection are improved, effectively reducing operating costs and improving overall business efficiency.

[0091] S6, based on the optimized space-time graph convolution network model, calculating the optimal inventory of each region, and generating a replenishment instruction, and generating an operation management report based on the replenishment instruction.

[0092] Specifically comprising the following steps,

[0093] Based on the optimized space-time graph convolution network model, the data of inventory, order and environment are fused by using the time window sliding alignment mechanism. In the specific operation, the data collected from different sources has different collection frequency and format. Through the time window sliding mechanism, it is ensured that the multi-source heterogeneous data is processed and compared in the same time period. This process uses Apache Flink and other stream processing frameworks to realize data synchronization, which guarantees the consistency and accuracy of multi-source heterogeneous data. At the same time, the safety inventory prediction value is input into the dynamic programming algorithm to calculate the optimal inventory value under the space-time constraint. The dynamic programming algorithm can effectively solve the multi-stage decision problem and help to determine the best inventory level of each grid unit in a specific time period, so as to minimize the cost;

[0094] Next, the replenishment path is generated and the AGV is dispatched to execute the task. In order to find the optimal replenishment path, A algorithm and ant colony algorithm are integrated. A algorithm can quickly find the shortest path, while ant colony algorithm can find the global optimal solution by simulating the foraging behavior of ants. The combination of the two can balance the needs of local optimization and global optimization, and generate an efficient replenishment path. The generated replenishment path dynamically schedules AGV to execute tasks through priority queue. The priority queue sorts tasks according to factors such as replenishment time limit and goods category, to ensure that urgent tasks are given priority. For example, for items that need to be replenished immediately, AGV can be quickly arranged for scheduling through high priority queue. In addition, the replenishment time limit and goods category are injected into the instruction template to generate standardized replenishment instructions, which clearly specify the specific content, execution time and priority of replenishment, facilitating AGV to efficiently execute tasks;

[0095] Finally, based on the replenishment instruction, a detailed operation management report is generated. The historical replenishment records and inventory fluctuation curves are aggregated through the time series database to comprehensively understand the past replenishment situation and its effect. The time series database is specially used to store and query data with timestamps, which is very suitable for processing such time series data. Using the Grafana tool, information can be extracted from the aggregated historical replenishment data to generate multi-dimensional analysis dashboards. Not only does it show the basic information of inventory level and replenishment frequency, but also in-depth analysis of the performance of different regions, time periods or item categories to assist decision-making. For example, by viewing the inventory fluctuation curve of a certain region in the past month, potential inventory shortage or surplus problems can be identified. Based on the above analysis results, a detailed operation management report is finally generated, covering inventory status, replenishment efficiency, AGV running status and other aspects of each region, providing comprehensive operation insights. This report can not only serve as an important reference for internal management, but also share information with other departments or partners to promote overall business collaboration.

[0096] The embodiment also provides an operation management system of a big data intelligent warehouse, comprising: an acquisition module, a prediction module, an anomaly positioning module, a deviation calculation module, a model optimization module and a report generation module.

[0097] The acquisition module is used to acquire multi-source heterogeneous data and divide them into grid units, and simultaneously fuse time, space and environmental characteristics to generate a spatio-temporal data matrix.

[0098] The prediction module is used to input the spatio-temporal data matrix into a spatio-temporal graph convolution network model, analyze the correlation between grid units through multiple layers of spatio-temporal convolution, and obtain a safety inventory prediction value of each region.

[0099] The anomaly positioning module is used to obtain the current inventory of each grid unit by using an IoT sensor network, compare it with the safety inventory prediction value, and obtain an abnormal candidate region.

[0100] The deviation calculation module is used to construct a generative adversarial network model, analyze the distribution characteristics of the abnormal candidate region in the generative adversarial network model space, and calculate the deviation degree. When the deviation value exceeds a statistical threshold, the coordinates of the abnormal region and the deviation level are output.

[0101] The model optimization module is used to start different response processes according to the deviation level, feed the deviation data in the deviation level to the feedback spatio-temporal graph convolution network model, dynamically adjust the parameters, and obtain an optimized spatio-temporal graph convolution network model.

[0102] The report generation module is used to calculate the optimal inventory of each region based on the optimized spatio-temporal graph convolution network model, generate a replenishment instruction, and generate an operation management report based on the replenishment instruction.

[0103] The embodiment also provides a computer device suitable for the operation management method of the big data intelligent warehouse, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the operation management method of the big data intelligent warehouse proposed in the above embodiment.

[0104] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, an operator network, NFC (Near Field Communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0105] The embodiment also provides a storage medium having a computer program stored thereon, the program being executed by a processor to realize the operation management method of the big data intelligent warehouse proposed in the above embodiment. The storage medium can be realized by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0106] To sum up, by integrating multi-source heterogeneous data and using a spatio-temporal graph convolution network model to analyze the correlation between the data, the application can more accurately obtain the safety stock prediction value of each region. This not only improves the accuracy of inventory management decisions, but also enables the warehouse to meet customer demand while reducing the occurrence of excessive inventory or out-of-stock phenomena. By using a generative adversarial network model to deeply analyze the distribution characteristics of abnormal candidate regions and calculating the deviation degree to accurately locate the abnormal regions and their deviation levels, the false positive rate is greatly reduced, reducing the dependence on manual review, saving human resources, and speeding up the problem response speed.

[0107] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application, which should be covered by the scope of the claims of the present application.

Claims

1. A method for the operation and management of a big data intelligent warehouse, characterized in that: include, Collect heterogeneous data from multiple sources and divide it into grid cells, while fusing temporal, spatial and environmental features to generate a spatiotemporal data matrix; The spatiotemporal data matrix is ​​input into the spatiotemporal graph convolutional network model, and the correlation between grid cells is analyzed through multi-layer spatiotemporal convolution to obtain the safety stock prediction value for each region. The steps for obtaining the safety stock forecast value for each region are as follows: The spatiotemporal data matrix is ​​input into the spatiotemporal graph convolutional network model. The spatial correlation features between adjacent grids are extracted through the first layer of spatiotemporal convolution to analyze the collaborative change patterns of geographically adjacent areas. By using a second-layer temporal convolution to capture the dynamic trend characteristics of each grid cell in the time series, long-term and short-term fluctuation patterns can be identified. An attention mechanism is used to dynamically adjust the weights of spatial correlation features and dynamic trend features, and outputs the future safety stock prediction value for each grid cell. The current inventory level of each grid cell is obtained by using an IoT sensor network and compared with the safety stock prediction value to identify anomaly candidate areas; Construct a generative adversarial network model, analyze the distribution characteristics of abnormal candidate regions in the generative adversarial network model space, and calculate the degree of deviation. When the deviation value exceeds the statistical threshold, output the coordinates of the abnormal region and the deviation level. Different response processes are initiated based on the deviation level, and the deviation data in the deviation level is fed to the feedback spatiotemporal graph convolutional network model. The parameters are dynamically adjusted to obtain the optimized spatiotemporal graph convolutional network model. Based on the optimized spatiotemporal graph convolutional network model, the optimal inventory level for each region is calculated, and replenishment instructions are generated. Based on the replenishment instructions, an operations management report is generated.

2. The operation and management method of the big data intelligent warehouse according to claim 1, characterized in that: The multi-source heterogeneous data includes geographic location data, environmental monitoring data, and historical inventory data.

3. The operation and management method of the big data intelligent warehouse according to claim 1, characterized in that: The steps for obtaining the abnormal candidate region are as follows: The system utilizes IoT sensor networks to collect warehouse data in real time, generates raw inventory logs, and eliminates interference through multi-read filtering and signal compensation methods to obtain inventory data. The inventory data is dynamically matched with the safety stock prediction value obtained by the spatiotemporal graph convolutional network model through real-time streaming window calculation and dynamic time warping algorithm, and the grid cells exceeding the safety stock threshold are marked as initial screening anomalies using the stream processing engine. The initial anomaly screening is combined with a rule engine to eliminate false alarms through context-aware rule chains. Density clustering algorithm combined with dynamic neighborhood radius is used to merge the initial anomalies into anomaly candidate regions. After cross-validation by multi-source sensors, the coordinates and anomaly type are output.

4. The operation and management method of the big data intelligent warehouse according to claim 1, characterized in that: The output of abnormal region coordinates and deviation levels includes the following steps. Based on the generator and discriminator, a conditional generative adversarial network model is constructed through contrastive learning supervision and dual-path adversarial training. Based on the generative adversarial network model, the spatiotemporal features of abnormal candidate regions are mapped and matched with the features of normal distribution through the spatiotemporal graph attention mechanism, and the degree of spatial deviation is calculated based on the triplet network. Deviation values ​​are generated based on the degree of spatial deviation. When the deviation value exceeds the statistical threshold, the grid coordinates are located by inverse mapping, and the coordinates of the abnormal area and the deviation level are output according to the proportion of exceeding the limit.

5. The operation and management method of the big data intelligent warehouse according to claim 4, characterized in that: The step of feeding back the deviation data to the spatiotemporal graph convolutional network model, dynamically adjusting the parameters, and obtaining the optimized spatiotemporal graph convolutional network model specifically includes the following steps. After aligning the bias data based on the Flink time window, periodic and trend features are extracted by Prophet residual decomposition, and then fed back to the spatiotemporal graph convolutional network model using a gated attention mechanism. The spatiotemporal graph convolutional network model is based on bias data. It updates the direction of the parameters by EWC constraint, dynamically adjusts the spatiotemporal attention weights by combining mini-batch incremental training, separates the EWC importance mask and momentum term in gradient clipping, and uses the AdamW optimizer to update the convolution kernel parameters, generating an optimized spatiotemporal graph convolutional network model.

6. The operation and management method of the big data intelligent warehouse according to claim 5, characterized in that: The optimized spatiotemporal graph convolutional network model calculates the optimal inventory level for each region and generates replenishment instructions. Based on these instructions, an operations management report is generated. The specific steps include the following: Based on the optimized spatiotemporal graph convolutional network model, the data of inventory, order and environment are fused through the time window sliding alignment mechanism, and combined with the safety stock prediction value, the data are input into the dynamic programming algorithm to calculate the optimal inventory value under spatiotemporal constraints. By integrating the A algorithm and the ant colony algorithm to generate replenishment paths, and dynamically scheduling AGVs to execute tasks through a priority queue, replenishment timeliness requirements and goods categories are injected into the instruction template to generate replenishment instructions. Based on replenishment orders, historical replenishment records and inventory fluctuation curves are aggregated through a time-series database, and Grafana tools are used to automatically generate multi-dimensional analysis dashboards and output operation management reports.

7. An operation and management system for a big data intelligent warehouse, based on the operation and management method for a big data intelligent warehouse according to any one of claims 1 to 6, characterized in that: It includes a data acquisition module, a prediction module, an anomaly localization module, a deviation calculation module, a model optimization module, and a report generation module; The acquisition module is used to collect multi-source heterogeneous data and divide it into grid cells, while fusing temporal, spatial and environmental features to generate a spatiotemporal data matrix. The prediction module is used to input the spatiotemporal data matrix into the spatiotemporal graph convolutional network model, analyze the correlation between grid cells through multi-layer spatiotemporal convolution, and obtain the predicted safety stock value for each region. The anomaly location module is used to obtain the current inventory of each grid cell using the IoT sensor network and compare it with the safety stock prediction value to obtain anomaly candidate areas; The deviation calculation module is used to construct a generative adversarial network model, analyze the distribution characteristics of abnormal candidate regions in the generative adversarial network model space, and calculate the degree of deviation. When the deviation value exceeds the statistical threshold, the coordinates of the abnormal region and the deviation level are output. The model optimization module is used to initiate different response processes based on the deviation level, and feed the deviation data in the deviation level to the feedback spatiotemporal graph convolutional network model, dynamically adjust the parameters, and obtain the optimized spatiotemporal graph convolutional network model. The report generation module is used to calculate the optimal inventory level for each region based on the optimized spatiotemporal graph convolutional network model, generate replenishment instructions, and generate an operations management report based on the replenishment instructions.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the operation and management method of the big data intelligent warehouse according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the operation and management method of the big data intelligent warehouse according to any one of claims 1 to 6.

Images