Agricultural greenhouse comprehensive environment regulation system with fault self-diagnosis

Abstract

The present application relates to the technical field of agricultural greenhouse environment control, and discloses a comprehensive environment control system for agricultural greenhouse with self-diagnosis of faults, comprising a main controller, a sensing node, a main actuator and a perturbation assembly comprising a micro-mist device and a circulating fan. When the tracking error integral is out of bounds, the main controller switches to the diagnosis mode and locks the main actuator, extracts the historical data of the target diagnosis sensing node to establish the baseline of natural weather evolution, dynamically selects the start of the micro-mist device or the circulating fan according to the saturated water vapor pressure difference, extracts the asymmetric ratio of the multiphase state feature evolution after deducting the above baseline, introduces the cross decoupling operator to diagnose mechanical faults or soft faults wrapped on the surface of the sensing node, and when the soft fault wrapped on the surface is diagnosed, the node is isolated and parameter compensation reconstruction and disease prevention control are performed. By introducing the perturbation signal and eliminating the background meteorological drift disturbance, the present application accurately identifies the hidden faults and fills the data blind area during the fault period.

Description

Technical Field

[0001] This invention relates to the field of agricultural greenhouse environmental control technology, specifically to an agricultural greenhouse integrated environmental control system with self-diagnosis of faults. Background Technology

[0002] Modern agricultural greenhouses are typically equipped with integrated environmental control systems. These systems use a network of sensors to collect meteorological parameters such as temperature and humidity in real time. The controller then processes these parameters to drive actuators such as ventilation windows, shading nets, or misting devices to maintain a suitable microclimate for crop growth. To ensure the reliability of the control system, existing systems usually have built-in basic fault alarm programs, primarily relying on setting safe upper and lower thresholds for environmental variables or equipment operating parameters to monitor the working status of the underlying hardware.

[0003] The microenvironment system inside a greenhouse exhibits thermodynamic characteristics of high inertia and strong coupling, and its environmental parameters are constantly affected by the superposition of external solar radiation and natural weather changes. Conventional static threshold diagnostic methods have limitations when dealing with latent faults in the system. When sensing node probes are coated with pesticide residue or dust due to long-term agricultural operations, or when actuators experience slight jamming leading to sluggish operation, environmental parameters often only show a slow accumulation of deviations. Existing diagnostic logic struggles to separate the actual physical anomalies of the equipment from the slow, natural weather drift within the greenhouse. It is not only insufficiently sensitive to detecting these latent soft faults and unable to effectively distinguish between actuator mechanical faults and sensor probe malfunctions, but also prone to misinterpreting normal environmental fluctuations as equipment failures when external weather conditions change abruptly, resulting in false alarms.

[0004] Some proactive fault diagnosis solutions fail to fully consider the thermodynamic state of the current greenhouse microenvironment when injecting test signals into the environment. Under high humidity conditions where the air is nearly saturated, fixed-mode water mist spraying tests result in ineffective evaporation of moisture, which instead condenses directly on the sensor probe surface. This not only hinders the transmission of thermodynamic pulse signals but also exacerbates sensor measurement lag. Furthermore, when the system confirms the failure of a local sensing node and physically isolates it, a blind spot in environmental data monitoring will appear in that area. Existing control systems lack effective data reconstruction and linkage intervention mechanisms in such situations. During periods of unmonitored local microclimate, temperature and humidity imbalances are highly likely to occur, leading to the formation of a persistent water film on the crop canopy and increasing the risk of fungal diseases growing inside the greenhouse. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an integrated environmental control system for agricultural greenhouses with self-diagnostic fault capabilities. This system solves the problem of sensor probes being prone to surface coating from pesticide residues and dust due to prolonged exposure to spraying and high humidity during greenhouse environmental control, which can lead to mechanical jamming of actuators in such environments. Existing diagnostic methods based on static thresholds or purely data-driven approaches cannot effectively isolate background drift disturbances caused by natural weather evolution in greenhouses, resulting in low sensitivity to latent faults and a high susceptibility to false alarms during sudden environmental changes.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of the present invention provides an integrated environmental control system for agricultural greenhouses with self-diagnosis of faults, including a main controller, sensing nodes distributed in different protection zones of the greenhouse, a main actuator for controlling the environment, and a perturbation component; the perturbation component includes a micro-fogging device and a circulating fan. The main controller is connected to the sensing nodes, the main actuator, and the perturbation component via a communication bus.

[0007] The main controller is configured to execute the following control logic:

[0008] When the tracking error integral of the main control loop exceeds the limit, the main actuator state is locked and the diagnostic mode is switched to select the target diagnostic sensing node.

[0009] Extracting historical data sequences from target diagnostic sensing nodes to establish a baseline for natural meteorological evolution;

[0010] Based on the calculated saturated water vapor pressure difference, physical field modal splitting is performed, and either a micro-mist device or a circulating fan is selected.

[0011] During the forced operation period of the perturbation component and the natural recovery period after shutdown, data were collected and the multiphase characteristic evolution asymmetry ratio after deducting the natural meteorological evolution baseline was extracted.

[0012] By using cross-decoupling operators to perform calculations on multi-dimensional features, mechanical faults in the main actuator or soft faults in the surface covering of sensing nodes can be diagnosed.

[0013] When a soft fault in the surface coating is diagnosed, the corresponding sensing node is isolated and spatial environmental parameter compensation reconstruction and closed-loop intervention control for preventing the spread of the fault are performed.

[0014] The innovative principle of the first aspect of this invention lies in changing the conventional diagnostic method of analyzing static sensor data. By introducing a perturbation component, the system can apply a physical forcing field of phase change cooling or boundary layer stripping to the local defense zone based on the current atmospheric physical carrying capacity state of the greenhouse. Combined with a pre-established baseline of natural meteorological evolution, the system eliminates data disturbances caused by macroscopic meteorological drift in the computational dimension. By extracting the asymmetric hysteresis characteristics of the probe surface microclimate in the stages of absorbing forced energy and dissipating energy, the difference in thermodynamic response is transformed into a discriminant index, and the failure modes of actuator mechanical displacement failure and sensor surface material encapsulation are distinguished based on physical mechanisms.

[0015] Furthermore, the micro-mist device is used to spray water mist into the local space where the target diagnostic sensing node is located under the call of the main controller, to establish a forced field for latent heat phase change cooling; the circulating fan is used to blow directional airflow under the call of the main controller, to establish a forced field for static boundary layer stripping.

[0016] Furthermore, the main controller integrates a ring buffer, which stores discrete sampling data sequences of sensing nodes within a set sliding time window, providing an independent data source for extracting the baseline of natural weather evolution.

[0017] Furthermore, the main controller also includes an independent alarm unit, which is used to send a data frame containing fault object identification information, geographic grid coordinates, and a confirmed fault type code to the monitoring center after a fault is diagnosed.

[0018] A second aspect of this invention provides a method for comprehensive environmental control of agricultural greenhouses with self-diagnosis of faults, applied to the aforementioned control system. The method includes the following steps:

[0019] S10, calculate the tracking error integral of the main control loop. When the tracking error integral is greater than the set trigger threshold, pause the main control loop, keep the current control state of the main actuator locked, and select the target diagnostic sensing node to switch to diagnostic mode.

[0020] S20: Extract the historical data sequence of the target diagnostic sensing node within a set time window, calculate the baseline drift slope of dry-bulb temperature and absolute humidity, and establish the baseline of natural weather evolution.

[0021] S30: Calculate the saturated water vapor pressure difference based on the current temperature and relative humidity collected by the target diagnostic sensing node. When the saturated water vapor pressure difference is greater than the critical phase change threshold, call the micro-mist device. When the saturated water vapor pressure difference is less than or equal to the critical phase change threshold, call the circulating fan.

[0022] S40: Data is collected during the forced operation period of the called perturbation component and the natural recovery period after shutdown. The first-order evolutionary derivatives of the variables are calculated and the asymmetric ratio of multiphase characteristic evolution is constructed.

[0023] S50 introduces the first-order evolutionary derivative and the asymmetric ratio of multiphase characteristic evolution into the cross-decoupling operator to diagnose the fault type;

[0024] S60: When a mechanical fault is diagnosed in the main actuator, the actuator safety handling branch is initiated; when a soft fault is diagnosed in the surface coating, the spatial environment parameters are reconstructed based on the surrounding normal sensing nodes, and the condensation risk index is assessed based on the reconstructed data. When the risk exceeds the standard, the main actuator is forced to perform dehumidification.

[0025] The innovative principle of the second aspect of this invention lies in the following: After diagnosing a soft fault on the surface of a sensing node, the system isolates the failed node to prevent interference with the main control loop, and reconstructs the meteorological elements of the blind zone by extracting data from adjacent valid nodes using an inverse distance weighting operator. This method introduces the reconstructed environmental parameters into the disease assessment model, using the hysteresis characteristics of water evaporation on the sensor surface as a physical mapping parameter for the residence time of water film on the crop canopy leaves. When the calculated condensation conditions reach the critical state for fungal disease occurrence, the system forcibly triggers a dehumidification command, establishing a linkage control mechanism between hardware fault handling and agronomic disease prevention.

[0026] Furthermore, in step S10, the formula for calculating the tracking error integral is as follows:

[0027] ;

[0028] In the formula, Indicates the current computation time The obtained integral value of the tracking error; This is the absolute system time of the current operation; This indicates the width of the pre-defined sliding time window; For continuous-time variables in integration, representing the sliding time window interval. Each historical time point within; Indicates in The instantaneous tracking error value is calculated by using feedback data from sensing node 200 or, in diagnostic mode, from feedback data from the target diagnostic sensing node; the absolute value sign is used to eliminate the mutual cancellation effect of positive and negative deviations during the integration process.

[0029] Furthermore, the target diagnostic sensing node selected in step S10 is the node with the largest error contribution value in the corresponding defense zone at the trigger time. The error contribution value is calculated by weighting the absolute values ​​of the current temperature deviation and humidity deviation of each candidate node.

[0030] Furthermore, in step S20, before calculating the baseline drift slope of dry-bulb temperature and absolute humidity, the actual measured relative humidity and dry-bulb temperature are converted into the absolute humidity of the air using the Magnus empirical formula, thus eliminating the influence of temperature changes on the moisture content calculation.

[0031] Furthermore, in step S20, the dry-bulb temperature baseline drift slope is calculated using a least-squares fitting algorithm based on discrete sampling points within a historical time window to obtain the natural evolution rate of the environment.

[0032] Furthermore, in step S30, the saturated water vapor pressure difference is calculated based on the product relationship between the current air saturated water vapor pressure and relative humidity.

[0033] Furthermore, in step S40, the calculated first-order evolutionary derivative includes the net forced evolutionary derivative and the net recovery evolutionary derivative; the net forced evolutionary derivative is obtained by subtracting the corresponding baseline drift slope of the natural meteorological evolutionary baseline from the first-order evolutionary derivative during the forced period; the net recovery evolutionary derivative is obtained by subtracting the corresponding baseline drift slope of the natural meteorological evolutionary baseline from the first-order evolutionary derivative during the natural recovery period.

[0034] Furthermore, in step S40, the formula for calculating the asymmetric ratio of multiphase characteristic evolution is as follows:

[0035] ;

[0036] In the formula, It represents the calculated asymmetric ratio of the multiphase characteristic evolution of dry-bulb temperature; the absolute value symbol is used to unify the scalar measurement dimension of positive heating and negative cooling, eliminating the interference of the direction of change on the asymmetric evaluation; This is a minimum error-proofing constant pre-written into the controller.

[0037] Furthermore, in step S50, importing the first-order evolutionary derivative and the multiphase characteristic evolution asymmetry ratio into the cross-decoupling operator includes: comparing the net forced evolutionary derivative and the multiphase characteristic evolution asymmetry ratio with the corresponding preset benchmark thresholds to generate a binarized current state feature vector; calculating the Manhattan distance between the current state feature vector and each typical fault template vector in the preset fault cross-diagnosis matrix; and taking the fault type corresponding to the smallest Manhattan distance as the final diagnosis result.

[0038] Furthermore, when it is calculated that there are two or more identical minimum Manhattan distances, the corresponding fault mode is selected as the diagnostic output according to the pre-set equipment hazard level list from high to low, thus avoiding decision conflicts.

[0039] Furthermore, in step S60, the spatial environment parameter compensation reconstruction is calculated using the inverse distance weighting principle; the reconstruction parameters are obtained by weighted summation using the reciprocal relationship between the measured values ​​of adjacent sensing nodes operating normally within the preset spatial radius and the square of the Euclidean distance between the nodes.

[0040] Furthermore, in step S60, the condensation risk index is calculated based on the relative humidity obtained from the compensated reconstruction and the exponential decay relationship of the difference between the compensated reconstruction dry-bulb temperature and the dew point temperature, and is used to quantify the potential for condensation of environmental moisture.

[0041] Furthermore, when a soft surface coating fault is diagnosed, the fault node is configured as a virtual leaf water film target, and the latent heat lock-in tail dissipation time presented during the natural recovery period is extracted as a physical mapping benchmark for the residence time of crop canopy leaf water film. Dehumidification actions are then performed in conjunction with the condensation risk index.

[0042] Furthermore, in step S60, when a mechanical fault is diagnosed in the main actuator, the main controller keeps the main actuator in a safe locked state or switches to a preset safe position, and calls the backup control strategy to take over the environmental control tasks of the corresponding zone.

[0043] This invention provides an integrated environmental control system for agricultural greenhouses with self-diagnosis capabilities. It offers the following advantages:

[0044] 1. This invention establishes a natural meteorological evolution baseline by extracting historical data sequences from target diagnostic sensing nodes through the main controller. During the forced operation period of the perturbation component and the natural recovery period after shutdown, it extracts the multiphase characteristic evolution asymmetry ratio after deducting the baseline. This mechanism, at the data processing level, removes background disturbances caused by the slow, natural drift of the greenhouse environment, enabling objective differentiation between mechanical faults of actuators and soft faults on the surface of sensing probes. It solves the problems of conventional static threshold diagnostic methods being insensitive to latent faults and prone to false alarms during sudden environmental changes.

[0045] 2. This invention calculates the saturated vapor pressure difference in the local area where the target diagnostic sensing node is located and executes a physical field mode splitting strategy. Based on the current atmospheric physical carrying capacity of the greenhouse, the system activates a micro-mist device to cool the air when the saturated vapor pressure difference exceeds the critical phase transition threshold, and activates a circulating fan to strip away the airflow boundary layer when the humidity is less than or equal to this threshold. This dynamic selection of perturbation methods avoids forced spraying under high humidity conditions, which could lead to liquid accumulation on the probe surface, ensuring the effectiveness of the thermodynamic pulse signal injection and the safety of the diagnostic process.

[0046] 3. After diagnosing a soft fault on the surface of a sensing node, this invention isolates the failed node and uses data from adjacent normal nodes within a preset spatial radius to perform inverse distance-weighted parameter compensation reconstruction. The system further uses the reconstructed environmental factors to calculate a condensation risk index, and forces the actuator to dehumidify when the index exceeds the limit. This structure fills the data monitoring blind spot caused by local hardware failures, establishes a closed-loop linkage control logic between underlying equipment failure handling and agronomic disease prevention, and reduces the risk of fungal diseases growing in the crop canopy during sensor control failures. Attached Figure Description

[0047] Figure 1 This is a system framework diagram of the present invention;

[0048] Figure 2 This is a flowchart of the method of the present invention;

[0049] Figure 3 This is a timing diagram of steady-state error triggering and system state switching in this invention;

[0050] Figure 4 This is a schematic diagram illustrating the principle of historical data extraction and meteorological evolution baseline establishment in this invention.

[0051] Figure 5 This is a schematic diagram of the physical field modal current splitting control principle of the present invention;

[0052] Figure 6 This is a time-series diagram of the dual-field perturbation injection and multiphase feature evolution extraction of the present invention;

[0053] Figure 7 This is a schematic diagram illustrating the principle of cross-matrix decoupling and system fault diagnosis in this invention.

[0054] Figure 8 This is a schematic diagram of the spatial compensation reconstruction and disease prevention closed-loop principle of the present invention;

[0055] Figure 9 This is a three-dimensional manifold topology distribution diagram of the fault diagnosis features of the present invention;

[0056] Figure 10 This is a spatial heat map for the compensation reconstruction of the microenvironment in the perception blind zone of the present invention.

[0057] Among them, 100 is the main controller; 200 is the sensing node; 300 is the main actuator; 400 is the perturbation component; 410 is the micro-mist device; and 420 is the circulating fan. Detailed Implementation

[0058] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] See attached document Figure 1 , Figure 1 This is an architecture diagram of an adaptive fault diagnosis system for agricultural greenhouses according to an embodiment of the present invention. The present invention provides an adaptive fault diagnosis system for agricultural greenhouses, including a main controller 100, a sensing node 200, a main actuator 300, and a perturbation component 400.

[0060] The main controller 100 is connected to the sensing node 200, the main actuator 300, and the perturbation component 400 via a communication bus. The main controller 100 executes the environmental closed-loop control algorithm and the multiphase cross-decoupling diagnostic logic. The main controller 100 has built-in canopy disease prevention logic, configuring the target diagnostic sensing node 200, which is diagnosed with a soft surface coating fault, as a virtual leaf water film target. It uses the evaporation lag characteristics of the coating layer microclimate to inversely determine the disease risk of greenhouse crops. The virtual leaf water film target is a physical mapping object in the disease risk assessment, used to characterize the possible leaf water film retention trend in the local canopy, rather than equating the sensing node measurement with the actual measured value of the leaf surface.

[0061] Sensing nodes 200 are distributed in different protection zones inside the greenhouse, collecting environmental temperature and humidity data and transmitting it to the main controller 100. In diagnostic mode, the main controller 100 selects a target diagnostic sensing node from the sensing nodes 200 for subsequent fault diagnosis. When multiple candidate nodes exist, the diagnostic judgment is performed sequentially according to a preset priority order. The main actuator 300 receives control commands from the main controller 100 to adjust the mechanical opening. The perturbation component 400 includes a mist device 410 and a circulating fan 420. The mist device 410 establishes a latent heat phase change forced field in the local space; the circulating fan 420 establishes a forced convection boundary layer in the local space.

[0062] See attached document Figure 2 , Figure 2 This is a flowchart of an adaptive fault diagnosis method for agricultural greenhouses according to an embodiment of the present invention. The present invention provides an adaptive fault diagnosis method for agricultural greenhouses, comprising the following steps:

[0063] S10, calculate the tracking error integral of the main control loop, pause the main control loop when the tracking error integral exceeds the limit, keep the current control state of the main actuator 300 locked, and switch to diagnostic mode;

[0064] S20, extract the historical data sequence of the target diagnostic sensing node 200 within a set time window, calculate the baseline drift slope of dry-bulb temperature and absolute humidity, and establish the natural meteorological evolution baseline; wherein, the target diagnostic sensing node 200 is preferably the node with the largest error contribution value in the corresponding defense zone at the trigger time, or the node with the highest correlation with the out-of-bounds main control variable; when there are multiple candidate nodes, the main controller 100 sequentially switches to the diagnostic mode for discrimination according to the preset priority order; wherein, the error contribution value is preferably calculated by weighting the absolute values ​​of each candidate node to the current temperature deviation and humidity deviation, the correlation is preferably determined according to the consistency or correlation coefficient between the candidate node measurement value and the change trend of the out-of-bounds main control variable, and the preset priority order is preferably determined based on at least one of the following: the importance of the defense zone, the frequency of historical anomalies of the node, and the spatial proximity to the action area of ​​the main actuator 300;

[0065] S30, calculate the saturated water vapor pressure difference based on the current temperature and relative humidity collected by the target diagnostic sensing node 200, perform physical field mode splitting based on the saturated water vapor pressure difference value, and choose to call the micro-mist device 410 to construct a phase change cooling environment or call the circulating fan 420 to perform static boundary layer stripping.

[0066] S40, during the forced operation period of the called perturbation component 400 and the natural recovery period after shutdown, continuously collect data from the target diagnostic sensing node 200, calculate the first-order evolutionary derivatives of physical variables at different phase evolution stages, and construct the multi-phase characteristic evolution asymmetric ratio.

[0067] S50 introduces the first-order evolutionary derivative and the asymmetric ratio of multiphase feature evolution into the cross-decoupling operator, and diagnoses the mechanical fault of the main actuator 300 or the soft surface covering fault of the target diagnostic sensing node 200 based on the topological distribution state of the multidimensional features; when the main actuator 300 is diagnosed as having a mechanical fault, the system switches to the actuator fault safety handling branch; when the target diagnostic sensing node 200 is diagnosed as having a soft surface covering fault, the system switches to the sensing node compensation and fault prevention branch.

[0068] S60, when the confirmed result is a soft fault on the surface of the target diagnostic sensing node 200, the spatial environment parameter compensation and reconstruction algorithm is triggered based on the confirmed soft fault on the surface of the target diagnostic sensing node 200 to perform parameter compensation. At the same time, the latent heat locking tail dissipation time of the target diagnostic sensing node 200 during the natural recovery period is extracted. The latent heat locking tail dissipation time is used as the physical mapping benchmark for the residence time of the crop canopy leaf water film. When the physical mapping benchmark exceeds the critical time for fungal disease reproduction, the conventional agronomic instructions are intercepted and the main actuator 300 is forced to perform dehumidification action.

[0069] When the diagnosis result is a mechanical fault in the main actuator 300, the main controller 100 keeps the main actuator 300 in a safe locked state or switches to a preset safe position, while outputting a maintenance signal and fault result, and the corresponding zone control is taken over by a backup control strategy or manual intervention; wherein, the backup control strategy is preferably at least one of the following: a simplified closed-loop control strategy executed based on the reconstruction result of adjacent normal sensing nodes, a preset safe opening control strategy, or a backup environment control strategy executed according to a time period.

[0070] After completing the corresponding fault handling, the main controller 100 exits the diagnostic mode according to the fault type: for soft faults on the surface of the target diagnostic sensing node 200, the fault node is kept isolated and the compensation result participates in subsequent control; for mechanical faults of the main actuator 300, after exiting the diagnostic mode, it continues to maintain the safety lock or safety position state until manual reset, maintenance is completed or the standby actuator takes over and the fault handling state is lifted.

[0071] The following will provide a detailed explanation of the specific implementation process and physical logic of each step in the above method.

[0072] See attached document Figure 3 , Figure 3 This is a timing diagram of steady-state error triggering and system state switching according to an embodiment of the present invention. In this embodiment, during the normal operation phase of executing the environmental closed-loop control algorithm, the main controller 100 reads the set target value of the greenhouse environment and the actual measured value fed back by the sensing node 200 according to the set discrete sampling period; after triggering the diagnostic mode, the target diagnostic sensing node 200 is selected to perform subsequent diagnostic calculations. By performing difference calculation between the set target value and the actual measured value, the system can generate instantaneous tracking error.

[0073] The greenhouse environment control includes a temperature and humidity decoupling algorithm. The instantaneous tracking error is preferably defined as the weighted combined error of temperature and humidity errors, or as the error of the master control variable currently participating in the main closed-loop control. As a preferred approach, the instantaneous tracking error... It can be represented as:

[0074] ;

[0075] In the formula, and For the preset weighting coefficients, and These represent the temperature setting and humidity setting, respectively. and These represent the actual temperature and humidity values ​​fed back by the sensing nodes, respectively. The instantaneous tracking error is specifically characterized as the deviation of the current master control variable from the agronomic target environment. For the discrete-time sampling logic and basic control parameter reading logic in conventional environmental closed-loop control algorithms, those skilled in the art can perform conventional configurations based on the underlying architecture of standard programmable logic controllers. The instruction execution and register call methods are well-known technologies in this field and will not be elaborated upon here.

[0076] To detect latent anomalies in the system, the main controller 100 establishes a continuous observation region that shifts backward along the time axis, i.e., a sliding time window. The system uses an integral algorithm to mathematically integrate the instantaneous tracking error accumulated within the sliding time window to obtain the integral value of the tracking error.

[0077] From the perspective of the physical mechanism of agricultural greenhouses, the greenhouse thermodynamic system exhibits large inertia and large hysteresis. Sudden changes in external meteorological factors can easily cause transient fluctuations in the feedback values ​​of sensing node 200. In diagnostic mode, this specifically manifests as transient fluctuations in the feedback values ​​of the target diagnostic sensing node 200. Introducing an integral operation on the absolute value of the instantaneous tracking error can effectively smooth and filter high-frequency instantaneous disturbances, reflecting the true cumulative deviation of energy and material exchange in the microenvironment over time, and avoiding frequent false triggers of the fault diagnosis program due to occasional disturbances. The specific formula for calculating the integral value of the tracking error is as follows:

[0078] ;

[0079] In the formula, Indicates the current computation time The obtained integral value of the tracking error; This is the absolute system time of the current operation; This indicates the width of the pre-defined sliding time window; For continuous-time variables in integration, representing the sliding time window interval. Each historical time point within; Indicates in The instantaneous tracking error value is calculated by using feedback data from sensing node 200 or, in diagnostic mode, from feedback data from the target diagnostic sensing node; the absolute value sign is used to eliminate the mutual cancellation effect of positive and negative deviations during the integration process.

[0080] As a preferred method, the width of the sliding time window The value is calibrated by combining the mechanical stroke cycle of the main actuator 300 with the volume of the greenhouse space. The value is the duration of 3 to 5 complete closed-loop control cycles to ensure that a complete environmental response hysteresis band can be covered.

[0081] After acquiring the integral characteristics, the main controller 100 inputs the real-time acquired tracking error integral value into the comparator and performs a numerical comparison logic operation with the trigger threshold pre-written in the memory. The trigger threshold characterizes the system's tolerance limit to environmental deviations, and its specific value is determined by looking up a table based on the sensitivity of the planted crop and the agronomic requirements of its growth stage.

[0082] When the integral value of the tracking error is less than or equal to the trigger threshold, the main controller 100 determines that the control system is operating within a reasonable fault tolerance range and continues to maintain the original environmental closed-loop control algorithm. Conversely, when the integral value of the tracking error is greater than the trigger threshold, the main controller 100 determines that the system may have a hidden physical fault such as slow actuator response or sensor link deviation, and then triggers a high-level interrupt event.

[0083] Preferably, before entering the diagnostic mode, the main controller 100 further determines whether the current environmental parameters exceed the preset absolute safety threshold; if they exceed the absolute safety threshold, it prioritizes the execution of safety interlock control or minimum intervention protection control, instead of directly entering the full lock diagnostic mode.

[0084] Upon triggering an interrupt event, the main controller 100 performs a low-level system state switch. Specifically, the main controller 100 suspends regular main control loop operation instructions, stops outputting control levels for dynamically adjusting environmental elements, and simultaneously issues a state lock command to the main actuator 300. Provided that an absolute safety interlock is not triggered, the main actuator 300 remains in the control state at the time of the out-of-bounds trigger, providing a relatively stable observation boundary for subsequent perturbation diagnosis. The command essentially disconnects the electrical link between the main actuator 300 and subsequent regular control laws, ensuring that the mechanical components of the main actuator 300 remain physically in the same position as at the time of the out-of-bounds trigger.

[0085] After completing the state locking operation, the system enters diagnostic mode. Through the state locking operation, the main control loop is blocked from continuously intervening in the microenvironmental meteorological elements of the greenhouse in terms of control logic. This eliminates the thermodynamic fluctuations introduced by external large-scale control variables, and builds a relatively stable observation basis with no obvious artificial interference for the subsequent injection of characteristic physical signals into the perturbation component 400.

[0086] See attached document Figure 4 , Figure 4 This is a schematic diagram illustrating the principle of historical data extraction and meteorological evolution baseline establishment according to an embodiment of the present invention. In this embodiment, the specific implementation process of historical data extraction and natural meteorological evolution baseline establishment in the adaptive fault diagnosis method for agricultural greenhouses can be divided into the following steps:

[0087] The moment the main controller 100 enters diagnostic mode, it accesses its internal circular buffer to extract discrete sampled data sequences from the target diagnostic sensing nodes 200 within a set historical time window. If the system adopts a zoned control architecture, it extracts data from the target diagnostic sensing nodes in the corresponding zone that triggered the anomaly; if the system adopts a multi-node collaborative diagnostic architecture, it executes subsequent steps on each candidate node in a preset order.

[0088] Inside agricultural greenhouses, environmental parameters are not absolutely static, but rather exhibit a slow, natural drift trend influenced by diurnal variations in external solar radiation and crop physiological transpiration. The main purpose of obtaining this historical sequence is to calculate the system's baseline rate of change without external intervention, thereby preventing the misinterpretation of normal microclimate drift as an anomalous response to physical perturbation signals. As a preferred approach, the length of the historical time window is set to 10 to 15 minutes, with a sampling period of 10 seconds. The time window length is chosen to cover enough data points to fit a smooth evolutionary trend while avoiding excessively large time spans that could introduce high-frequency interference from large-scale weather abrupt changes.

[0089] The target diagnostic sensing node 200 directly measures and outputs data including dry-bulb temperature and relative humidity. Relative humidity is significantly affected by changes in dry-bulb temperature, making it difficult to independently reflect the absolute moisture content in the air. To accurately assess the evaporation potential of moisture and the latent heat phase transition state of the probe surface in subsequent physical field diagnostics, the main controller 100 uses Magnus's empirical formula to convert relative humidity into absolute humidity. This conversion process involves thermodynamic state decoupling calculations in engineering, providing a reference standard unaffected by temperature fluctuations for subsequent mass exchange analysis. The specific conversion formula is as follows:

[0090] ;

[0091] ;

[0092] In the formula, Indicates in The dry-bulb temperature measured by the target diagnostic sensing node 200 at any given time, in degrees Celsius; Indicates in The saturated vapor pressure of air at any given time, expressed in hectopascals (hPa). is the base of the natural logarithm; Indicates in The percentage value of relative humidity measured at any given time; The absolute humidity calculated after conversion is expressed in grams per cubic meter. The constants 6.112, 17.67, 243.5, and 2.1667 are fixed standard empirical coefficients in Magnus's formula and the ideal gas equation of state. For the instruction calls and floating-point conversions performed by the underlying microprocessor unit for logarithmic or exponential calculations, those skilled in the art can use standard mathematical function libraries for compilation and configuration; the program implementation methods are well-known in the field and will not be elaborated upon here.

[0093] After acquiring the pre-processed absolute humidity sequence and the original dry-bulb temperature sequence, the main controller 100 uses a least-squares fitting algorithm to calculate the baseline drift slope of dry-bulb temperature and absolute humidity within the historical time window. The baseline drift slope, in a physical sense, represents the natural evolution rate of the greenhouse microenvironment under conditions of no external forced intervention. The specific fitting calculation formula is as follows:

[0094] ;

[0095] ;

[0096] In the formula, The reference drift slope representing the dry-bulb temperature; The baseline drift slope representing absolute humidity; This represents the total number of sampled data points included within the historical time window. To ensure the effectiveness of the fitting calculation, the denominator should not be zero. The lower limit for the value is set to 10; Indicates the first The timestamps corresponding to each sampling point are in seconds; This represents the arithmetic mean of the timestamps of all sample points; and They represent the first Dry bulb temperature measurement and absolute humidity conversion values ​​at each sampling point; and These represent the arithmetic mean of dry-bulb temperature and absolute humidity within the historical time window, respectively.

[0097] Calculated and Together, these constitute the baseline characteristics of the current microenvironment's natural meteorological evolution. Establishing the baseline drift slope provides a reference value for computational comparison throughout the adaptive diagnostic logic. In the subsequent stage of injecting forced evolution signals through the perturbation component 400, the main controller 100 subtracts the corresponding baseline drift slope from the actually observed rate of change of variables, thereby numerically stripping away the superimposed interference of the natural evolution of the environment and extracting the net physical response parameters purely caused by the perturbation component.

[0098] See attached document Figure 5 , Figure 5 This is a schematic diagram of the physical field modal flow splitting control principle according to an embodiment of the present invention. In this embodiment, the specific implementation process of saturated water vapor pressure difference calculation and physical field modal flow splitting in the adaptive fault diagnosis method for agricultural greenhouses can be divided into the following:

[0099] After establishing the baseline of natural meteorological evolution, the main controller 100 needs to extract the latest dry-bulb temperature and relative humidity values ​​sampled by the target diagnostic sensing node 200 at the current moment to calculate the saturated vapor pressure difference of the local microenvironment. The saturated vapor pressure difference is a key thermodynamic state parameter, its physical meaning being to characterize the remaining potential of the current air to hold additional water vapor. The saturated vapor pressure difference provides data support for the system to determine the atmospheric carrying capacity of the microenvironment and decide on the physical forcing methods selected by the subsequent perturbation component 400. The main controller 100 completes the calculation using the difference between the current saturated vapor pressure and the actual vapor pressure. The specific formula for calculating the saturated vapor pressure difference is as follows:

[0100] ;

[0101] In the formula, Indicates the current moment The calculated saturated water vapor pressure difference is expressed in hectopascals. Indicates in The current air saturation vapor pressure is calculated using the Magnus empirical formula, and the unit is hectopascals. Indicates that the target diagnosis sensing node 200 is in The actual measured percentage of relative humidity at any given time.

[0102] After obtaining the current saturated vapor pressure difference, in order to construct the decision boundary for the physical field modes, the main controller 100 calculates the saturated vapor pressure difference. The value is compared with a preset critical phase transition threshold. This critical phase transition threshold defines the thermodynamic critical point at which ambient air transitions from an evaporative state to a water vapor condensation state. If water mist is forcibly injected into a relatively humid, near-saturated air environment, the droplets will be unable to absorb ambient heat through flash evaporation and may instead directly adhere to the probe surface, forming a liquid water film, increasing the risk of sensor detection lag or data distortion. As a preferred approach, the critical phase transition threshold is set to a range of 3 hPa to 5 hPa. Those skilled in the art can calibrate specific threshold constants within this range based on the internal spatial structure of the greenhouse and the ventilation frequency.

[0103] Based on the numerical comparison results above, the main controller 100 executes a bidirectional flow splitting strategy for the physical field modes. The system selects one of the two non-overlapping control branches to conduct, based on the current atmospheric physical carrying capacity of the greenhouse.

[0104] When the saturated water vapor pressure difference When the latent heat threshold is exceeded, the main controller 100 determines that the ambient air is relatively dry and has sufficient space for latent heat exchange. At this time, the system enters the phase change drive branch, and the main controller 100 outputs a control level to activate the micro-mist device 410. After the micro-mist device 410 is activated, it sprays micron-sized water mist into the local space where the target diagnostic sensing node 200 is located. The tiny droplets vaporize in the unsaturated air, absorbing the sensible heat from the surrounding air and the probe surface of the target diagnostic sensing node 200, thus creating a phase change cooling environment in the space. This cooling effect provides a thermodynamic pulse signal for subsequent evaluation of the probe response rate.

[0105] When the saturated water vapor pressure difference When the humidity is less than or equal to the critical phase transition threshold, the main controller 100 determines that the current microenvironment is in a high-humidity state, making it difficult for liquid moisture to evaporate effectively. At this time, the system enters the pneumatic drive branch, and the main controller 100 calls the circulating fan 420. After the circulating fan 420 starts, it blows a directional airflow into the area where the target diagnostic sensing node 200 is located. The airflow disrupts and peels off the stagnant air layer attached to the probe shell surface of the target diagnostic sensing node 200 from a hydrodynamic perspective, performing a static boundary layer stripping operation. This flow field reconstruction exposes the probe to the mainstream air, introducing temperature disturbances by enhancing convective heat transfer, thereby reducing the adverse effects of phase transition inhibition on the diagnostic process under high humidity conditions.

[0106] See attached document Figure 6 , Figure 6 This is a time-series diagram of dual-field perturbation injection and multiphase feature evolution extraction according to an embodiment of the present invention. In this embodiment, the specific implementation process of dual-field perturbation injection and multiphase feature evolution extraction in the adaptive fault diagnosis method for agricultural greenhouses is described below:

[0107] Based on the physical field mode splitting results executed previously, the main controller 100 issues a fixed-pulse-width drive command to the selected perturbation component 400. The selected perturbation component 400 receives the command and starts, establishing a physical perturbation field in the local space. The continuous time interval during which the perturbation component 400 operates is defined as the forcing period in the system logic. As a preferred approach, the pulse width of the drive command is set to 60 to 120 seconds to ensure that the microclimate on the probe surface can absorb or dissipate sufficient heat, thereby generating quantifiable thermodynamic response characteristics. During the forcing period, the main controller 100 continuously acquires dry-bulb temperature and relative humidity data fed back by the target diagnostic sensing node 200 according to a set discrete sampling period, converts the relative humidity to absolute humidity using the Magnus formula, and thus forms the forcing period state sequence.

[0108] After obtaining the state sequence during the forced period, the main controller 100 calls its internal computing unit to calculate the first-order evolutionary derivatives of dry-bulb temperature and absolute humidity during the forced period using a linear regression algorithm. To obtain the physical response characteristics purely induced by the perturbation component 400, the main controller 100 subtracts the slope of the previously established natural meteorological evolution baseline from the first-order evolutionary derivative during the forced period. This operation essentially performs mathematical stripping of the environmental background drift, eliminating error interference caused by the slow changes in the greenhouse microclimate itself, and ultimately extracts the net forced evolutionary derivative. The specific formula for calculating the net forced evolutionary derivative of dry-bulb temperature is as follows:

[0109] ;

[0110] In the formula, This represents the net forced evolution derivative of dry-bulb temperature after deducting the natural meteorological background. This represents the original first derivative of the dry-bulb temperature obtained by fitting a sequence of measured data during the forcing period; This represents the established dry-bulb temperature baseline drift slope. The calculation logic for the net forced evolution derivative of absolute humidity is consistent with this. For the instruction calculation process of linear difference and linear regression fitting executed by the underlying microprocessor, those skilled in the art can call standard mathematical function libraries to complete the configuration. The fitting implementation principle is a well-known technology in this field and will not be elaborated here.

[0111] At the end of the pulse width time set by the drive command, the main controller 100 cuts off the power to the perturbation component 400. The physical forcing effect of the local space disappears, and the microclimate on the probe surface begins to undergo natural heat and mass exchange under the convection and mass transfer of the surrounding mainstream air. The free evolution time interval after the perturbation stops is defined as the natural recovery period. During the natural recovery period, the main controller 100 continuously acquires discrete data sequences from the target diagnostic sensing node 200. As a preferred approach, the data acquisition duration during the natural recovery period is set to 300 to 600 seconds to cover the physical phase transition hysteresis process of surface water film evaporation and boundary layer reconstruction.

[0112] The main controller 100 uses the same linear regression logic to calculate the original first derivative of the recovery period and subtracts the baseline drift slope to obtain the net recovery evolution derivative. The formula for calculating the net recovery evolution derivative of dry-bulb temperature is as follows:

[0113] ;

[0114] In the formula, This represents the net recovery evolution derivative of dry-bulb temperature; This represents the original first derivative of the dry-bulb temperature obtained by fitting a sequence of measured data during the natural recovery period.

[0115] After extracting derivatives at different stages, the main controller 100 imports the net forced evolution derivative and the net recovery evolution derivative into a preset mathematical model to construct the multiphase characteristic evolution asymmetry ratio. From a thermodynamic perspective, the multiphase characteristic evolution asymmetry ratio reflects the rate difference between the microclimate on the probe surface during the forced energy absorption stage and the energy dissipation stage. When there is a soft coating layer such as dust or mud on the external dust cover of the target diagnostic sensing node 200 or on the probe surface, moisture penetrates into the coating layer during the forced period and slowly evaporates during the recovery period due to capillary action, causing the evolution derivative during the recovery period to decay. At this time, the constructed asymmetry ratio value will deviate from the normal range. The specific construction formula for the dry-bulb temperature multiphase characteristic evolution asymmetry ratio is as follows:

[0116] ;

[0117] In the formula, It represents the calculated asymmetric ratio of the multiphase characteristic evolution of dry-bulb temperature; the absolute value symbol is used to unify the scalar measurement dimension of positive heating and negative cooling, eliminating the interference of the direction of change on the asymmetric evaluation; This is a minimum error-proofing constant pre-written into the controller. The error-proofing constant is introduced to prevent the net forced evolution derivative from approaching zero due to electrical disconnection or mechanical jamming of the perturbation component (400°), ensuring that the microprocessor's divider will not overflow and crash. As a preferred method, The value is set to 0.0001.

[0118] See attached document Figure 7 , Figure 7 This is a schematic diagram illustrating the principle of cross-matrix decoupling and system fault diagnosis according to an embodiment of the present invention. In this embodiment, the specific implementation process of cross-matrix decoupling and system fault diagnosis in the adaptive fault diagnosis method for agricultural greenhouses can be divided into the following steps:

[0119] After completing the dual-field perturbation injection and multiphase feature evolution extraction, in order to reduce the computational load on the underlying microprocessor and improve the robustness of pattern matching, the main controller 100 needs to convert the acquired continuous physical response features into discrete logic states that are easy for the computer to process. Specifically, the main controller 100 calls its internal comparator logic to compare the net forced evolution derivative of dry-bulb temperature, the net forced evolution derivative of absolute humidity, the asymmetric ratio of multiphase feature evolution of dry-bulb temperature, and the asymmetric ratio of multiphase feature evolution of absolute humidity with their respective preset benchmark thresholds to generate a binary current state feature vector. The specific vector construction logic formula is as follows:

[0120] ;

[0121] ;

[0122] ;

[0123] In the formula, The current state feature vector is represented by a four-dimensional column vector. and These represent the Boolean determination results for the dry-bulb temperature-related characteristics, respectively. and The Boolean determination result representing the absolute humidity-related characteristics, and its determination calculation logic are similar to... and Complete correspondence, among which, The comparison results are used to characterize the evolutionary derivative of net absolute humidity forcing relative to the corresponding derivative response threshold. The comparison results are used to characterize the asymmetric ratio of the multiphase feature evolution of absolute humidity relative to the corresponding asymmetric hysteresis threshold; This represents the preset derivative response threshold; This represents the preset asymmetric hysteresis threshold.

[0124] As a preferred approach, the derivative response threshold The determination is based on calibration tests conducted in a standard greenhouse environment on brand-new, fault-free sensing node samples, or standard nodes of the same type as the target diagnostic sensing node, subjecting them to physical forcing with the same pulse width. The statistical average of multiple test data is taken and then lowered by 20% as the benchmark; asymmetric hysteresis threshold. The value range is set to 1.5 to 2.5 to determine whether the probe has severe moisture adsorption and evaporation inhibition.

[0125] After constructing the current state feature vector, it needs to be mapped to a specific physical fault space. The main controller 100 reads the pre-set fault cross-diagnosis matrix in the memory, decouples the current state feature vector from the fault cross-diagnosis matrix, and thus determines the specific fault mode. In this embodiment, the row directions of the fault cross-diagnosis matrix contain theoretical feature template vectors of various typical physical faults. The fault cross-diagnosis matrix is ​​preferably established through offline calibration and includes at least a normal state template, a main actuator 300 mechanical jamming template, a main actuator 300 slow response template, and a sensing node surface-wrapped soft fault template; each template vector is generated by statistically analyzing the mode or averaging of binary feature vectors obtained from repeated experiments under the corresponding fault conditions and then thresholding them, and is pre-written into the memory of the main controller 100.

[0126] The system quantifies the degree of matching by calculating the Manhattan distance between the current state feature vector and each fault template vector. The specific distance calculation formula is as follows:

[0127] ;

[0128] In the formula, Represents the relationship between the current state feature vector and the first... Manhattan distance between typical fault templates; Represents the current state feature vector The first in One element; Represents the cross-diagnosis matrix of faults In the middle, corresponding to the first The first type of fault template vector One theoretical element.

[0129] The main controller performs 100 iterations to calculate all... The system calculates the fault type corresponding to the smallest distance as the final diagnosis. At the system code implementation level, in the extreme case of equal matching distances, direct output might lead to system decision confusion. Therefore, the main controller 100 selects the corresponding fault mode for output according to a pre-set list of equipment hazard levels, from highest to lowest. This matching mechanism, which introduces priority determination, effectively avoids the algorithm's decision-making dead zone under equal distance conditions and improves the completeness of the diagnostic logic under all operating conditions.

[0130] Once the fault is accurately identified, the system enters the safety protection process. The alarm module is configured as an integrated functional unit of the main controller 100 or as an independent alarm unit. When the main controller 100 confirms that the target diagnostic sensing node 200 has an abnormal fault mode, it triggers the alarm module to execute fault alarm, control isolation, and alternative takeover response; when the main controller 100 confirms that the main actuator 300 has a mechanical fault mode, it triggers the alarm module to execute actuator shutdown alarm and safety position hold response. The main controller 100 generates a data frame containing fault object identification information, geographic grid coordinates, and a confirmed fault type code. After receiving the data frame, the alarm module transmits it to the host computer in the monitoring center or the mobile terminal of the maintenance personnel via a wireless radio frequency channel.

[0131] To prevent distorted data output by the target diagnostic sensing node 200 from causing malfunctions in actuators such as shading nets and ventilation windows inside the greenhouse, the main controller 100 will temporarily suspend all environmental closed-loop control strategies involving the faulty node at the software level, allowing neighboring normal nodes to take over the data acquisition tasks for the local area. For network transmission encapsulation, checksum generation, and wireless routing distribution of data frames, those skilled in the art can use standard IoT protocol stack function libraries for configuration. The underlying network communication driver mechanism is well-known in the field and will not be elaborated upon here.

[0132] See attached document Figure 8 , Figure 8This is a schematic diagram illustrating the spatial compensation reconstruction and disease prevention closed-loop principle according to an embodiment of the present invention. In this embodiment, the specific implementation process of the spatial compensation reconstruction and disease prevention closed-loop in the adaptive fault diagnosis method for agricultural greenhouses is described below:

[0133] After suspending and isolating the faulty target diagnostic sensing node 200, the main controller 100 needs to restore the environmental data monitoring capability of the local physical space where the faulty node is located. To fill the data acquisition blind spot caused by hardware failure, the main controller 100 calls a spatial interpolation algorithm to perform spatial environmental parameter compensation reconstruction. From the spatial distribution characteristics of the greenhouse microclimate, meteorological parameters in nearby locations have strong spatial correlation. Therefore, the main controller 100 extracts data from multiple normally operating adjacent sensing nodes within a preset spatial radius centered on the faulty node, and calculates the reconstructed dry-bulb temperature and reconstructed relative humidity at the blind spot location based on the inverse distance weighting principle; the adjacent sensing nodes are sensing nodes in normal working condition other than the target diagnostic sensing node. The reconstruction mechanism is weighted according to geographical proximity; the closer the effective node is to the faulty node, the greater the weight of its collected data in the blind spot estimation. The specific formula for calculating the spatial compensation reconstruction of dry-bulb temperature is as follows:

[0134] ;

[0135] In the formula, This represents the reconstructed dry-bulb temperature of the calculated fault location; This represents the total number of adjacent sensing nodes operating normally within a preset spatial radius; Indicates the first The dry-bulb temperature currently measured by each adjacent sensing node; Indicates the first The linear Euclidean distance between each adjacent sensing node and the faulty node. The calculation logic for reconstructing relative humidity corresponds exactly to this. For reading the coordinate data of adjacent sensing nodes and geometrically solving the Euclidean distance, those skilled in the art can call the standard spatial calculation function library to complete the configuration. The coordinate calculation process is a well-known technology in this field and will not be described in detail here.

[0136] After acquiring the reconstructed environmental data, considering the risk of crop diseases induced by microclimate fluctuations due to the temporary loss of real-time monitoring from the actual physical probes in the target area, the main controller 100 initiates a disease latent assessment logic. The system uses the reconstructed dry-bulb temperature and relative humidity to assess the probability of water vapor condensation in local blind spots. In agricultural greenhouse environments, continuous condensation on plant leaf surfaces is a significant factor in inducing fungal diseases such as downy mildew and gray mold. The main controller 100 quantifies potential hazards by constructing a condensation risk index. The specific formula for calculating the condensation risk index is as follows:

[0137] ;

[0138] In the formula, This represents the calculated local condensation risk index. This represents the percentage of relative humidity obtained from the reconstruction. Indicates the reconstructed dry-bulb temperature; This represents the dew point temperature calculated using the Magnus empirical formula based on the reconstructed dry-bulb temperature and reconstructed relative humidity. A preset temperature sensitivity coefficient is used to adjust the effect of temperature difference on exponential decay. As a preferred method, the temperature sensitivity coefficient... The value is set to 5.0, which is based on the thermodynamic cold damage conduction characteristics of greenhouse crops.

[0139] To prevent the alternation of healthy and diseased physical environmental conditions, the main controller 100 compares the calculated condensation risk index with a preset anti-condensation control threshold. The anti-condensation control threshold defines the boundary parameter for the microenvironment to evolve from a safe state to a condensation state. As a preferred approach, the anti-condensation control threshold is set to a value between 0.85 and 0.92.

[0140] When the condensation risk index is greater than or equal to the anti-fouling control threshold, the main controller 100 determines that the target blind zone has a high risk of causing disease and then generates an anti-fouling control command. The main controller 100 outputs a control level to a local actuator deployed above the area where the fault node is located. The local actuator is preferably the corresponding area execution unit in the main actuator 300 and / or an auxiliary execution unit linked with the main actuator 300, such as turning on the circulating fan or adjusting the opening of the top window to perform directional ventilation and dehumidification. When the microenvironment improves and the condensation risk index drops below the anti-fouling control threshold, the main controller 100 cancels the anti-fouling control command and restores the normal control state. Through the closed-loop intervention mechanism, the system avoids the risk of alternating between disease and health caused by missing data nodes to a certain extent.

[0141] The following section uses a modern multi-span greenhouse tomato planting scenario to objectively explain the underlying control process, diagnostic efficiency, and experimental data of the agricultural greenhouse adaptive fault diagnosis system under complex multi-dimensional working conditions, ensuring that the data characteristics conform to the laws of agricultural thermodynamics and physics.

[0142] The greenhouse interior is divided into multiple physical defense zones, with a main controller 100, multiple sets of sensing nodes 200, a main actuator 300 controlling ventilation windows and shading nets, and a micro-disturbance component 400 consisting of a micro-mist device 410 and a circulating fan 420 deployed simultaneously. During the tomato fruiting period, intensive spraying and foliar fertilization are carried out inside the greenhouse. In some defense zones, the probes of the sensing nodes 200 become covered with a mixture of pesticide solution and dust, gradually developing into a soft surface malfunction. The mechanical hinges of the top ventilation windows are exposed to high humidity for extended periods, posing a physical risk of corrosion and jamming.

[0143] During the normal control phase of the system, the main controller 100 polls the meteorological parameters of each defense zone via the bus. When the underlying logic detects that the integral value of the humidity tracking error in defense zone 3 exceeds the set trigger threshold, the main controller 100 suspends the main control loop of that defense zone, locks the current mechanical opening of the main actuator 300 through electrical commands, and extracts the node with the highest error contribution value as the target diagnostic sensing node 200 to enter the diagnostic mode.

[0144] The main controller 100 retrieves the discrete data sequence from the 15 minutes prior to the triggering of node 3, and fits it to obtain the current natural meteorological evolution baseline slope, establishing a thermodynamic reference benchmark for subsequent background meteorological drift stripping. The system calculates the saturated vapor pressure difference based on the current measured dry-bulb temperature and relative humidity. Numerical comparison shows that the saturated vapor pressure difference is strictly less than the critical phase transition threshold, indicating that the local microclimate atmospheric carrying capacity is approaching saturation. Based on this, the main controller 100 activates the aerodynamic drive branch, driving the circulating fan 420 to blow directional airflow towards node 3, executing the static boundary layer stripping mechanism.

[0145] During the alternating forced and natural recovery periods, the main controller 100 frequently acquires the dynamic response sequence of node 3 under the flow field reconstruction, subtracting the slope of the natural meteorological evolution baseline from the time dimension. The computational unit extracts the multiphase feature evolution asymmetry ratio, revealing that the absolute humidity dissipation rate of this node exhibits an abnormal physical hysteresis after the fan stops. The main controller 100 inputs the constructed binary state feature vector into the fault cross-diagnosis matrix, and through multidimensional Manhattan distance optimization decoupling, determines that node 3 has a surface-wrapped soft fault.

[0146] After the fault was identified, the main controller 100 blocked the upload link of distorted data from node 3, and scheduled the matrix of adjacent normal sensing nodes. It then reconstructed the dry-bulb temperature and relative humidity at the spatial coordinates of node 3 using an inverse distance weighting operator. Based on the reconstructed meteorological elements, the system calculated that the local canopy condensation risk index had risen to 0.89, approaching the set anti-deposition control threshold of 0.92. To prevent the potential for water film retention on the leaves due to the microclimate mapping of the physical canopy layer, the main controller 100 took over the local defense zone, forcing the surrounding main actuators 300 to activate directional ventilation and dehumidification until the condensation risk index fell to a safe control range.

[0147] To quantitatively evaluate the technical indicators of this invention, a three-month cross-comparison experiment was conducted in a standard agricultural greenhouse. The experiment incorporated actual agricultural operations, recording real-world conditions such as natural mud encapsulation of the probe, slight wear and jamming of the actuator gears, and sudden changes in internal and external weather conditions. The control group used a traditional static threshold alarm method and a model-based diagnostic method driven by pure data. Experimental statistical data excluded explicit faults such as extreme sensor disconnections, summarizing only latent faults.

[0148] The experimental data are shown in the table below:

[0149] Comparison Table of Fault Diagnosis Performance and Disease Control Effect

[0150] Latent soft fault detection rate 32.4% 76.5% 89.6% False alarm rate in environments with strong interference 28.6% 14.2% 5.4% Average time to fault diagnosis 45.0 minutes 18.5 minutes 11.5 minutes Closed-loop reconstruction compensation error This function is not available. This function is not available. ±0.7 degrees Celsius and ±4.2% relative humidity Incidence of gray mold in tomatoes 12.5% 8.3% 4.5%

[0151] The experimental data demonstrates the objective physical boundaries of the technical solution in real agricultural scenarios. Traditional static threshold alarm methods lack the ability to capture early-stage, latent soft faults and frequently induce false alarms when external meteorological conditions undergo abrupt changes. Purely data-driven diagnostic methods lack physical support for the thermodynamic evolution mechanism of the greenhouse microenvironment, suffer from accuracy bottlenecks under multivariate cross-coupling conditions, and lack a post-fault spatial meteorological reconstruction compensation mechanism.

[0152] The adaptive diagnostic method of this invention achieved a latent soft fault detection rate of 89.6%. The reason this index did not reach the theoretical maximum value is that some extremely thin water-soluble drug coating layers underwent phase transition and detachment under forced airflow, leading to feature extraction deviation. However, the overall detection efficiency was significantly higher than the control group. The average fault diagnosis time was 11.5 minutes, which strictly matched the physical phase transition time required by the sum of the forced period and the natural recovery period set in the scheme, verifying the rationality of the diagnostic logic's time scale. The closed-loop reconstruction compensation error remained within ±0.7 degrees Celsius and ±4.2% relative humidity, consistent with the natural gradient decay law of the three-dimensional spatial meteorological distribution in multi-span greenhouses. The built-in closed-loop intervention strategy for latent disease control reduced the incidence of tomato gray mold in the targeted prevention area from 12.5% ​​to 4.5%. Due to the synergistic influence of biological factors such as airborne spore density on agricultural diseases, the incidence rate was not completely eliminated, but the pathogenic environment caused by water vapor condensation was substantially destroyed.

[0153] See attached document Figure 9 , Figure 9This is a three-dimensional manifold topology distribution map of fault diagnosis features according to an embodiment of the present invention. The three orthogonal axes of the three-dimensional coordinate system map the net forced evolution derivative of dry-bulb temperature, the net forced evolution derivative of absolute humidity, and the asymmetric ratio of multiphase feature evolution, respectively. Different operating states generate independent topological manifolds in the three-dimensional feature space due to differences in physical response. Normal state node data exhibits convergent spherical aggregation; mechanical opening-limited faults undergo directional spatial offset along a single temperature axis; surface-wrapped soft faults exhibit stretching distortion in the asymmetric ratio time dimension. The cross-matrix decoupling algorithm delineates the hyperplane decision boundary of each heterogeneous manifold, completing the stripping and classification of latent soft fault features in the multidimensional vector space. This map truly reflects the physical response distribution of measured discrete data points after the introduction of perturbations.

[0154] See attached document Figure 10 , Figure 10 This is a spatial thermogram of microenvironmental compensation reconstruction in the perception blind zone according to an embodiment of the present invention. The map presents a continuous spatial meteorological interpolation surface constructed by the system based on the effective perception matrix after the target node is physically isolated. The contour lines mark the boundary of the condensation risk zone output by the compensation algorithm. The gradient fluctuations of the spatial surface reflect the mathematical fitting of the inverse distance weighting control law to the attenuation characteristics of heat and mass exchange at the microscale of the greenhouse, confirming the estimation status of environmental elements within the hardware blind zone by the spatial compensation reconstruction logic. The interpolation surface in real space exhibits an asymmetric temperature gradient decrease due to the cold bridging effect at the greenhouse edge, which is completely consistent with the thermodynamic heat transfer laws.

[0155] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An agricultural greenhouse comprehensive environment regulation system with fault self-diagnosis, characterized in that, It includes a sensing node connected to the main controller, a main actuator, and a perturbation component; the perturbation component includes a micro-mist device and a circulating fan; the main controller is configured to: When the integral of the tracking error is greater than the trigger threshold, the main actuator is locked, and a target diagnostic sensing node is selected from the sensing nodes; Extract the historical data sequence of the target diagnostic sensing nodes and calculate the natural meteorological evolution baseline; The saturated water vapor pressure difference is calculated based on the data collected by the target diagnostic sensing node; the micro-mist device is activated when the saturated water vapor pressure difference is greater than the critical phase change threshold, and the circulating fan is activated when it is not greater than the critical phase change threshold. Data were collected during the forced period and natural recovery period of the perturbation component, and the asymmetric ratio of multiphase characteristic evolution after deducting the natural meteorological evolution baseline was calculated. The asymmetric ratio of the multiphase feature evolution is introduced into the cross-decoupling operator to output the diagnosis result of the mechanical fault of the main actuator or the soft surface wrapping fault of the target diagnostic sensing node; When a soft fault is diagnosed on the surface, the target diagnostic sensing node is isolated and spatial environmental parameter compensation reconstruction and fault prevention control are performed.

2. The agricultural greenhouse integrated environmental control system with fault self-diagnosis as described in claim 1, characterized in that, The main controller is configured to perform an integral operation on the absolute value of the instantaneous tracking error within the sliding time window to obtain the tracking error integral; the target diagnostic sensing node is the node with the largest error contribution value in the corresponding defense zone at the trigger time.

3. The agricultural greenhouse integrated environmental control system with fault self-diagnosis as described in claim 1, characterized in that, The historical data sequence includes a dry-bulb temperature sequence and a relative humidity sequence; the main controller is configured to convert the relative humidity sequence into an absolute humidity sequence using the Magnus empirical formula, calculate the baseline drift slope of the dry-bulb temperature sequence and the absolute humidity sequence within the corresponding time window using the least squares fitting algorithm, and use the baseline drift slope as the natural weather evolution baseline.

4. The agricultural greenhouse integrated environmental control system with fault self-diagnosis as described in claim 1, characterized in that, The micro-mist device is configured to spray water mist into the local space where the target diagnostic sensing node is located; the circulating fan is configured to blow directional airflow into the target diagnostic sensing node.

5. The agricultural greenhouse integrated environmental control system with fault self-diagnosis according to claim 3, characterized in that, The main controller is configured to calculate the first-order evolutionary derivatives of dry-bulb temperature and absolute humidity during the forced period and the natural recovery period; and to subtract the corresponding baseline drift slope from the first-order evolutionary derivatives during the forced period to obtain the net forced evolutionary derivatives. The net recovery evolution derivative is obtained by subtracting the corresponding baseline drift slope from the first-order evolution derivative during the natural recovery period.

6. The agricultural greenhouse integrated environmental control system with fault self-diagnosis according to claim 5, characterized in that, The main controller is configured to divide the net recovery evolution derivative by the sum of the net forced evolution derivative and a preset constant, and take the absolute value of the division result to obtain the multiphase characteristic evolution asymmetry ratio.

7. The agricultural greenhouse integrated environmental control system with fault self-diagnosis according to claim 6, characterized in that, The execution logic of the cross-decoupling operator is as follows: compare the net forced evolution derivative and the multiphase feature evolution asymmetry ratio with the corresponding preset benchmark thresholds to generate a binarized current state feature vector; calculate the Manhattan distance between the binarized current state feature vector and each typical fault template vector in the preset fault cross-diagnosis matrix; and take the fault type corresponding to the minimum Manhattan distance as the diagnosis result.

8. The agricultural greenhouse integrated environmental control system with fault self-diagnosis according to claim 1, characterized in that, The main controller is configured to extract the measurement values ​​of adjacent sensing nodes that are in normal working condition within a preset spatial radius when performing the spatial environment parameter compensation reconstruction, and to perform a weighted summation operation using the relationship between the measurement values ​​and the reciprocal of the square of the Euclidean distance between the corresponding adjacent sensing node and the target diagnostic sensing node, to obtain the reconstructed dry-bulb temperature and reconstructed relative humidity.

9. The agricultural greenhouse integrated environmental control system with fault self-diagnosis according to claim 8, characterized in that, The main controller is configured to calculate a condensation risk index based on the reconstructed relative humidity and the difference between the reconstructed dry-bulb temperature and the dew point temperature when performing the disease prevention and control. When the condensation risk index is greater than or equal to the preset anti-submersion control threshold, the main actuator is controlled to perform a dehumidification action.

10. The agricultural greenhouse integrated environmental control system with fault self-diagnosis according to claim 1, characterized in that, The main controller is configured to, when a mechanical fault is diagnosed in the main actuator, control the main actuator to remain locked or switch to a preset safe position, and invoke a backup control strategy to take over the control of the defense zone where the target diagnostic sensing node is located.

Images