Warehouse data analysis system and methods for intelligent monitoring of agricultural product warehouses

Abstract

This invention discloses a warehouse data analysis system and method for intelligent monitoring of agricultural product warehouses, relating to the field of agricultural product monitoring technology. The analysis method includes the following steps: based on the warehouse layout, several nodes of the warehouse are obtained, and a warehouse network model is generated; warehouse data of each node in the warehouse network model is periodically collected; correlation analysis is performed on the warehouse data between each node to obtain the causal relationship between different nodes, and several influence paths are generated in the warehouse network model; abnormal nodes in any unit period are identified, several traceability paths are generated, and the optimal traceability path for each abnormal node is obtained; abnormal nodes in the warehouse network model are identified in real time, and the expected abnormal situation of the warehouse is analyzed; breaking through the limitations of traditional threshold alarms, it realizes pre-event prediction and in-event traceability, greatly improving the initiative and accuracy of warehouse management.

Description

Technical Field

[0001] This invention relates to the field of agricultural product monitoring technology, specifically to a storage data analysis system and method for intelligent monitoring of agricultural product warehouses. Background Technology

[0002] In agricultural product storage management, maintaining stable environmental parameters is key to ensuring quality and reducing losses; existing warehouse monitoring systems typically rely on deploying various sensors for data collection and employ threshold-based alarm mechanisms.

[0003] However, such traditional methods have significant drawbacks: First, their alarms are severely delayed, and by the time environmental parameters exceed the threshold, the deterioration of agricultural products is often irreversible. Second, a warehouse is a complex system with tightly coupled links, and a local anomaly can propagate over time and space, eventually triggering anomalies at a distant location. Existing threshold alarm systems can only indicate "where the anomaly occurred," but cannot answer the two key questions of "what is the root cause of the anomaly" and "how did it propagate to this location." This leads to difficulties for maintenance personnel in troubleshooting, low efficiency, and unnecessary economic losses. Summary of the Invention

[0004] The purpose of this invention is to provide a storage data analysis system and method for intelligent monitoring of agricultural product warehouses, in order to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a warehouse data analysis method for intelligent monitoring of agricultural product warehouses, the analysis method comprising the following steps:

[0006] Step S100: Based on the warehouse layout, obtain several nodes of the warehouse and generate a warehouse network model; several monitoring devices are pre-installed to periodically collect the storage data of each node in the warehouse network model.

[0007] Step S200: Extract the warehouse data of each node in any unit period, perform correlation analysis on the warehouse data between each node to obtain the causal relationship between different nodes; generate several influence paths in the warehouse network model based on the causal relationship between different nodes.

[0008] Step S300: Identify several abnormal nodes in any unit period based on warehouse data; generate several traceability paths based on the distribution of all abnormal nodes in the warehouse network model; evaluate the suitability of each traceability path to obtain the optimal traceability path for each abnormal node.

[0009] Step S400: Identify abnormal nodes in the warehouse network model in real time; identify the tracing path based on the distribution of abnormal nodes, and analyze the expected abnormal situation of the warehouse according to the causal relationship of the other nodes in the tracing path.

[0010] Furthermore, step S100 includes the following steps:

[0011] Step S101: Pre-build a warehouse management system. The warehouse management system stores a complete set of warehouse layout information, including warehouse area division information, control equipment information, and fixed structure information. Set any layout information as a node in the warehouse, generate several nodes of the warehouse, and summarize them to obtain the warehouse network model. Among them, the warehouse area division information includes functional areas such as receiving area, storage area, picking area, and aisle; the control equipment includes equipment such as air conditioners, refrigeration units, humidifiers, dehumidifiers, and ventilation systems; and the fixed structure information includes fixed structures such as doors and windows in the warehouse.

[0012] Step S102: Several monitoring devices are installed in the warehouse to collect storage data at any node in the warehouse. For each node, a corresponding monitoring dimension is preset. The monitoring data collected by the monitoring devices at each unit cycle is summarized to obtain the storage data set of any node in any unit cycle. The monitoring dimension for doors and windows is whether they are open, the monitoring dimension for air conditioners is temperature, and the monitoring dimension for functional areas is the degree of abnormality of crops, etc.

[0013] Furthermore, step S200 includes the following steps:

[0014] Step S201: Arbitrarily select a unit period, and arbitrarily select a node from the selected unit period to obtain the warehouse data set of the selected node; sort the monitoring data in the warehouse data set according to the order of the collection time points, and arbitrarily select the a-th monitoring data D. a and the (a+1)th monitoring data D a+1 The trend ΔD between the two monitoring data points is obtained. (a,a+1) =D a+1 -D a If ΔD (a,a+1) If >0, then for monitoring data D a+1 If ΔD is set as the first feature data, then... (a,a+1) <0, then for monitoring data D a+1 If ΔD is set as the second feature data, then... (a,a+1) =0, then the monitoring data D a+1 With monitoring data D aThe characteristic data types remain the same; the first characteristic data indicates that the monitored data is in a state of growth, and the second characteristic data indicates that it is in a state of decrease.

[0015] Step S202: Obtain the characteristic data type of the last monitoring data, and count the number of data in the warehouse data set that have the same characteristic data type as the last monitoring data as m. sim Set the total amount of data in the warehouse dataset to M. total The calculated data volume ratio is γ=m sim / M total Preset a data volume percentage threshold γ th If γ≥γ th If the last monitored data is selected, the feature data type of the selected node will be set to the feature data type of the selected node; otherwise, the selected node will be set to the fluctuation data type. The fluctuation data type indicates that the dimension data is in a fluctuating state, such as the continuous opening and closing of doors and windows.

[0016] Step S203: Obtain the monitoring dimensions and data types of each node, establish a logical database in advance, and set corresponding causal relationships between monitoring dimensions of different data types in the logical database; obtain the monitoring devices corresponding to each node, extract the monitoring range of each monitoring device, if the monitoring devices corresponding to two nodes are the same, then set the two nodes as associated nodes; if the monitoring devices corresponding to two nodes are different and there are no other monitoring ranges between the monitoring ranges of the two monitoring devices, then set the two nodes as associated nodes as well.

[0017] Step S204: Randomly select two nodes. If the two selected nodes are related nodes, obtain the monitoring dimensions and data types of the two nodes respectively, and call the logical database to perform causal relationship matching on the two selected nodes. If there is a causal relationship between the two selected nodes in the logical database, set the two selected nodes as a causal node group, and set the two selected nodes as the cause node and the effect node respectively.

[0018] Step S205: Extract causal node groups between different nodes in the selected unit period of the warehouse network model. Randomly select two causal node groups. If the cause node in one causal node group is the effect node in the other causal node group, then connect the two causal node groups through the same node to obtain the influence path of the two causal node groups. Connect the influence paths between different causal node groups to obtain several complete influence paths.

[0019] Furthermore, step S300 includes the following steps:

[0020] Step S301: Preset corresponding anomaly identification thresholds for the monitoring dimensions of each node. Randomly select a unit period, and randomly select a node from that unit period. Extract the warehouse data set of the selected node. If any monitoring data in the warehouse data set exceeds the anomaly identification threshold, then that monitoring data is set as an anomaly. Count the number of anomaly data in the warehouse data set as m. yc Let the total amount of data in the warehouse data set be M1. Calculate the percentage of abnormal data in the warehouse data set η = m. yc / M1; Preset an abnormality percentage threshold η th If η≥η th If so, the selected node is set as an abnormal node;

[0021] Step S302: Obtain all abnormal nodes in the warehouse network model within the selected unit period, acquire several influence paths of the warehouse network model, and arbitrarily select one influence path. Compare each abnormal node with each node in the selected influence path to obtain the number of abnormal nodes in the selected influence path as p. Set the total number of abnormal nodes in the selected unit period as Q, and calculate the fit degree C = p / Q of the selected influence path; preset a fit degree threshold C. th If C≥C th Then the selected influence path will be set to the expected source tracing path of the unit period;

[0022] Step S303: Obtain several expected tracing paths for the selected unit period. If the fit degree C=1 of an expected tracing path exists, set that expected tracing path as the optimal tracing path. Otherwise, select the expected tracing path with the highest fit degree and set it as the target tracing path. Remove the abnormal nodes in the target tracing path to obtain a set of remaining abnormal nodes. Regenerate a new expected tracing path from the set of remaining abnormal nodes, and again select the expected tracing path with the highest fit degree as the target tracing path, until all abnormal nodes are included or no new expected tracing path is generated. Summarize all target tracing paths to generate the optimal tracing path for the selected unit period.

[0023] Furthermore, step S400 includes the following steps:

[0024] Step S401: Monitor and collect data from each node in the warehouse network model in real time to obtain the real-time warehouse data set of each node. Compare the real-time warehouse data set with the corresponding anomaly identification threshold to obtain several real-time abnormal nodes.

[0025] Step S402: Obtain the distribution of the several real-time abnormal nodes in the warehouse network model, generate several expected traceability paths, and obtain the adaptation degree of each expected traceability path; accumulate the adaptation degree of each expected traceability path to obtain a comprehensive adaptation value; preset an abnormal adaptation threshold, and if the comprehensive adaptation value exceeds the abnormal adaptation threshold, send an abnormal reminder to the warehouse storage data.

[0026] To better implement the above methods, a warehouse data analysis system is also proposed. The analysis system includes a warehouse data analysis module, a causal anomaly analysis module, a traceability path analysis module, and a real-time anomaly prediction module.

[0027] The warehouse data analysis module is used to obtain several nodes of the warehouse based on the warehouse layout and generate a warehouse network model; several monitoring devices are pre-installed to periodically collect warehouse data of each node in the warehouse network model;

[0028] The causal anomaly analysis module is used to extract warehouse data from each node in any unit period, perform correlation analysis on the warehouse data between each node, obtain the causal relationship between different nodes, and generate several influence paths in the warehouse network model based on the causal relationship between different nodes.

[0029] The traceability path analysis module is used to identify several abnormal nodes in any unit period based on warehouse data, generate several traceability paths based on the distribution of all abnormal nodes in the warehouse network model, evaluate the adaptability of each traceability path, and obtain the optimal traceability path for each abnormal node.

[0030] The real-time anomaly prediction module is used to identify abnormal nodes in the warehouse network model in real time; it identifies the tracing path based on the distribution of abnormal nodes, and analyzes the expected anomalies in the warehouse based on the causal relationships of other nodes in the tracing path.

[0031] Furthermore, the warehouse data analysis module includes a warehouse network modeling unit and a warehouse data extraction unit;

[0032] The warehouse network modeling unit is used to obtain several nodes of the warehouse based on the warehouse layout and generate a warehouse network model; the warehouse data extraction unit is used to periodically collect warehouse data of each node in the warehouse network model, which is pre-installed with several monitoring devices.

[0033] Furthermore, the causal anomaly analysis module includes a causal node association unit and an influence path generation unit;

[0034] The causal node association unit is used to extract warehouse data from each node in any unit period, perform correlation analysis on the warehouse data between each node, and obtain the causal relationship between different nodes; the influence path generation unit is used to generate several influence paths in the warehouse network model based on the causal relationship between different nodes.

[0035] Furthermore, the source tracing path analysis module includes a source tracing path identification unit and a path probability evaluation unit;

[0036] The traceability path identification unit is used to identify several abnormal nodes in any unit period based on warehouse data, and generate several traceability paths based on the distribution of all abnormal nodes in the warehouse network model; the path probability evaluation unit is used to evaluate the suitability of each traceability path and obtain the optimal traceability path for each abnormal node.

[0037] Furthermore, the real-time anomaly prediction module includes an anomaly node identification unit and an anomaly storage judgment unit;

[0038] The abnormal node identification unit is used to identify abnormal nodes in the warehouse network model in real time; the abnormal storage judgment unit is used to identify the traceability path based on the distribution of abnormal nodes, and analyze the expected abnormal situation of the warehouse based on the causal relationship of other nodes in the traceability path.

[0039] Compared with the prior art, the beneficial effects of the present invention are:

[0040] 1. This invention, by constructing a warehouse network model and realizing the correlation analysis of multi-node data, can accurately identify the generation and propagation path of environmental anomalies, significantly improving the ability to trace the source of complex anomalies in the agricultural product storage environment; it realizes the leap from "post-event alarm" to "pre-event prediction and in-event traceability", greatly improving the initiative and accuracy of warehouse management.

[0041] 2. Based on the causal relationship-based influence path generation mechanism, this invention can dynamically construct the relationship network between nodes, thereby quickly locating the root node and its propagation path when an anomaly occurs; it improves the system's adaptability in complex warehousing environments, provides clear decision support for operation and maintenance personnel, and reduces the cost of manual troubleshooting;

[0042] 3. Through adaptation assessment and multi-path fusion analysis, this invention can optimize the selection and dynamic update of anomaly tracing paths, thereby maintaining a high diagnostic accuracy rate even in scenarios with multiple concurrent anomalies, and providing continuous and stable intelligent monitoring support for agricultural product storage. Attached Figure Description

[0043] Figure 1A schematic diagram illustrating the steps of a warehouse data analysis method for intelligent monitoring of agricultural product warehouses;

[0044] Figure 2 This is a schematic diagram of a warehouse data analysis system used for intelligent monitoring of agricultural product warehouses. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] Example: Figures 1 to 2 As shown, this invention provides a warehouse data analysis method for intelligent monitoring of agricultural product warehouses. The analysis method includes the following steps:

[0047] Step S100: Based on the warehouse layout, obtain several nodes of the warehouse and generate a warehouse network model; several monitoring devices are pre-installed to periodically collect the storage data of each node in the warehouse network model.

[0048] Step S100 includes the following steps:

[0049] Step S101: Pre-build a warehouse management system, which stores a complete set of warehouse layout information, including warehouse area division information, control equipment information, and fixed structure information; set any layout information corresponding to a structure in the warehouse as a node of the warehouse, generate several nodes of the warehouse, and summarize them to obtain the warehouse network model.

[0050] Example 1: Set up an agricultural product warehouse divided into three storage areas: a cold storage area, a normal temperature area, and a ventilated area. Each storage area is equipped with temperature and humidity sensors and shelving structures. Define each sensor and each row of shelves as a node, for example: Node 1: Temperature and humidity sensor in the cold storage area, Node 2: Shelf A in the cold storage area, Node 3: CO2 sensor in the ventilated area, etc., to generate a warehouse network model containing a total of several nodes.

[0051] Step S102: Several monitoring devices are installed in the warehouse to collect storage data at any node in the warehouse. For each node, a corresponding monitoring dimension is preset. The monitoring data collected by the monitoring devices at each unit cycle is summarized to obtain the storage data set of any node in any unit cycle.

[0052] Step S200: Extract the warehouse data of each node in any unit period, perform correlation analysis on the warehouse data between each node to obtain the causal relationship between different nodes; generate several influence paths in the warehouse network model based on the causal relationship between different nodes.

[0053] Step S200 includes the following steps:

[0054] Step S201: Arbitrarily select a unit period, and arbitrarily select a node from the selected unit period to obtain the warehouse data set of the selected node; sort the monitoring data in the warehouse data set according to the order of the collection time points, and arbitrarily select the a-th monitoring data D. a and the (a+1)th monitoring data D a+1 The trend ΔD between the two monitoring data points is obtained. (a,a+1) =D a+1 -D a If ΔD (a,a+1) If >0, then for monitoring data D a+1 If ΔD is set as the first feature data, then... (a,a+1) <0, then for monitoring data D a+1 If ΔD is set as the second feature data, then... (a,a+1) =0, then the monitoring data D a+1 With monitoring data D a The characteristic data types remain the same;

[0055] Step S202: Obtain the characteristic data type of the last monitoring data, and count the number of data in the warehouse data set that have the same characteristic data type as the last monitoring data as m. sim Set the total amount of data in the warehouse dataset to M. total The calculated data volume ratio is γ=m sim / M total Preset a data volume percentage threshold γ th If γ≥γ th If so, the feature data type of the last monitored data is set to the feature data type of the selected node; otherwise, the selected node is set to the fluctuation data type.

[0056] Step S203: Obtain the monitoring dimensions and data types of each node, establish a logical database in advance, and set corresponding causal relationships between monitoring dimensions of different data types in the logical database; obtain the monitoring devices corresponding to each node, extract the monitoring range of each monitoring device, if the monitoring devices corresponding to two nodes are the same, then set the two nodes as associated nodes; if the monitoring devices corresponding to two nodes are different and there are no other monitoring ranges between the monitoring ranges of the two monitoring devices, then set the two nodes as associated nodes as well.

[0057] Step S204: Randomly select two nodes. If the two selected nodes are related nodes, obtain the monitoring dimensions and data types of the two nodes respectively, and call the logical database to perform causal relationship matching on the two selected nodes. If there is a causal relationship between the two selected nodes in the logical database, set the two selected nodes as a causal node group, and set the two selected nodes as the cause node and the effect node respectively.

[0058] Step S205: Extract causal node groups between different nodes in the selected unit period of the warehouse network model. Randomly select two causal node groups. If the cause node in one causal node group is the effect node in the other causal node group, then connect the two causal node groups through the same node to obtain the influence path of the two causal node groups. Connect the influence paths between different causal node groups to obtain several complete influence paths.

[0059] Example 2: Select node 1, the temperature and humidity sensor in the cold storage area. In the temperature data sequence of period T1, if most data shows an upward trend, then the node is determined to be an increasing node in this period. The logic database is preset with the following: if node A: temperature decreases, then node B: CO2 concentration increases; if node C: refrigeration power increases, then node A: temperature decreases, thus obtaining an influence path: refrigeration, temperature, CO2 concentration.

[0060] Step S300: Identify several abnormal nodes in any unit period based on warehouse data; generate several traceability paths based on the distribution of all abnormal nodes in the warehouse network model; evaluate the suitability of each traceability path to obtain the optimal traceability path for each abnormal node.

[0061] Step S300 includes the following steps:

[0062] Step S301: Preset corresponding anomaly identification thresholds for the monitoring dimensions of each node. Randomly select a unit period, and randomly select a node from that unit period. Extract the warehouse data set of the selected node. If any monitoring data in the warehouse data set exceeds the anomaly identification threshold, then that monitoring data is set as an anomaly. Count the number of anomaly data in the warehouse data set as m. yc Let the total amount of data in the warehouse data set be M1. Calculate the percentage of abnormal data in the warehouse data set η = m. yc / M1; Preset an abnormality percentage threshold η th If η≥η th If so, the selected node is set as an abnormal node;

[0063] Step S302: Obtain all abnormal nodes in the warehouse network model within the selected unit period, acquire several influence paths of the warehouse network model, and arbitrarily select one influence path. Compare each abnormal node with each node in the selected influence path to obtain the number of abnormal nodes in the selected influence path as p. Set the total number of abnormal nodes in the selected unit period as Q, and calculate the fit degree C = p / Q of the selected influence path; preset a fit degree threshold C. th If C≥C th Then the selected influence path will be set to the expected source tracing path of the unit period;

[0064] Step S303: Obtain several expected tracing paths for the selected unit period. If the fit degree C=1 of an expected tracing path exists, set that expected tracing path as the optimal tracing path. Otherwise, select the expected tracing path with the highest fit degree and set it as the target tracing path. Remove the abnormal nodes in the target tracing path to obtain a set of remaining abnormal nodes. Regenerate a new expected tracing path from the set of remaining abnormal nodes, and again select the expected tracing path with the highest fit degree as the target tracing path, until all abnormal nodes are included or no new expected tracing path is generated. Summarize all target tracing paths to generate the optimal tracing path for the selected unit period.

[0065] Step S400: Real-time identification of abnormal nodes in the warehouse network model; identification of the tracing path based on the distribution of abnormal nodes; and analysis of the expected abnormal situation of the warehouse based on the causal relationship of the remaining nodes in the tracing path.

[0066] Step S400 includes the following steps:

[0067] Step S401: Monitor and collect data from each node in the warehouse network model in real time to obtain the real-time warehouse data set of each node. Compare the real-time warehouse data set with the corresponding anomaly identification threshold to obtain several real-time abnormal nodes.

[0068] Step S402: Obtain the distribution of the several real-time abnormal nodes in the warehouse network model, generate several expected traceability paths, and obtain the adaptation degree of each expected traceability path; accumulate the adaptation degree of each expected traceability path to obtain a comprehensive adaptation value; preset an abnormal adaptation threshold, and if the comprehensive adaptation value exceeds the abnormal adaptation threshold, send an abnormal reminder to the warehouse storage data.

[0069] A warehouse data analysis system, comprising a warehouse data analysis module, a causal anomaly analysis module, a traceability path analysis module, and a real-time anomaly prediction module;

[0070] The warehouse data analysis module is used to obtain several nodes of the warehouse based on the warehouse layout and generate a warehouse network model; several monitoring devices are pre-installed to periodically collect warehouse data of each node in the warehouse network model;

[0071] The causal anomaly analysis module is used to extract warehouse data from each node in any unit period, perform correlation analysis on the warehouse data between each node, obtain the causal relationship between different nodes, and generate several influence paths in the warehouse network model based on the causal relationship between different nodes.

[0072] The traceability path analysis module is used to identify several abnormal nodes in any unit period based on warehouse data, generate several traceability paths based on the distribution of all abnormal nodes in the warehouse network model, evaluate the adaptability of each traceability path, and obtain the optimal traceability path for each abnormal node.

[0073] The real-time anomaly prediction module is used to identify abnormal nodes in the warehouse network model in real time; it identifies the tracing path based on the distribution of abnormal nodes, and analyzes the expected anomalies in the warehouse based on the causal relationships of other nodes in the tracing path.

[0074] The warehouse data analysis module includes a warehouse network modeling unit and a warehouse data extraction unit.

[0075] The warehouse network modeling unit is used to obtain several nodes of the warehouse based on the warehouse layout and generate a warehouse network model; the warehouse data extraction unit is used to periodically collect warehouse data of each node in the warehouse network model, which is pre-installed with several monitoring devices.

[0076] The causal anomaly analysis module includes a causal node association unit and an influence path generation unit.

[0077] The causal node association unit is used to extract warehouse data from each node in any unit period, perform correlation analysis on the warehouse data between each node, and obtain the causal relationship between different nodes; the influence path generation unit is used to generate several influence paths in the warehouse network model based on the causal relationship between different nodes.

[0078] The source tracing path analysis module includes a source tracing path identification unit and a path probability evaluation unit.

[0079] The traceability path identification unit is used to identify several abnormal nodes in any unit period based on warehouse data, and generate several traceability paths based on the distribution of all abnormal nodes in the warehouse network model; the path probability evaluation unit is used to evaluate the suitability of each traceability path and obtain the optimal traceability path for each abnormal node.

[0080] The real-time anomaly prediction module includes an anomaly node identification unit and an anomaly storage judgment unit.

[0081] The abnormal node identification unit is used to identify abnormal nodes in the warehouse network model in real time; the abnormal storage judgment unit is used to identify the traceability path based on the distribution of abnormal nodes, and analyze the expected abnormal situation of the warehouse based on the causal relationship of other nodes in the traceability path.

[0082] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A warehouse data analysis method for intelligent monitoring of agricultural product warehouses, characterized by: The analytical method includes the following steps: Step S100: Based on the warehouse layout, obtain several nodes of the warehouse and generate a warehouse network model; several monitoring devices are pre-installed to periodically collect the storage data of each node in the warehouse network model. Step S200: Extract the warehouse data of each node in any unit period, perform correlation analysis on the warehouse data between each node to obtain the causal relationship between different nodes; generate several influence paths in the warehouse network model based on the causal relationship between different nodes. Step S300: Identify several abnormal nodes in any unit period based on warehouse data; generate several traceability paths based on the distribution of all abnormal nodes in the warehouse network model; evaluate the suitability of each traceability path to obtain the optimal traceability path for each abnormal node. Step S400: Real-time identification of abnormal nodes in the warehouse network model; identification of the tracing path based on the distribution of abnormal nodes; and analysis of the expected abnormal situation of the warehouse based on the causal relationship of the remaining nodes in the tracing path. Step S300 includes the following steps: Step S301: Preset corresponding anomaly identification thresholds for the monitoring dimensions of each node. Randomly select a unit period, and randomly select a node from that unit period. Extract the warehouse data set of the selected node. If any monitoring data in the warehouse data set exceeds the anomaly identification threshold, then that monitoring data is set as an anomaly. Count the number of anomaly data in the warehouse data set as m. yc Let the total amount of data in the warehouse data set be M1. Calculate the percentage of abnormal data in the warehouse data set η = m. yc / M1; Preset an abnormality percentage threshold η th If η≥η th If so, the selected node is set as an abnormal node; Step S302: Obtain all abnormal nodes in the warehouse network model within the selected unit period, acquire several influence paths of the warehouse network model, and arbitrarily select one influence path. Compare each abnormal node with each node in the selected influence path to obtain the number of abnormal nodes in the selected influence path as p. Set the total number of abnormal nodes in the selected unit period as Q, and calculate the fit degree C = p / Q of the selected influence path; preset a fit degree threshold C. th If C≥C th Then the selected influence path will be set to the expected source tracing path of the unit period; Step S303: Obtain several expected tracing paths for the selected unit period. If the fit degree C=1 of an expected tracing path exists, set that expected tracing path as the optimal tracing path. Otherwise, select the expected tracing path with the highest fit degree and set it as the target tracing path. Remove the abnormal nodes in the target tracing path to obtain a set of remaining abnormal nodes. Regenerate a new expected tracing path from the set of remaining abnormal nodes, and again select the expected tracing path with the highest fit degree as the target tracing path, until all abnormal nodes are included or no new expected tracing path is generated. Summarize all target tracing paths to generate the optimal tracing path for the selected unit period.

2. The warehouse data analysis method for intelligent monitoring of agricultural product warehouses according to claim 1, characterized in that: Step S100 includes the following steps: Step S101: Pre-build a warehouse management system, which stores a complete set of warehouse layout information, including warehouse area division information, control equipment information, and fixed structure information; set any layout information corresponding to a structure in the warehouse as a node of the warehouse, generate several nodes of the warehouse, and summarize them to obtain the warehouse network model. Step S102: Several monitoring devices are installed in the warehouse to collect storage data at any node in the warehouse. For each node, a corresponding monitoring dimension is preset. The monitoring data collected by the monitoring devices at each unit cycle is summarized to obtain the storage data set of any node in any unit cycle.

3. The warehouse data analysis method for intelligent monitoring of agricultural product warehouses according to claim 2, characterized in that: Step S200 includes the following steps: Step S201: Arbitrarily select a unit period, and arbitrarily select a node from the selected unit period to obtain the warehouse data set of the selected node; sort the monitoring data in the warehouse data set according to the order of the collection time points, and arbitrarily select the a-th monitoring data D. a and the (a+1)th monitoring data D a+1 The trend ΔD between the two monitoring data points is obtained. (a,a+1) =D a+1 -D a If ΔD (a,a+1) If >0, then for monitoring data D a+1 If ΔD is set as the first feature data, then... (a,a+1) <0, then for monitoring data D a+1 If ΔD is set as the second feature data, then... (a,a+1) =0, then the monitoring data D a+1 With monitoring data D a The characteristic data types remain the same; Step S202: Obtain the characteristic data type of the last monitoring data, and count the number of data in the warehouse data set that have the same characteristic data type as the last monitoring data as m. sim Set the total amount of data in the warehouse dataset to M. total The calculated data volume ratio is γ=m sim / M total Preset a data volume percentage threshold γ th If γ≥γ th If so, the feature data type of the last monitored data is set to the feature data type of the selected node; otherwise, the selected node is set to the fluctuation data type. Step S203: Obtain the monitoring dimensions and data types of each node, establish a logical database in advance, and set corresponding causal relationships between monitoring dimensions of different data types in the logical database; obtain the monitoring devices corresponding to each node, extract the monitoring range of each monitoring device, if the monitoring devices corresponding to two nodes are the same, then set the two nodes as associated nodes; if the monitoring devices corresponding to two nodes are different and there are no other monitoring ranges between the monitoring ranges of the two monitoring devices, then set the two nodes as associated nodes as well. Step S204: Randomly select two nodes. If the two selected nodes are related nodes, obtain the monitoring dimensions and data types of the two nodes respectively, and call the logical database to perform causal relationship matching on the two selected nodes. If there is a causal relationship between the two selected nodes in the logical database, set the two selected nodes as a causal node group, and set the two selected nodes as the cause node and the effect node respectively. Step S205: Extract causal node groups between different nodes in the selected unit period of the warehouse network model. Randomly select two causal node groups. If the cause node in one causal node group is the effect node in the other causal node group, then connect the two causal node groups through the same node to obtain the influence path of the two causal node groups. Connect the influence paths between different causal node groups to obtain several complete influence paths.

4. The warehousing data analysis method for intelligent monitoring of agricultural product warehouses according to claim 3, characterized in that: Step S400 includes the following steps: Step S401: Monitor and collect data from each node in the warehouse network model in real time to obtain the real-time warehouse data set of each node. Compare the real-time warehouse data set with the corresponding anomaly identification threshold to obtain several real-time abnormal nodes. Step S402: Obtain the distribution of the several real-time abnormal nodes in the warehouse network model, generate several expected traceability paths, and obtain the adaptation degree of each expected traceability path; accumulate the adaptation degree of each expected traceability path to obtain a comprehensive adaptation value; preset an abnormal adaptation threshold, and if the comprehensive adaptation value exceeds the abnormal adaptation threshold, send an abnormal reminder to the warehouse storage data.

5. A warehouse data analysis system, used to execute the warehouse data analysis method for intelligent monitoring of agricultural product warehouses as described in any one of claims 1-4, characterized in that: The analysis system includes a warehouse data analysis module, a cause-and-effect anomaly analysis module, a source tracing path analysis module, and a real-time anomaly prediction module. The warehouse data analysis module is used to obtain several nodes of the warehouse based on the warehouse layout and generate a warehouse network model; several monitoring devices are pre-installed to periodically collect warehouse data of each node in the warehouse network model. The causal anomaly analysis module is used to extract warehouse data from each node in any unit period, perform correlation analysis on the warehouse data between each node, obtain the causal relationship between different nodes, and generate several influence paths in the warehouse network model based on the causal relationship between different nodes. The traceability path analysis module is used to identify several abnormal nodes in any unit period based on warehouse data, generate several traceability paths based on the distribution of all abnormal nodes in the warehouse network model, evaluate the adaptability of each traceability path, and obtain the optimal traceability path for each abnormal node. The real-time anomaly prediction module is used to identify abnormal nodes in the warehouse network model in real time; identify the tracing path based on the distribution of abnormal nodes; and analyze the expected anomaly situation of the warehouse based on the causal relationship of other nodes in the tracing path.

6. A warehouse data analysis system according to claim 5, characterized in that: The warehouse data analysis module includes a warehouse network modeling unit and a warehouse data extraction unit; The warehouse network modeling unit is used to obtain several nodes of the warehouse based on the layout of the warehouse and generate a warehouse network model; the warehouse data extraction unit is used to periodically collect warehouse data of each node in the warehouse network model by pre-installing several monitoring devices.

7. A warehouse data analysis system according to claim 5, characterized in that: The causal anomaly analysis module includes a causal node association unit and an influence path generation unit; The causal node association unit is used to extract the storage data of each node in any unit period, perform correlation analysis on the storage data between each node, and obtain the causal relationship between different nodes; the influence path generation unit is used to generate several influence paths in the warehouse network model according to the causal relationship between different nodes.

8. A warehouse data analysis system according to claim 5, characterized in that: The source tracing path analysis module includes a source tracing path identification unit and a path probability evaluation unit; The traceability path identification unit is used to identify several abnormal nodes in any unit period based on warehouse data, and generate several traceability paths based on the distribution of all abnormal nodes in the warehouse network model; the path probability evaluation unit is used to evaluate the suitability of each traceability path and obtain the optimal traceability path for each abnormal node.

9. A warehouse data analysis system according to claim 5, characterized in that: The real-time anomaly prediction module includes an anomaly node identification unit and an anomaly storage judgment unit. The abnormal node identification unit is used to identify abnormal nodes in the warehouse network model in real time; the abnormal warehouse judgment unit is used to identify the traceability path based on the distribution of abnormal nodes, and analyze the expected abnormal situation of the warehouse according to the causal relationship of other nodes in the traceability path.

Images