An intelligent monitoring and early warning system for agricultural product preservation and storage warehouses
By introducing a feedforward pre-compensation collaborative mechanism and sparse sensing technology into agricultural product preservation and storage facilities, the problems of temperature and humidity control lag and monitoring blind spots have been solved, enabling real-time reconstruction and hierarchical early warning of the entire temperature field, extending the shelf life of agricultural products and reducing energy consumption.
Patent Information
- Application Number
- CN202611250101.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-18
- Publication Date
- 2026-11-17
AI Technical Summary
The existing temperature and humidity control response in agricultural product preservation and storage warehouses is lagging, leading to localized quality deterioration. Furthermore, point sensors cannot cover the entire temperature field, creating monitoring blind spots and making it impossible to achieve proactive prevention and comprehensive perception.
Establish a feedforward and pre-compensation collaborative mechanism between warehousing operation plan information and environmental control system. Combine sparse sensing and lightweight computational fluid dynamics digital twin technology to achieve real-time reconstruction and hierarchical early warning of the global temperature field through edge computing and cloud offline modeling.
It enables proactive pre-cooling before door opening disturbances, shortens temperature and humidity recovery time, eliminates monitoring blind spots, significantly extends the shelf life of agricultural products, and reduces refrigeration energy consumption.
Smart Images

Figure CN123114170A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent warehousing and cold chain environmental control technology, specifically to an intelligent monitoring and early warning system for agricultural product preservation and storage warehouses. Background Technology
[0002] Agricultural product preservation and storage warehouses are the core infrastructure of modern cold chain logistics and smart agriculture. Their core function is to stably control key parameters such as temperature and humidity within the warehouse within the optimal preservation range required by agricultural products, so as to inhibit the respiratory and metabolic rate of agricultural products and delay spoilage. In the field of refrigeration environment control, the industry currently widely adopts the proportional-integral-derivative feedback control method: several temperature and humidity sensors are arranged at selected locations within the warehouse, and the collected measured values of temperature and humidity are compared with set thresholds. Based on the deviation, the variable frequency compressor and solenoid valve are driven to perform refrigeration regulation actions. This control logic is mature and reliable, but it is essentially a passive response mechanism with an inherent response lag defect.
[0003] In the field of warehouse automation, modern intelligent warehouses generally deploy warehouse management systems and warehouse control systems to manage the inbound and outbound of goods, inventory counting, and scheduling of automated conveying equipment. These two systems store high-precision operation plan data, including inbound and outbound time windows, goods batches, and goods volume and weight information. However, in the existing solutions, the above data and the environmental control system are completely independent of each other and fail to form a coordinated linkage.
[0004] The existing technical solutions have the following two fundamental flaws:
[0005] First, the temperature and humidity control response is severely delayed, leading to localized deterioration in the quality of agricultural products. During daily entry and exit operations, the frequent opening of doors in storage warehouses introduces a large amount of hot and humid air from the outside, creating strong heat and humidity load disturbances. Traditional refrigeration control systems based on threshold feedback must wait for sensors to detect temperature and humidity deviations from the set values before passively starting compensation. The sensor response delay, combined with the heat conduction time constant inside the warehouse, results in a severe lag in system response. The recovery time of the warehouse environment can be as long as several hours. Agricultural products near the warehouse door are continuously subjected to heat shock, which accelerates respiratory and metabolic rates, exacerbates cell membrane damage, and significantly shortens shelf life. The more fundamental limitation is that the warehousing scheduling system and the environmental control system are completely independent of each other. The environmental control system is completely unaware of entry and exit actions and can only respond passively after disturbances occur and temperature and humidity have deviated, making it impossible to achieve proactive prevention.
[0006] Secondly, point sensors cannot cover the entire temperature field, and the early warning system has structural blind spots. In large storage warehouses, goods are densely stacked, and the stacks physically block the cooling airflow, creating ventilation dead spots and local hot spots inside the stacks. The existing system can only rely on point sensors around the walls and shelves to obtain discrete data, which cannot restore the true three-dimensional temperature field distribution inside the warehouse. This means that the hidden local overheated areas inside the stacks are in the monitoring blind spot for a long time. Agricultural products rot prematurely due to the accumulation of local respiratory heat, rendering the early warning system ineffective. The rot spreads to the entire batch of goods, causing significant losses to the stored grain.
[0007] In the field of temperature field simulation, computational fluid dynamics can accurately simulate three-dimensional temperature fields under complex boundary conditions such as stack obstruction and fan layout. However, full-scale simulation requires extremely high computing power, and a single simulation often takes several hours or even days, which is far from meeting the real-time requirements of industrial online applications. Intrinsic orthogonal decomposition is a mature model reduction method. Its core idea is to perform singular value decomposition on a large amount of offline simulation data and extract several orthogonal characteristic basis functions containing the main flow field information, thereby reducing the original high-dimensional equation set to a set of low-dimensional algebraic equations, making online real-time solution possible. However, there is still no industrial implementation plan for applying it to the real-time reconstruction and hierarchical early warning of the entire temperature field in agricultural product preservation and storage warehouses. To this end, we propose an intelligent monitoring and early warning system for agricultural product preservation and storage warehouses. Summary of the Invention
[0008] To address the aforementioned deficiencies in existing technologies, the present invention aims to provide an intelligent monitoring and early warning system for agricultural product preservation and storage warehouses. This system establishes a feedforward and pre-compensation collaborative mechanism between storage operation plan information and environmental control system. Furthermore, by integrating sparse sensing and lightweight computational fluid dynamics digital twin three-dimensional temperature field real-time reconstruction technology, it achieves full-domain temperature field perception and hierarchical early warning, fundamentally solving the control lag problem under dynamic disturbances and the perception blind spot problem in dense stacking environments.
[0009] To achieve the above objectives, the present invention provides the following technical solution: an intelligent monitoring and early warning system for agricultural product preservation and storage warehouses, comprising a storage information access layer, an edge computing and control layer, a physical sensing and execution layer, a cloud offline modeling layer, and a visualization and early warning layer, wherein each layer achieves real-time data communication through industrial Ethernet and standard application programming interfaces;
[0010] The warehouse information access layer includes an API gateway that connects to the warehouse management system interface, the warehouse control system interface, and the automated guided vehicle dispatching platform interface, and is used to obtain inbound and outbound operation plan instructions in real time.
[0011] The physical sensing and execution layer includes an array of temperature and humidity sensors, a CO2 concentration sensor, an external meteorological sensor, a variable frequency refrigeration compressor unit, a fan array that is independently controlled according to spatial zones, and a humidification and dehumidification device.
[0012] The edge computing and control layer includes an industrial edge computing gateway, which integrates a feedforward model prediction controller module and a reduced-order model online fast solver module.
[0013] The cloud-based offline modeling layer includes a high-fidelity computational fluid dynamics offline simulation server, an intrinsic orthogonal decomposition mode extraction module, and a reduced-order model database.
[0014] The visualization and early warning layer includes a three-dimensional temperature field cloud map visualization screen, a hierarchical alarm terminal, and an energy consumption monitoring dashboard.
[0015] The industrial edge computing gateway reads the inbound and outbound operation instructions issued by the warehouse control system through the feedforward model prediction controller module, extracts the planned door opening time, estimated door opening duration, the quality of this batch of goods and the initial temperature of the goods entering the warehouse, and combines the outside temperature and relative humidity collected by the external meteorological sensor to calculate the sensible heat disturbance, latent heat disturbance and heat carried in by the goods, and summarizes the total heat and humidity load disturbance, converts it into feedforward pre-cooling compensation power, and adds the feedforward pre-cooling compensation power to the feedback control output within a predetermined time window before the planned door opening time, driving the variable frequency refrigeration compressor unit to complete the pre-cooling and cold storage in advance;
[0016] The industrial edge computing gateway, through the online fast solver module of the reduced-order model, based on the reduced-order model data package issued by the cloud offline modeling layer, and taking the measured temperature of the temperature and humidity sensor array as input, reconstructs the three-dimensional temperature field of the entire database within 100 milliseconds through least squares matrix operations, and performs node-by-node three-level threshold comparison based on the reconstructed temperature field to trigger graded early warning and directional fan subarray control.
[0017] Furthermore, the sensible heat disturbance amount Latent heat disturbance Heat carried in by goods The calculation formulas are as follows:
[0018] ;
[0019] ;
[0020] ;
[0021] in, The density of the outside air is taken as the standard value of 1.293 kg / m³.
[0022] The specific heat capacity of air at constant pressure is taken as 1005 J / (kg·K).
[0023] To estimate the volume of outside air entering the warehouse during the opening period, the calculation is based on an empirical model of the warehouse door size, the duration of the opening, and the airflow velocity through the doorway, with the unit being m³.
[0024] The ambient temperature outside the warehouse is expressed in Kelvin (K).
[0025] The target temperature inside the warehouse, in Kelvin (K).
[0026] The moisture content of the outside air is expressed in kg / kg and is calculated from the outside temperature and relative humidity using the formula for the physical properties of moist air.
[0027] The target moisture content in the warehouse is expressed in kg / kg.
[0028] For the latent heat of vaporization of water, take J / kg;
[0029] The mass of this batch of goods is expressed in kg.
[0030] The average specific heat capacity of the cargo is expressed in J / (kg·K).
[0031] The initial temperature of the goods upon entering the warehouse, expressed in Kelvin (K).
[0032] Total thermal and moisture load disturbance and feedforward precooling compensation power The calculation formula is:
[0033] ;
[0034] ;
[0035] in, The energy efficiency coefficient of the refrigeration unit under the current operating conditions is obtained by looking up the table in real time from the unit's characteristic curve, and its dimension is one.
[0036] The pre-cooling intervention time window, measured in seconds, is used to send compressor control commands to the variable frequency refrigeration compressor unit. The calculation formula is:
[0037] ;
[0038] in, The feedback control output of the model predictive controller is calculated based on the current temperature and humidity deviation in the warehouse, and the unit is W;
[0039] This refers to the feedforward precooling compensation power, measured in watts (W). The compressor unit control command must be issued before the planned door opening time. It is issued at all times to ensure that the compressor unit completes pre-cooling and cold storage before the door opening action occurs.
[0040] Furthermore, the real-time reconstruction of the three-dimensional temperature field is achieved based on a two-stage method of offline modeling and online simulation.
[0041] In the offline modeling phase, a high-fidelity computational fluid dynamics model is constructed based on the actual geometric configuration of the target reservoir, targeting a space that covers the actual operational parameters. Each operating condition (parameter combination includes ambient temperature, fan speed vector, and silo door opening / closing status) A full-scale simulation was performed using 100 to 500 operating conditions, discretizing the three-dimensional temperature field of the entire warehouse under each operating condition into a subset containing... A column vector with degrees of freedom ( , Usually to (on a certain order of magnitude), combined to form a snapshot matrix:
[0042] ;
[0043] For snapshot matrix Perform singular value decomposition ,in column vectors ( This refers to the modal basis functions of the eigenorthogonal decomposition space;
[0044] For a singular value diagonal matrix, the diagonal elements Reflecting the contribution of each mode to the total energy, the cutoff criterion is that the cumulative energy share should not be less than 99.9%.
[0045] ;
[0046] Before selection The first mode forms a reduced basis matrix ,in The typical value does not exceed 20th order, assuming the layout within the library is... The set of grid node indices corresponding to each physical sensor is: ( Define the sensor observation matrix. , of which Behavior in the The unit row vectors for each node value are pre-calculated in the offline phase, resulting in a reduced-order observation matrix. and its pseudo-inverse matrix Together with the reduced-order basis matrix The data is packaged into a reduced-order model data package and sent to the persistent storage of the industrial edge computing gateway.
[0047] During the online simulation phase, the edge computing gateway reads data from the database at a 5-second sampling period. Each sensor measures the actual temperature, and a measured temperature vector is constructed. (superscript) (Representing the current time), the pre-stored pseudo-inverse matrix is called to solve for the eigenorthogonal decomposition mode coefficient vector using least squares operation. :
[0048] ;
[0049] Then, the global three-dimensional temperature field is reconstructed:
[0050] ;
[0051] Will The data is mapped back to the three-dimensional geometric space of the reservoir to generate a visualized three-dimensional temperature field cloud map. The entire online calculation process takes no more than 100 milliseconds, meeting the requirements for real-time monitoring.
[0052] Furthermore, the tiered early warning system adopts a three-level threshold system:
[0053] When the reconstruction temperature value of any spatial node does not exceed the first warning threshold Maintain a level one normal condition at all times;
[0054] When exceeding And not exceeding the second warning threshold When a Level II yellow alert is triggered, an alert notification and the precise location coordinates of the hotspot are pushed to the management terminal;
[0055] When exceeding When a level three red alarm is triggered, the hotspot coordinates are mapped to the wind turbine array partition map, and a speed increase command is sent to the corresponding frequency converter driver of the wind turbine subarray responsible for covering the hotspot area to achieve precise local cooling supplementation. Set as the upper limit of the optimal preservation temperature for agricultural products. Set as the critical temperature for the deterioration of agricultural product quality.
[0056] Compared with the prior art, the beneficial effects of the present invention by adopting the above technical solution are as follows:
[0057] 1. This invention establishes a cross-system collaborative control mechanism by using warehousing operation plan information as a feedforward input variable for the environmental control system. It actively pre-cools the product before the door opening disturbance occurs, shortening the recovery time of transient temperature and humidity changes by more than 80% compared to the traditional passive feedback scheme. This fundamentally eliminates the harm of thermal shock to the quality of agricultural products and significantly extends the shelf life of agricultural products.
[0058] 2. This invention integrates sparse sensing and lightweight computational fluid dynamics digital twin technology, using a small number of peripheral physical sensors as input, to achieve millisecond-level real-time reconstruction of the three-dimensional temperature field of the entire warehouse, completely eliminating blind spots in the monitoring inside the stack, realizing early graded early warning of agricultural product quality deterioration, upgrading from post-event alarm to pre-event prevention, and upgrading from point-based discrete sensing to full-domain continuous sensing.
[0059] 3. This invention replaces the approach of significantly increasing the global cooling capacity with directional fan subarray control, reducing overcooling caused by lag compensation and significantly reducing cooling energy consumption. The system architecture is based on mature industrial hardware and standard communication protocols, has high industrial deployment feasibility, and is suitable for the construction and intelligent upgrading of large-scale agricultural product preservation and storage facilities. Attached Figure Description
[0060] Figure 1 This is a layered block diagram of the overall architecture of the intelligent monitoring and early warning system for agricultural product preservation and storage warehouses in this embodiment of the invention;
[0061] Figure 2 This is a flowchart of the pre-compensation collaborative control mechanism based on warehousing operation plan information in an embodiment of the present invention;
[0062] Figure 3 This is a flowchart illustrating the hierarchical early warning and directional wind turbine array control logic in an embodiment of the present invention. Detailed Implementation
[0063] The following is in conjunction with the appendix Figure 1-3 The specific embodiments of the present invention will be further described below. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.
[0064] I. System Overall Architecture
[0065] The intelligent monitoring and early warning system for agricultural product preservation and storage proposed in this invention consists of five functional layers.
[0066] The warehouse information access layer includes interfaces for the warehouse management system, the warehouse control system, and the automated guided vehicle (AGV) dispatching platform. It continuously pushes a feedforward data stream of the operation plan to the edge computing layer in the form of an API gateway. The data content includes the planned door opening time, the quantity of goods in this batch, and batch information. This layer is the key information entry point that distinguishes this invention from traditional environmental control systems, transforming the warehouse operation plan into a feedforward driving signal that the environmental control system can perceive and utilize.
[0067] The physical sensing and execution layer includes an array of temperature and humidity sensors inside and outside the warehouse, a CO2 concentration sensor, a weather sensor outside the warehouse, a variable frequency refrigeration compressor unit, a fan array, and a humidification and dehumidification device. It interacts with the edge computing layer in a two-way communication manner to exchange sensing data and control commands, forming the underlying interface for the system to interact with the physical world.
[0068] The core of the edge computing and control layer is the industrial edge computing gateway, which integrates a feedforward model predictive controller module and a reduced-order model online fast solver module. These modules respectively carry out the online computing tasks of the two core innovative mechanisms of this invention. The industrial edge computing gateway interacts with each layer in real time through industrial Ethernet and is the execution center for intelligent decision-making in the entire system.
[0069] The cloud-based offline modeling layer includes a high-fidelity computational fluid dynamics offline simulation server, an intrinsic orthogonal decomposition mode extraction module, and a reduced-order model database. It is responsible for offline modeling calculations and regularly distributes the reduced-order model data packets to the edge computing gateway via industrial Ethernet for updates. The offline calculation results of this layer form the knowledge base for online real-time reconstruction, compressing hours of high-fidelity simulation calculations into millisecond-level online matrix operations.
[0070] The visualization and early warning layer includes a 3D temperature field cloud map visualization screen, a hierarchical alarm terminal, and an energy consumption monitoring dashboard. It receives monitoring data and alarm commands uploaded from the edge computing layer, provides managers with an intuitive interface for perceiving the environmental situation within the warehouse, and supports hierarchical notifications through multiple channels such as SMS push, application messages, and large screen audio-visual alarms.
[0071] II. Specific Configuration of the Physical Perception and Execution Layer
[0072] An array of temperature and humidity sensors is deployed inside the warehouse. The sensors are installed according to the principle of uniform distribution along the boundaries and densification at key nodes: they are evenly distributed along the four walls and ceiling of the warehouse, and densified near the warehouse door, the air outlet and return air inlet of the refrigeration fan. The distance between adjacent sensor nodes does not exceed one-third of the characteristic dimension of the warehouse to ensure the accuracy of the boundary conditions required by the reduced-order model algorithm. Each sensor node synchronously collects temperature, relative humidity and CO2 concentration, with a sampling period of 5 seconds. The data from each node is aggregated to the industrial edge computing gateway via industrial Ethernet. Meteorological sensors are deployed outside the warehouse to collect the ambient temperature and relative humidity as meteorological input parameters for the feedforward calculation of heat and humidity load.
[0073] The refrigeration side is equipped with a variable frequency compressor unit, which receives speed control commands from the industrial edge computing gateway through a 0 to 10V analog signal channel to achieve continuous stepless adjustment of the cooling capacity. The response time meets the timing requirements of the feedforward pre-cooling command. On the airflow organization side, the circulating fans in the air conditioning unit are divided into several independently controlled fan sub-arrays according to spatial partitions. Each fan is equipped with a variable frequency driver to support independent speed adjustment, so as to achieve directional air supply to local hot spots without global increase of cooling capacity. Real-time data communication between the functional layers is achieved through industrial Ethernet. The feedforward control data stream, the sensing data uplink stream, and the control command downlink stream are transmitted independently and do not interfere with each other.
[0074] III. Core Mechanism 1: Pre-compensation Collaborative Control Based on Feedforward of Warehouse Operation Plan Information
[0075] The core mechanism works as follows: the feedforward model predictive controller module within the system's industrial edge computing gateway continuously polls the API interface of the warehouse control system. When a new inbound / outbound operation instruction is detected, the feedforward heat and humidity load calculation module is immediately triggered to execute the following calculation process.
[0076] Extract planned door opening times from warehouse control system operation instructions. Estimated duration of door opening (Estimated from the volume of the batch of goods and the operating speed of the automated guided vehicle), the weight of this batch of goods ;
[0077] Read the current outside temperature from the external weather sensor. and relative humidity ;
[0078] Read the target temperature in the library from the system settings. and target relative humidity ;
[0079] Initial temperature of goods entering the warehouse Information is obtained from cargo information records in the warehouse management system or from actual measurements taken by an infrared thermometer before the goods are stored.
[0080] Calculate the sensible heat disturbance introduced by opening the door. This refers to the heat that causes the temperature to rise in the storage room when hot outside air enters.
[0081] ;
[0082] in, The density of the outside air is taken as the standard value of 1.293 kg / m³.
[0083] The specific heat capacity of air at constant pressure is taken as 1005 J / (kg·K).
[0084] To estimate the volume of outside air entering the warehouse during the opening period, the warehouse door size and opening duration are used as factors. The airflow velocity in the doorway was calculated using an empirical model, and the unit is m³.
[0085] and All units are in K.
[0086] Calculate the latent heat disturbance introduced by opening the door. That is, the heat released by the condensation of water vapor carried by the moist air in the outside:
[0087] ;
[0088] in, Moisture content of ambient air, expressed in kg / kg. and Calculated using the formula for the physical properties of moist air;
[0089] The target moisture content in the warehouse, expressed in kg / kg, is determined by... and calculate;
[0090] For the latent heat of vaporization of water, take J / kg.
[0091] Calculate the initial heat carried by this batch of goods entering the warehouse. This refers to the heat released into the warehouse when the temperature of the goods is higher than the warehouse temperature.
[0092] ;
[0093] in, The average specific heat capacity of the cargo is expressed in J / (kg·K).
[0094] The initial temperature of the goods upon entering the warehouse, expressed in Kelvin (K).
[0095] The total thermal and humidity load disturbance is obtained by summarizing. :
[0096] ;
[0097] The total heat and humidity load disturbance is converted into feedforward precooling compensation power. :
[0098] ;
[0099] in, The energy efficiency coefficient of the refrigeration unit under the current operating conditions is obtained by looking up the table in real time from the unit's characteristic curve, and its dimension is one.
[0100] The pre-cooling intervention time window is set by the system based on empirical parameters of the heat capacity of the storage tank, with a typical value of 5 to 15 minutes before opening the door.
[0101] The model predictive controller within the industrial edge computing gateway will feedforward compensation amount As a pre-compensation term, it is added to the original feedback control output. This generates control commands for the compressor unit.
[0102] ;
[0103] in, This is the refrigeration power setting value sent to the variable frequency refrigeration compressor unit, in watts (W).
[0104] This is the feedback control output of the model predictive controller, calculated based on the current temperature and humidity deviation within the warehouse, in W units. This instruction must be executed within... The system sends out signals at all times to ensure that the compressor unit completes pre-cooling and cold storage before the door opening action occurs, realizing a fundamental shift from lag compensation to feedforward pre-cooling. When the warehouse control system reports the end of the operation signal, the feedforward superposition component exits, and the system resumes the pure model predictive control feedback control loop.
[0105] The core innovation of this mechanism lies in using the warehouse operation plan information as the feedforward input variable of the environmental control system, realizing cross-system collaboration between the warehouse scheduling system and the refrigeration environmental control system. Compared with traditional passive feedback control, this mechanism has completed the compensation and cold storage before the disturbance occurs, eliminating the double lag caused by sensor detection delay and the heat conduction time constant in the warehouse. It fundamentally solves the problem of large fluctuations in temperature and humidity in the warehouse caused by inbound and outbound operations, and shortens the temperature and humidity recovery time by more than 80%.
[0106] IV. Core Mechanism Two: Real-time Reconstruction of 3D Temperature Field by Integrating Sparse Sensing and Lightweight Computational Fluid Dynamics Digital Twin
[0107] This core mechanism is divided into an offline modeling stage and an online simulation stage. The complete algorithm flow is shown in the diagram, with vertical dividing lines separating the cloud offline stage (left) and the edge gateway online stage (right).
[0108] (a) Offline modeling stage (executed on a high-performance cloud server)
[0109] Based on the actual geometric dimensions of the target storage warehouse (length, width, height, door location and size, shelving layout, fan installation location and orientation), a three-dimensional computational fluid dynamics computational domain mesh model is constructed, and boundary conditions are set: the air outlet of the refrigeration fan is set as the velocity inlet boundary, the return air outlet is set as the pressure outlet boundary, the warehouse wall is set as the thermally insulating solid wall boundary, and the stack is set as a porous medium region. Porosity and permeability coefficient are set to characterize the resistance characteristics of the stack to airflow.
[0110] Full-scale computational fluid dynamics simulations were performed on multiple sets of operating parameters covering the actual operational range. These parameter combinations included ambient temperature, fan speed vector, and gate opening / closing status. Each working condition is used to discretize the three-dimensional temperature field of the entire warehouse under each working condition into a single data structure. A column vector with degrees of freedom ( ), combined to form a snapshot matrix:
[0111] ;
[0112] For snapshot matrix Perform singular value decomposition:
[0113] ;
[0114] in, column vectors ( This refers to the intrinsic orthogonal decomposition spatial modal basis functions, where each basis function is a spatial distribution vector defined on all grid nodes;
[0115] For a singular value diagonal matrix, the diagonal elements It reflects the degree to which each mode contributes to the total flow field energy;
[0116] It is the transpose of the right singular vector matrix.
[0117] The selection is based on the cutoff criterion that the cumulative energy share is not less than 99.9%. First mode:
[0118] ;
[0119] in, To preserve the number of intrinsic orthogonal decomposition modes, the typical value is no more than 20, and the realization dimension is from (Typically reaching millions) to Significant compression, retaining the former The first-order modes form a reduced-order basis matrix. .
[0120] Assume the layout inside the warehouse The set of grid node indices corresponding to each physical sensor location is ( Define the sensor observation matrix. , of which Behavior in the A unit row vector containing values at each node position is used to extract the temperature value at the sensor location from the global temperature field vector. The global temperature field can be approximately expressed as a linear combination of intrinsic orthogonal decomposition modes:
[0121] ;
[0122] in, is the column vector of the global three-dimensional temperature field to be reconstructed;
[0123] Let be the vector of eigenorthogonal decomposition mode coefficients to be solved.
[0124] The reduced-order observation matrix is pre-calculated during the offline phase. and its pseudo-inverse matrix Together with the reduced-order basis matrix The data is packaged into a reduced-order model data package, distributed to the industrial edge computing gateway via industrial Ethernet and persistently stored. The entire offline modeling stage is executed on a high-performance server in the cloud. A single full-scale working condition simulation takes several hours, but the resulting reduced-order model data package can be used for online stages for a long time. Only when there are major changes in the warehouse layout is offline modeling required again.
[0125] (II) Online simulation phase (executed in real time by industrial edge computing gateway)
[0126] Industrial edge computing gateways sample data from the database at 5-second intervals. Each temperature and humidity sensor reads the current measured temperature value and constructs a measured temperature vector. superscript Indicates the current sampling time.
[0127] The pseudo-inverse matrix is pre-stored offline, and the eigenorthogonal decomposition mode coefficient vector at the current time step is solved using the least squares method. :
[0128] ;
[0129] Since the pseudo-inverse matrix has been pre-calculated and stored in the gateway during the offline phase, there is no need to repeat the matrix inversion operation during online invocation; only one matrix multiplication is required. The modal coefficient vector dimension is... ( (No more than 20 orders), all matrix multiplication operations can be completed in milliseconds, with a computational load far lower than that of full-scale computational fluid dynamics simulation.
[0130] The obtained modal coefficients Substituting the reduced-order basis matrix, we reconstruct the three-dimensional temperature field of the entire library:
[0131] ;
[0132] Output This is the reconstruction result of the high-precision three-dimensional temperature field of the entire warehouse at the current moment, including temperature information of all grid nodes in the entire warehouse, covering locations that sensors cannot directly reach, such as ventilation dead spots inside the stacks. After mapping back to the three-dimensional geometric space of the reservoir, a visualized three-dimensional temperature field cloud map is formed and uploaded to the large screen of the visualization and early warning layer for display. The entire online calculation process (collecting sensor data, constructing measured vectors, solving modal coefficients, reconstructing the temperature field, and mapping three-dimensional space) takes no more than 100 milliseconds, which meets the requirements of real-time monitoring. Compared with offline simulation, which takes several hours, the calculation efficiency is improved by more than 10,000 times. The online simulation results also serve as the input for graded early warning judgment and enter the graded early warning and directional fan control links.
[0133] V. Tiered Early Warning and Directional Wind Turbine Control
[0134] The logical flow of graded early warning and directional wind turbine control.
[0135] The system reconstructs the global three-dimensional temperature field. Perform a node-by-node scan and record the temperature values of each node. ( It is compared one by one with the preset three temperature thresholds.
[0136] When the temperature value of any space node does not exceed the first warning threshold When the system maintains a Level 1 green normal state, it continues the normal cooling control cycle. When the temperature value of any space node exceeds the first warning threshold... When the system triggers a Level 2 yellow alert, it records the hotspot coordinates and highlights the hotspot location in yellow on the 3D temperature field cloud map. Simultaneously, it pushes an alert notification and the precise 3D coordinates of the hotspot to the management terminal for analysis and handling by management personnel. If the temperature value of any spatial node further exceeds the second alert threshold... When the system triggers a Level 3 red alarm, it automatically maps the hotspot coordinates to the fan array partition map, identifies the fan subarray responsible for covering the hotspot area, and sends a speed increase command to the corresponding frequency converter driver. This causes the cooling airflow to be directionally and strongly delivered to the hotspot area, achieving precise localized supplemental cooling and avoiding energy waste caused by a significant increase in global cooling capacity. The first warning threshold... Set as the upper limit of the optimal preservation temperature for agricultural products, and the second warning threshold. The critical temperature for the deterioration of agricultural product quality is set, and both thresholds can be parameterized according to the preservation requirements of different agricultural product varieties.
[0137] After each directional control action is completed, the system waits for the next sampling cycle (5 seconds) to re-enter the temperature field reconstruction cycle, performs a verification scan on the hot spot area, and confirms whether the hot spot has been eliminated. If the hot spot is eliminated, the corresponding fan subarray will restore the normal speed setting.
[0138] If the hot spot persists, the directional speed increase command will be maintained until the hot spot is eliminated, forming a closed-loop control verification mechanism. Compared with the traditional early warning system that only issues alarm notifications, this mechanism deeply integrates early warning with automatic handling, upgrading from alarm notifications to automatic closed-loop intervention.
[0139] VI. Overall Workflow of System Collaboration
[0140] The complete collaborative workflow of the system is presented in the form of a swimlane diagram, which includes four swimlanes: the warehouse management and control layer, the feedforward pre-compensation control module, the reduced-order model real-time temperature field reconstruction module, and the graded early warning and fan control module. The action sequence relationship of each module is marked with the time axis as the horizontal axis.
[0141] The warehouse management and control layer continuously outputs operational plan instructions;
[0142] After receiving the command, the feedforward pre-compensation control module immediately calculates the total thermal and humidity load disturbance. The system continuously sends feedforward compensation commands to the model predictive controller to drive the variable frequency compressor unit to pre-cool the warehouse. After the door is opened, it continuously monitors the recovery process of the warehouse environment. Once the warehouse control system reports the end of the operation signal, it resumes pure feedback control.
[0143] The reduced-order model real-time temperature field reconstruction module collects sensor data every 5 seconds, completes the three-dimensional temperature field reconstruction, and continuously outputs the three-dimensional temperature field cloud map and hotspot coordinates.
[0144] The graded early warning and wind turbine control module receives hotspot coordinates, determines the early warning level, and implements targeted control, feeding back the control effect to the reconstruction module for verification.
[0145] Two core mechanisms operate in parallel, independent yet coordinated: the feedforward pre-compensation control mechanism operates before the door opening action occurs. Pre-cooling and cold storage are completed within the time window to cope with macroscopic heat and humidity load disturbances caused by inbound and outbound operations;
[0146] The reduced-order model's real-time temperature field reconstruction mechanism continuously updates the full-domain three-dimensional temperature field every 5 seconds, accurately identifying hidden hot spots inside the stack and triggering targeted cooling to address local overheating caused by the accumulation of respiratory heat in the goods. The two mechanisms work together to ensure the overall quality of the preservation environment for agricultural products in the warehouse, achieving deep collaboration between the warehousing system and the environmental control system.
[0147] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of the present invention. Therefore, any modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope defined by the claims of the present invention.
Claims
1. An intelligent monitoring and early warning system for agricultural product preservation and storage warehouses, characterized in that, It includes a warehouse information access layer, an edge computing and control layer, a physical sensing and execution layer, a cloud-based offline modeling layer, and a visualization and early warning layer; The warehouse information access layer includes an API gateway that connects to the warehouse management system interface, the warehouse control system interface, and the automated guided vehicle dispatching platform interface. The physical sensing and execution layer includes a temperature and humidity sensor array, a CO2 concentration sensor, an external weather sensor, a variable frequency refrigeration compressor unit, a fan array, and a humidification and dehumidification device; The edge computing and control layer includes an industrial edge computing gateway, which integrates a feedforward model prediction controller module and a reduced-order model online fast solver module. The cloud-based offline modeling layer includes a high-fidelity computational fluid dynamics offline simulation server, an intrinsic orthogonal decomposition mode extraction module, and a reduced-order model database. The visualization and early warning layer includes a three-dimensional temperature field cloud map visualization screen, a hierarchical alarm terminal, and an energy consumption monitoring dashboard. The industrial edge computing gateway reads the inbound and outbound operation instructions issued by the warehouse control system through the feedforward model prediction controller module and the API gateway. It extracts the planned door opening time, the estimated door opening duration, the quality of the current batch of goods, and the initial temperature of the goods upon entering the warehouse. Combined with the outside temperature and relative humidity collected by the external meteorological sensor, it calculates the sensible heat disturbance, the latent heat disturbance, and the heat carried in by the goods. It summarizes the total heat and humidity load disturbance, converts it into feedforward pre-cooling compensation power, and adds the feedforward pre-cooling compensation power to the feedback control output within a predetermined time window before the planned door opening time, driving the variable frequency refrigeration compressor unit to complete the pre-cooling and cold storage in advance. The industrial edge computing gateway, through the online fast solver module of the reduced-order model, based on the reduced-order model data package issued by the cloud offline modeling layer, and taking the measured temperature of the temperature and humidity sensor array as input, reconstructs the three-dimensional temperature field of the entire database within 100 milliseconds through least squares matrix operations, and performs node-by-node three-level threshold comparison based on the reconstructed temperature field to trigger graded early warning and directional fan subarray control.
2. The intelligent monitoring and early warning system for agricultural product preservation and storage warehouses according to claim 1, characterized in that, The temperature and humidity sensor array is evenly distributed along the four walls and ceiling of the warehouse, and is densely arranged near the warehouse door, the air outlet and return air outlet of the refrigeration fan. The distance between adjacent sensor nodes does not exceed one-third of the characteristic dimension of the warehouse. Each sensor node synchronously collects temperature, relative humidity and CO2 concentration, with a sampling period of 5 seconds. The fan array is divided into several independently controlled fan sub-arrays according to spatial partitions. Each fan is equipped with a frequency converter driver and supports independent speed adjustment.
3. The intelligent monitoring and early warning system for agricultural product preservation and storage warehouses according to claim 1, characterized in that, The sensible heat disturbance Latent heat disturbance Heat carried in by goods and total heat and humidity load disturbance The calculation method is as follows: ; ; ; ; in, The density of the outside air is taken as the standard value of 1.293 kg / m³. The specific heat capacity of air at constant pressure is taken as 1005 J / (kg·K). To estimate the volume of outside air entering the warehouse during the opening period; The ambient temperature outside the warehouse; The target temperature inside the warehouse; The humidity of the outside air; The target moisture content within the warehouse; For the latent heat of vaporization of water, take J / kg; For the quality of this batch of goods; The average specific heat capacity of the goods; The initial temperature of the goods upon entering the warehouse.
4. The intelligent monitoring and early warning system for agricultural product preservation and storage warehouses according to claim 3, characterized in that, The feedforward precooling compensation power The calculation method is as follows: ; in, The coefficient of performance (COP) of the refrigeration unit under the current operating conditions is obtained by looking up the table in real time from the unit's characteristic curve. To allow for early intervention during the pre-cooling period; The feedforward precooling compensation power is superimposed on the feedback control output of the model prediction controller. Then, compressor unit control commands are generated. Send to the variable frequency refrigeration compressor unit; The compressor unit control command is given before the planned door opening time. Send it out immediately; Once the warehouse control system reports the completion of the operation, the system reverts to pure model predictive control feedback control.
5. The intelligent monitoring and early warning system for agricultural product preservation and storage warehouses according to claim 1, characterized in that, The reduced-order model data package is generated through the following offline modeling process: a three-dimensional computational fluid dynamics computational domain mesh model is constructed based on the actual geometric configuration of the target storage warehouse. The velocity inlet boundary, pressure outlet boundary, adiabatic solid wall boundary, and porous media region correspond to the fan outlet, return air outlet, warehouse wall, and cargo stack, respectively. Full-scale simulations are performed for multiple sets of operating conditions covering the actual operating parameter space. The three-dimensional temperature field of the entire warehouse under each operating condition is discretized into column vectors and combined to form a snapshot matrix. The snapshot matrix is subjected to singular value decomposition to extract the intrinsic orthogonal decomposition spatial mode basis functions. The first few orders of modes are selected to form the reduced-order basis matrix according to the truncation criterion that the cumulative energy ratio is not less than 99.9%. The reduced-order observation matrix and its pseudo-inverse matrix are pre-calculated, and the reduced-order model data package is sent to the industrial edge computing gateway.
6. The intelligent monitoring and early warning system for agricultural product preservation and storage warehouses according to claim 5, characterized in that, The cumulative energy percentage threshold for the cutoff criterion is set at 99.9%. The number of retained intrinsic orthogonal decomposition modes does not exceed 20. The total number of operating conditions in the offline simulation ranges from 100 to 500, and the parameter combinations include the ambient temperature, the speed vector of each fan subarray, and the opening and closing status of the silo door.
7. The intelligent monitoring and early warning system for agricultural product preservation and storage warehouses according to claim 5, characterized in that, The pseudo-inverse matrix of the reduced-order observation matrix is pre-calculated and persistently stored in the industrial edge computing gateway during the offline phase. During the online simulation phase, the measured temperature vector from the temperature and humidity sensors inside the warehouse was used. As input, a pre-stored pseudo-inverse matrix is called to perform matrix multiplication to solve for the eigenorthogonal decomposition mode coefficient vector. Then, by reducing the basis matrix Reconstructing the global three-dimensional temperature field The online calculation process takes no more than 100 milliseconds.
8. The intelligent monitoring and early warning system for agricultural product preservation and storage warehouses according to claim 1, characterized in that, The tiered early warning system adopts a three-level threshold system: When the reconstruction temperature value of any spatial node does not exceed the first warning threshold Maintain a level one normal condition at all times; When the reconstruction temperature value of any spatial node exceeds the first warning threshold And not exceeding the second warning threshold When a level-two yellow alert is triggered, an alert notification and the precise location coordinates of the hotspot are pushed to the management terminal, and the hotspot is marked on the three-dimensional temperature field cloud map; When the reconstruction temperature value of any spatial node exceeds the second warning threshold When a level three red alarm is triggered, the hot spot coordinates are mapped to the wind turbine array partition map, the wind turbine sub-array responsible for covering the hot spot area is identified, and a speed increase command is sent to the corresponding strain frequency driver to achieve precise local cooling. Among them, the first warning threshold The upper limit of the optimal preservation temperature for agricultural products, and the second warning threshold. This refers to the critical temperature at which agricultural products deteriorate in quality.
9. The intelligent monitoring and early warning system for agricultural product preservation and storage warehouses according to claim 8, characterized in that, After the directional fan subarray is adjusted, the system waits for the next 5-second sampling cycle to re-enter the three-dimensional temperature field reconstruction cycle and perform a verification scan on the hot spot area. When the hot spot is eliminated, the corresponding wind turbine subarray resumes its normal speed setting; when the hot spot persists, the directional speed increase command is maintained until the hot spot is eliminated, forming a closed-loop control verification mechanism.
10. A method for intelligent monitoring and early warning of agricultural product preservation and storage facilities, characterized in that, include: The system reads the inbound and outbound operation instructions of the warehouse control system through the API gateway, extracts the planned door opening time, estimated door opening duration, the quality of the current batch of goods, and the initial temperature of the goods upon entering the warehouse. Combined with the outside temperature and relative humidity collected by the weather sensors outside the warehouse, the system calculates the sensible heat disturbance, latent heat disturbance, and heat carried in by the goods introduced by the door opening, and summarizes them into the total heat and humidity load disturbance. This is converted into feedforward pre-cooling compensation power. Within a predetermined time window before the planned door opening time, the feedforward pre-cooling compensation power is superimposed on the feedback control output of the model predictive controller to drive the variable frequency refrigeration compressor unit to complete the pre-cooling and cold storage in advance. When the warehouse control system reports the end of the operation signal, it resumes pure feedback control. as well as In the offline cloud phase, a high-fidelity computational fluid dynamics model is constructed based on the actual geometric configuration of the target storage facility. Multiple sets of full-scale simulations under various operating conditions are performed to generate snapshot matrices. Intrinsic orthogonal decomposition modal basis functions are extracted through singular value decomposition. The reduced-order basis matrix is formed by truncating the model according to the criterion that the cumulative energy ratio is not less than 99.9%. The reduced-order observation matrix and pseudo-inverse matrix are pre-calculated, and the reduced-order model data package is sent to the industrial edge computing gateway. In the online simulation phase, the measured temperature values collected every 5 seconds by the temperature and humidity sensors in the storage facility are used as input. The pre-stored pseudo-inverse matrix is called to reconstruct the three-dimensional temperature field of the entire storage facility within 100 milliseconds through least squares matrix operations. The reconstructed temperature field is subjected to a node-by-node three-level threshold comparison. When the node temperature exceeds the first warning threshold, a warning notification and hotspot coordinates are pushed to the management terminal. When the node temperature exceeds the second warning threshold, a speed increase command is issued to the fan subarray responsible for covering the hotspot area to achieve precise local cooling. The hotspot area is then verified in the next sampling cycle.
