Intelligent ventilation control methods, devices, equipment and media for tobacco storage warehouses

Abstract

This invention provides an intelligent ventilation control method, device, equipment, and medium for tobacco storage warehouses, comprising: triggering environmental data collection based on a preset cycle; collecting tobacco storage environment data sets through multi-dimensional sensing components deployed in different storage areas and storage shelves of each storage area; sequentially performing data preprocessing, feature construction, state recognition, and strategy generation operations for each tobacco storage environment data set to generate a local ventilation control strategy; performing collaborative analysis and conflict resolution based on the local ventilation control strategy and the priority weights of each storage area to generate a global ventilation strategy for the warehouse; and controlling the independent ventilation equipment deployed in each storage area and the centralized ventilation equipment of the entire warehouse based on the local ventilation control strategy and the global ventilation strategy to perform precise ventilation operations, thereby solving the problems of localized tobacco mold and abnormal fermentation caused by the strong heterogeneity of the environmental space and the crude and lagging nature of traditional ventilation methods in large tobacco warehouses.

Description

Technical Field

[0001] This invention relates to the field of intelligent warehousing and industrial automation technology, and in particular to an intelligent ventilation control method, device, equipment and medium for tobacco storage warehouses. Background Technology

[0002] As a continuously bioactive agricultural product, tobacco storage is essentially a meticulous biochemical management process. During storage, tobacco leaves continuously respire and ferment, releasing heat, moisture, and carbon dioxide. Improper management of this process can easily lead to problems such as localized heat buildup, excessive humidity, and insufficient oxygen, resulting in mold, carbonization, or quality deterioration. Therefore, maintaining a stable storage environment, especially precise and timely ventilation control, is crucial for ensuring the quality of stored tobacco leaves and minimizing losses.

[0003] Currently, ventilation control in tobacco warehouses mainly relies on two traditional modes: one is timed and scheduled global ventilation based on fixed cycles or manual experience; the other is based on temperature and humidity sensors deployed at a few fixed points within the warehouse, triggering full ventilation by simple threshold judgments, such as temperature exceeding a certain value or humidity exceeding a certain value. However, tobacco storage warehouses are typically characterized by large volumes, high shelving, and complex internal airflow organization. Different areas, such as sunny and shady sides, and different vertical levels, such as the top and bottom shelves, will form significantly heterogeneous microenvironments due to external environmental penetration, differences in tobacco leaf metabolism, and airflow inertia. The aforementioned traditional methods have obvious drawbacks: firstly, manual or timed modes have slow response times and cannot cope with sudden deterioration of local environments; secondly, the single-point threshold-triggered control method is indiscriminate, activating full ventilation whenever a problem occurs in a local area, which not only consumes a lot of energy but may also cause secondary interference to areas that do not require intervention, such as causing excessive water loss in dry areas, and failing to achieve differentiated and precise control.

[0004] In recent years, with the development of IoT technology, monitoring solutions involving the deployment of more sensors in warehouses have emerged. Some studies have also attempted to introduce simple data models for prediction. However, these improved solutions still have significant shortcomings: First, the utilization of sensor data is relatively superficial, mostly limited to threshold alarms for independent parameters, lacking the ability to comprehensively analyze vertical spatial gradient changes, multi-parameter coupling relationships, and internal heat distribution in tobacco piles; second, existing methods have failed to effectively construct a mapping model from environmental data to the physiological state of tobacco leaves, making it impossible to identify key risk states such as mold tendency and fermentation activity in advance; finally, at the control level, there is still a lack of a closed-loop intelligent control system that can automatically generate and coordinate the execution of local regional strategies and global warehouse strategies based on multi-dimensional, three-dimensional sensing results.

[0005] Therefore, there is an urgent need for an intelligent ventilation control method for tobacco storage warehouses to solve the technical problems of localized tobacco mold growth, abnormal fermentation, and quality decline caused by the strong heterogeneity of the environmental space and the crude and lagging nature of traditional ventilation methods in large tobacco warehouses. Summary of the Invention

[0006] To overcome the problems existing in the related technologies, this disclosure provides an intelligent ventilation control method, device, equipment and medium for tobacco storage warehouses, in order to solve the technical problems of localized tobacco mold growth, abnormal fermentation and quality decline caused by the strong heterogeneity of the environmental space and the crude and lagging traditional ventilation methods in large tobacco warehouses.

[0007] This specification provides one or more embodiments of an intelligent ventilation control method for a tobacco storage warehouse, including the following steps: Based on the preset cycle of triggering environmental data collection, multi-dimensional sensing components deployed in different storage areas of the warehouse and different vertical levels of the storage racks in each storage area are used to collect environmental parameter sets of each storage area and each level, forming a tobacco leaf storage environment data set that corresponds to each storage rack. For each of the tobacco storage environment data groups, data preprocessing, feature construction, state recognition and strategy generation operations are performed sequentially to generate a local ventilation control strategy; Based on the local ventilation control strategies of all storage areas and the priority weights of each storage area, a collaborative analysis and conflict resolution are performed to generate a global ventilation strategy for the warehouse. Based on the local ventilation control strategy and the warehouse global ventilation strategy, the independent ventilation equipment deployed in each storage area and the centralized ventilation equipment of the warehouse as a whole are controlled respectively to perform precise ventilation operations.

[0008] Preferably, for each of the tobacco storage environment data groups, the steps of performing data preprocessing, feature construction, state identification, and strategy generation operations in sequence to generate a local ventilation control strategy specifically include the following steps: The data set of tobacco storage environment is preprocessed and features are constructed to generate multi-dimensional environmental features that characterize the overall and layered environmental conditions of the shelves. The multidimensional environmental features are input into a pre-trained tobacco storage state recognition model, and the tobacco storage state information corresponding to each shelf level is output. The tobacco storage state information includes at least the heat accumulation state, the moisture accumulation state, the fermentation activity state, and the condensation risk state. Based on the tobacco storage status information and combined with the preset state-policy mapping rules, a local ventilation control strategy for each storage area is generated.

[0009] Preferably, the step of preprocessing and feature construction of the tobacco storage environment data set to generate multidimensional environmental features characterizing the overall and layered environmental conditions of the shelving specifically includes the following steps: Sensor type splitting, data calibration, and outlier handling were performed on the raw environmental data set; For each level of the shelf, extract directly observable features including temperature, humidity, and gas concentration within the shelf from the sensor data. Calculate the dew point temperature and relative humidity saturation characteristics of the current layer based on the temperature and humidity within the layer; Based on infrared temperature distribution data, spatial thermal distribution characteristics that characterize the temperature uniformity and hot spot distribution inside the smoke pile are extracted. Calculate the gradient differences in temperature, humidity, and gas concentration between adjacent vertical layers to form the interlayer environmental gradient characteristics; By combining current data with historical data from the same period, the changing trends of key environmental parameters are calculated to form time-series variation characteristics; All features at the same level are concatenated to form a hierarchical feature vector, and the hierarchical feature vectors of all levels of the shelf are concatenated in spatial order to form the multidimensional environmental features.

[0010] Preferably, the step of inputting the multidimensional environmental features into a pre-trained tobacco storage state recognition model and outputting tobacco storage state information corresponding to each shelf level specifically includes the following processing steps: The multidimensional environmental features received as input, wherein the tobacco storage state recognition model is a multi-task learning model based on a deep neural network; The first feature extraction branch encodes the hierarchical observation features arranged in time sequence to capture the temporal dependencies of environmental states. The spatial heat distribution features are encoded through the second feature extraction branch to capture the two-dimensional thermal field spatial pattern inside the smoke pile. The feature fusion module performs cross-modal alignment and fusion on the features extracted from the two branches to form a unified environment state code that includes spatiotemporal semantics. The environmental state code is input into the multi-task decoding head, and the quantitative scores of each level on multiple predetermined storage state dimensions are output in parallel as the tobacco leaf storage state information.

[0011] Preferably, the step of generating a local ventilation control strategy for each storage area based on the tobacco storage status information and in conjunction with a preset state-policy mapping rule specifically includes the following steps: The multiple state scores of each level in the tobacco storage state information are aggregated according to the preset inter-level influence weights to obtain the comprehensive state vector of the storage area. The integrated state vector is input to a policy trigger consisting of multiple logic judgment units; Each of the aforementioned logical judgment units is configured to identify a specific composite storage anomaly pattern and output a trigger strength value; The vector composed of multiple trigger strength values ​​output by the strategy trigger is converted into a specific device control parameter vector through a learnable strategy mapping matrix, thus forming the local ventilation control strategy. The equipment control parameters include at least the regional air supply intensity, air supply direction angle, local exhaust air intensity, and dehumidification intensity.

[0012] Preferably, the step of generating a global warehouse ventilation strategy based on the local ventilation control strategies of all storage areas and the priority weights of each storage area through collaborative analysis and conflict resolution specifically includes the following steps: Each storage area is assigned a dynamic priority weight, which is determined based on at least one of the following factors: the varietal value of the stored tobacco leaves, the severity of the current condition, and the frequency of historical problems. Based on the local ventilation control strategy and dynamic priority weight of each storage area, the basic control parameters of the global ventilation equipment are calculated in a weighted manner. Identify the dominant storage risk types that are prevalent or reach a certain severity level across all regions; If a dominant storage risk type exists, the basic control parameters are modified based on the global control rules corresponding to the storage risk type to generate the final warehouse global ventilation strategy.

[0013] Preferably, the precise ventilation operation, based on the local ventilation control strategy and the warehouse global ventilation strategy, involves controlling the independent ventilation equipment deployed in each storage area and the centralized ventilation equipment of the entire warehouse to perform precise ventilation operations, specifically including the following steps: For each storage area, under the premise of meeting the safety operation constraints of the area equipment, the opening and angle of the independent air valves, the speed of the circulating fan and the working level of the dehumidifier in each area are adjusted according to the local ventilation control strategy. Under the premise of meeting the overall energy consumption and wind pressure balance constraints of the entire warehouse system, the operating parameters of the main supply fan, main exhaust fan and centralized dehumidification system of the warehouse are adjusted according to the overall ventilation strategy of the warehouse. The local ventilation control strategy has a higher execution priority than the global warehouse ventilation strategy. However, when the two conflict in terms of resource usage, the constraints of the global warehouse ventilation strategy shall prevail for coordination.

[0014] This specification provides one or more embodiments of an intelligent ventilation control device for a tobacco storage warehouse, comprising: The environmental data acquisition module is used to trigger environmental data acquisition based on a preset cycle. Through multi-dimensional sensing components deployed in different storage areas of the warehouse and different vertical levels of the storage racks in each storage area, it collects environmental parameter sets of each storage area and each level, forming a tobacco leaf storage environment data set that corresponds to each storage rack. The local strategy generation module is used to sequentially perform data preprocessing, feature construction, state recognition and strategy generation operations for each of the tobacco storage environment data groups to generate a local ventilation control strategy. The global strategy generation module is used to perform collaborative analysis and conflict resolution based on the local ventilation control strategies of all storage areas and the priority weights of each storage area to generate a global ventilation strategy for the warehouse. The ventilation execution module is used to control the independent ventilation equipment deployed in each storage area and the centralized ventilation equipment of the entire warehouse based on the local ventilation control strategy and the warehouse global ventilation strategy, and to perform precise ventilation operations.

[0015] This specification provides one or more embodiments of a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the intelligent ventilation control method for the tobacco storage warehouse described above.

[0016] This specification provides one or more embodiments of a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the intelligent ventilation control method for the tobacco storage warehouse described above.

[0017] This disclosure provides an intelligent ventilation control method, device, equipment, and medium for tobacco storage warehouses. Its advantages lie in its ability to collect environmental data based on a preset cycle. By deploying multi-dimensional sensing components in different storage areas and vertical levels of the storage shelves within each area, it collects environmental parameter sets for each storage area and level, forming a tobacco storage environment data set corresponding to each storage shelf. The deployment of multi-dimensional sensors in each storage area and shelf level, along with periodic synchronous data collection, completely solves the problems of large blind spots and inability to reflect spatial differences in traditional single-point monitoring, providing a comprehensive and reliable data foundation for precise control. For each tobacco storage environment data set, data preprocessing, feature construction, state recognition, and strategy generation operations are sequentially performed to generate a local ventilation control strategy. The data for each shelf is serialized and analyzed using AI, achieving a deep diagnosis from raw data to the "tobacco storage state." The system automatically generates quantitative local ventilation strategies based on the local ventilation control strategies of all storage areas, achieving an intelligent leap from "environmental perception" to "preliminary decision-making." Based on the local ventilation control strategies of all storage areas and the priority weights of each storage area, a collaborative analysis and conflict resolution process is performed to generate a global ventilation strategy for the warehouse. This integrates all local strategies and, based on regional priorities, resolves conflicts and schedules resources to generate a consistent global strategy, resolving potential system conflicts caused by zoning control and ensuring overall stability and efficiency. Based on the local ventilation control strategies and the global ventilation strategy, the system controls the independent ventilation equipment deployed in each storage area and the centralized ventilation equipment of the entire warehouse, executing precise ventilation operations. The strategy is decomposed and distributed to the corresponding regional and centralized equipment, driving their coordinated actions. Ultimately, this achieves precise ventilation through "zoning, layering, and differentiation," forming a complete automated closed loop of "perception-decision-execution." Attached Figure Description

[0018] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating an intelligent ventilation control method for a tobacco storage warehouse provided in one or more embodiments of this specification; Figure 2 This is a schematic diagram of the distribution of storage areas and storage racks within a tobacco storage warehouse provided in one or more embodiments of this specification. Figure 3A schematic diagram illustrating the characteristics of the tobacco storage environment, tobacco storage status information, and regional state vectors provided in one or more embodiments of this specification; Figure 4 A schematic diagram of the model structure for the tobacco leaf storage state output model provided in one or more embodiments of this specification; Figure 5 A schematic diagram of the structure of an intelligent ventilation control device for a tobacco storage warehouse provided for one or more embodiments of this specification; Figure 6 This is a schematic diagram of the structure of a computer device provided for one or more embodiments of this specification. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this invention.

[0021] The present invention will now be described in detail with reference to specific embodiments and accompanying drawings.

[0022] Method Implementation Examples According to embodiments of the present invention, an intelligent ventilation control method for a tobacco storage warehouse is provided, such as... Figure 1 The diagram shown is a flowchart illustrating the intelligent ventilation control method for a tobacco storage warehouse provided in this embodiment. The intelligent ventilation control method for a tobacco storage warehouse according to this embodiment includes the following steps: S110. Based on a preset cycle, environmental data collection is triggered. Through multi-dimensional sensing components deployed in different storage areas of the warehouse and different vertical levels of the storage racks in each storage area, environmental parameter sets of each storage area and each level are collected to form a tobacco leaf storage environment data set corresponding to each storage rack.

[0023] In some embodiments, the execution entity (e.g., a computing device) of the ventilation control method applied to a tobacco storage warehouse can, in response to a periodic time condition being met via wired or wireless connection, collect environmental data from various storage areas within the tobacco storage warehouse using multi-dimensional sensors positioned in different regions and at different levels, thereby obtaining a tobacco storage environmental data set. The aforementioned periodic time condition can be that the time interval between the current time and the last environmental data collection is greater than or equal to a preset time. For example, the preset time could be 30 minutes. Figure 2 The diagram shows the distribution of storage areas and the storage racks within the tobacco storage warehouse provided in this embodiment. 201 represents the perimeter wall of the tobacco storage warehouse, which contains multiple storage areas as indicated by reference numeral 202 (the areas within the black squares). Each storage area has a multi-layered storage rack, such as a three-layered storage rack as indicated by reference numeral 203. The number of racks in each storage area is the same, and the rack height is also the same. The number of rack layers used in the tobacco storage warehouse is not limited here. Each tobacco storage environment data group corresponds to a storage rack within the storage area; that is, the tobacco storage environment data in a tobacco storage environment data group are collected from different tobacco storage layers within the same storage rack.

[0024] The aforementioned multi-dimensional sensors may include temperature and humidity sensors, CO2 sensors, O2 sensors, VOC (volatile organic compound) gas sensors, and area array infrared temperature sensors. Specifically, each shelf in each storage area is equipped with one temperature and humidity sensor, one CO2 sensor, one O2 sensor, and one VOC sensor to monitor the temperature, humidity, respiratory metabolic intensity, oxygen content, and volatile organic compound concentration near the tobacco piles within the tobacco storage layer, thereby identifying the storage environment, degree of mold, or degree of fermentation of the tobacco in that layer. At the top of each tobacco storage layer, i.e., at the bottom of the shelf above, is an area array infrared temperature sensor to periodically capture the surface temperature distribution of the tobacco in that layer and output the temperature at each location within that area in a two-dimensional matrix to identify heat accumulation or fermentation hotspots. All point sensors within each shelf, i.e., each tobacco storage layer, including temperature and humidity sensors, CO2 sensors, O2 sensors, and VOC gas sensors, are installed at the front or middle of the tobacco stack, approximately 15-20 cm from the shelf surface, avoiding direct contact with the tobacco or interference from air vents. The area array infrared temperature sensor covers the entire tobacco storage layer from above, enabling non-contact temperature distribution monitoring.

[0025] The aforementioned computing devices can be hardware or software. When the computing device is hardware, it can be implemented as a distributed cluster of multiple servers or terminal devices, or as a single server or terminal device. When the computing device is software, it can be installed on the hardware devices listed above. It can be implemented as, for example, multiple software programs or software modules used to provide distributed services, or as a single software program or software module. No specific limitations are made here. It should be understood that the number of computing devices can be arbitrary, depending on the implementation requirements.

[0026] It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultra wideband) connections, and other currently known or future wireless connection methods.

[0027] It should be noted that each storage area is equipped with independently controllable branch air valves, miniature circulating fans or supply fans, and exhaust fans. Branch air valves regulate the airflow direction from the main air supply duct of the warehouse and are installed at the inlet port of each storage area. Miniature circulating fans or supply fans are installed on the sides of each tobacco storage layer or inter-layer uprights of the storage racks to promote local airflow within the area and quickly address situations such as moisture accumulation at the bottom or insufficient oxygen concentration in the middle layers. Exhaust fans are located at the top of the storage area to promote air circulation and exhaust air. In addition, each storage area is equipped with a split-type dehumidifier, located on the side or bottom of the storage racks to address the risk of sudden humidity increases or condensation. The entire tobacco storage warehouse is equipped with a main supply fan and return air system, located at the inlet and outlet of the main air duct, to provide basic supply air and temperature and humidity balanced return air to all storage areas. Furthermore, the entire tobacco storage warehouse is equipped with a main dehumidification unit to centrally handle the moisture load in the incoming or returning air, which can improve the overall dehumidification efficiency of the warehouse when multiple storage areas experience humidity or condensation risks. The top or side walls of the tobacco storage warehouse are also equipped with air ducts and exhaust vents to optimize airflow direction and effectively remove air from the top heat accumulation or high humidity areas.

[0028] S120. For each tobacco storage environment data group, perform data preprocessing, feature construction, state identification, and strategy generation operations in sequence to generate a local ventilation control strategy, specifically including the following steps: Data preprocessing and feature construction were performed on the tobacco storage environment data set to generate multidimensional environmental features that characterize the overall and layered environmental conditions of the shelving.

[0029] Multidimensional environmental features are input into a pre-trained tobacco storage status recognition model, which outputs tobacco storage status information for each shelf level. The tobacco storage status information includes at least heat accumulation status, moisture accumulation status, fermentation activity status, and condensation risk status.

[0030] Based on tobacco storage status information and combined with preset status-policy mapping rules, a local ventilation control strategy is generated for each storage area.

[0031] S130. Based on the local ventilation control strategies of all storage areas and the priority weights of each storage area, a collaborative analysis and conflict resolution are performed to generate a global ventilation strategy for the warehouse.

[0032] S140. Based on local ventilation control strategies and warehouse global ventilation strategies, control the independent ventilation equipment deployed in each storage area and the centralized ventilation equipment of the entire warehouse to perform precise ventilation operations.

[0033] The method provided in this embodiment, by triggering environmental data collection based on a preset cycle, collects environmental parameter sets for each storage area and each level of the warehouse through multi-dimensional sensing components deployed in different storage areas and at different vertical levels of the storage shelves in each storage area. This forms a data set of tobacco storage environment data corresponding to each storage shelf. By deploying multi-dimensional sensors in each storage area and on each shelf and collecting data synchronously and periodically, the method completely solves the problems of large blind spots and inability to reflect spatial differences in traditional single-point monitoring, providing a comprehensive and reliable data foundation for precise control. For each tobacco storage environment data set, data preprocessing, feature construction, state recognition, and strategy generation operations are performed sequentially to generate a local ventilation control strategy. The data of each shelf is serialized and analyzed by AI, realizing a deep diagnosis from raw data to "tobacco storage status," and automatically generating quantitative data accordingly. The local ventilation strategy achieves an intelligent leap from "environmental perception" to "preliminary decision-making." Based on the local ventilation control strategies of all storage areas and the priority weights of each storage area, a collaborative analysis and conflict resolution are performed to generate a global ventilation strategy for the warehouse. By integrating all local strategies and resolving conflicts and scheduling resources according to regional priorities, a coordinated global strategy is generated, resolving system conflicts that may be caused by zoning control and ensuring the stability and efficiency of overall operation. Based on the local ventilation control strategy and the global ventilation strategy, the independent ventilation equipment deployed in each storage area and the centralized ventilation equipment of the entire warehouse are controlled respectively to execute precise ventilation operations. The strategy is decomposed and distributed to the corresponding regional equipment and centralized equipment to drive their coordinated actions, ultimately achieving precise ventilation of "zoning, layering, and differentiation," forming a complete automated closed loop of "perception-decision-execution."

[0034] In one embodiment, the aforementioned executing entity can perform feature construction on the tobacco storage environment data sequence to generate tobacco storage environment features. The tobacco storage environment data sequence is obtained after preprocessing the tobacco storage environment data set.

[0035] In some optional implementations of certain embodiments, before performing feature construction on the tobacco storage environment data sequence to generate tobacco storage environment features, the execution entity may perform the following steps: preprocessing and feature construction on the tobacco storage environment data set to generate multidimensional environmental features characterizing the overall and layered environmental conditions of the shelving, specifically including the following steps: Sensor type splitting, data calibration, and outlier handling were performed on the raw environmental data set.

[0036] For sensor data at each level of the shelving unit, directly observable features, including temperature, humidity, and gas concentration within each level, are extracted. Specifically, the tobacco storage environment data is parameterized according to sensor type to obtain data on storage temperature, humidity, carbon dioxide concentration, oxygen concentration, volatile organic compound (VOC) concentration, and temperature distribution within the stack. In practice, the implementing entity can perform parameterized analysis on the collected tobacco storage environment data according to sensor type, using the collected temperature, humidity, carbon dioxide concentration, oxygen concentration, VOC concentration, and temperature distribution data as the data for storage temperature, humidity, carbon dioxide concentration, oxygen concentration, VOC concentration, and temperature distribution within the stack. The temperature distribution data within the stack can be the temperature values ​​of various monitoring points within the monitored area, represented in matrix form.

[0037] Spatial smoothing is performed on the aforementioned in-pile temperature distribution data to update it. In practice, the execution entity can use a 3×3 sliding kernel to perform Gaussian smoothing on the in-pile temperature distribution data, thereby enhancing the continuity of temperature values ​​within the region, highlighting temperature distribution changes, and updating the in-pile temperature distribution data.

[0038] The storage temperature, humidity, carbon dioxide concentration, oxygen concentration, and volatile organic compound concentration within the aforementioned layers are calibrated. In practice, the aforementioned actuator can perform linear calibration on parameter values ​​collected by sensors exhibiting numerical drift (i.e., those requiring correction coefficients), according to the corresponding correction coefficients. The calibrated value is calculated as: Corrected value = a × Collected value + b. Here, a and b are correction coefficients set for sensors exhibiting numerical drift; each sensor has a different coefficient to improve the accuracy of the environmental data collected during sensor use.

[0039] The updated tobacco storage environment data set is hierarchically sorted to obtain a tobacco storage environment data sequence. Each data point in this sequence has a corresponding hierarchical label. The hierarchical label indicates the specific shelf number within the storage rack from which the data originates (where shelf numbers are counted from low to high by default, with the bottom shelf being shelf number 1). In practice, the executing entity can sort the updated tobacco storage environment data from high to low according to the shelf number indicated by the corresponding hierarchical label to obtain the tobacco storage environment data sequence.

[0040] The dew point temperature and relative humidity saturation characteristics of the current layer are calculated based on the temperature and humidity within the layer.

[0041] Based on infrared temperature distribution data, spatial thermal distribution characteristics that characterize the temperature uniformity and hot spot distribution inside the smoke pile are extracted.

[0042] By combining current data with historical data from the same period, the changing trends of key environmental parameters are calculated to form time-series change characteristics.

[0043] All features at the same level are concatenated to form a hierarchical feature vector, and the hierarchical feature vectors of all levels of the shelf are concatenated in spatial order to form a multidimensional environmental feature.

[0044] Specifically, it includes the following steps: Based on the pile temperature distribution data included in the aforementioned tobacco storage environment data, a pile heat distribution characteristic is generated. This pile heat distribution characteristic can be the average temperature, maximum temperature, minimum temperature, and 0.95 quantile temperature value from the aforementioned pile temperature distribution data.

[0045] Based on the above-mentioned tobacco storage environment data, including the storage temperature and humidity within the storage layer, the dew point temperature within the layer is determined. In practice, the implementing entity can substitute the storage temperature and humidity within the storage layer into the Magnus dew point model to determine the dew point temperature within the layer, i.e., the critical temperature at which dew droplets form within the layer.

[0046] Based on the hierarchical labels corresponding to the aforementioned tobacco storage environment data and adjacent tobacco storage environment data, inter-layer gradient environmental features are generated. These inter-layer gradient environmental features can characterize the differences in environmental states between adjacent layers. In practice, the implementing entity can use the differences obtained by subtracting the tobacco storage environment data corresponding to the higher-level tobacco storage layer from the various parameters included in the aforementioned storage environment data—namely, the storage temperature, humidity, carbon dioxide concentration, oxygen concentration, and volatile organic compound concentration—according to the corresponding storage shelf layer number represented by the hierarchical label, as the inter-layer gradient environmental features.

[0047] It should be noted that the difference obtained after subtracting the parameters can be either positive or negative, representing the temperature change in the vertical direction within the storage space. The last tobacco storage environment data in the above tobacco storage environment data sequence, i.e., the tobacco storage environment data corresponding to the highest level of the storage rack, does not require the construction of inter-layer gradient environmental features.

[0048] A temporal distribution environmental feature is generated using the corresponding historical tobacco storage environment data. This temporal distribution environmental feature characterizes the changing trend of the tobacco storage environment at the corresponding level over time. The target historical tobacco storage environment data can be tobacco storage environment data from the previous environmental data collection, processed through the aforementioned preprocessing steps, and belonging to the same storage area and storage layer as the aforementioned tobacco storage environment data. In practice, the executing entity can use the differences obtained by subtracting the parameters included in the aforementioned tobacco storage environment data—namely, the storage temperature within the layer, the storage humidity within the layer, the carbon dioxide concentration within the layer, the oxygen concentration within the layer, and the volatile organic compound concentration within the layer—from the parameters included in the target historical tobacco storage environment data as the temporal distribution environmental feature.

[0049] It should be noted that the ventilation control method applied to tobacco storage warehouses is executed periodically under the aforementioned cycle time conditions. Therefore, the relevant information generated by each step in the method during each execution is stored in the database, providing historical data support for the next ventilation control.

[0050] The gradient differences in temperature, humidity, and gas concentration between adjacent vertical layers are calculated to form interlayer environmental gradient features. Specifically, the aforementioned tobacco storage environment data, the aforementioned intralayer dew point temperature, the aforementioned pile heat distribution characteristics, the aforementioned interlayer gradient environmental features, and the aforementioned time distribution environmental features are horizontally concatenated to obtain hierarchical storage environment features. For example, the aforementioned hierarchical storage environment features can be (A1, A2, A3, A4, A5). Here, A1 can be tobacco storage environment data, A2 can be intralayer dew point temperature, A3 can be pile heat distribution characteristics, A4 can be interlayer gradient environmental features, and A5 can be time distribution environmental features. The generated hierarchical storage environment features are then vertically concatenated to obtain the tobacco storage environment features. In practice, the executing entity performs vector concatenation of the generated hierarchical storage environment features in the vertical direction, that is, concatenating them from high to low according to the corresponding layers to obtain the tobacco storage environment features. Thus, the aforementioned tobacco storage environment features can characterize both the overall environmental state of each tobacco storage layer in a storage area and the independent environmental state of each tobacco storage layer.

[0051] In one embodiment, multidimensional environmental features are input into a pre-trained tobacco storage status recognition model, which outputs tobacco storage status information corresponding to each shelf level. The specific processing steps include the following: The multi-dimensional environmental features received as input, namely the temporal features of the aforementioned hierarchical storage environment, are input into the aforementioned tobacco leaf storage state output model. The tobacco leaf storage state recognition model is a multi-task learning model built based on a deep neural network. The aforementioned hierarchical tobacco leaf storage states include heat accumulation state, moisture accumulation state, fermentation activity state, condensation state, mold tendency state, hotspot distribution state, and carbon-oxygen state. The aforementioned tobacco leaf storage state output model can be a neural network model that takes the temporal features of the environment of each tobacco storage layer as input and the storage state as output. Figure 3 The diagram shown is a schematic representation of the tobacco storage environment features 301, tobacco storage state information 302, and region state vector 303 provided in this embodiment. Figure 3 Taking the first layer of data (shown by reference numeral 302) as an example, the hierarchical tobacco storage state can include seven state parameters, namely s11, s12, s13, s14, s15, s16, and s17, which are used to characterize the heat accumulation state, wet accumulation state, fermentation activity state, condensation state, mold tendency state, hot spot distribution state, and carbon-oxygen state of the corresponding tobacco storage layer at the current time point.

[0052] Specifically, the aforementioned heat accumulation state refers to a sustained increase in temperature or localized high temperatures in a certain storage area or layer of tobacco leaves and the surrounding air, often caused by tobacco leaf respiration and metabolism, fermentation heat release, or poor airflow. When the parameter value corresponding to the aforementioned heat accumulation state is high (e.g., s11 > 0.7), it indicates that the temperature in the corresponding tobacco storage layer is consistently ≥ 3 degrees Celsius higher than the average layer temperature, and localized high-temperature points persist in infrared thermal imaging, indicating heat accumulation within the tobacco pile or poor ventilation in this area, requiring increased air supply. When the parameter value is in the middle range (e.g., s11 within the range of [0.4, 0.7]), it indicates a slight increase in temperature or upper-layer thermal stratification in the corresponding tobacco storage layer, requiring continued basic air supply. When the parameter value is low (e.g., s11 < 0.4), it indicates that the corresponding tobacco storage layer environment is dry and stable.

[0053] The aforementioned humidity accumulation indicates abnormally high humidity levels in certain storage areas or layers, exhibiting a cumulative trend. This may be due to insufficient ventilation leading to moisture retention, or localized condensation caused by temperature differences causing moisture to adhere to the tobacco leaves or shelf surfaces, thereby increasing the probability of mold growth. When the parameter value corresponding to the aforementioned humidity accumulation state is high (e.g., s12 > 0.7), it indicates that the humidity in the corresponding tobacco storage layer is ≥10% higher than the environmental average, and the humidity sensor shows a significant vertical gradient, or that the relative humidity in the bottom layer or near the wall is >80%, easily leading to condensation and mold growth, requiring the activation of dehumidification and circulating ventilation. When the parameter value is in the middle range (e.g., s12 within the range of [0.4, 0.7]), it indicates that the humidity in the corresponding tobacco storage layer is high but has not reached the dew point temperature, requiring light dehumidification. When the parameter value is low (e.g., s12 < 0.4), it indicates that the corresponding tobacco storage layer environment is dry and stable.

[0054] The aforementioned fermentation activity state refers to the degree to which tobacco leaves have entered an active fermentation stage in a localized area. This is reflected by changes in gas (such as increased CO2 and decreased O2) and temperature. Excessive fermentation can lead to off-flavors or a rapid decline in tobacco quality. When the parameter value corresponding to the fermentation activity state is high (e.g., s13 > 0.7), it indicates that the temperature and CO2 concentration in the corresponding tobacco storage layer are rising synchronously, while the O2 concentration is decreasing, indicating enhanced spontaneous respiration of the tobacco leaves. The infrared hot zone overlaps with the CO2 hot zone, requiring increased ventilation and gas exchange to prevent localized overheating or excessive fermentation. When the parameter value is in the middle range (e.g., s13 within the range of [0.4, 0.7]), it indicates mild biological activity within the corresponding tobacco storage layer, requiring moderate ventilation. When the parameter value is low (e.g., s13 < 0.4), it indicates that fermentation in the corresponding tobacco storage layer is stabilizing or in a dormant stage, requiring no additional ventilation intervention.

[0055] The aforementioned condensation state refers to the phenomenon where air reaches relative humidity saturation and liquid water precipitates when it encounters a cold surface or low-temperature environment, such as at the bottom or corners of warehouse shelves. When the parameter value corresponding to this condensation state is high (e.g., s14 > 0.7), it indicates that the temperature inside the corresponding tobacco storage layer is close to the dew point (temperature difference ≤ 2°C) and the humidity is ≥ 85%, often found at the bottom or top of the shelves, requiring enhanced dehumidification and ventilation. When the parameter value is in the middle range (e.g., s14 in the range [0.4, 0.7]), it indicates that the temperature inside the corresponding tobacco storage layer is close to the dew point, requiring light dehumidification. When the parameter value is low (e.g., s14 < 0.4), it indicates that there is no significant risk of condensation inside the corresponding tobacco storage layer, and no additional ventilation intervention is needed.

[0056] The aforementioned mold-prone state refers to a condition in a certain area of ​​the storage facility where, due to factors such as temperature and humidity conditions, stagnant air, and microclimate deviations, there is a high risk of mold growth. This can be determined through temperature, temperature distribution, and humidity. When the parameter value corresponding to the mold-prone state is high (e.g., s15 > 0.7), it indicates that the temperature and humidity conditions within the corresponding tobacco storage layer are both high and sustained for a relatively long period (e.g., 4 cycles), providing conditions for mold growth. Enhanced dehumidification and accelerated air circulation are required. When the parameter value is in the middle range (e.g., s15 within the range of [0.4, 0.7]), it indicates short-term fluctuations in high humidity and temperature within the corresponding tobacco storage layer, requiring light ventilation. When the parameter value is low (e.g., s15 < 0.4), it indicates that the environment within the corresponding tobacco storage layer is dry and stable, lacking conditions for mold growth, and no additional ventilation intervention is needed.

[0057] The aforementioned hotspot distribution state refers to the presence of high-temperature points, hot spots, or uneven temperature areas in the temperature field captured by the infrared array on the surface of the tobacco stack. When the parameter value corresponding to the hotspot distribution state is high (e.g., s16 > 0.7), it indicates the presence of obvious hotspot areas (e.g., area ratio > 5%, temperature difference > 3°C) in the infrared array of the corresponding tobacco storage layer. This indicates uneven air circulation in the tobacco storage layer, requiring enhanced ventilation. When the parameter value is in the middle range (e.g., s16 within the range of [0.4, 0.7]), it indicates that the infrared array hotspots in the corresponding tobacco storage layer are small and dispersed, requiring continued basic ventilation. When the parameter value is low (e.g., s16 < 0.4), it indicates a uniform temperature field within the corresponding tobacco storage layer, with no obvious hotspots, requiring no additional ventilation intervention.

[0058] When the parameter values ​​corresponding to the carbon-oxygen state are high (e.g., s17 > 0.7), it indicates that the CO2 concentration in the corresponding tobacco storage layer is >2000 ppm and O2 < 18%, indicating insufficient gas exchange and requiring enhanced ventilation. When the parameter values ​​are in the middle range (e.g., s17 is in the range [0.4, 0.7]), it indicates that the CO2 concentration in the corresponding tobacco storage layer is showing a slight upward trend, requiring moderate ventilation. When the parameter values ​​are low (e.g., s17 < 0.4), it indicates that the gas balance in the corresponding tobacco storage layer is normal, and no additional ventilation intervention is required.

[0059] The obtained tobacco leaf storage status at each level is defined as tobacco leaf storage status information. In practice, the aforementioned executing entity can sort the obtained tobacco leaf storage status at each level from high to low, and define the sorted tobacco leaf storage status at each level as tobacco leaf storage status information.

[0060] The first feature extraction branch encodes the hierarchical observation features arranged in time sequence to capture the temporal dependencies of environmental states.

[0061] The spatial thermal distribution features are encoded through the second feature extraction branch to capture the two-dimensional thermal field spatial pattern inside the smoke pile.

[0062] Specifically, the in-pile temperature distribution data, which includes the time-series characteristics of the above-mentioned hierarchical storage environment, is input into the infrared feature encoding module included in the above-mentioned tobacco storage state output model to obtain infrared feature encoding information.

[0063] In practice, existing neural network models often focus on feature learning for single-modality or single-dimensional data. For example, Convolutional Neural Networks (CNNs) can only extract spatial features such as infrared distribution data, while Recurrent Neural Networks (RNNs / Long Short-Term Memory, LSTMs) or Temporal Transformers are typically suitable for time series modeling. Therefore, directly using conventional neural network models makes it difficult to capture the spatial differences between different tobacco storage layers along the vertical direction of storage racks, the coupling of heat and moisture gases, and the dynamic correlations between multimodal signals. Secondly, in tobacco storage environments, states such as heat accumulation, moisture accumulation, and condensation often occur concurrently and are significantly affected by inter-layer airflow coupling. Traditional network structures struggle to explicitly establish dependencies between layers, easily leading to confusion in state judgments. In particular, when infrared hotspot signals and environmental sensor signals fluctuate asynchronously, single-branch models struggle to identify the source of abnormal signals, resulting in poor representation of the hierarchical tobacco storage state. Based on this, a composite network structure combining infrared data processing branch, temporal backbone network, cross-layer attention and gating fusion is combined with conventional deep networks. This enables joint feature learning of spatial hierarchy, temporal dynamics and multimodal data, thereby improving the ability to determine the storage status of each tobacco leaf storage layer in the warehouse rack.

[0064] like Figure 4 The diagram shown is a schematic of the model structure of the tobacco storage state output model provided in this embodiment. The tobacco storage state output model may include a temporal compression module, a semantic encoding module, an infrared feature encoding module, an alignment pooling module, a cross-layer attention fusion module, a guided gating module, a multi-scale splicing module, and a multi-head decoder.

[0065] The aforementioned temporal compression module can be composed of a one-dimensional convolutional layer, a one-dimensional depthwise separable convolutional layer, a one-dimensional point convolutional layer, and a temporal pooling layer connected sequentially. The one-dimensional convolutional layer consists of a 3×3 convolutional kernel with padding of 1, a normalization layer, and a GeLU activation function. The one-dimensional depthwise separable convolutional layer uses a 3×3 convolutional kernel with padding of 1. The one-dimensional point convolutional layer uses a 1×1 convolutional kernel and forms a residual connection with the output of the one-dimensional convolutional layer. The temporal pooling layer can be an attention pooling layer. The aforementioned semantic encoding module can be composed of a normalization layer, a first linear layer, a GeLU activation function, a second linear layer, and a Dropout layer connected sequentially, with temporal feature encoding information s1 as the output.

[0066] The aforementioned infrared feature encoding module takes the in-pile temperature distribution data, including the time-series features of the hierarchical storage environment, as input, and includes a trunk encoding unit and a top-down fusion unit. The trunk encoding unit is composed of a first trunk block, a second trunk block, and a third trunk block connected sequentially. The first trunk block consists of a two-dimensional convolutional layer with a 3×3 kernel, a stride of 2, and padding of 1, a two-dimensional normalization layer (BatchNorm2D layer), and a SiLU activation function, taking the in-pile temperature distribution data as input and outputting P1. The second trunk block consists of a DCNv2 (Deformable Convolutional Networks v2) deformable convolutional layer with a 3×3 kernel, a stride of 2, and padding of 1, an SE attention module, and a two-dimensional convolutional layer with a stride of 2, taking P1 as input and outputting P2. The aforementioned third backbone block consists of a DCNv2 (Deformable Convolutional Networks v2) deformable convolutional layer with a 3×3 kernel, a stride of 2, and padding of 1, an SE attention module, and a 2D convolutional layer with a stride of 2. It takes P2 as input and P3 as output. The top-down fusion unit first upsamples P3 and concatenates it with P2 along the channel dimension. The concatenated information is then processed by a 2D convolutional layer with a 3×3 kernel, a BN layer, and a SiLU activation function to obtain F2. Next, F2 is upsampled and concatenated with P1 along the channel dimension. The concatenated information is then processed by a 2D convolutional layer with a 3×3 kernel, a BN layer, and a SiLU activation function to obtain F1. Finally, P3 is convolved by a 2D convolutional layer with a 3×3 kernel to obtain F3. The three-scale feature maps of F1, F2, and F3 are then used as the output of the aforementioned infrared feature encoding module. The alignment pooling module described above may include a geometric mapping and ROI indexing unit, used to map the physical region of each layer to the corresponding coordinates of the three-scale feature maps. The geometric mapping and ROI indexing unit can perform two-dimensional adaptive average pooling (AdaptiveAvgPool2D) on the ROI at each scale, and apply linear layer and layer normalization (i.e., LayerNorm) processing to the pooling results. The alignment pooling module can take the three-scale feature maps F1, F2, and F3 as input and output three sets of infrared feature vectors R1, R2, and R3 arranged by layer.

[0067] The aforementioned cross-layer attention fusion module takes the temporal feature encoding information s1 and the heat vector R2 as inputs and the fusion feature U0 as output. It comprises a concatenation unit, a linear projection layer, a multi-head self-attention layer, and a feed-forward network (FFN). The concatenation unit aligns the temporal feature encoding information s1 and the heat vector R2 along the channel dimension as the fusion input. The multi-head self-attention layer employs a multi-head self-attention mechanism and is modeled along the layer axis Z. The feed-forward network consists of a residual connection between the output of the concatenation unit and the input of the multi-head self-attention layer, and a linear layer. The output of the multi-head self-attention layer also has a residual connection with the input of the feed-forward network, and a linear layer.

[0068] The aforementioned guiding gating module takes the fused feature U0 and the three-scale infrared feature vectors R1, R2, and R3 as inputs and outputs the main feature U1. First, the guiding gating module concatenates the infrared feature vectors (R1, R2, R3) corresponding to each layer and obtains the scalar gate value gIR(z) for each layer axis through a fully connected network. Here, gIR(z) = Sigmoid(W2.GeLU(W1.LN(IRz))). IRz can be the concatenation result of the infrared feature vectors (R1, R2, R3) in a certain channel. LN() refers to layer normalization. W1 and W2 are two linear layers. GeLU() and Sigmoid() refer to the GeLU activation function and the Sigmoid activation function, respectively. Then, the guiding gating module can adjust the fused feature U0 of the corresponding layer using the scalar gate value gIR(z) for each layer axis. Specifically, the fusion feature U0 can be multiplied channel by channel by the adjusted scalar threshold gIR(z) of each layer axis to obtain the main feature U1. Specifically, the amplification factor (usually between 0.5 and 1) can be multiplied by the scalar threshold gIRz of each layer axis and then increased by 1 to adjust the scalar threshold gIRz.

[0069] The aforementioned multi-scale stitching module takes the main feature U1 and the three-scale infrared feature vectors R1, R2, and R3 as inputs and outputs the stitched feature H1. It is composed of stitching units, a normalization layer, a gated sensing network, and a dropout layer connected sequentially. The stitching unit aligns the main feature U1 and the three-scale infrared feature vectors R1, R2, and R3 in parallel along the channel dimension to obtain the initial stitched feature H0. The gated sensing network can be a Sigmoid-gated MLP network. The initial stitched feature H0 is processed sequentially by the normalization layer, the gated sensing network, and the dropout layer to obtain the stitched feature H1.

[0070] The aforementioned multi-head decoder comprises seven parallel decoding heads, each corresponding to one of seven storage states (i.e., the aforementioned heat accumulation state, the aforementioned moisture accumulation state, the aforementioned fermentation activity state, the aforementioned condensation state, the aforementioned mold tendency state, the aforementioned hotspot distribution state, and the aforementioned carbon-oxygen state). Specifically, the input to the k-th decoding head is the aforementioned concatenated feature H1. Each decoding head can be composed of a normalization layer, a first linear layer (64 dimensions reduced to 32 dimensions) and a GeLU activation function, a second linear layer (32 dimensions reduced to 1 dimension) and a Sigmoid function, ultimately outputting the corresponding layer-by-layer confidence as the corresponding state parameter. The outputs of the seven decoding heads are concatenated in parallel along the layer axis to form the final output hierarchical tobacco leaf storage state. Furthermore, the seven decoding heads can share the first linear layer, while the weights of the remaining layers are independent of each other.

[0071] The feature fusion module performs cross-modal alignment and fusion on the features extracted from the two branches to form a unified environment state code containing spatiotemporal semantics. Based on the infrared feature coding information (i.e., the three-scale feature maps F1, F2, F3) and the alignment pooling module, aligned infrared feature coding information (i.e., the three-scale infrared feature vectors R1, R2, R3) is generated. For example, the infrared feature coding information can be input into the alignment pooling module to obtain aligned infrared feature coding information. Based on the hierarchical storage environment temporal features and the temporal compression module, compressed storage environment temporal features are generated. For example, the hierarchical storage environment temporal features can be input into the temporal compression module to obtain compressed storage environment temporal features. Based on the compressed storage environment temporal features and the semantic coding module, temporal feature coding information, i.e., temporal feature coding information s1, is generated. For example, the compressed storage environment temporal features can be input into the semantic coding module to obtain temporal feature coding information. Based on the aligned infrared feature encoding information, the temporal feature encoding information, and the cross-layer attention fusion module, fused encoding information, i.e., fused feature U0, is generated. For example, the aligned infrared feature encoding information and the temporal feature encoding information can be input into the cross-layer attention fusion module to obtain the fused encoding information. Based on the fused encoding information and the multi-scale stitching module, stitched encoding information, i.e., stitched feature H1, is generated. For example, the fused encoding information can be input into the multi-scale stitching module to obtain the stitched encoding information.

[0072] The environmental state code is input into the multi-task decoding head, and the quantitative scores of each level on multiple predetermined storage state dimensions are output in parallel as tobacco leaf storage state information.

[0073] In one embodiment, based on tobacco storage status information and combined with preset state-policy mapping rules, a local ventilation control strategy for each storage area is generated, specifically including the following steps: The multiple state scores of each level in the tobacco storage state information are aggregated according to a preset inter-layer influence weight to obtain a comprehensive state vector of the storage area. The aforementioned storage state triggers include a first state trigger, a second state trigger, a third state trigger, a fourth state trigger, and a fifth state trigger. These triggers are used to determine the upper layer. In practice, the executing entity can use the reference weights corresponding to different tobacco storage layers under each storage state to perform a weighted summation of the state parameter values ​​corresponding to each storage state in the tobacco storage state information, and determine the weighted summation results as the regional storage state characteristics. For example, in a three-layer storage rack, the reference weight coefficients corresponding to each tobacco storage layer under heat accumulation state can be T = (0.1, 0.3, 0.6), representing a heat accumulation state reference weight of 0.1 for the lower storage layer, 0.3 for the middle storage layer, and 0.6 for the upper storage layer. The thermal accumulation state parameter values ​​for the upper, middle, and lower tobacco leaf storage states in the above tobacco leaf storage state information are 0.81, 0.64, and 0.25, respectively. It can be determined that the thermal accumulation state parameter value for the regional storage state characteristics is 0.703, which indicates that there is thermal accumulation in the tobacco leaves stored in the corresponding storage area.

[0074] The integrated state vector is input to a policy trigger consisting of multiple logic decision units. Specifically,

[0075] Each logic decision unit is configured to identify a specific composite storage exception pattern and output a trigger strength value.

[0076] Specifically, the aforementioned regional storage state characteristics are input to the aforementioned first state trigger to generate a first state value. This first state value characterizes whether there is upper-layer heat and moisture accumulation in the tobacco leaves stored on the corresponding storage area's shelves. In practice, the aforementioned first state trigger can generate the first state value using a sigmoid function (W1×S1+W2×S2+W3×△T+W4×△RH+W5×Hotpoint(t)+b1). Here, S1 and S2 are the heat accumulation state parameter value and moisture accumulation state parameter value in the aforementioned regional storage state characteristics, respectively. △T and △RH are the temperature difference and humidity difference between the top and bottom tobacco storage layers within the corresponding storage area, respectively. Hotpoint(t) can be the presence of a hot spot in the aforementioned pile temperature distribution data at the current time t, i.e., a temperature value exceeding a temperature threshold. For example, the temperature threshold can be 30 degrees Celsius. When a hot spot exists in the aforementioned pile temperature distribution data, Hotpoint(t) can be 1; otherwise, it is 0. b1 can be a bias term. The parameters W1, W2, W3, W4, and W5 above represent the weighting weights for different parameters in determining the superposition of heat and humidity. The sigmoid() function can be represented as described above.

[0077] The aforementioned regional storage state characteristics are input into the aforementioned second state trigger to generate a second state value. This second state value indicates whether there is moisture accumulation at the bottom of the tobacco leaves in the corresponding storage area's shelving. In practice, the aforementioned second state trigger can generate the second state value using a sigmoid function (W6×S2+W7×S4+W8×S7+W9×(-△RH)+W10×(-△O2)+b2). ​​Here, S4 and S7 are the condensation state parameter value and the carbon-oxygen state parameter value, respectively, in the aforementioned regional storage state characteristics. △RH and △O2 are the humidity difference and oxygen concentration difference between the top and bottom tobacco storage layers in the corresponding storage area, respectively. When the humidity of the bottom tobacco storage layer is higher, (-△RH) is positive. When the oxygen concentration of the bottom tobacco storage layer is higher, (-△O2) is positive. b2 can be a bias term. W6, W7, W8, W9, and W10 are the weighted weights of different parameters for determining the moisture accumulation at the bottom.

[0078] The storage state characteristics of the aforementioned region are input into the third state trigger to generate a third state value. This third state value indicates whether fermentation gases are rising in the tobacco leaves stored on the corresponding storage shelves. In practice, the third state trigger can generate the third state value using a sigmoid function (W11×S3+W12×S7+W13×S1+W14×ΔCO2+W15×δCO2+W16×(-δO2)+b3). Here, S3 is the fermentation activity state parameter value in the storage state characteristics of the aforementioned region. ΔCO2 is the carbon dioxide concentration difference between the top and bottom tobacco storage layers in the corresponding storage area. δCO2 and (-δO2) are the rates of increase in carbon dioxide concentration and decrease in oxygen concentration between the top and bottom tobacco storage layers in the corresponding storage area, respectively. b3 can be a bias term. W11, W12, W13, W14, W15, and W16 are the weighted weights of different parameters for determining the rise of fermentation gases.

[0079] The aforementioned regional storage state characteristics are input into the fourth state trigger to generate a fourth state value, which characterizes whether the tobacco leaves in the storage shelves of the corresponding storage area experience a continuous temperature increase. In practice, the fourth state trigger can generate the fourth state value using sigmoid(W17×S1+W18×S6+W19×δT+W20×avgT+W21×Hotarea+b4). Here, S6 is the hotspot distribution state parameter value in the aforementioned regional storage state characteristics. ΔCO2 is the carbon dioxide concentration difference between the top and bottom tobacco storage layers in the corresponding storage area. δT and avgT are the temperature change rate between the top and bottom tobacco storage layers in the corresponding storage area and the average temperature within the hotspot region in the pile temperature distribution data, respectively. Hotspot region can be the number of hotspot regions in the pile temperature distribution data. The hotspot region can refer to the region formed by connecting all temperatures greater than or equal to the aforementioned temperature threshold (the number of temperatures greater than the threshold). b4 can be a bias term. The above W17, W18, W19, W20, and W21 are the weighted weights of different parameters for judging whether the temperature continues to rise.

[0080] The storage state characteristics of the aforementioned region are input into the fifth state trigger to generate a fifth state value, which characterizes whether water vapor condensation exists in the tobacco leaves on the storage shelves of the corresponding storage region. In practice, the fifth state trigger can generate the fifth state value using sigmoid(W22×S4+W23×S2+W24×S5+W25×avgRH+W26×△Tm+b5). Here, S5 is the mold tendency state parameter value in the aforementioned storage state characteristics. avgRH is the average humidity of each tobacco storage layer in the corresponding storage region. avgRH is the temperature difference between the bottom tobacco storage layer and the determined dew point temperature in the corresponding storage region. b5 can be a bias term. W22, W23, W24, W25, and W26 are weighted weights for different parameters in determining the continuous temperature increase.

[0081] A vector composed of multiple trigger intensity values ​​output by the strategy trigger is converted into a specific device control parameter vector through a learnable strategy mapping matrix, forming a local ventilation control strategy. The device control parameters include at least the area air supply intensity, air supply direction angle, local exhaust intensity, and dehumidification intensity. The area ventilation strategy includes four parameters: Qa air supply intensity, Qd airflow direction, Qh exhaust volume, and Qc dehumidification intensity. Qa air supply intensity refers to the fan speed level, used to control the fan speed setting (e.g., 30%, 60%, 90%) within the corresponding storage area. Qd airflow direction indicates the deflection direction of the fan's guide vane or damper, used to control the air supply direction of the fan or damper in the corresponding storage area (e.g., -1 to 0 downward airflow, 0 to 1 upward airflow; the larger the absolute value, the larger the deflection angle). The Qh exhaust volume indicates the operating intensity of the exhaust vent or top-level exhaust fan, used to control the fan speed level of the exhaust fan or top-level exhaust fan in the corresponding storage area (e.g., less than 0.4 indicates low speed, 0.4 to 0.7 indicates medium speed, and greater than 0.7 indicates high speed). The Qc dehumidification intensity indicates the operating intensity of the dehumidification equipment, used to control the operating power of the dehumidification equipment in the corresponding storage area (e.g., less than 0.4 indicates low speed, 0.4 to 0.7 indicates medium speed, and greater than 0.7 indicates high speed).

[0082] In practice, firstly, the aforementioned executing entity can determine the first, second, third, fourth, and fifth state values ​​as the region state vector. For example... Figure 3The first, second, third, fourth, and fifth state values ​​mentioned above can be represented by E1, E2, E3, E4, and E5, respectively, to form the area state vector 303. Then, the executing entity can multiply the pre-constructed mapping matrix M with the area state vector (as a column vector), and process the result using the sigmoid function to obtain the area ventilation strategy. The mapping matrix has a size of 4×5. This mapping matrix is ​​used to map the area state vector to the area ventilation strategy. For example, in the area state vector, E1, E2, E3, E4, and E5 are 0.8, 0.3, 0.7, 0.6, and 0.2, respectively. The resulting area ventilation strategy includes Qa=0.78 (high-level air supply), Qd=0.65 (upward airflow at a large angle), Qh=0.8 (high-level exhaust volume), and Qc=0.25 (low-power dehumidification).

[0083] In one embodiment, a global ventilation strategy for the warehouse is generated by performing collaborative analysis and conflict resolution based on the local ventilation control strategies of all storage areas and the priority weights of each storage area. This includes the following steps: Each storage area is assigned a dynamic priority weight, which is determined based on at least one of the following factors: the varietal value of the stored tobacco leaves, the severity of the current condition, and the frequency of historical problems.

[0084] Based on the local ventilation control strategy and dynamic priority weight of each storage area, the basic control parameters of the global ventilation equipment are calculated in a weighted manner.

[0085] Identify the dominant storage risk types that are prevalent or reach a certain severity level across all regions.

[0086] If a dominant storage risk type exists, the basic control parameters are modified based on the global control rules corresponding to the risk type to generate the final warehouse global ventilation strategy.

[0087] Specifically, the region priority for each storage area is determined. In practice, different storage areas within a warehouse may contain different varieties of tobacco leaves, some of which may be more sensitive to temperature and humidity. Therefore, a corresponding region priority is set for each storage area. This allows for ventilation operations to be performed on the storage area with higher region confidence and priority when there are differences in storage conditions between two storage areas, such as heat accumulation on the sunny side while the shady side does not require additional ventilation intervention. The region priority values ​​range from [1, 3], with higher values ​​indicating higher priority.

[0088] Based on the corresponding regional priorities, the generated regional status information is identified to generate global warehouse storage status information. In practice, the aforementioned execution entity can determine the corresponding high-value (i.e., >0.7) status parameter values ​​in each regional status information. Then, according to the status type, the number of high-value status parameter values ​​under each status type is determined as the number of risk states. The number of high-value values ​​for that status type in each regional status information is counted according to the corresponding regional priority. For example, a regional priority of 3 is recorded as 3 high-value values ​​for that status type. Then, the determined number of risk states is grouped into three groups: wet accumulation state, condensation state, and mold tendency state; heat accumulation state, fermentation activity state, and hotspot distribution state; and carbon-oxygen state is grouped separately. The storage status tags corresponding to the maximum number of risk states after merging are then used as the global warehouse storage status information.

[0089] It should be noted that the ventilation operations governed by each state group are different. The state groups corresponding to the aforementioned moisture accumulation, condensation, and mold-prone states will primarily involve dehumidification throughout the warehouse. The state groups corresponding to the aforementioned heat accumulation, fermentation activity, and hotspot distribution states will primarily involve supply air throughout the warehouse. The state groups corresponding to the aforementioned carbon-oxygen states will primarily involve exhaust air throughout the warehouse.

[0090] For each of the aforementioned regional ventilation strategies, a strategy write-back is performed to update the regional ventilation strategy. In practice, the implementing entity can improve the ventilation strategy represented by the aforementioned regional ventilation strategy according to the dominant operation corresponding to the aforementioned regional status information. For example, if the dominant operation corresponding to the aforementioned regional status information is air supply operation, and Qa in the aforementioned regional ventilation strategy represents low to medium air supply, then the Qa value is increased by a preset percentage. As an example, the preset percentage could be 15%. Qa represents air supply intensity, Qd represents airflow direction, Qh represents exhaust volume, and Qc represents dehumidification intensity.

[0091] Based on the updated ventilation strategies for each area and the aforementioned global storage status information, a warehouse ventilation strategy is generated. In practice, for the supply air intensity Qa, the executing entity can determine the supply air level represented by the ventilation strategy for each area, using the average of these supply air levels as the supply air level for the warehouse ventilation strategy. The setting of dehumidification intensity Qc and exhaust volume Qh follows the same method as determining the supply air intensity Qa, and will not be elaborated further here. For the Qd airflow direction, the warehouse ventilation equipment uses a default polling method for airflow guidance and supply. This default polling method means that the warehouse ventilation equipment adjusts the airflow guidance and supply direction at regular intervals, rotating the airflow guidance and supply in each direction.

[0092] In one embodiment, based on local ventilation control strategies and warehouse-wide ventilation strategies, the independent ventilation equipment deployed in each storage area and the centralized ventilation equipment of the entire warehouse are controlled respectively to perform precise ventilation operations, specifically including the following steps: For each storage area, under the premise of meeting the safety operation constraints of the area equipment, the opening and angle of the independent air valves, the speed of the circulating fan, and the working level of the dehumidifier in each area are adjusted according to the local ventilation control strategy.

[0093] Under the premise of meeting the overall energy consumption and wind pressure balance constraints of the entire warehouse system, the operating parameters of the main supply fan, main exhaust fan and centralized dehumidification system of the warehouse are adjusted according to the overall ventilation strategy of the warehouse.

[0094] Among them, the execution priority of the local ventilation control strategy is higher than that of the warehouse global ventilation strategy. However, when the two conflict in terms of resource usage, the constraints of the warehouse global ventilation strategy shall prevail for coordination.

[0095] Specifically, the aforementioned implementing entity can execute ventilation equipment control operations corresponding to the ventilation strategies for each of the aforementioned areas and the ventilation strategies for the aforementioned warehouse through the associated ventilation equipment in each area and warehouse.

[0096] The aforementioned implementing entity can execute ventilation equipment control operations corresponding to the ventilation strategies for each of the aforementioned areas and the aforementioned warehouse ventilation strategies through the following steps: The first step is to adjust the equipment parameters of the associated ventilation equipment for each of the aforementioned ventilation strategies, based on the ventilation constraints and the equipment adjustment method represented by the ventilation strategy. These ventilation constraints can be conditions that limit the upper and lower limits of the equipment parameter adjustments for the ventilation equipment. For example, a ventilation strategy for a storage area might have Qa=0.75, Qd=+0.6, Qh=0.8, and Qc=0.25. The ventilation constraints could be an upper limit of 0.8 for the supply fan, an upper limit of 0.9 for the exhaust fan, and a minimum dehumidification value of 0.2. Then, the executing entity can read from the controller that the output power of the supply fan in the storage area is 60%, and Qa=0.75 satisfies the ventilation constraints. The frequency converter of the supply fan is then set to 75%, increasing the airflow speed by approximately 15%. The current air valve guide plate angle is 0° (horizontal), and the maximum adjustment angle is 30 degrees. Q... d =+0.6 corresponds to a target upward blowing angle of 18°. Rotate the control valve mechanism by 18° to deflect the airflow upwards. The current exhaust fan speed is medium (0.6), Qh=0.8, meeting the above ventilation limits. Adjust the exhaust fan frequency to increase the top exhaust rate. The current dehumidifier power is 0.2 (low setting), Qc=0.25, meeting the above ventilation limits. Adjust the dehumidifier power to 0.25.

[0097] The second step involves adjusting the equipment parameters of the relevant warehouse ventilation equipment according to the overall warehouse ventilation restrictions and the equipment adjustment method represented by the aforementioned warehouse ventilation strategy. These overall warehouse ventilation restrictions can be conditions that limit the upper and lower limits of the equipment parameter adjustments for the warehouse ventilation equipment. For example, the equipment adjustment method represented by the aforementioned warehouse ventilation strategy might be: main supply air setting (Qa) 0.8, main return air setting (Qh) 0.7, main dehumidification setting (Qc) 0.65, and a "bottom supply, top exhaust" airflow mode. The overall warehouse ventilation restrictions would have an upper limit of 0.8 for the supply fan, an upper limit of 0.8 for the exhaust fan, and a minimum dehumidification value of 0.4. Then, the executing entity can read through the controller that the output power of the supply fan in the warehouse ventilation equipment is 40%, and Qa=0.7 satisfies the aforementioned overall warehouse ventilation restrictions, setting the frequency converter of the supply fan to 70%. This adjusts the main air duct system in the warehouse to a bottom supply, top exhaust mode. The current exhaust fan speed in the warehouse is at low (0.3), Qh=0.7, which meets the above-mentioned ventilation limit conditions for the entire warehouse. Adjust the frequency of the exhaust fan to high to increase the exhaust rate from the top. The current dehumidifier power is 0.5 (medium), Qc=0.65, which meets the above-mentioned ventilation limit conditions. Adjust the dehumidifier power to 0.65 (still at medium speed, but with increased power).

[0098] Device Examples According to embodiments of the present invention, an intelligent ventilation control device for a tobacco storage warehouse is provided, such as... Figure 5 The diagram shown is a structural schematic of the intelligent ventilation control device for a tobacco storage warehouse provided in this embodiment. The intelligent ventilation control device for a tobacco storage warehouse according to this embodiment includes: The environmental data acquisition module 51 is used to trigger environmental data acquisition based on a preset period. Through multi-dimensional sensing components deployed in different storage areas of the warehouse and different vertical levels of the storage racks in each storage area, it collects environmental parameter sets of each storage area and each level, forming a tobacco leaf storage environment data set corresponding to each storage rack.

[0099] The local strategy generation module 52 is used to sequentially perform data preprocessing, feature construction, state recognition and strategy generation operations for each tobacco storage environment data group to generate a local ventilation control strategy.

[0100] The global policy generation module 53 is used to perform collaborative analysis and conflict resolution based on the local ventilation control policies of all storage areas and the priority weights of each storage area to generate a global ventilation policy for the warehouse.

[0101] The ventilation execution module 54 is used to control the independent ventilation equipment deployed in each storage area and the centralized ventilation equipment of the entire warehouse based on the local ventilation control strategy and the warehouse global ventilation strategy, so as to perform precise ventilation operations.

[0102] The device provided in this embodiment collects environmental data based on a preset cycle. Through multi-dimensional sensing components deployed in different storage areas and vertical levels of the storage shelves within each storage area, it collects environmental parameter sets for each storage area and level, forming a data set of tobacco storage environment data corresponding to each storage shelf. By deploying multi-dimensional sensors in each storage area and on each shelf level and periodically collecting data synchronously, it completely solves the problems of large blind spots and inability to reflect spatial differences in traditional single-point monitoring, providing a comprehensive and reliable data foundation for precise control. For each tobacco storage environment data set, it sequentially performs data preprocessing, feature construction, state recognition, and strategy generation operations to generate a local ventilation control strategy. It performs serialization processing and AI analysis on the data of each shelf, achieving a deep diagnosis from raw data to the "tobacco storage state," and automatically generating quantitative data accordingly. The local ventilation strategy achieves an intelligent leap from "environmental perception" to "preliminary decision-making." Based on the local ventilation control strategies of all storage areas and the priority weights of each storage area, a collaborative analysis and conflict resolution are performed to generate a global ventilation strategy for the warehouse. By integrating all local strategies and resolving conflicts and scheduling resources according to regional priorities, a coordinated global strategy is generated, resolving system conflicts that may be caused by zoning control and ensuring the stability and efficiency of overall operation. Based on the local ventilation control strategy and the global ventilation strategy, the independent ventilation equipment deployed in each storage area and the centralized ventilation equipment of the entire warehouse are controlled respectively to execute precise ventilation operations. The strategy is decomposed and distributed to the corresponding regional equipment and centralized equipment to drive their coordinated actions, ultimately achieving precise ventilation of "zoning, layering, and differentiation," forming a complete automated closed loop of "perception-decision-execution."

[0103] The embodiments of the present invention are device embodiments corresponding to the above method embodiments. The specific operations of each module processing step can be understood with reference to the description of the method embodiments, and will not be repeated here.

[0104] like Figure 6 As shown, the present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it implements the intelligent ventilation control method for the tobacco storage warehouse in the above embodiments, or when the computer program is executed by a processor, it implements the intelligent ventilation control method for the tobacco storage warehouse in the above embodiments.

[0105] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0106] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for apparatus or system embodiments, since they are basically similar to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. The apparatus and system embodiments described above are merely illustrative. Units described as separate components may or may not be physically separate. Components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0107] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and the contents not described in detail in the specification of the present invention are known to those skilled in the art.

Claims

1. A method for intelligent ventilation control in a tobacco storage warehouse, characterized in that, Includes the following steps: Based on the preset cycle of triggering environmental data collection, multi-dimensional sensing components deployed in different storage areas of the warehouse and different vertical levels of the storage racks in each storage area are used to collect environmental parameter sets of each storage area and each level, forming a tobacco leaf storage environment data set that corresponds to each storage rack. For each of the tobacco storage environment data groups, data preprocessing, feature construction, state recognition and strategy generation operations are performed sequentially to generate a local ventilation control strategy; Based on the local ventilation control strategies of all storage areas and the priority weights of each storage area, a collaborative analysis and conflict resolution are performed to generate a global ventilation strategy for the warehouse. Based on the local ventilation control strategy and the warehouse global ventilation strategy, the independent ventilation equipment deployed in each storage area and the centralized ventilation equipment of the warehouse as a whole are controlled respectively to perform precise ventilation operations.

2. The intelligent ventilation control method for a tobacco storage warehouse as described in claim 1, characterized in that, For each group of tobacco storage environment data, data preprocessing, feature construction, state identification, and strategy generation operations are performed sequentially to generate a local ventilation control strategy, specifically including the following steps: The data set of tobacco storage environment is preprocessed and features are constructed to generate multi-dimensional environmental features that characterize the overall and layered environmental conditions of the shelves. The multidimensional environmental features are input into a pre-trained tobacco storage state recognition model, and the tobacco storage state information corresponding to each shelf level is output. The tobacco storage state information includes at least the heat accumulation state, the moisture accumulation state, the fermentation activity state, and the condensation risk state. Based on the tobacco storage status information and combined with the preset state-policy mapping rules, a local ventilation control strategy for each storage area is generated.

3. The intelligent ventilation control method for a tobacco storage warehouse as described in claim 2, characterized in that, The process of preprocessing and feature construction of the tobacco storage environment data set to generate multidimensional environmental features characterizing the overall and layered environmental conditions of the shelving includes the following steps: Sensor type splitting, data calibration, and outlier handling were performed on the raw environmental data set; For each level of the shelf, extract directly observable features including temperature, humidity, and gas concentration within the shelf from the sensor data. Calculate the dew point temperature and relative humidity saturation characteristics of the current layer based on the temperature and humidity within the layer; Based on infrared temperature distribution data, spatial thermal distribution characteristics that characterize the temperature uniformity and hot spot distribution inside the smoke pile are extracted. Calculate the gradient differences in temperature, humidity, and gas concentration between adjacent vertical layers to form the interlayer environmental gradient characteristics; By combining current data with historical data from the same period, the changing trends of key environmental parameters are calculated to form time-series variation characteristics; All features at the same level are concatenated to form a hierarchical feature vector, and the hierarchical feature vectors of all levels of the shelf are concatenated in spatial order to form the multidimensional environmental features.

4. The intelligent ventilation control method for a tobacco storage warehouse as described in claim 3, characterized in that, The process of inputting the multidimensional environmental features into a pre-trained tobacco storage status recognition model and outputting tobacco storage status information corresponding to each shelf level specifically includes the following steps: The multidimensional environmental features received as input, wherein the tobacco storage state recognition model is a multi-task learning model based on a deep neural network; The first feature extraction branch encodes the hierarchical observation features arranged in time sequence to capture the temporal dependencies of environmental states. The spatial heat distribution features are encoded through the second feature extraction branch to capture the two-dimensional thermal field spatial pattern inside the smoke pile. The feature fusion module performs cross-modal alignment and fusion on the features extracted from the two branches to form a unified environment state code that includes spatiotemporal semantics. The environmental state code is input into the multi-task decoding head, and the quantitative scores of each level on multiple predetermined storage state dimensions are output in parallel as the tobacco leaf storage state information.

5. The intelligent ventilation control method for a tobacco storage warehouse as described in claim 2, characterized in that, The step of generating a local ventilation control strategy for each storage area based on the tobacco storage status information and a preset state-policy mapping rule includes the following steps: The multiple state scores of each level in the tobacco storage state information are aggregated according to the preset inter-level influence weights to obtain the comprehensive state vector of the storage area. The integrated state vector is input to a policy trigger consisting of multiple logic judgment units; Each of the aforementioned logical judgment units is configured to identify a specific composite storage anomaly pattern and output a trigger strength value; The vector composed of multiple trigger strength values ​​output by the strategy trigger is converted into a specific device control parameter vector through a learnable strategy mapping matrix, thus forming the local ventilation control strategy. The equipment control parameters include at least the regional air supply intensity, air supply direction angle, local exhaust air intensity, and dehumidification intensity.

6. The intelligent ventilation control method for a tobacco storage warehouse as described in claim 1, characterized in that, The process of generating a global warehouse ventilation strategy based on the local ventilation control strategies of all storage areas and the priority weights of each storage area through collaborative analysis and conflict resolution includes the following steps: Each storage area is assigned a dynamic priority weight, which is determined based on at least one of the following factors: the varietal value of the stored tobacco leaves, the severity of the current condition, and the frequency of historical problems. Based on the local ventilation control strategy and dynamic priority weight of each storage area, the basic control parameters of the global ventilation equipment are calculated in a weighted manner. Identify the dominant storage risk types that are prevalent or reach a certain severity level across all regions; If a dominant storage risk type exists, the basic control parameters are modified based on the global control rules corresponding to the storage risk type to generate the final warehouse global ventilation strategy.

7. The intelligent ventilation control method for a tobacco storage warehouse as described in claim 1, characterized in that, The method, based on the local ventilation control strategy and the warehouse global ventilation strategy, controls the independent ventilation equipment deployed in each storage area and the centralized ventilation equipment of the entire warehouse to perform precise ventilation operations, specifically including the following steps: For each storage area, under the premise of meeting the safety operation constraints of the area equipment, the opening and angle of the independent air valves, the speed of the circulating fan and the working level of the dehumidifier in each area are adjusted according to the local ventilation control strategy. Under the premise of meeting the overall energy consumption and wind pressure balance constraints of the entire warehouse system, the operating parameters of the main supply fan, main exhaust fan and centralized dehumidification system of the warehouse are adjusted according to the overall ventilation strategy of the warehouse. The local ventilation control strategy has a higher execution priority than the global warehouse ventilation strategy. However, when the two conflict in terms of resource usage, the constraints of the global warehouse ventilation strategy shall prevail for coordination.

8. An intelligent ventilation control device for a tobacco storage warehouse, characterized in that, include: The environmental data acquisition module is used to trigger environmental data acquisition based on a preset cycle. Through multi-dimensional sensing components deployed in different storage areas of the warehouse and different vertical levels of the storage racks in each storage area, it collects environmental parameter sets of each storage area and each level, forming a tobacco leaf storage environment data set that corresponds to each storage rack. The local strategy generation module is used to sequentially perform data preprocessing, feature construction, state recognition and strategy generation operations for each of the tobacco storage environment data groups to generate a local ventilation control strategy. The global strategy generation module is used to perform collaborative analysis and conflict resolution based on the local ventilation control strategies of all storage areas and the priority weights of each storage area to generate a global ventilation strategy for the warehouse. The ventilation execution module is used to control the independent ventilation equipment deployed in each storage area and the centralized ventilation equipment of the entire warehouse based on the local ventilation control strategy and the warehouse global ventilation strategy, and to perform precise ventilation operations.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the intelligent ventilation control method for the tobacco storage warehouse as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the intelligent ventilation control method for the tobacco storage warehouse as described in any one of claims 1 to 7.

Images