Tobacco leaf storage data early warning method and system

Abstract

The application discloses a kind of tobacco storage data early warning method and system, the method includes:1 by environmental sensing node obtains environmental parameter, and obtains tobacco bale core parameter by bale core embedded sensing node, environmental parameter and tobacco bale core parameter are preprocessed, obtain the environmental parameter after preprocessing, tobacco bale core parameter;2 calculate humidity risk index, tobacco moisture content and oxygen influence factor, calculate alcoholization rate, calculate basic threshold, calculate final threshold;3 train light weight LSTM model, obtain early warning model;4 calculate the moisture content of the tobacco to be processed, the bale core temperature and humidity of the tobacco to be processed, bale core oxygen concentration, moisture content and storage days of the tobacco to be processed are input into early warning model, obtain early warning threshold, and start hierarchical alarm according to early warning threshold.The application can solve the problems of the existing tobacco storage monitoring system, such as single monitoring dimension, static early warning mechanism, weak multi-source data fusion capability and lack of linkage control.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for early warning of tobacco leaf storage data. Background Technology

[0002] In cigarette production, tobacco leaves, as the core raw material, undergo a maturation and storage period lasting several months or even years to improve their sensory quality, combustion characteristics, and chemical stability. The maturation effect is highly dependent on the coordinated control of multiple parameters, including temperature and humidity, oxygen concentration in the storage environment, and temperature, humidity, and oxygen content inside the tobacco leaf wrapper. If these parameters deviate from their optimal range, problems such as mold growth, excessive oxidation, carbonization, or insufficient maturation can easily occur, leading to raw material spoilage and economic losses.

[0003] To ensure the safety of tobacco leaf storage, the industry has gradually introduced sensor monitoring and data acquisition technologies. However, existing technologies still have the following significant problems: (1) Incomplete monitoring dimensions: Most systems only focus on the temperature and humidity of the warehouse environment, lacking synchronous collection of key internal state parameters such as temperature, humidity and oxygen concentration inside the tobacco leaf core, making it difficult to accurately assess the actual aging state of the tobacco leaves.

[0004] (2) The early warning mechanism is static and isolated: fixed thresholds are generally used to trigger alarms, without combining dynamic factors such as tobacco quality, storage period, seasonal changes and historical aging curves for intelligent judgment, which is prone to false alarms or missed alarms.

[0005] (3) Weak ability to fuse multi-source data: Environmental parameters and core parameters are usually processed independently, lacking the ability to perform spatiotemporal alignment, correlation analysis and trend prediction of multi-channel heterogeneous data such as temperature, humidity and oxygen concentration.

[0006] (4) Lack of linkage control mechanism: Even if an anomaly is detected, the system often only provides alarm information and fails to form a closed loop linkage with storage control equipment such as ventilation, dehumidification and nitrogen filling, so it cannot achieve active intervention. Summary of the Invention

[0007] The present invention addresses the shortcomings of the existing technology by proposing a data early warning method and system for tobacco leaf storage. This system aims to achieve real-time data collection, intelligent analysis, dynamic early warning, and equipment linkage control, thereby improving the safety, controllability, and intelligence level of raw material storage. It also solves the problems of existing tobacco leaf storage monitoring systems, such as single monitoring dimensions, static early warning mechanisms, and lack of linkage control.

[0008] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: The present invention provides a data early warning method for tobacco leaf storage, characterized by the following steps: Step 1: Obtain environmental parameters, including ambient temperature and humidity, and ambient oxygen concentration, through environmental sensor nodes inside the tobacco storage warehouse; The parameters of the tobacco leaf wrapping are obtained through embedded sensing nodes at the center of the wrapping, including: wrapping temperature and humidity, and wrapping oxygen concentration. A filtering algorithm is used to preprocess the environmental parameters and tobacco leaf core-wrapping parameters, and the preprocessed environmental parameters and preprocessed tobacco leaf core-wrapping parameters are obtained accordingly. Step 2: Based on the pre-treated tobacco leaf core parameters, calculate the temperature and humidity risk index, tobacco leaf moisture content and oxygen influence factor, and calculate the aging rate according to the temperature and humidity risk index and oxygen influence factor. Then, calculate the basic threshold according to the aging rate and the pre-treated tobacco leaf core parameters, and finally calculate the final threshold based on the basic threshold and the deviation threshold. Step 3: Using the preprocessed tobacco leaf core parameters, tobacco leaf moisture content, and tobacco leaf storage days as inputs, and the final threshold as the output label, train a lightweight LSTM model to obtain the early warning model. Step 4: Obtain the core temperature and humidity, and core oxygen concentration of the tobacco leaves to be treated. Calculate the moisture content of the tobacco leaves to be treated based on the core temperature and humidity. Input the core temperature and humidity, core oxygen concentration, moisture content, and storage days of the tobacco leaves to be treated into the early warning model for processing to obtain the early warning threshold. Then, activate the graded alarm based on the early warning threshold.

[0009] The characteristic of the tobacco leaf storage data early warning method of the present invention is that, in step 1, the environmental parameters or tobacco leaf core parameters are preprocessed using formula (1) to obtain the current time. Environmental parameters or tobacco leaf core parameters after pretreatment : (1) In equation (1), Indicates the previous moment The corresponding pre-treated environmental parameters or tobacco leaf core parameters; when season For the current moment Collected environmental parameters or tobacco leaf core parameters, For the current moment Collected environmental parameters or tobacco leaf core parameters, Indicates the current time Kalman filter gain.

[0010] Furthermore: the current time can be obtained using equation (2). Kalman filter gain : (2) In equation (2), Indicates the current time The uncertainty of the estimate, and ,in, Indicates the previous moment The uncertainty of the estimate, when season =1; For process noise, It is a constant value.

[0011] Furthermore: Calculate the current time using equation (3). Temperature and humidity risk index : (3) In equation (3), For the current moment Pre-treated core temperature, For the current moment Humidity of the pre-treated core.

[0012] Furthermore: Calculate the current time using equation (4). Tobacco leaf moisture content : (4) In equation (4), The core temperature after pretreatment. , , All have 3 empirical coefficients.

[0013] Furthermore: Calculate the current time using equation (5). alcoholization rate : (5) In equation (5), Indicates the current time Oxygen-affecting factors, The number of days the tobacco leaves are stored.

[0014] Furthermore, in step 2, the final threshold is obtained through the following process: Calculate the current time using equation (7) Base threshold : (7) In equation (7), Indicates the current time Oxygen concentration in the treated core. Indicates the total collection time; Calculate the current time using equation (8) Deviation threshold : (8) In equation (8), Indicates the current time The pre-treated core temperature is the same as the previous moment. The change in core temperature after pretreatment Indicates the current time The humidity of the pre-treated core is the same as the previous moment. Changes in humidity of the pre-treated core package Indicates the current time The oxygen concentration of the pretreated core was the same as that at the previous time. The change in oxygen concentration in the core after pretreatment , , These represent the weighting coefficients for temperature, humidity, and oxygen concentration, respectively.

[0015] Calculate the current time final threshold .

[0016] Furthermore, in step 4, the tiered alarm is initiated according to the following process: When the warning threshold is lower than the first set threshold, a level one warning is activated, and an alarm message is pushed to the management terminal to prompt manual verification. When the warning threshold is greater than the first set threshold and less than the second set threshold, a level two warning is activated, and ventilation or dehumidification equipment is started. When the warning threshold is greater than the second set threshold but less than the third set threshold, a level three warning is activated, and the nitrogen filling equipment is started. No action is required when the warning threshold is greater than the third set threshold. The first set threshold is less than the second set threshold, and the second set threshold is less than the third set threshold.

[0017] The present invention provides a data early warning system for tobacco leaf storage, characterized by comprising: The data acquisition module acquires environmental parameters through environmental sensor nodes, including environmental temperature and humidity, and environmental oxygen concentration. Simultaneously, it acquires tobacco leaf core-wrapping parameters through core-embedded sensor nodes, including core-wrapping temperature and humidity, and core-wrapping oxygen concentration. The module then preprocesses the environmental parameters and tobacco leaf core-wrapping parameters to obtain preprocessed environmental parameters and tobacco leaf core-wrapping parameters. The data analysis module calculates the temperature and humidity risk index, tobacco moisture content, and oxygen impact factor based on the pre-treated tobacco leaf core parameters. It then calculates the aging rate based on the temperature and humidity risk index and oxygen impact factor, and calculates the basic threshold based on the aging rate and the pre-treated tobacco leaf core parameters. Finally, it calculates the final threshold based on the basic threshold and the deviation threshold. The training module takes the pre-processed tobacco leaf core parameters, tobacco leaf moisture content and tobacco leaf storage days as input, and the final threshold as the output label to train a lightweight LSTM model to obtain the early warning model. The early warning module acquires the core temperature and humidity, and core oxygen concentration of the tobacco leaves to be treated, and calculates the moisture content of the tobacco leaves based on the core temperature and humidity. Then, it inputs the core temperature and humidity, core oxygen concentration, moisture content, and number of days the tobacco leaves have been stored into the early warning model to obtain the early warning threshold, and initiates tiered alarms based on the early warning threshold.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention acquires environmental parameters through environmental sensing nodes and tobacco leaf core parameters through core-embedded sensing nodes, achieving simultaneous monitoring of four parameters: temperature, humidity, and oxygen, across two dimensions of tobacco storage environment and core. A lightweight LSTM model is trained using pre-processed tobacco leaf core parameters, tobacco leaf moisture content, tobacco leaf storage days, and a final threshold as training data. This trained LSTM model serves as an early warning model, predicting the early warning threshold for the tobacco leaves to be processed. This overcomes the limitations of fixed thresholds by introducing a dynamic model and trend prediction, significantly reducing false alarm and false negative rates. Finally, tiered alarms are activated based on the early warning threshold, realizing a closed-loop control system of "monitoring-early warning-response."

[0019] 2. This invention achieves synchronous and high-precision acquisition of multiple parameters such as ambient temperature and humidity and oxygen concentration during the tobacco leaf aging process by setting up ambient temperature and humidity sensors and ambient oxygen concentration sensors in the tobacco leaf storage warehouse, and setting up core temperature and humidity sensors and core oxygen concentration sensors in the center of the tobacco leaf core.

[0020] 3. This invention constructs a dual-dimensional, four-parameter homogeneous data acquisition system of "environment + core" to supplement and integrate data sources, and uses Kalman filtering for unified preprocessing to achieve noise reduction and normalization fusion of multi-source data; it also establishes a multi-parameter correlation calculation model to achieve feature-level deep fusion.

[0021] 4. The tobacco leaf storage data early warning system of this invention has strong scalability. The system architecture supports flexible access to new types of sensors or control devices and is suitable for raw material storage scenarios of cigarette enterprises of different sizes. Attached Figure Description

[0022] Figure 1 A flowchart of the tobacco leaf storage data early warning method provided in Embodiment 1 of the present invention; Figure 2 This is an architecture diagram of the lightweight LSTM model in the tobacco storage data early warning method provided in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the tobacco leaf storage data early warning system provided in Embodiment 2 of the present invention; Figure 4 This is an architecture diagram of the tobacco leaf storage data early warning system provided in Embodiment 3 of the present invention; Figure 5 This is a schematic diagram of the digital centralized control hub for raw material storage and maintenance of the tobacco storage data early warning system provided in Embodiment 3 of the present invention. Detailed Implementation

[0023] Example 1: like Figure 1 As shown, this embodiment provides a data early warning method for tobacco leaf storage. It is an integrated data early warning method for tobacco leaf aging and storage scenarios in cigarette manufacturing enterprises, supporting real-time acquisition, intelligent analysis, dynamic early warning, and equipment linkage control of multi-source sensor data (including environmental and core temperature, humidity, oxygen concentration, etc.). The method includes the following steps: Step 1: Set up environmental sensing nodes in the tobacco storage warehouse, including: environmental temperature and humidity sensors and environmental oxygen concentration sensors, so as to collect environmental parameters in the tobacco storage warehouse in real time through the environmental sensing nodes, including: environmental temperature and humidity, and environmental oxygen concentration. An embedded sensing node is set at the center of the tobacco leaf core, including a core temperature and humidity sensor and a core oxygen concentration sensor. The embedded sensing node acquires tobacco leaf core parameters, including the temperature and humidity of the tobacco leaf core and the oxygen concentration of the tobacco leaf core. The environmental parameters and tobacco leaf core parameters are preprocessed to obtain preprocessed environmental parameters and tobacco leaf core parameters. The environmental sensing node and the embedded sensing node upload the data to the edge computing gateway via a low-power wide area network (LoRa) to ensure high synchronization and low transmission latency.

[0024] By installing environmental temperature and humidity sensors and ambient oxygen concentration sensors in the tobacco storage warehouse, and core temperature and humidity sensors and core oxygen concentration sensors at the center of the tobacco leaf core, synchronous and high-precision data acquisition of multiple parameters, including environmental temperature and humidity and oxygen concentration, as well as the internal temperature and humidity and oxygen concentration of the tobacco leaf core, is achieved during the tobacco leaf aging process. This is the first time that simultaneous monitoring of four parameters—environmental and core dimensions, and temperature, humidity, and oxygen—has been achieved, truly reflecting the aging state of the tobacco leaves and enabling comprehensive perception.

[0025] A filtering algorithm is used to preprocess the environmental parameters and tobacco leaf core-wrapping parameters. The preprocessed environmental parameters and tobacco leaf core-wrapping parameters specifically include: Using equation (1) to preprocess environmental parameters or tobacco leaf core parameters, the current time is obtained. Environmental parameters or tobacco leaf core parameters after pretreatment : (1) In equation (1), Indicates the previous moment The corresponding pre-treated environmental parameters or tobacco leaf core parameters; when season For the current moment Collected environmental parameters or tobacco leaf core parameters, For the current moment Collected environmental parameters or tobacco leaf core parameters, Indicates the current time The Kalman filter gain is obtained from equation (2): (2) In equation (2), Indicates the current time The uncertainty of the estimate , Indicates the previous moment The uncertainty of the estimate, when season =1; This refers to process noise. Adjustments are made in real time based on prediction results, process noise. The range is 0.01-0.1. As a constant value, in this invention, the constant value is... The value range is 0.1-3, and the value in this embodiment is a fixed value. Take 1.

[0026] This invention obtains more accurate environmental parameters (ambient temperature and humidity, ambient oxygen concentration) and tobacco leaf core-coating parameters (core-coating temperature and humidity, core-coating oxygen concentration) through filtering, avoiding data abrupt changes and errors, and improving the accuracy of subsequent calculations.

[0027] Step 2: Calculate the temperature and humidity risk index, tobacco moisture content, and oxygen impact factor based on the pretreated tobacco leaf core-wrapping parameters. Calculate the aging rate based on the temperature and humidity risk index and oxygen impact factor. Calculate the basic threshold based on the aging rate and the pretreated tobacco leaf core-wrapping parameters. Calculate the final threshold based on the basic threshold and the deviation threshold. Specifically, this includes: Step 2.1: Calculate the current time using equation (3). Temperature and humidity risk index : (3) In equation (3), For the current moment Pre-treated core temperature, For the current moment Humidity of the pre-treated core.

[0028] Step 2.2: Calculate the current time using equation (4). Tobacco leaf moisture content : (4) In equation (4), The core temperature after pretreatment. , , Each has 3 empirical coefficients. , , Regular values ​​can be obtained through AI large-scale model data analysis, and then adjusted in real time through continuous data accumulation.

[0029] Step 2.3: Calculate the current time using equation (5). alcoholization rate : (5) In equation (5), Indicates the current time Oxygen-affecting factors, The number of days the tobacco leaves are stored, and the following information is provided: (6) In equation (6), Indicates the current time Oxygen concentration of the core after pretreatment; Step 2.4: Calculate the current time using equation (7). Base threshold : (7) In equation (7), Indicates the current time Oxygen concentration in the treated core. Indicates the total collection time; Step 2.5: Calculate the current time using equation (8). Deviation threshold : (8) In equation (8), Indicates the current time The pre-treated core temperature is the same as the previous moment. The change in core temperature after pretreatment Indicates the current time The humidity of the pre-treated core is the same as the previous moment. Changes in humidity of the pre-treated core package Indicates the current time The oxygen concentration of the pretreated core was the same as that at the previous time. The change in oxygen concentration in the core after pretreatment , , These represent the weighting coefficients for temperature, humidity, and oxygen concentration, respectively. , , Adjustments are made in real time based on large AI models and accumulated data analysis.

[0030] Step 2.6: Calculate the current time using equation (9). final threshold : (9) Step 3: Using the pre-processed tobacco leaf core parameters, tobacco leaf moisture content, and tobacco leaf storage days as inputs, and the final threshold as the output label, train a lightweight LSTM model. When the loss function is minimized, the trained lightweight LSTM model is obtained. The trained lightweight LSTM model is used as an early warning model to predict the early warning threshold of the tobacco leaves to be processed. It breaks through the fixed threshold limitation, introduces dynamic models and trend prediction, and significantly reduces the false alarm rate and false negative rate. The intelligent early warning lead time can reach 12 to 48 hours.

[0031] See Figure 2 The lightweight LSTM model in this embodiment includes a Dropout layer, a fully connected layer, and an output layer connected in sequence. The lightweight LSTM model used in this invention is small in size, has low computational cost, fast inference speed, strong real-time performance, simple deployment, and strong compatibility.

[0032] Step 4: Obtain the core temperature and humidity, and core oxygen concentration of the tobacco leaves to be treated. Calculate the moisture content of the tobacco leaves based on the core temperature and humidity. Input the core temperature and humidity, core oxygen concentration, moisture content, and storage days of the tobacco leaves to the early warning model to obtain the early warning threshold. Then, activate tiered alarms based on the early warning threshold, specifically including: When the warning threshold is lower than the first set threshold, a level one warning is activated, and an alarm message is pushed to the management terminal to prompt manual verification. When the warning threshold is greater than the first set threshold and less than the second set threshold, a level two warning is activated, and ventilation or dehumidification equipment is started. When the warning threshold is greater than the second set threshold but less than the third set threshold, a level three warning is activated, and the nitrogen filling equipment is started. No action is required when the warning threshold is greater than the third set threshold. The first set threshold is less than the second set threshold, and the second set threshold is less than the third set threshold. In this embodiment, the first set threshold is 50, the second set threshold is 80, and the third set threshold is 90.

[0033] In summary, this invention acquires tobacco leaf core-packing parameters (core-packing temperature and humidity, core-packing oxygen concentration), calculates the temperature and humidity risk index, tobacco leaf moisture content, and oxygen influencing factor, calculates the aging rate based on the temperature and humidity risk index and oxygen influencing factor, calculates a basic threshold based on the aging rate and pre-treated tobacco leaf core-packing parameters, and calculates the final threshold based on the basic threshold and the deviation threshold. Simultaneously, it utilizes an AI model (lightweight LSTM model) to compare, statistically analyze, and summarize daily real-time parameters, predicting subsequent risks. By detecting multiple environmental parameters (collected core-packing temperature and humidity, oxygen, etc.), it fully considers the impact of each environmental parameter, increases data accumulation and analysis, dynamically reflects the entire storage process, and adds a risk prediction function to assess storage risks in advance and make timely strategic adjustments, greatly improving the safety and stability of warehousing and minimizing the risk of material storage. The data collected by this invention, as well as the calculated parameters and thresholds, are stored in real-time on a server for subsequent retrieval.

[0034] This invention completes the data source fusion by constructing a dual-dimensional, four-parameter homogeneous data acquisition system of "environment + core," and adopts Kalman filtering for unified preprocessing to achieve noise reduction and normalization fusion of multi-source data. It also establishes a multi-parameter correlation calculation model to achieve feature-level deep fusion. At the same time, through spatiotemporal alignment and feature fusion of multi-source data, it improves the accuracy and robustness of anomaly identification, thereby realizing integrated closed-loop control of "perception-analysis-decision-execution," significantly improving the automation of warehouse management and emergency response capabilities.

[0035] See Figure 3 In this embodiment, a tobacco leaf storage data early warning system includes: The data acquisition module is used to acquire environmental parameters, including ambient temperature and humidity and ambient oxygen concentration, through environmental sensor nodes, and to acquire tobacco leaf core-wrapping parameters, including core-wrapping temperature and humidity and core-wrapping oxygen concentration, through core-wrapping embedded sensor nodes. The environmental parameters and tobacco leaf core-wrapping parameters are preprocessed to obtain preprocessed environmental parameters and tobacco leaf core-wrapping parameters.

[0036] The environmental sensing nodes include environmental temperature and humidity sensors and environmental oxygen concentration sensors, which are all installed in the tobacco storage warehouse. The core-embedded sensing nodes include core-temperature and humidity sensors and core-oxygen concentration sensors, which are all located at the center of the tobacco core.

[0037] The data analysis module is used to calculate the temperature and humidity risk index, tobacco moisture content, and oxygen influence factor based on the pre-treated tobacco leaf core parameters. It also calculates the aging rate based on the temperature and humidity risk index and oxygen influence factor, calculates the basic threshold based on the aging rate and the pre-treated tobacco leaf core parameters, and calculates the final threshold based on the basic threshold and the deviation threshold.

[0038] The training module is used to train a lightweight LSTM model with preprocessed tobacco leaf core parameters, tobacco leaf moisture content and tobacco leaf storage days as inputs and the final threshold as the output label to obtain the early warning model. The early warning module is used to obtain the core temperature and humidity and core oxygen concentration of the tobacco leaves to be treated, calculate the moisture content of the tobacco leaves to be treated based on the core temperature and humidity, input the core temperature and humidity, core oxygen concentration, moisture content and the number of days the tobacco leaves to be treated are stored into the early warning model to obtain the early warning threshold, and activate the graded alarm based on the early warning threshold.

[0039] The process of initiating tiered alarms based on warning thresholds includes: when the warning threshold is less than the first set threshold, a Level 1 warning is initiated, and alarm information is pushed to the management terminal, prompting manual verification; when the warning threshold is greater than the first set threshold but less than the second set threshold, a Level 2 warning is initiated, and ventilation or dehumidification equipment is activated; when the warning threshold is greater than the second set threshold but less than the third set threshold, a Level 3 warning is initiated, and nitrogen filling equipment is activated; when the warning threshold is greater than the third set threshold, no action is required, wherein the first set threshold is less than the second set threshold, and the second set threshold is less than the third set threshold.

[0040] The tobacco leaf storage data early warning system of this invention also includes a data storage module, a data transmission module, and a linkage module. The data storage module stores parameters collected by the data acquisition module and data calculated by the data analysis module. The data transmission module receives parameters collected by the data acquisition module, summarizes and formats the parameters, and reports them to the server (data analysis module) to complete the data reporting. The linkage module interfaces with the data analysis module via industrial communication protocols such as Modbus TCP and OPC UA. The linkage module includes a dehumidifier, a ventilator, and a nitrogen filling device. The dehumidifier, ventilator, and nitrogen filling device are controlled according to the reported data to control the environmental temperature, humidity, oxygen levels, etc., thereby achieving a stable storage environment. Specifically, when a level two early warning is activated, one of the dehumidifier and ventilator operates, or both operate simultaneously. When a level three early warning is activated, the nitrogen filling device operates to reduce the oxygen concentration in the storage room to inhibit mold growth or oxidation reactions.

[0041] The tobacco leaf storage data early warning system of this invention has strong scalability, and the system architecture supports flexible access to new types of sensors or control devices, making it suitable for raw material storage scenarios of cigarette enterprises of different sizes.

[0042] Example 3: This embodiment takes a cigarette factory as an example to introduce the application of this embodiment 2 in the raw material aging warehouse of the cigarette factory.

[0043] System Overall Architecture: The overall system architecture of this invention adopts an integrated "end-edge-cloud-control" design. Core-embedded sensor nodes and environmental sensor nodes deployed within the tobacco aging warehouse serve as the perception layer; the edge computing gateway acts as the local processing hub; the cloud platform server is responsible for long-term data storage and advanced analysis; control execution units (including fans, dehumidifiers, and nitrogen filling devices) provide physical responses; and the management terminal is used for human-machine interaction and alarm reception. All modules are interconnected via communication protocols such as LoRa, NB-IoT, Industrial Ethernet, and OPC UA, forming a closed-loop system from data acquisition and intelligent analysis to automatic control, ensuring real-time, accurate, and proactive management of the entire tobacco aging process.

[0044] Sensor Deployment and Hardware Configuration: The core-embedded sensor node uses a custom-designed miniature probe (Φ15 mm × 80 mm) integrating a Sensirion SHT45 high-precision temperature and humidity sensor (temperature accuracy ±0.1℃, humidity accuracy ±1.8% RH) and an SST LS-0101 oxygen sensor (range 0-25% vol, accuracy ±0.1%). It is powered by a CR2477 button battery and uploads data via a LoRa module (470-510 MHz). During installation, it is inserted into the center of the tobacco bale before packing, with one node deployed for every 100 bales. The environmental sensor node has the capability to monitor temperature and humidity (±0.3℃ / ±2% RH) and oxygen (±0.2%), with a coverage area of ​​approximately 500 m. 3 One unit is deployed in the warehouse, positioned 1.5m above the ground and more than 0.5m above the tobacco stack, and connected to the network via a Huawei ME909s-821 NB-IoT module. The edge computing gateway uses an Advantech UNO-2484G industrial host (Intel Core i5-8365U, 8 GB RAM, 256 GB SSD), running Ubuntu 20.04, and supports Docker containerized deployment for local data alignment, interpolation, and model inference.

[0045] Regarding the early warning model and algorithm parameters: The system constructs a three-layer dynamic early warning mechanism: The basic threshold layer automatically adjusts the safety range based on the tobacco variety, storage stage, and season. For example, the ambient temperature and humidity are set to 18℃-28℃ / 60%RH-68%RH, the upper limit of the core temperature is 32℃, and the oxygen concentration is maintained at 18%-21%. The trend analysis layer uses a 6-hour sliding window to calculate the core temperature change rate. If the slope of two consecutive windows exceeds 0.5℃ / h, it is judged as an abnormal temperature rise. The prediction layer is based on a lightweight LSTM model, trained using 36 million historical aging data points from the plant over the past 3 years. It can predict key parameters for the next 24 hours, triggering a high-level early warning when the probability of the core temperature exceeding the limit is ≥80%. All models are deployed on the edge gateway in ONNX format to achieve low-latency inference.

[0046] Linked Control: Implements a three-level response: Level 1 (Attention) – Risk index 60-75: Only pushes WeChat or SMS alarms to the management terminal and records logs; Level 2 (Warning) – 75-90: Automatically starts the ventilation fan for 30 minutes; if the ambient humidity >70%, the dehumidifier is simultaneously turned on; Level 3 (Emergency) – >90: Immediately closes the warehouse air inlet valve, starts the nitrogen filling device to reduce the oxygen concentration to ≤2% and continues for 2 hours. All control commands are generated by the edge gateway, verified by the Siemens S7-1200 PLC, and executed. The execution status is fed back to the cloud platform in real time to ensure safe and reliable operation.

[0047] Figure 4 and Figure 5This demonstrates two different aspects of a complete digital management system for raw material storage and maintenance: Figure 4 This is an architecture diagram of a tobacco leaf storage data early warning system, illustrating the overall system architecture design. Figure 5 This diagram illustrates the digital centralized control hub for raw material storage and maintenance in the tobacco leaf storage data early warning system, showcasing the system's actual operation interface (visualized centralized control hub).

[0048] Figure 4 It showcases the hardware components, software layers, and data flow of the entire intelligent warehousing system. It explains how the system "works."

[0049] The core hierarchy includes: Visualized Central Control Hub (Top): This is the "brain" of the system, responsible for displaying data on a large screen for managers to monitor the overall situation.

[0050] Desktop terminal control center: This allows for specific management and operation via a computer.

[0051] Data Acquisition and Control Service (Middle Layer): This is the core bridge connecting hardware and software, responsible for handling all data interactions.

[0052] The perception and execution layers (bottom and sides) include: IoT environmental sensing: This is the system's "eyes and nose." It includes wireless core temperature monitors, wireless warehouse oxygen concentration monitors, and insect infestation visual monitoring devices. They transmit data via LoRa (a long-range wireless technology) gateways.

[0053] Equipment control: The system can control actual warehouse equipment, including dehumidifiers, air curtains, and electric roller shutters. These devices are connected to the system via a PLC (Programmable Logic Controller).

[0054] Alarms and Interaction: When environmental data is abnormal, the audible and visual alarm device will be activated; staff can use WiFi mobile work devices (such as handheld PDAs and tablets) to operate on-site.

[0055] Figure 4 This describes the complete closed-loop logic of the system, which collects environmental data (temperature, humidity, oxygen, pests) through IoT sensors, processes it through industrial control services, displays it in the central control center, and can control warehouse equipment (such as automatically turning on dehumidifiers).

[0056] Figure 5 yes Figure 1 This is a display of the specific software interface of the "Visualized Central Control Hub." It shows the view that managers see daily.

[0057] Interface functionality.

[0058] Real-time monitoring: The "Data Update Time" displayed at the top of the interface indicates that the system is updating in real time or near real-time.

[0059] Multi-warehouse area management: Different warehouses (such as warehouse 201#, warehouse 202#) and different floors (first to fifth floors) can be selected.

[0060] Visual Layout: The center of the interface displays a floor plan of the warehouse, with each small cell representing a specific storage location or area. Data display includes: Environmental parameters: Figure 4 Key data is directly labeled on the screen, such as temperature (23.1℃), humidity (61.4%), and oxygen concentration (20.9%). Equipment status: Passed. Figure 4 As can be seen, different colored icons represent different device statuses or types (e.g., green represents normal, and red may represent an alarm).

[0061] Figure 5 The bottom section provides a detailed explanation of the icons, including wireless core temperature monitors, wireless warehouse temperature and humidity monitors, wireless cargo location oxygen concentration monitors, insect infestation visual monitoring equipment, and audible and visual alarm devices. Figure 5 It demonstrates how users can use the system. Figure 5 Administrators can see at a glance whether the temperature, humidity, or oxygen concentration of a specific storage area or location exceeds the standard, as well as whether there is a risk of pests, thus achieving "digital maintenance".

Claims

1. A method for early warning of tobacco leaf storage data, characterized in that, The steps include the following: Step 1: Obtain environmental parameters, including ambient temperature and humidity, and ambient oxygen concentration, through environmental sensor nodes inside the tobacco storage warehouse; The parameters of the tobacco leaf wrapping are obtained through embedded sensing nodes at the center of the wrapping, including: wrapping temperature and humidity, and wrapping oxygen concentration. A filtering algorithm is used to preprocess the environmental parameters and tobacco leaf core-wrapping parameters, and the preprocessed environmental parameters and preprocessed tobacco leaf core-wrapping parameters are obtained accordingly. Step 2: Based on the pre-treated tobacco leaf core parameters, calculate the temperature and humidity risk index, tobacco leaf moisture content and oxygen influence factor, and calculate the aging rate according to the temperature and humidity risk index and oxygen influence factor. Then, calculate the basic threshold according to the aging rate and the pre-treated tobacco leaf core parameters, and finally calculate the final threshold based on the basic threshold and the deviation threshold. Step 3: Using the preprocessed tobacco leaf core parameters, tobacco leaf moisture content, and tobacco leaf storage days as inputs, and the final threshold as the output label, train a lightweight LSTM model to obtain the early warning model. Step 4: Obtain the core temperature and humidity, and core oxygen concentration of the tobacco leaves to be treated. Calculate the moisture content of the tobacco leaves to be treated based on the core temperature and humidity. Input the core temperature and humidity, core oxygen concentration, moisture content, and storage days of the tobacco leaves to be treated into the early warning model for processing to obtain the early warning threshold. Then, activate the graded alarm based on the early warning threshold.

2. The tobacco leaf storage data early warning method according to claim 1, characterized in that, In step 1, equation (1) is used to preprocess environmental parameters or tobacco leaf core parameters to obtain the current time. Environmental parameters or tobacco leaf core parameters after pretreatment : (1) In equation (1), Indicates the previous moment The corresponding pre-treated environmental parameters or tobacco leaf core parameters; when season For the current moment Collected environmental parameters or tobacco leaf core parameters, For the current moment Collected environmental parameters or tobacco leaf core parameters, Indicates the current time Kalman filter gain.

3. The tobacco leaf storage data early warning method according to claim 2, characterized in that, The current time is obtained using equation (2). Kalman filter gain : (2) In equation (2), Indicates the current time The uncertainty of the estimate, and ,in, Indicates the previous moment The uncertainty of the estimate, when season =1; For process noise, It is a constant value.

4. The tobacco leaf storage data early warning method according to claim 1, characterized in that: Calculate the current time using equation (3) Temperature and humidity risk index : (3) In equation (3), For the current moment Pre-treated core temperature, For the current moment Humidity of the pre-treated core.

5. The tobacco leaf storage data early warning method according to claim 4, characterized in that: Calculate the current time using equation (4) Tobacco leaf moisture content : (4) In equation (4), The core temperature after pretreatment. , , All have 3 empirical coefficients.

6. The tobacco leaf storage data early warning method according to claim 4, characterized in that: Calculate the current time using equation (5) alcoholization rate : (5) In equation (5), Indicates the current time Oxygen-affecting factors, The number of days the tobacco leaves are stored.

7. The tobacco leaf storage data early warning method according to claim 6, characterized in that, In step 2, the final threshold is obtained through the following process: Calculate the current time using equation (7) Base threshold : (7) In equation (7), Indicates the current time Oxygen concentration in the treated core. Indicates the total collection time; Calculate the current time using equation (8) Deviation threshold : (8) In equation (8), Indicates the current time The pre-treated core temperature is the same as the previous moment. The change in core temperature after pretreatment Indicates the current time The humidity of the pre-treated core is the same as the previous moment. Changes in humidity of the pre-treated core package Indicates the current time The oxygen concentration of the pretreated core was the same as that at the previous time. The change in oxygen concentration in the core after pretreatment , , These represent the weighting coefficients for temperature, humidity, and oxygen concentration, respectively. Calculate the current time final threshold .

8. The tobacco leaf storage data early warning method according to claim 1, characterized in that: Step 4 involves initiating a tiered alarm system as follows: When the warning threshold is lower than the first set threshold, a level one warning is activated, and an alarm message is pushed to the management terminal to prompt manual verification. When the warning threshold is greater than the first set threshold and less than the second set threshold, a level two warning is activated, and ventilation or dehumidification equipment is started. When the warning threshold is greater than the second set threshold but less than the third set threshold, a level three warning is activated, and the nitrogen filling equipment is started. No action is required when the warning threshold is greater than the third set threshold. The first set threshold is less than the second set threshold, and the second set threshold is less than the third set threshold.

9. A tobacco leaf storage data early warning system, characterized in that: include: The data acquisition module acquires environmental parameters through environmental sensor nodes, including environmental temperature and humidity, and environmental oxygen concentration. Simultaneously, it acquires tobacco leaf core-wrapping parameters through core-embedded sensor nodes, including core-wrapping temperature and humidity, and core-wrapping oxygen concentration. The module then preprocesses the environmental parameters and tobacco leaf core-wrapping parameters to obtain preprocessed environmental parameters and tobacco leaf core-wrapping parameters. The data analysis module calculates the temperature and humidity risk index, tobacco moisture content, and oxygen impact factor based on the pre-treated tobacco leaf core parameters. It then calculates the aging rate based on the temperature and humidity risk index and oxygen impact factor, and calculates the basic threshold based on the aging rate and the pre-treated tobacco leaf core parameters. Finally, it calculates the final threshold based on the basic threshold and the deviation threshold. The training module takes the pre-processed tobacco leaf core parameters, tobacco leaf moisture content and tobacco leaf storage days as input, and the final threshold as the output label to train a lightweight LSTM model to obtain the early warning model. The early warning module acquires the core temperature and humidity, and core oxygen concentration of the tobacco leaves to be treated, and calculates the moisture content of the tobacco leaves based on the core temperature and humidity. Then, it inputs the core temperature and humidity, core oxygen concentration, moisture content, and number of days the tobacco leaves have been stored into the early warning model to obtain the early warning threshold, and initiates tiered alarms based on the early warning threshold.

Images