A data-driven, refined operation and maintenance management system for the entire life cycle of agricultural greenhouses

Abstract

This invention relates to the technical field of intelligent agricultural greenhouse management. It discloses a data-driven, precision operation and maintenance management system for the entire lifecycle of agricultural greenhouses. The system includes defining light intensity and air humidity as drive and response parameters, respectively, collecting their time-series data, calculating the time-lag cross-correlation of their characteristic waveforms to determine a drive-response disconnection index, and generating system vitality warnings based on the evolutionary trend of this index. This system breaks away from traditional steady-state monitoring methods. By analyzing the dynamic response relationships between environmental parameters, it enables non-intrusive assessment of crop physiological status. This system can predict system functional decline even when environmental parameters are normal, providing a new decision-making dimension for precision agricultural management.

Description

Technical Field

[0001] The present invention relates to a data-driven, refined operation and maintenance management system for the entire life cycle of an agricultural greenhouse, belonging to the technical field of intelligent management of agricultural greenhouses. Background Art

[0002] Environmental threshold monitoring systems are commonly used in the field of intelligent management technology for agricultural greenhouses. Their core logic relies on maintaining a static stable range for parameters such as temperature and humidity. Such systems compare environmental data with preset thresholds and trigger an alarm once a deviation is detected. However, this approach has fundamental limitations: in typical scenarios where the physiological response ability of crops slowly declines, environmental parameters often remain in an ideal state for a long time, and the system cannot perceive the decline in the crop's response efficiency to external driving forces. For example, when root vitality decreases, resulting in a delayed transpiration response, traditional systems cannot identify this hidden functional degradation because they only monitor the absolute value of humidity.

[0003] More notably, existing improvement plans attempt to improve accuracy by increasing sensor types or density, but this increases hardware costs and data redundancy, and still fails to address the lack of dynamic assessment of system vitality. These underlying contradictions lie in: 1. Relying on static snapshots of environmental parameters, they ignore the dynamic relationship between driving sources and physiological responses, resulting in a blind spot for early detection of intrinsic system functional degradation; 2. Intervention is limited to the detection of physical anomalies, failing to predict progressive functional decline and missing the golden opportunity for intervention; 3. Improving performance requires the addition of high-cost sensors, but this does not enhance the ability to address the core issue (system vitality assessment). Therefore, the industry urgently needs a low-cost management solution that breaks through steady-state monitoring methods and focuses on dynamic functional assessment. Summary of the Invention

[0004] The present invention provides a data-driven, refined operation and maintenance management system for the entire life cycle of agricultural greenhouses. Its main purpose is to solve the problem that traditional monitoring systems cannot perceive the degradation of crop physiological response capabilities due to their reliance on static thresholds of environmental parameters, resulting in a lack of early warning capabilities for the decline of system functional vitality.

[0005] To achieve the above objectives, the present invention provides a data-driven, refined operation and maintenance management system for the entire life cycle of agricultural greenhouses, the system comprising:

[0006] A driving parameter sensor unit for collecting time series data of driving parameters inside the agricultural greenhouse, where the driving parameters represent external energy input;

[0007] At least two response parameter sensor units are deployed in different spatial areas inside the agricultural greenhouse to collect time series data of response parameters corresponding to each spatial area, where the response parameters represent the collective physiological activities of the crops;

[0008] An edge data processing unit is connected to the drive parameter sensor unit and the at least two response parameter sensor units, and is configured to: extract and generate a drive characteristic waveform based on the drive parameter time series data; extract and generate a response characteristic waveform corresponding to each spatial region based on the response parameter time series data of each spatial region; calculate, for each spatial region, a time lag cross-correlation between the drive characteristic waveform and the response characteristic waveform of the corresponding spatial region to determine a drive-response disconnection index, the drive-response disconnection index representing the strength and delay of the time lag cross-correlation; and instantaneously increase the data sampling frequency of the drive parameter sensor unit and the at least two response parameter sensor units when the rate of change of the drive-response disconnection index exceeds a preset system shock threshold;

[0009] A central decision-making and control unit is connected to the edge data processing unit. The central decision-making and control unit is configured to: generate predictive warning information on the functional vitality of the agricultural greenhouse system based on the long-term evolution trend of the drive-response disconnection degree indicator; construct a time-series topological fingerprint representing the instantaneous impact based on the drive parameter time-series data, response parameter time-series data and high-sampling frequency data of at least one other environmental parameter collected at a high sampling frequency; and generate emergency warning information or emergency management control instructions corresponding to a specific fault source based on the time-series topological fingerprint; and generate visual management instructions for the vitality status of the agricultural greenhouse system based on the spatial distribution trend presented by the drive-response disconnection degree indicator corresponding to each spatial area.

[0010] Preferably, the driving parameter represents light intensity, and the response parameter represents air humidity.

[0011] Preferably, the edge data processing unit is configured to generate a driving characteristic waveform and a response characteristic waveform within a preset sliding time window, and the generation method includes calculating the arithmetic mean, peak count and average rise time from trough to peak of the timing data.

[0012] Preferably, the edge data processing unit is configured such that the calculation result of the time-delay cross-correlation includes a maximum correlation coefficient and an optimal time-delay time, and the drive-response disconnection degree index is calculated based on the maximum correlation coefficient and the optimal time-delay time.

[0013] Preferably, the drive-response disconnection index Calculated according to the following formula:

[0014] ,

[0015] in, represents the maximum correlation coefficient, represents the optimal lag time, is the weight factor, It is the benchmark time lag when the agricultural greenhouse system is in a healthy state.

[0016] Preferably, the central decision and control unit is configured to automatically learn and calibrate the normal fluctuation range of the drive-response disconnection degree indicator of the agricultural greenhouse in a healthy state by collecting data for more than one week during the initial operation stage of the system.

[0017] Preferably, the edge data processing unit is configured to switch to an emergency detection mode when the rate of change of the drive-response disconnection degree indicator exceeds the system shock threshold. The emergency detection mode includes increasing the data sampling frequency of the drive parameter sensor unit and at least two response parameter sensor units to their physical limit frequency, and recording high-frequency time series data.

[0018] Preferably, the edge data processing unit is configured to, within a preset short time window after high sampling frequency data collection, not perform complex numerical calculations, but to construct a timing topological fingerprint based on the timing changes of driving parameters, response parameters and other environmental parameters at high sampling frequency. The timing topological fingerprint includes the relative order and direction of the drastic changes in each parameter.

[0019] Preferably, the central decision and control unit is configured to have a built-in impact type knowledge base for fault classification and response based on temporal topological fingerprints. The impact type knowledge base includes: topological fingerprints in which humidity drops sharply, temperature drops sharply, and light remains stable, indicating structural damage to the shed; topological fingerprints in which humidity rises sharply, temperature changes insignificantly, and light drops sharply, indicating abnormal closure of the external shielding device; and topological fingerprints in which only a single sensor data changes dramatically while the other two sensor data remain stable, indicating a fault in the corresponding sensor itself.

[0020] Preferably, each of the at least two response parameter sensor units is a modular response probe, which in hardware only includes a humidity sensing element and a microcontroller for wireless data transmission. The modular response probe is configured to periodically measure the air humidity of the spatial area in which it is located and wirelessly send the air humidity data to the central decision and control unit.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] 1. By defining light and humidity parameters as the driving source and physiological response carrier, respectively, the system is freed from its dependence on the absolute values ​​of environmental parameters. A time-delayed cross-correlation analysis of the driving characteristic waveform and the response characteristic waveform is constructed at the edge node, enabling the management system to capture the dynamic response efficiency changes of crops to external energy input for the first time. When this response correlation is delayed or weakened, even if the temperature and humidity readings are still within the ideal range, the system can provide early warning of potential system function decline, thereby shifting management decisions from passively maintaining environmental homeostasis to actively maintaining the vitality of the life system.

[0023] 2. When the drive-response disconnection index fluctuates violently, the system achieves millisecond-level event tracing by instantaneously increasing the sampling frequency of multiple parameters and capturing their changing time series topology. The combination of the changing order and direction of basic parameters such as light, temperature, and humidity under high-frequency sampling naturally forms a fingerprint feature with physical interpretability. This mechanism does not require the addition of new sensors, but only reuses existing data channels to distinguish the essential differences between emergencies such as shed damage and equipment failure, thereby transforming chaotic signals classified as unknown anomalies in traditional monitoring systems into clear diagnostic basis for classification and disposal.

[0024] 3. The asymmetric architecture of a single global illumination driving source combined with distributed simple humidity probes enables the system to obtain spatial diagnostic capabilities at almost zero hardware upgrade cost. The central node naturally forms a heat map reflecting the vitality distribution inside the greenhouse by parallel calculation of the humidity response and light-driven loss index of each area. Managers can intuitively identify local functional abnormalities (based on irrigation blockage points), avoiding the missed diagnosis of local lesions caused by the spatial averaging effect of traditional solutions, and realizing a transition from overall health assessment to precise spatial intervention. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is the closed-loop control flow chart of edge computing for agricultural greenhouses of the present invention;

[0026] Figure 2 This is a timing diagram of abnormal changes in environmental parameters of the present invention;

[0027] Figure 3 This is a flow chart of the agricultural greenhouse drive-response loss degree index calculation and response mechanism of the present invention. DETAILED DESCRIPTION

[0028] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0029] The present application provides a data-driven, refined operation and maintenance management system for the entire life cycle of an agricultural greenhouse, the system comprising:

[0030] A driving parameter sensor unit for collecting time series data of driving parameters inside the agricultural greenhouse, where the driving parameters represent external energy input;

[0031] At least two response parameter sensor units are deployed in different spatial areas inside the agricultural greenhouse to collect time series data of response parameters corresponding to each spatial area, where the response parameters represent the collective physiological activities of the crops;

[0032] An edge data processing unit is connected to the drive parameter sensor unit and the at least two response parameter sensor units, and is configured to: extract and generate a drive characteristic waveform based on the drive parameter time series data; extract and generate a response characteristic waveform corresponding to each spatial region based on the response parameter time series data of each spatial region; calculate, for each spatial region, a time lag cross-correlation between the drive characteristic waveform and the response characteristic waveform of the corresponding spatial region to determine a drive-response disconnection index, the drive-response disconnection index representing the strength and delay of the time lag cross-correlation; and instantaneously increase the data sampling frequency of the drive parameter sensor unit and the at least two response parameter sensor units when the rate of change of the drive-response disconnection index exceeds a preset system shock threshold;

[0033] A central decision-making and control unit is connected to the edge data processing unit. The central decision-making and control unit is configured to: generate predictive warning information on the functional vitality of the agricultural greenhouse system based on the long-term evolution trend of the drive-response disconnection degree indicator; construct a time-series topological fingerprint representing the instantaneous impact based on the drive parameter time-series data, response parameter time-series data and high-sampling frequency data of at least one other environmental parameter collected at a high sampling frequency; and generate emergency warning information or emergency management control instructions corresponding to a specific fault source based on the time-series topological fingerprint; and generate visual management instructions for the vitality status of the agricultural greenhouse system based on the spatial distribution trend presented by the drive-response disconnection degree indicator corresponding to each spatial area.

[0034] Preferably, the driving parameter represents light intensity, and the response parameter represents air humidity.

[0035] Preferably, the edge data processing unit is configured to generate a driving characteristic waveform and a response characteristic waveform within a preset sliding time window, and the generation method includes calculating the arithmetic mean, peak count and average rise time from trough to peak of the timing data.

[0036] Preferably, the edge data processing unit is configured such that the calculation result of the time-delay cross-correlation includes a maximum correlation coefficient and an optimal time-delay time, and the drive-response disconnection degree index is calculated based on the maximum correlation coefficient and the optimal time-delay time.

[0037] Preferably, the drive-response disconnection index Calculated according to the following formula:

[0038] ,

[0039] in, represents the maximum correlation coefficient, represents the optimal lag time, is the weight factor, is the benchmark time lag when the agricultural greenhouse system is in a healthy state; specifically, the weight factor Used to balance the maximum correlation coefficient and the optimal lag time In the calculation of the drive-response disconnection index, during the initial operation of the system, the central decision and control unit will automatically learn and calibrate the normal fluctuation range of the drive-response disconnection index in a healthy agricultural greenhouse by collecting data for more than one week. During this learning process, the system will simultaneously conduct multiple rounds of engineering exploration and data analysis, combining the physiological response characteristics of the target crop and the system's sensitivity to response delays to adaptively determine the weighting factor. The optimal value of the weight factor is determined by taking into account the crop type, growth stage and specific environmental conditions of the greenhouse to ensure that the weight factor can accurately reflect the coupling relationship characteristics under different application scenarios. The system supports joint updates based on rule templates and expert feedback to ensure its good adaptability and maintainability in practical applications.

[0040] Preferably, the central decision and control unit is configured to automatically learn and calibrate the normal fluctuation range of the drive-response disconnection degree indicator of the agricultural greenhouse in a healthy state by collecting data for more than one week during the initial operation stage of the system.

[0041] Preferably, the edge data processing unit is configured to switch to emergency detection mode when the rate of change of the drive-response disconnection index exceeds a system shock threshold. This emergency detection mode involves increasing the data sampling frequency of the drive parameter sensor unit and at least two response parameter sensor units to their physical limit frequency and recording high-frequency time series data. The system shock threshold, used to trigger emergency detection mode, is not a fixed value. Instead, it is determined during the system's initial operation phase by the central decision and control unit, which automatically learns and calibrates the normal fluctuation range of the drive-response disconnection index under healthy conditions by collecting at least one week of historical data. Specifically, the system analyzes historical fluctuation data of the drive-response disconnection index under healthy conditions and determines the upper limit of its normal rate of change based on statistical methods (e.g., calculating the standard deviation or setting a confidence interval). Any rate of change exceeding this upper limit is defined as the system shock threshold. Different crops, greenhouse environments, and the initial system state may cause this threshold to vary. The system has adaptive adjustment capabilities to ensure the dynamic and accurate nature of this threshold.

[0042] Preferably, the edge data processing unit is configured to, within a preset short time window after high sampling frequency data collection, not perform complex numerical calculations, but to construct a timing topological fingerprint based on the timing changes of driving parameters, response parameters and other environmental parameters at high sampling frequency. The timing topological fingerprint includes the relative order and direction of the drastic changes in each parameter.

[0043] Preferably, the central decision and control unit is configured to have a built-in impact type knowledge base for fault classification and response based on temporal topological fingerprints. The impact type knowledge base includes: topological fingerprints in which humidity drops sharply, temperature drops sharply, and light remains stable, indicating structural damage to the shed; topological fingerprints in which humidity rises sharply, temperature changes insignificantly, and light drops sharply, indicating abnormal closure of the external shielding device; and topological fingerprints in which only a single sensor data changes dramatically while the other two sensor data remain stable, indicating a fault in the corresponding sensor itself.

[0044] Preferably, each of the at least two response parameter sensor units is a modular response probe, which in hardware only includes a humidity sensing element and a microcontroller for wireless data transmission. The modular response probe is configured to periodically measure the air humidity of the spatial area in which it is located and wirelessly send the air humidity data to the central decision and control unit.

[0045] Example 1: The system of this embodiment is constructed around the drive-response dynamic coupling analysis mechanism, and the whole system includes a drive parameter sensor unit, multiple response parameter sensor units, an edge data processing unit and a central decision and control unit. The above units work together to form a closed-loop structure of data acquisition, processing and decision-making. The system is deployed in an agricultural greenhouse scene, aiming to achieve non-intrusive prediction of potential functional decline of the system by modeling and trend monitoring the dynamic coupling characteristics between light and humidity; wherein, the drive parameter sensor unit is mainly used to collect representative data of energy input inside and outside the agricultural greenhouse. In this embodiment, light intensity is preferably used as the drive parameter. The sensor is set in an area with sufficient light and has a stable and adjustable sampling capability. Its initial sampling interval is set to between 30 seconds and 5 minutes according to the type of greenhouse crop and climate characteristics, and the data is transmitted to the edge data processing unit by wired or wireless means; the response parameter sensor unit The preferred design is a modular response probe, each probe containing an air humidity sensor element and a microcontroller module for wireless data transmission. Multiple probes are arranged in typical spatial areas inside the greenhouse to periodically collect humidity data in their respective spaces and upload them synchronously for further processing. The sampling time is synchronized with the drive parameters to ensure that the time series data can be used for coupled analysis. The edge data processing unit is mainly responsible for feature extraction and drive-response relationship calculation tasks. Within a preset sliding time window (for example, 30 minutes or 1 hour), the unit extracts representative characteristic waveforms from the collected light and humidity time series data. The extraction method includes: calculating the arithmetic mean within the time window to reflect the baseline level, counting the number of peaks to estimate the periodic frequency, and measuring the average rise time from the trough to the peak to characterize the response speed. This processing method can extract the dynamic behavior characteristics between energy drive and physiological response without significantly increasing the computational burden.At the same time, although this system uses light intensity as the driving parameter and air humidity as the response parameter for core analysis, the actual collection and analysis process fully considers the linkage characteristics between multi-sensor data, including the restrictive role of environmental factors such as soil moisture. For example, when constructing the time-series topological fingerprint, the edge data processing unit conducts a comprehensive analysis based on the time-series changes of the driving parameter, the response parameter, and at least one other environmental parameter (such as soil moisture, temperature, and carbon dioxide concentration) at a high sampling frequency. When determining the response hysteresis of the crop physiological transpiration process, the system further determines the response hysteresis of the physiological transpiration process based on the characteristics such as air humidity fluctuations accompanied by slowing local temperature changes and being out of sync with light intensity changes, thereby accurately corresponding to the changing trend of the crop's water regulation function. This method, which focuses on humidity and combines other environmental information such as soil moisture and temperature for comprehensive judgment, effectively avoids the interpretation bias caused by a single parameter and improves the stability and versatility of the system. Through multi-factor coupling analysis, the system can more comprehensively evaluate the response efficiency and response pattern of crops to light, improving the decision-making dimension of precision agriculture management.Specifically, in the evaluation and emergency detection mode of the drive-response loss of contact index, the system's comprehensive analysis of multiple environmental parameters is not a simple superposition, but a layered and fusion mechanism. First, for environmental parameters other than light and air humidity, such as soil moisture, temperature, and carbon dioxide concentration, the edge data processing unit will perform independent feature extraction, which includes calculating the average value, fluctuation amplitude, change rate, and trend direction of its time series data to form their respective feature vectors. Secondly, when constructing the time series topological fingerprint, the system will take its high-frequency sampling data into consideration together with the time series data of light and air humidity. The system will analyze the relative order and change direction of these parameters that change dramatically within a preset short time window. For example, when the rate of change of the drive-response disconnection index exceeds the system shock threshold, the system switches to emergency detection mode and increases the sampling frequency of parameters such as light intensity, air humidity, soil moisture, and temperature to their physical limit frequency. In this high-frequency data stream, the system does not perform complex numerical calculations, but captures the starting point, relative order, rising or falling trend, and duration of drastic changes in each parameter to form a time-series fingerprint map with physical interpretability. For example, a topological fingerprint in which humidity and temperature drop sharply while light remains stable indicates structural damage to the shed. If this fingerprint is accompanied by a simultaneous sharp drop in soil moisture, it further strengthens the judgment that structural damage leads to a rapid exchange of internal and external environments. Furthermore, the system uses the event knowledge base module to categorize and analyze the topological fingerprints formed by these multiple parameters, thereby identifying potential sources of faults or types of crop stress. For example, if the driver-response disconnection index indicates a decrease in response efficiency, but soil moisture data is within the normal range, this may indicate a decline in root vitality rather than a simple water shortage. This multi-parameter cross-validation and pattern recognition allows the system to delve beyond superficial parameter anomalies to the underlying causes of functional decline, enabling more accurate fault diagnosis and management intervention.

[0046] After obtaining the driving characteristic waveform and the response characteristic waveform, the system calculates the correlation coefficient between the two at different time delays through the sliding cross-correlation analysis method, and determines the maximum correlation coefficient and its corresponding optimal time delay. Based on this, a driving-response disconnection degree index is constructed to characterize the system coupling state. , and its calculation formula is: ,in, Used to measure the synchronization strength between drive and response, Reflects the degree of response delay, The average value of the system in its initial healthy state is taken as the indicator, which comprehensively reflects the coupling strength and response efficiency, and can sensitively reflect the potential degradation trend of the system's physiological function. When the rate of change exceeds the preset system shock threshold, the system immediately activates emergency detection mode. In this mode, the sampling frequency of the drive and response parameters is rapidly increased to the maximum frequency allowed by the sensor's physics (for example, a 1-second interval), and high-density data is continuously collected over the following short period of time (for example, 60 seconds). The high-sampling data at this stage does not enter the conventional numerical analysis process, but is directly used to construct a temporal topological fingerprint. This fingerprint is based on the order and direction (i.e., rise or fall) of drastic changes in multiple key parameters (light, temperature, humidity) within the same time window, extracting physically meaningful event behavior patterns. This mechanism can use existing sensor data to realize event recognition without the need for additional sensor hardware resources.

[0047] The central decision-making and control unit is responsible for continuous monitoring If a stable offset or continuous deterioration trend is found, predictive warning information of the system's functional vitality will be generated. At the same time, the central unit also has an embedded impact type knowledge base, which can judge the event type based on the constructed temporal topological fingerprint and generate corresponding fault warnings or management instructions. The knowledge base contains a variety of matching rules between typical topological fingerprints and their corresponding events. For example, a sharp drop in humidity and temperature at the same time but stable light indicates that the shed structure is damaged; a sudden rise in humidity and a sudden drop in light indicate abnormal closure of the sunshade device; a sudden change in a single sensor data while the others remain stable indicates a corresponding sensor failure; in addition, the system supports spatial visualization management functions. The central unit processes the corresponding The numerical values ​​are mapped on the spatial distribution map of agricultural greenhouses in the form of a color scale diagram to form a thermal map of the greenhouse vitality status. This map can intuitively reveal the potential response anomalies or functional degradation phenomena in local areas, and provide an early diagnosis basis for spatial problems such as local irrigation pipeline blockage and soil compaction. Regarding the identification of crop physiological responses, although this system uses air humidity as the main observation parameter, the actual collection and analysis process takes into account the linkage characteristics between multi-sensor data. For example, when air humidity fluctuations are accompanied by a slowdown in local temperature changes and are not synchronized with changes in light intensity, it can be further determined to be a response hysteresis of the physiological transpiration process, thereby accurately corresponding to the changing trend of the water regulation function in the crop body. This method of comprehensive judgment based on humidity and combined with environmental information such as temperature effectively avoids the interpretation bias caused by a single parameter and improves the stability and versatility of the system.

[0048] Example 2: The system of this embodiment can be deployed in an agricultural greenhouse scenario, and its core functions are constructed around the light-humidity and drive-response coupling relationships. The overall structure consists of a drive parameter sensor unit, multiple response parameter sensor units, an edge data processing unit and a central decision and control unit. The units form a closed-loop collaborative system of collection, processing and control through data links, aiming to achieve non-intrusive dynamic evaluation of the functional status of the greenhouse system and provide early warning signals in the early stages of functional degradation. Among them, the drive parameter sensor unit is installed on the top of the agricultural greenhouse or in an area with stable lighting conditions, and is used to periodically collect light intensity data. The initial sampling period of the unit is set to a range of 30 seconds to 5 minutes based on the type of greenhouse crop and local climate characteristics. The specific value can be reasonably set and optimized by technical personnel in this field according to actual application requirements. The data is transmitted to the edge data processing unit via a serial bus or a short-range wireless module, and a fault-tolerant mechanism is added to the transmission link to ensure data integrity. The response parameter sensor unit is designed as a modular response probe structure. Each probe includes an air humidity sensor element and a microcontroller module with wireless communication capabilities. Multiple probes are distributed in typical areas inside the greenhouse to collect humidity data at corresponding spatial positions. To ensure the time alignment requirements of subsequent drive-response coupling analysis, all response probes and drive parameter sensors have a unified sampling period, and a synchronization marking mechanism is used to manage timing consistency. The edge data processing unit sets a sliding time window of fixed length (for example, 30 minutes or 60 minutes), and performs normalization preprocessing and feature extraction on the original time series data of light and humidity in each window period. The characteristic parameters include: the arithmetic mean within the sampling interval, which is used to measure the steady-state level; the local peak count, which reflects the periodic fluctuation characteristics; the average rise time from the trough to the peak, which is used to quantify the response rate. The above method is a common data statistical processing strategy in this field with low computational intensity and is suitable for lightweight deployment of edge nodes.

[0049] After completing the waveform feature extraction, the system uses the sliding cross-correlation analysis method to calculate the time-lag correlation between the driving characteristic waveform and the humidity waveform corresponding to each response probe. The analysis results include: the maximum cross-correlation coefficient and its corresponding optimal lag time, based on which the drive-response disconnection index is constructed. , which is calculated as follows: above The weight relationship used to balance the strength of correlation and response delay, is the average hysteresis value of the system based on multi-cycle observations in the initial healthy state. This indicator is used to comprehensively characterize the response efficiency and coupling stability of the system. Its changing trend has the sensitivity to identify functional degradation. If an indicator shows a significant increase in two consecutive time windows and the rate of change exceeds the preset system shock threshold, the system triggers the emergency detection mode. In this mode, the sampling frequency of the driving and response parameters is simultaneously increased to the upper limit rate supported by the sensor, for example, once per second, and a complete high-frequency data sequence is collected within a continuous period of 60 seconds. To improve response efficiency, the data at this stage does not enter the regular analysis process, but is used to construct event topology features; this event topology feature is generated based on the order and direction of drastic changes in multiple parameters within the same time window, including the starting time, relative order and duration of the rise or fall of key variables such as light intensity, air humidity and temperature, to form a time series fingerprint map with physical interpretability. This mechanism does not require additional hardware sensing resources and can use existing sampling data to identify different types of system anomalies, providing a basis for subsequent classification response.

[0050] The central decision and control unit is automatically calibrated within the first 7 days after system deployment using stable data during operation. Normal fluctuation range, used as a basis for subsequent judgment, the system continuously tracks the corresponding Evolution trajectory, if a trend shift or spatial imbalance is found, a predictive warning prompt of the system's functional vitality is generated; in addition, the unit integrates an event knowledge base module, which can classify and analyze the topological fingerprints generated in the high-frequency sampling stage, and output preliminary fault judgment and disposal instructions accordingly. Typical matching rules include: humidity and temperature drop sharply at the same time while the light remains stable, which is judged to be a damaged shed; humidity rises accompanied by a sharp drop in light, which is judged to be an abnormal closure of the sunshade device; a single probe data mutates while the others are stable, which is judged to be a failure of the corresponding sensor. The system also supports spatial visualization display function, and the central unit can display the corresponding The indicator is projected onto the two-dimensional topological map of the agricultural greenhouse to form a heat map reflecting the local vitality state. The map uses color coding to distinguish the degree of coupling response of each region, which can be used to identify areas of functional degradation caused by local soil compaction, drip irrigation blockage and other reasons, and provide managers with a basis for precise spatial intervention. In actual applications, the evaluation of system functional vitality is based on the coupling state between the humidity response of multiple spatial regions and the unified light drive. Its changes are manifested as delays in response time and weakening of response amplitude. In order to accurately define the functional vitality state, the system uses the normal operating data within seven consecutive days after the initial deployment as the benchmark, and counts the time lag change amplitude between the humidity response and light drive of each region on an hourly basis as a representative indicator of response efficiency. If the response delay time in a certain area continuously exceeds twice the benchmark average, and the synchronous decline is significantly smaller than the historical average level, it can be determined that there are signs of system functional vitality decline in the area.

[0051] Example 3: The core of the present invention is to avoid the traditional environmental parameter threshold monitoring method and instead focus on the dynamic response relationship between environmental parameters within the agricultural greenhouse. By non-invasively assessing the physiological state of crops, early prediction of system functional decline can be achieved, providing a new decision-making dimension for precision agricultural management. This method is fundamentally different from traditional static monitoring and is more in line with the actual needs and technological trends of the current development of intelligent agriculture. In current agricultural greenhouse management, even if environmental parameters such as light, temperature, and humidity are maintained within the set range, crop growth conditions may still show sub-health or hidden decline. This is usually due to slight changes or delays in the dynamic coupling relationship between driving factors and physiological responses. Traditional monitoring systems lack the ability to perceive such potential functional degradation. For example, when the vitality of crop roots decreases, resulting in a delay in the response of transpiration to light, a system that relies only on absolute humidity monitoring cannot identify this inherent degradation. This experiment aims to construct a test platform that simulates a real agricultural greenhouse environment and focus on verifying how the system can effectively quantify the drive-response disconnection degree index ( ), and based on this, generate system vitality warnings before environmental parameters become significantly abnormal. This experiment also explores how, in emergency situations, the system can quickly identify and locate fault sources through high-frequency data acquisition and temporal topological fingerprint construction. A typical artificial climate chamber was selected as the test platform. This climate chamber has the ability to precisely control light intensity, temperature, air humidity, and carbon dioxide concentration, simulating agricultural greenhouse environments in different seasons and geographical regions. A distributed sensor network is deployed within the climate chamber, including a high-precision light intensity sensor (driving parameter sensor unit) installed at the top center and six modular humidity probes (response parameter sensor units) evenly distributed across different spatial regions to simulate the spatial heterogeneity of environmental parameters within a real greenhouse. All sensors have high-speed data acquisition capabilities, with an initial sampling frequency of 30 seconds. The edge data processing unit, an embedded single-board computer equipped with a high-performance processor, receives sensor data in real time, extracts characteristic waveforms, and calculates the drive-response disconnection degree indicator. A central decision and control unit, deployed on a host server, is responsible for long-term trend analysis of the drive-response disconnection degree indicator, generating warnings, and matching impact type knowledge bases.

[0052] During the experiment, we introduced a typical crop for observation to ensure the representativeness of the experimental results. The light intensity range was set from 0 lux to 50,000 lux to simulate the change of sunlight and artificial lighting. The monitoring range of air humidity was 20% relative humidity to 95% relative humidity. In the experiment, the weight factor After multiple rounds of engineering exploration and data analysis, combined with the physiological response characteristics of the crop and the system's sensitivity to response delay, it was finally determined to be 0.5, the benchmark lag time under healthy conditions. The system was automatically learned and calibrated after a period of initial stable operation, with an average value of approximately 8 minutes. This value reflects the typical physiological response delay of transpiration to light changes in the crop under stable environmental conditions. During the three-week continuous monitoring period, the environmental parameters of the climate chamber simulated the typical scenarios of the early, middle and mild stress periods of healthy crop growth. No obvious external abnormal shocks were introduced during the experiment. Instead, the physiological process of gradually decreasing transpiration response efficiency due to a slow decline in crop root vitality was simulated by gradually and slightly adjusting the substrate moisture content. The edge data processing unit continuously extracted the arithmetic mean, peak count, and average rise time from trough to peak of the light time series data within an hourly sliding time window. The window size was determined based on the trade-off between the analysis of the crop's physiological response cycle and computational efficiency to form a driving characteristic waveform. At the same time, the response parameter time series data collected by each humidity probe was similarly processed to generate its own response characteristic waveform. Subsequently, the time-lag cross-correlation between the driving characteristic waveform and each response characteristic waveform was calculated to obtain the maximum correlation coefficient. and the optimal lag time , the calculation formula of the drive-response disconnection index is: We continuously recorded the drive-response disconnection index value and its change rate corresponding to each probe. The observation results showed that in the first two weeks, as the substrate moisture content slightly decreased, the drive-response disconnection index value reflected by each humidity probe showed a slow and steady upward trend, but did not reach the preset alarm threshold of the traditional absolute humidity value; especially at the end of the second week, although the air humidity readings remained within the appropriate range, the change rate of the drive-response disconnection index exceeded the system's preset functional decline warning threshold. This threshold is determined based on the normal fluctuation range of the drive-response disconnection index under long-term monitoring of crop health. The central decision and control unit immediately generated a system vitality predictive warning, indicating that the transpiration response efficiency of crops in some areas has continued to decline, and it is recommended to pay attention to the health status of the root system; this warning was generated earlier than any obvious environmental parameter anomaly, which clearly verified the ability of the present invention to predict system functional decline when no environmental parameter anomalies occurred, providing a new decision-making dimension for precision agricultural management.

[0053] In the third week after the above-mentioned warning was issued, we randomly selected an area where a humidity probe was located and simulated two typical transient shock events: one scenario was achieved by partially opening the ventilation holes of the climate chamber, causing the air humidity and temperature in the area to drop sharply at the same time, but the light intensity remained stable, which was intended to simulate structural damage to the greenhouse; the other scenario was achieved by temporarily partially blocking the light sensor, causing the light intensity in the area to drop sharply, while the local humidity rose sharply due to the obstruction of crop transpiration, and the temperature change was not obvious, which was intended to simulate the abnormal closure of the external shading device; when the change rate of the drive-response disconnection degree indicator exceeded the preset system shock threshold during the above-mentioned simulated shock, the edge data processing unit immediately triggered the emergency detection mode, increased the data sampling frequency of the drive parameter sensor unit and the response parameter sensor unit to its physical limit frequency (for example, once per second), and continuously collected high-density time series data for the next 60 seconds. This high-sampling data does not enter the conventional numerical analysis process, but is directly used to construct the time series topological fingerprint; the central decision and control unit constructed a characterization of the transient shock based on the collected high-frequency data. For example, in the scenario of simulated structural damage to a shed, the temporal topological fingerprint is characterized by a sharp drop in humidity and temperature, while light remains stable. The central decision and control unit, based on a built-in impact type knowledge base built by summarizing typical fault data and expert experience, accurately matches the fault type of the shed's structural damage and generates an emergency warning message: "Signs of shed structural damage have been detected in area 3. Please check the vents or sealing structure immediately." In the scenario of simulated abnormal closure of external shading equipment, the temporal topological fingerprint is characterized by a sharp rise in humidity, no obvious temperature change, and a sharp drop in light. The system then matches the fault type to the abnormal closure of the external shading equipment and generates an emergency warning message: "Area 5: An abnormal drop in light accompanied by a sudden rise in humidity. A suspected shading system failure is suspected. Please check immediately." The test results clearly show that this system can achieve millisecond-level event tracing and effectively distinguish different types of emergencies through high-frequency sampling and temporal topological fingerprint construction, transforming signals that may be classified as unknown anomalies in traditional monitoring systems into clear diagnostic evidence that can be classified and handled.During the aforementioned experiment, the central decision-making and control unit continuously generated visual management instructions for the vitality status of the agricultural greenhouse system based on the spatial distribution of the drive-response disconnection index corresponding to each spatial region. By mapping the drive-response disconnection index values ​​of each probe onto the two-dimensional spatial distribution map of the climate chamber in the form of a color scale, an intuitive heat map of the greenhouse vitality status was formed. When the crops experienced mild physiological stress, the heat map clearly indicated areas with high drive-response disconnection index values. Even areas with normal absolute humidity values ​​were highlighted with different color scales, visually revealing areas of local functional abnormalities. During the transient impact simulation, the drive-response disconnection index values ​​of the affected areas suddenly surged, appearing highlighted on the heat map, allowing managers to quickly identify the specific spatial location of the fault. This confirms that the system's asymmetric architecture, combining a single global illumination driver with distributed, simple humidity probes, achieves refined spatial diagnostic capabilities at a low hardware cost. This effectively avoids the missed diagnosis of localized lesions caused by the spatial averaging effect of traditional solutions, and achieves a transition from holistic health assessment to precise spatial intervention.

[0054] Example 4: This example combines Figures 1 to 3 , this paper describes the implementation of a data-driven, sophisticated operation and maintenance management system for the entire life cycle of agricultural greenhouses. Figure 1 As shown in Figure 2, the edge computing layer consists of a waveform processor and a dynamic coupling analyzer. After receiving the above data stream, the waveform processor first extracts the characteristic waveform set, and then the dynamic coupling analyzer calculates the time-delay cross-correlation, and then generates the drive-response disconnection degree index ( indicator); the The indicator is updated regularly in a 5-minute loop. When its value does not exceed the threshold, it enters the normal processing flow. When the threshold is exceeded, the alt branch switches to high-frequency sampling mode to enhance monitoring sensitivity. The edge computing layer also The indicators are passed to the central controller in the decision-making layer, which queries the matching pattern, returns the health baseline, and cooperates with the knowledge base to perform further judgment and control.

[0055] like Figure 2As shown in the figure, the horizontal axis is uniformly labeled as time (seconds), and the vertical axis is uniformly labeled as parameter change (normalized). The environmental parameters involved include light intensity, ambient temperature, and air humidity, which are represented by dashed lines, solid lines, and dotted lines in the figure, respectively. In scenario 1: structural damage to the shed, the light intensity, ambient temperature, and air humidity all show a significant downward trend after about 15 seconds, indicating that the external environment directly affects the interior of the shed, typically reflecting the linkage response characteristics caused by structural damage. In scenario 2: abnormal closure of the external shading device, the light intensity drops rapidly after about 10 seconds, while the air humidity shows a significant upward trend, while the ambient temperature does not change significantly. This is a typical scenario in which the internal transpiration process is weakened due to the abnormal closure of the shading device, the humidity rises, but the temperature does not drop significantly. In scenario 3: single sensor failure (taking the humidity sensor as an example), only the air humidity signal fluctuates violently, while the light intensity and ambient temperature signals remain stable. This accurately depicts the abnormal characteristic curve of the humidity sensor when it fails without external interference.

[0056] like Figure 3 As shown, first, at the data acquisition layer, the system obtains the corresponding response parameters and driving parameter data through the response parameter sensor unit (modular humidity probe) and the driving parameter sensor unit (light intensity), and sends them to the edge data processing unit for analysis. In the edge processing stage, the system performs feature waveform extraction, and the extracted content includes average value, peak count, and rise time, which are used to construct the basic feature waveform. Then, the time-lag cross-correlation calculation is performed, including the two key indicators of maximum correlation coefficient and optimal time lag, to evaluate the dynamic coupling relationship between the driving parameters and the response parameters. Based on the above calculation results, the system further generates a drive-response disconnection degree index , and its calculation formula is , which comprehensively considers synchronization strength and response delay; then, the system enters the central decision and control unit to make response decisions. The unit has three types of response capabilities: one is system vitality warning, when The second is the emergency detection mode (high frequency sampling), when It is triggered when a sudden change occurs, entering the temporal topological fingerprint construction process, and further calling the impact type knowledge base (typical fault classification) for event identification; the third is the spatial vitality heat map (precise spatial intervention), which generates a visual vitality map by parallel analysis of spatial distribution data to achieve refined intervention deployment; in terms of response speed and spatial intervention, this system realizes rapid identification and classification of abnormal situations through joint analysis of the start time of the change, the upward or downward trend, and the duration. The spatial intervention recommendation is based on the two-dimensional regional response map updated daily by the system. By comparing the historical and current status of each region, if a region shows an abnormal response in two consecutive detection cycles, the system will list the region as a spatial unit that needs priority processing and trigger corresponding operation and maintenance recommendations, such as adjusting the irrigation rhythm or strengthening manual inspections. The above mechanism does not rely on external input, and can achieve fine-grained spatial judgment and rapid response only through dynamic analysis of existing data.

[0057] Example 5: In a typical application scenario, this system is deployed in a medium-sized agricultural greenhouse group located in the coastal area of ​​Liaoning Province to support the digital planting and refined operation and maintenance management of spring tomato crops. In view of the characteristics of the region's changeable spring climate, significant temperature difference between day and night, and high humidity environment that easily induces pests and diseases, the system realizes dynamic adaptive multimodal data fusion and process control strategy by constructing a full life cycle operation and maintenance model based on the growing season. In this application scenario, the system's multimodal perception subsystem obtains key data in the greenhouse through environmental sensor nodes and crop status visual acquisition units deployed in the greenhouse. These data include but are not limited to air temperature and humidity, soil temperature and humidity, light intensity, carbon dioxide concentration, wind speed and direction, crop leaf color and curling status, etc. In order to improve the system's sensitivity to signs of sudden diseases, the sampling frequency of the perception unit will be dynamically increased during the period when the temperature and humidity difference between morning and evening is large, thereby forming a high-density data sequence with time correlation; the above-mentioned collected data is first preprocessed in the edge computing node, which integrates a locally deployed lightweight model library and can be used for typical The system quickly matches and judges equipment anomalies with early image features of pests and diseases, and calculates the current physiological load index of the crop and the greenhouse loss index. The physiological load index is obtained by comparing the current state of the crop with the growth baseline model established in the early stage of planting, and is used to measure the degree of deviation between the current crop growth state and the ideal state. The loss index comprehensively evaluates the consistency between various sensor data and the degree of decrease in the confidence of the system reasoning model, and is used to determine whether the current data can form an effective closed-loop support in the system model. At the same time, in the design of the visual acquisition unit, the system preferably adopts a high-definition wide-angle camera module with automatic exposure and infrared fill light functions, which is deployed above the typical crop area in the greenhouse, and combined with a fixed-point shooting mechanism at a fixed time period to realize data collection. The unit uses an embedded edge algorithm module to extract feature values ​​of the color, curling degree, edge clarity and spot features of the crop leaves in the image. The feature extraction process includes: first, image denoising and standardization processing, and then using the preset HSV color space threshold to identify the chromaticity distribution characteristics of the green area of ​​the leaf, and statistically calculate the area and color difference distribution of each unit.Then, the edge morphology change information is extracted through the contour tracking algorithm, and the leaf status index is output in combination with the curling calculation model. The above image feature data is transmitted to the edge processing unit at a fixed period, and the correlation analysis is performed with the environmental parameter data of the same period. It is further used for physiological load index evaluation and automatic learning strategy for the normal fluctuation range in the initial operation stage. For example, the system is set to run around the clock for seven consecutive days after deployment, with a sampling interval of once every thirty seconds. During this stage, the system processes the humidity data obtained by each response probe and the unified driving parameter (light), and respectively counts the response delay time and the response delay time within each hour. The system then calculates the mean and standard deviation of the collected time delay data. After eliminating outliers, it sets the mean as the normal baseline lag time. A normal fluctuation range is defined as a range within 30 percent of the mean. The system uses this range as a reference for determining whether the coupling state has deviated during subsequent operations. This learning process requires no external intervention. All parameter ranges are automatically recorded in the system configuration file and can be manually modified through the user interface based on crop variety differences, ensuring the system's adaptability to a variety of crop scenarios. These are all extended implementations known to those skilled in the art.

[0058] When the physiological load index exceeds the warning threshold set by the system and the loss of connection index is within an acceptable range, the edge node will automatically trigger the first-level operation and maintenance recommendation generation process, which includes: retrieving processing records of similar scenarios from the historical operation and maintenance case database, combining the current state of the crop and the environmental trend prediction results to form a set of recommended operation actions, such as adjusting the opening and closing amplitude of the sunshade, adjusting the irrigation cycle, or recommending manual inspections, etc. The above recommendations are reported by the edge node to the central management platform, and further globally optimized through the crop life cycle model and target yield constraints on the management platform to generate the final scheduling instructions; in another typical scenario, when the loss of connection index exceeds the preset threshold and multiple sensor data show a nonlinear drift trend, such as temperature and humidity data continue to fluctuate abnormally without actual intervention, The system will enter a low-confidence fault-tolerant mode. In this mode, the edge node will no longer trigger routine processing suggestions, but will start an abnormal recovery mechanism based on time series data. This mechanism relies on the greenhouse-crop-season multidimensional model accumulated by the system over a long period of time, extracts the reference state interval under the corresponding conditions in the high-confidence historical window, and combines the currently available data to deduce the most likely real state, and use this as a temporary basis to generate transitional operation and maintenance instructions, such as limiting high-frequency ventilation to avoid potential water vapor imbalance, and requesting on-site personnel to review. It should be pointed out that the switching logic, judgment threshold and response mechanism of the above two operation and maintenance modes can be customized at the policy level according to specific crop varieties, growth stages, regional climatic conditions and other factors, and support joint updates based on rule templates and expert feedback to ensure its good adaptability and maintainability in practical applications.

[0059] Example 6: The system of this example is deployed in an agricultural greenhouse for the cultivation of high-value-added medicinal plants. This application scenario requires the system to be able to accurately quantify and respond to changes in physiological state determined by the specific photoperiod and metabolic rhythm of the plant. The system performs a fourteen-day fully automated offline calibration process during the initial deployment phase. During this period, the system collects global light intensity time series data and air humidity time series data from multiple modular response probes distributed inside the greenhouse at a frequency of once every thirty seconds. The system is used to calculate the time series data for the benchmark lag time. The system will calculate the best time delay time every hour within fourteen days. The data are aggregated into a single dataset and the 10th percentile trimmed mean of the dataset is calculated. This method can effectively eliminate extreme values ​​caused by early deployment or occasional weather disturbances, thereby establishing a statistically robust health status response delay benchmark.

[0060] In the calibration weight factor The system executes a self-optimization program based on maximizing sensitivity. At the end of the calibration period, the system guides the irrigation system to perform a preset, 12-hour micro-reduction in water supply through operation and maintenance instructions to form a mild drought stress. is the step length, to Traverse different value, and use each candidate The drive-response disconnection index of the healthy period and the mild stress period was calculated by The system will eventually automatically select and solidify the Mean and healthy period The ratio of the means reaches the maximum This method ensures that The indicator has the highest sensitivity to early signs of functional decline. Also in this offline calibration phase, the system calculates the Mean of the first derivative of the indicator time series and standard deviation , and set the system shock threshold in a programmed manner , providing a statistically based and reproducible trigger basis for the subsequent emergency detection mode.

[0061] For the construction of the impact type knowledge base, the system provides a human-machine collaborative template entry mode during the offline calibration phase. According to the plan, the operation and maintenance personnel artificially simulate specific faults, including briefly opening the side vents to simulate structural damage to the shed, and using sunshades to quickly block the light sensors to simulate abnormal closure of external shielding equipment. In each simulation, the system is manually triggered to enter the emergency detection mode, sampling at its physical limit frequency, and automatically performing the construction of the time series topology fingerprint. The construction process is as follows: First, the system monitors the high-frequency data streams of light, temperature, and humidity, and calculates its rate of change based on seconds; secondly, when the absolute value of the rate of change of any parameter exceeds three times its standard deviation in the quiet state, the system records an event tuple containing the parameter name, change direction and precise timestamp; finally, after an event cycle, the system sorts all event tuples by timestamp to form a standardized vectorized fingerprint. The operation and maintenance personnel label the fingerprint with a certain semantic label and store it in the knowledge base. After the system is put into online operation, any The unknown fingerprints generated by the automatic triggering of mutations are matched with all the labeled templates in the knowledge base by calculating the cosine similarity, so as to achieve rapid classification and diagnosis of emergencies.

[0062] After the system enters long-term online operation and maintenance, it will continue to The indicator generates and refreshes the space vitality heat map. The system performs an automatic recommendation of space intervention decision based on the map every day. Its judgment logic is: the system compares the current areas The value and its historical average over the past seven days, if a specific area If the value exceeds 1.5 times its historical average for two consecutive monitoring periods, the system automatically marks the area as potentially low-activity and generates a specific maintenance instruction, instructing inspection or adjustment of that specific area. This process, through data-driven, refined spatial analysis, transforms system maintenance from reactive response to proactive prediction.

[0063] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A data-driven, sophisticated operation and maintenance management system for the entire life cycle of agricultural greenhouses, characterized by: The system comprises: a driving parameter sensor unit for collecting driving parameter time series data; At least two response parameter sensor units, used to collect time series data of response parameters of each spatial area; an edge data processing unit, connected to the sensor unit, configured to: generate a characteristic waveform based on the time series data to determine a drive-response disconnection degree index for each spatial region; instantaneously increase the data sampling frequency when the rate of change of the index exceeds a system shock threshold, and construct a time series topology fingerprint based on the high-frequency data, the time series topology fingerprint including the relative order and direction of the drastic changes in each parameter; a central decision-making and control unit, connected to the edge data processing unit, and configured to: generate a predictive warning of functional vitality based on the long-term evolution trend of the drive-response disconnection degree indicator, and generate an emergency warning or emergency management control instruction for a specific fault source based on the temporal topological fingerprint; The edge data processing unit is configured such that the calculation result of the time-delay cross-correlation includes a maximum correlation coefficient and an optimal time-delay time, and the drive-response disconnection degree index is calculated based on the maximum correlation coefficient and the optimal time-delay time; Drive-response disconnection index Calculated according to the following formula: ， in, represents the maximum correlation coefficient, represents the optimal lag time, is the weight factor, It is the benchmark time lag when the agricultural greenhouse system is in a healthy state.

2. The data-driven, refined operation and maintenance management system for the entire life cycle of agricultural greenhouses according to claim 1 is characterized in that: The driving parameter represents the light intensity, and the response parameter represents the air humidity.

3. The data-driven, refined operation and maintenance management system for the entire life cycle of agricultural greenhouses according to claim 1 is characterized in that: The edge data processing unit is configured to generate a driving characteristic waveform and a response characteristic waveform within a preset sliding time window, and the generation method includes calculating the arithmetic mean, peak count and average rise time from trough to peak of the timing data.

4. The data-driven, refined operation and maintenance management system for the entire life cycle of agricultural greenhouses according to claim 1 is characterized in that: The central decision and control unit is configured to automatically learn and calibrate the normal fluctuation range of the drive-response loss degree indicator of the agricultural greenhouse in a healthy state by collecting data for more than one week during the initial operation stage of the system.

5. The data-driven, refined operation and maintenance management system for the entire life cycle of agricultural greenhouses according to claim 1 is characterized in that: The edge data processing unit is configured to switch to an emergency detection mode when the rate of change of the drive-response disconnection degree indicator exceeds the system shock threshold. The emergency detection mode includes increasing the data sampling frequency of the drive parameter sensor unit and at least two response parameter sensor units to their physical limit frequency, and recording high-frequency time series data.

6. The data-driven, refined operation and maintenance management system for the entire life cycle of agricultural greenhouses according to claim 1, characterized in that: The edge data processing unit is configured to construct a temporal topological fingerprint based on the temporal changes of driving parameters, response parameters and other environmental parameters at high sampling frequency within a preset short time window after high sampling frequency data acquisition, instead of performing complex numerical calculations.

7. A data-driven, sophisticated operation and maintenance management system for the entire life cycle of agricultural greenhouses according to claim 6, characterized in that: The central decision and control unit is configured to have a built-in shock type knowledge base for fault classification and response based on temporal topological fingerprints. The shock type knowledge base includes: topological fingerprints in which humidity drops sharply, temperature drops sharply, and light remains stable, indicating structural damage to the shed; topological fingerprints in which humidity rises sharply, temperature changes insignificantly, and light drops sharply, indicating abnormal closure of the external shielding device; and topological fingerprints in which only a single sensor data changes dramatically while the other two sensor data remain stable, indicating a fault in the corresponding sensor itself.

8. The data-driven, refined operation and maintenance management system for the entire life cycle of agricultural greenhouses according to claim 1, characterized in that: Each of the at least two response parameter sensor units is a modular response probe, which only includes a humidity sensing element and a microcontroller for wireless data transmission in hardware. The modular response probe is configured to periodically measure the air humidity of the spatial area where it is located and wirelessly send the air humidity data to the central decision and control unit.

Images